Upload folder using huggingface_hub

.gitattributes

@@ -42,3 +42,5 @@ video_mllm_swift/s1_declip_siglip2_qwen3_1.7b/v0-20260314-141147/checkpoint-2181
 video_mllm_swift/s1_siglip2_qwen3_1.7b/v11-20260314-090153/checkpoint-2000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 video_mllm_swift/s1_siglip2_qwen3_1.7b/v11-20260314-090153/checkpoint-2181/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 video_mllm_swift/s2_declip_siglip2_qwen3_1.7b_10pct/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text

video_mllm_swift/s2_image_only_10pct/v0-20260316-082051/args.json

@@ -0,0 +1,376 @@
+{
+  "output_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v0-20260316-082051",
+  "per_device_train_batch_size": 1,
+  "num_train_epochs": 3.0,
+  "max_steps": 500,
+  "learning_rate": 1e-05,
+  "lr_scheduler_type": "cosine",
+  "lr_scheduler_kwargs": null,
+  "warmup_steps": 0,
+  "optim": "adamw_torch_fused",
+  "optim_args": null,
+  "weight_decay": 0.1,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_epsilon": 1e-08,
+  "optim_target_modules": null,
+  "gradient_accumulation_steps": 8,
+  "average_tokens_across_devices": true,
+  "max_grad_norm": 1.0,
+  "label_smoothing_factor": 0.0,
+  "bf16": true,
+  "fp16": false,
+  "bf16_full_eval": false,
+  "fp16_full_eval": false,
+  "tf32": null,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": "{\"use_reentrant\": false}",
+  "torch_compile": false,
+  "torch_compile_backend": null,
+  "torch_compile_mode": null,
+  "use_liger_kernel": false,
+  "liger_kernel_config": null,
+  "use_cache": false,
+  "neftune_noise_alpha": null,
+  "torch_empty_cache_steps": null,
+  "auto_find_batch_size": false,
+  "logging_strategy": "steps",
+  "logging_steps": 1,
+  "logging_first_step": true,
+  "log_on_each_node": true,
+  "logging_nan_inf_filter": true,
+  "include_num_input_tokens_seen": false,
+  "log_level": "passive",
+  "log_level_replica": "warning",
+  "disable_tqdm": null,
+  "report_to": [
+    "none"
+  ],
+  "run_name": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v0-20260316-082051",
+  "project": "huggingface",
+  "trackio_space_id": "trackio",
+  "eval_strategy": "no",
+  "eval_steps": 100.0,
+  "eval_delay": 0,
+  "per_device_eval_batch_size": 1,
+  "prediction_loss_only": false,
+  "eval_on_start": false,
+  "eval_do_concat_batches": true,
+  "eval_use_gather_object": false,
+  "eval_accumulation_steps": null,
+  "include_for_metrics": [],
+  "batch_eval_metrics": false,
+  "save_only_model": false,
+  "save_strategy": "steps",
+  "save_steps": 100.0,
+  "save_on_each_node": false,
+  "save_total_limit": 2,
+  "enable_jit_checkpoint": false,
+  "push_to_hub": false,
+  "hub_token": null,
+  "hub_private_repo": null,
+  "hub_model_id": null,
+  "hub_strategy": "every_save",
+  "hub_always_push": false,
+  "hub_revision": null,
+  "load_best_model_at_end": false,
+  "metric_for_best_model": "loss",
+  "greater_is_better": false,
+  "ignore_data_skip": false,
+  "restore_callback_states_from_checkpoint": false,
+  "full_determinism": false,
+  "seed": 42,
+  "data_seed": 42,
+  "use_cpu": false,
+  "accelerator_config": "{\"dispatch_batches\": false}",
+  "parallelism_config": null,
+  "dataloader_drop_last": false,
+  "dataloader_num_workers": 4,
+  "dataloader_pin_memory": true,
+  "dataloader_persistent_workers": false,
+  "dataloader_prefetch_factor": 4,
+  "remove_unused_columns": true,
+  "label_names": null,
+  "train_sampling_strategy": "random",
+  "length_column_name": "length",
+  "ddp_find_unused_parameters": null,
+  "ddp_bucket_cap_mb": null,
+  "ddp_broadcast_buffers": null,
+  "ddp_backend": null,
+  "ddp_timeout": 36000,
+  "fsdp": [],
+  "fsdp_config": null,
+  "deepspeed": {
+    "fp16": {
+      "enabled": "auto",
+      "loss_scale": 0,
+      "loss_scale_window": 1000,
+      "initial_scale_power": 16,
+      "hysteresis": 2,
+      "min_loss_scale": 1
+    },
+    "bf16": {
+      "enabled": "auto"
+    },
+    "zero_optimization": {
+      "stage": 2,
+      "offload_optimizer": {
+        "device": "none",
+        "pin_memory": true
+      },
+      "allgather_partitions": true,
+      "allgather_bucket_size": 200000000.0,
+      "overlap_comm": false,
+      "reduce_scatter": true,
+      "reduce_bucket_size": 200000000.0,
+      "contiguous_gradients": true
+    },
+    "gradient_accumulation_steps": "auto",
+    "gradient_clipping": "auto",
+    "steps_per_print": 2000,
+    "train_batch_size": "auto",
+    "train_micro_batch_size_per_gpu": "auto",
+    "wall_clock_breakdown": false
+  },
+  "debug": null,
+  "skip_memory_metrics": true,
+  "do_train": false,
+  "do_eval": false,
+  "do_predict": false,
+  "resume_from_checkpoint": null,
+  "warmup_ratio": 0.05,
+  "logging_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v0-20260316-082051/runs",
+  "local_rank": 0,
+  "sortish_sampler": false,
+  "predict_with_generate": false,
+  "generation_max_length": null,
+  "generation_num_beams": null,
+  "generation_config": null,
+  "tuner_backend": "peft",
+  "vit_gradient_checkpointing": null,
+  "router_aux_loss_coef": 0.0,
+  "enable_dft_loss": false,
+  "enable_channel_loss": false,
+  "safe_serialization": true,
+  "max_shard_size": "5GB",
+  "check_model": true,
+  "acc_strategy": "token",
+  "train_dataloader_shuffle": true,
+  "group_by_length": false,
+  "max_epochs": null,
+  "aligner_lr": null,
+  "vit_lr": 1e-06,
+  "use_logits_to_keep": null,
+  "ds3_gather_for_generation": true,
+  "resume_only_model": false,
+  "optimizer": null,
+  "loss_type": null,
+  "eval_metric": null,
+  "callbacks": [],
+  "early_stop_interval": null,
+  "eval_use_evalscope": false,
+  "eval_dataset": [],
+  "eval_dataset_args": null,
+  "eval_limit": null,
+  "eval_generation_config": null,
+  "extra_eval_args": null,
+  "tuner_type": "full",
+  "use_galore": false,
+  "galore_target_modules": null,
+  "galore_rank": 128,
+  "galore_update_proj_gap": 50,
+  "galore_scale": 1.0,
+  "galore_proj_type": "std",
+  "galore_optim_per_parameter": false,
+  "galore_with_embedding": false,
+  "galore_quantization": false,
+  "galore_proj_quant": false,
+  "galore_proj_bits": 4,
+  "galore_proj_group_size": 256,
+  "galore_cos_threshold": 0.4,
+  "galore_gamma_proj": 2,
+  "galore_queue_size": 5,
+  "lisa_activated_layers": 0,
+  "lisa_step_interval": 20,
+  "use_flash_ckpt": false,
+  "use_ray": false,
+  "ray_exp_name": null,
+  "device_groups": null,
+  "model": "/opt/tiger/model_cache/checkpoint-2181",
+  "model_type": "llava_siglip2_qwen3",
+  "model_revision": null,
+  "task_type": "causal_lm",
+  "torch_dtype": "bfloat16",
+  "attn_impl": "flash_attn",
+  "experts_impl": null,
+  "new_special_tokens": [],
+  "num_labels": null,
+  "problem_type": null,
+  "rope_scaling": null,
+  "device_map": null,
+  "max_memory": {},
+  "max_model_len": null,
+  "local_repo_path": null,
+  "init_strategy": null,
+  "template": "llava_siglip2_qwen3",
+  "system": null,
+  "max_length": 16384,
+  "truncation_strategy": "delete",
+  "max_pixels": null,
+  "agent_template": null,
+  "norm_bbox": null,
+  "use_chat_template": true,
+  "padding_side": "right",
+  "padding_free": true,
+  "loss_scale": "default",
+  "sequence_parallel_size": 1,
+  "template_backend": "swift",
+  "response_prefix": null,
+  "enable_thinking": null,
+  "add_non_thinking_prefix": true,
+  "dataset": [
+    "vmllm_s2_image_10pct"
+  ],
+  "val_dataset": [],
+  "cached_dataset": [],
+  "cached_val_dataset": [],
+  "split_dataset_ratio": 0.0,
+  "dataset_num_proc": 64,
+  "load_from_cache_file": true,
+  "dataset_shuffle": true,
+  "val_dataset_shuffle": false,
+  "streaming": false,
+  "interleave_prob": null,
+  "stopping_strategy": "first_exhausted",
+  "shuffle_buffer_size": 1000,
+  "download_mode": "reuse_dataset_if_exists",
+  "columns": {},
+  "strict": false,
+  "model_name": null,
+  "model_author": null,
+  "custom_dataset_info": [],
+  "quant_method": null,
+  "quant_bits": null,
+  "hqq_axis": null,
+  "bnb_4bit_compute_dtype": "bfloat16",
+  "bnb_4bit_quant_type": "nf4",
+  "bnb_4bit_use_double_quant": true,
+  "bnb_4bit_quant_storage": null,
+  "max_new_tokens": 64,
+  "temperature": 0.0,
+  "top_k": null,
+  "top_p": null,
+  "repetition_penalty": null,
+  "num_beams": 1,
+  "stream": false,
+  "stop_words": [],
+  "logprobs": false,
+  "top_logprobs": null,
+  "structured_outputs_regex": null,
+  "train_type": null,
+  "adapters": [],
+  "external_plugins": [
+    "video_mllm/model_plugin.py",
+    "video_mllm/dataset_plugin.py"
+  ],
+  "custom_register_path": [],
+  "model_kwargs": {},
+  "load_args": false,
+  "load_data_args": false,
+  "packing": true,
+  "packing_length": 16384,
+  "packing_num_proc": 1,
+  "lazy_tokenize": false,
+  "use_hf": true,
+  "ignore_args_error": false,
+  "use_swift_lora": false,
+  "freeze_parameters": [],
+  "freeze_parameters_regex": null,
+  "freeze_parameters_ratio": 0.0,
+  "trainable_parameters": [
+    "model.multi_modal_projector"
+  ],
+  "trainable_parameters_regex": null,
+  "freeze_llm": false,
+  "freeze_vit": false,
+  "freeze_aligner": false,
+  "target_modules": [
+    "all-linear"
+  ],
+  "target_regex": null,
+  "target_parameters": null,
+  "modules_to_save": [],
+  "lora_rank": 8,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "lora_bias": "none",
+  "lora_dtype": null,
+  "lorap_lr_ratio": null,
+  "use_rslora": false,
+  "use_dora": false,
+  "lora_ga_batch_size": 2,
+  "lora_ga_iters": 2,
+  "lora_ga_max_length": 1024,
+  "lora_ga_direction": "ArB2r",
+  "lora_ga_scale": "stable",
+  "lora_ga_stable_gamma": 16,
+  "init_weights": true,
+  "fourier_n_frequency": 2000,
+  "fourier_scaling": 300.0,
+  "boft_block_size": 4,
+  "boft_block_num": 0,
+  "boft_n_butterfly_factor": 1,
+  "boft_dropout": 0.0,
+  "vera_rank": 256,
+  "vera_projection_prng_key": 0,
+  "vera_dropout": 0.0,
+  "vera_d_initial": 0.1,
+  "adapter_act": "gelu",
+  "adapter_length": 128,
+  "adalora_target_r": 8,
+  "adalora_init_r": 12,
+  "adalora_tinit": 0,
+  "adalora_tfinal": 0,
+  "adalora_deltaT": 1,
+  "adalora_beta1": 0.85,
+  "adalora_beta2": 0.85,
+  "adalora_orth_reg_weight": 0.5,
+  "llamapro_num_new_blocks": 4,
+  "llamapro_num_groups": null,
+  "reft_layer_key": null,
+  "reft_layers": null,
+  "reft_rank": 4,
+  "reft_intervention_type": "LoreftIntervention",
+  "reft_args": null,
+  "swanlab_token": null,
+  "swanlab_project": "ms-swift",
+  "swanlab_workspace": null,
+  "swanlab_exp_name": null,
+  "swanlab_notification_method": null,
+  "swanlab_webhook_url": null,
+  "swanlab_secret": null,
+  "swanlab_sender_email": null,
+  "swanlab_receiver_email": null,
+  "swanlab_smtp_server": null,
+  "swanlab_smtp_port": null,
+  "swanlab_email_language": "zh",
+  "swanlab_mode": "cloud",
+  "add_version": true,
+  "create_checkpoint_symlink": false,
+  "zero_hpz_partition_size": null,
+  "deepspeed_autotp_size": null,
+  "swift_version": "4.1.0.dev0",
+  "ckpt_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "rank": 0,
+  "global_world_size": 8,
+  "local_world_size": 8,
+  "model_suffix": "checkpoint-2181",
+  "model_info": "ModelInfo(model_type='llava_siglip2_qwen3', model_dir='/opt/tiger/model_cache/checkpoint-2181', torch_dtype=torch.bfloat16, max_model_len=40960, quant_method=None, quant_bits=None, rope_scaling=None, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
+  "model_meta": "ModelMeta(model_type='llava_siglip2_qwen3', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-0.6B', hf_model_id='Qwen/Qwen3-0.6B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-1.7B', hf_model_id='Qwen/Qwen3-1.7B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-4B', hf_model_id='Qwen/Qwen3-4B', model_path=None, ms_revision=None, hf_revision=None)], template=None, ignore_patterns=None, requires=None, tags=[])], loader=<class 'model_plugin.SigLIP2Qwen3Loader'>, template='llava_siglip2_qwen3', model_arch=MultiModelKeys(arch_name='llava_hf', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['model.language_model', 'lm_head'], aligner=['model.multi_modal_projector'], vision_tower=['model.vision_tower'], generator=[]), architectures=['LlavaOnevisionForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=[], tags=['vision', 'video'])",
+  "model_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "template_meta": "QwenTemplateMeta(template_type='llava_siglip2_qwen3', prefix=[], prompt=['<|im_start|>user\\n{{QUERY}}<|im_end|>\\n<|im_start|>assistant\\n'], chat_sep=['<|im_end|>\\n'], suffix=['<|im_end|>\\n'], template_cls=<class 'model_plugin.SigLIP2LlavaTemplate'>, system_prefix=['<|im_start|>system\\n{{SYSTEM}}<|im_end|>\\n'], default_system=None, auto_add_bos=False, stop_words=['<|endoftext|>'], agent_template='hermes', is_thinking=False, thinking_prefix='', non_thinking_prefix='', history_thinking_prefix='')",
+  "_val_dataset_exists": false,
+  "hub": "<class 'swift.hub.hub.HFHub'>",
+  "evaluation_strategy": "steps",
+  "training_args": "Seq2SeqTrainingArguments(output_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v0-20260316-082051', per_device_train_batch_size=1, num_train_epochs=3.0, max_steps=500, learning_rate=1e-05, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_steps=0.05, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, optim_target_modules=None, gradient_accumulation_steps=8, average_tokens_across_devices=None, max_grad_norm=1.0, label_smoothing_factor=0.0, bf16=True, fp16=False, bf16_full_eval=False, fp16_full_eval=False, tf32=None, gradient_checkpointing=True, gradient_checkpointing_kwargs={'use_reentrant': False}, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, use_liger_kernel=False, liger_kernel_config=None, use_cache=False, neftune_noise_alpha=None, torch_empty_cache_steps=None, auto_find_batch_size=False, logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_steps=1, logging_first_step=True, log_on_each_node=True, logging_nan_inf_filter=True, include_num_input_tokens_seen=None, log_level='passive', log_level_replica='warning', disable_tqdm=False, report_to=[], run_name='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v0-20260316-082051', project='huggingface', trackio_space_id='trackio', eval_strategy=<IntervalStrategy.NO: 'no'>, eval_steps=100.0, eval_delay=0, per_device_eval_batch_size=1, prediction_loss_only=False, eval_on_start=False, eval_do_concat_batches=True, eval_use_gather_object=False, eval_accumulation_steps=None, include_for_metrics=[], batch_eval_metrics=False, save_only_model=False, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=100, save_on_each_node=False, save_total_limit=2, enable_jit_checkpoint=False, push_to_hub=False, hub_token=None, hub_private_repo=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_always_push=False, hub_revision=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, restore_callback_states_from_checkpoint=False, full_determinism=False, seed=42, data_seed=42, use_cpu=False, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), parallelism_config=None, dataloader_drop_last=False, dataloader_num_workers=4, dataloader_pin_memory=True, dataloader_persistent_workers=False, dataloader_prefetch_factor=4, remove_unused_columns=False, label_names=None, train_sampling_strategy='random', length_column_name='length', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, ddp_backend=None, ddp_timeout=36000, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 2, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'allgather_partitions': True, 'allgather_bucket_size': 200000000.0, 'overlap_comm': False, 'reduce_scatter': True, 'reduce_bucket_size': 200000000.0, 'contiguous_gradients': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, debug=[], skip_memory_metrics=True, do_train=False, do_eval=False, do_predict=False, resume_from_checkpoint=None, warmup_ratio=0.05, logging_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v0-20260316-082051/runs', local_rank=0, sortish_sampler=False, predict_with_generate=False, generation_max_length=None, generation_num_beams=None, generation_config=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, safe_serialization=True, max_shard_size='5GB', check_model=True, acc_strategy='token', train_dataloader_shuffle=True, group_by_length=False, max_epochs=None, aligner_lr=None, vit_lr=1e-06, use_logits_to_keep=None, ds3_gather_for_generation=True, resume_only_model=False, optimizer='multimodal', loss_type=None, eval_metric=None, callbacks=[], early_stop_interval=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, tuner_type='full', use_galore=False, galore_target_modules=None, galore_rank=128, galore_update_proj_gap=50, galore_scale=1.0, galore_proj_type='std', galore_optim_per_parameter=False, galore_with_embedding=False, galore_quantization=False, galore_proj_quant=False, galore_proj_bits=4, galore_proj_group_size=256, galore_cos_threshold=0.4, galore_gamma_proj=2, galore_queue_size=5, lisa_activated_layers=0, lisa_step_interval=20, use_flash_ckpt=False)"
+}

video_mllm_swift/s2_image_only_10pct/v0-20260316-082051/logging.jsonl

@@ -0,0 +1 @@
+{"train_dataset": "15834.342530±556.264794, min=14612.000000, max=16384.000000, size=14860", "model_parameter_info": "LlavaOnevisionForConditionalGeneration: 2436.0474M Params (2436.0474M Trainable [100.0000%]), 0.0009M Buffers.", "last_model_checkpoint": null, "best_model_checkpoint": null, "best_metric": null, "global_step": 0, "log_history": [], "memory": null}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/args.json

@@ -0,0 +1,376 @@
+{
+  "output_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215",
+  "per_device_train_batch_size": 1,
+  "num_train_epochs": 3.0,
+  "max_steps": 500,
+  "learning_rate": 1e-05,
+  "lr_scheduler_type": "cosine",
+  "lr_scheduler_kwargs": null,
+  "warmup_steps": 0,
+  "optim": "adamw_torch_fused",
+  "optim_args": null,
+  "weight_decay": 0.1,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_epsilon": 1e-08,
+  "optim_target_modules": null,
+  "gradient_accumulation_steps": 8,
+  "average_tokens_across_devices": true,
+  "max_grad_norm": 1.0,
+  "label_smoothing_factor": 0.0,
+  "bf16": true,
+  "fp16": false,
+  "bf16_full_eval": false,
+  "fp16_full_eval": false,
+  "tf32": null,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": "{\"use_reentrant\": false}",
+  "torch_compile": false,
+  "torch_compile_backend": null,
+  "torch_compile_mode": null,
+  "use_liger_kernel": false,
+  "liger_kernel_config": null,
+  "use_cache": false,
+  "neftune_noise_alpha": null,
+  "torch_empty_cache_steps": null,
+  "auto_find_batch_size": false,
+  "logging_strategy": "steps",
+  "logging_steps": 1,
+  "logging_first_step": true,
+  "log_on_each_node": true,
+  "logging_nan_inf_filter": true,
+  "include_num_input_tokens_seen": false,
+  "log_level": "passive",
+  "log_level_replica": "warning",
+  "disable_tqdm": null,
+  "report_to": [
+    "none"
+  ],
+  "run_name": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215",
+  "project": "huggingface",
+  "trackio_space_id": "trackio",
+  "eval_strategy": "no",
+  "eval_steps": 100.0,
+  "eval_delay": 0,
+  "per_device_eval_batch_size": 1,
+  "prediction_loss_only": false,
+  "eval_on_start": false,
+  "eval_do_concat_batches": true,
+  "eval_use_gather_object": false,
+  "eval_accumulation_steps": null,
+  "include_for_metrics": [],
+  "batch_eval_metrics": false,
+  "save_only_model": false,
+  "save_strategy": "steps",
+  "save_steps": 100.0,
+  "save_on_each_node": false,
+  "save_total_limit": 2,
+  "enable_jit_checkpoint": false,
+  "push_to_hub": false,
+  "hub_token": null,
+  "hub_private_repo": null,
+  "hub_model_id": null,
+  "hub_strategy": "every_save",
+  "hub_always_push": false,
+  "hub_revision": null,
+  "load_best_model_at_end": false,
+  "metric_for_best_model": "loss",
+  "greater_is_better": false,
+  "ignore_data_skip": false,
+  "restore_callback_states_from_checkpoint": false,
+  "full_determinism": false,
+  "seed": 42,
+  "data_seed": 42,
+  "use_cpu": false,
+  "accelerator_config": "{\"dispatch_batches\": false}",
+  "parallelism_config": null,
+  "dataloader_drop_last": false,
+  "dataloader_num_workers": 4,
+  "dataloader_pin_memory": true,
+  "dataloader_persistent_workers": false,
+  "dataloader_prefetch_factor": 4,
+  "remove_unused_columns": true,
+  "label_names": null,
+  "train_sampling_strategy": "random",
+  "length_column_name": "length",
+  "ddp_find_unused_parameters": null,
+  "ddp_bucket_cap_mb": null,
+  "ddp_broadcast_buffers": null,
+  "ddp_backend": null,
+  "ddp_timeout": 7200,
+  "fsdp": [],
+  "fsdp_config": null,
+  "deepspeed": {
+    "fp16": {
+      "enabled": "auto",
+      "loss_scale": 0,
+      "loss_scale_window": 1000,
+      "initial_scale_power": 16,
+      "hysteresis": 2,
+      "min_loss_scale": 1
+    },
+    "bf16": {
+      "enabled": "auto"
+    },
+    "zero_optimization": {
+      "stage": 2,
+      "offload_optimizer": {
+        "device": "none",
+        "pin_memory": true
+      },
+      "allgather_partitions": true,
+      "allgather_bucket_size": 200000000.0,
+      "overlap_comm": false,
+      "reduce_scatter": true,
+      "reduce_bucket_size": 200000000.0,
+      "contiguous_gradients": true
+    },
+    "gradient_accumulation_steps": "auto",
+    "gradient_clipping": "auto",
+    "steps_per_print": 2000,
+    "train_batch_size": "auto",
+    "train_micro_batch_size_per_gpu": "auto",
+    "wall_clock_breakdown": false
+  },
+  "debug": null,
+  "skip_memory_metrics": true,
+  "do_train": false,
+  "do_eval": false,
+  "do_predict": false,
+  "resume_from_checkpoint": null,
+  "warmup_ratio": 0.05,
+  "logging_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/runs",
+  "local_rank": 0,
+  "sortish_sampler": false,
+  "predict_with_generate": false,
+  "generation_max_length": null,
+  "generation_num_beams": null,
+  "generation_config": null,
+  "tuner_backend": "peft",
+  "vit_gradient_checkpointing": null,
+  "router_aux_loss_coef": 0.0,
+  "enable_dft_loss": false,
+  "enable_channel_loss": false,
+  "safe_serialization": true,
+  "max_shard_size": "5GB",
+  "check_model": true,
+  "acc_strategy": "token",
+  "train_dataloader_shuffle": true,
+  "group_by_length": false,
+  "max_epochs": null,
+  "aligner_lr": null,
+  "vit_lr": 1e-06,
+  "use_logits_to_keep": null,
+  "ds3_gather_for_generation": true,
+  "resume_only_model": false,
+  "optimizer": null,
+  "loss_type": null,
+  "eval_metric": null,
+  "callbacks": [],
+  "early_stop_interval": null,
+  "eval_use_evalscope": false,
+  "eval_dataset": [],
+  "eval_dataset_args": null,
+  "eval_limit": null,
+  "eval_generation_config": null,
+  "extra_eval_args": null,
+  "tuner_type": "full",
+  "use_galore": false,
+  "galore_target_modules": null,
+  "galore_rank": 128,
+  "galore_update_proj_gap": 50,
+  "galore_scale": 1.0,
+  "galore_proj_type": "std",
+  "galore_optim_per_parameter": false,
+  "galore_with_embedding": false,
+  "galore_quantization": false,
+  "galore_proj_quant": false,
+  "galore_proj_bits": 4,
+  "galore_proj_group_size": 256,
+  "galore_cos_threshold": 0.4,
+  "galore_gamma_proj": 2,
+  "galore_queue_size": 5,
+  "lisa_activated_layers": 0,
+  "lisa_step_interval": 20,
+  "use_flash_ckpt": false,
+  "use_ray": false,
+  "ray_exp_name": null,
+  "device_groups": null,
+  "model": "/opt/tiger/model_cache/checkpoint-2181",
+  "model_type": "llava_siglip2_qwen3",
+  "model_revision": null,
+  "task_type": "causal_lm",
+  "torch_dtype": "bfloat16",
+  "attn_impl": "flash_attn",
+  "experts_impl": null,
+  "new_special_tokens": [],
+  "num_labels": null,
+  "problem_type": null,
+  "rope_scaling": null,
+  "device_map": null,
+  "max_memory": {},
+  "max_model_len": null,
+  "local_repo_path": null,
+  "init_strategy": null,
+  "template": "llava_siglip2_qwen3",
+  "system": null,
+  "max_length": 16384,
+  "truncation_strategy": "delete",
+  "max_pixels": null,
+  "agent_template": null,
+  "norm_bbox": null,
+  "use_chat_template": true,
+  "padding_side": "right",
+  "padding_free": true,
+  "loss_scale": "default",
+  "sequence_parallel_size": 1,
+  "template_backend": "swift",
+  "response_prefix": null,
+  "enable_thinking": null,
+  "add_non_thinking_prefix": true,
+  "dataset": [],
+  "val_dataset": [],
+  "cached_dataset": [
+    "/mnt/bn/strategy-mllm-train/common/datasets/vmllm_cached/image_10pct/train"
+  ],
+  "cached_val_dataset": [],
+  "split_dataset_ratio": 0.0,
+  "dataset_num_proc": 1,
+  "load_from_cache_file": false,
+  "dataset_shuffle": true,
+  "val_dataset_shuffle": false,
+  "streaming": false,
+  "interleave_prob": null,
+  "stopping_strategy": "first_exhausted",
+  "shuffle_buffer_size": 1000,
+  "download_mode": "reuse_dataset_if_exists",
+  "columns": {},
+  "strict": false,
+  "model_name": null,
+  "model_author": null,
+  "custom_dataset_info": [],
+  "quant_method": null,
+  "quant_bits": null,
+  "hqq_axis": null,
+  "bnb_4bit_compute_dtype": "bfloat16",
+  "bnb_4bit_quant_type": "nf4",
+  "bnb_4bit_use_double_quant": true,
+  "bnb_4bit_quant_storage": null,
+  "max_new_tokens": 64,
+  "temperature": 0.0,
+  "top_k": null,
+  "top_p": null,
+  "repetition_penalty": null,
+  "num_beams": 1,
+  "stream": false,
+  "stop_words": [],
+  "logprobs": false,
+  "top_logprobs": null,
+  "structured_outputs_regex": null,
+  "train_type": null,
+  "adapters": [],
+  "external_plugins": [
+    "video_mllm/model_plugin.py",
+    "video_mllm/dataset_plugin.py"
+  ],
+  "custom_register_path": [],
+  "model_kwargs": {},
+  "load_args": false,
+  "load_data_args": false,
+  "packing": true,
+  "packing_length": 16384,
+  "packing_num_proc": 1,
+  "lazy_tokenize": false,
+  "use_hf": true,
+  "ignore_args_error": false,
+  "use_swift_lora": false,
+  "freeze_parameters": [],
+  "freeze_parameters_regex": null,
+  "freeze_parameters_ratio": 0.0,
+  "trainable_parameters": [
+    "model.multi_modal_projector"
+  ],
+  "trainable_parameters_regex": null,
+  "freeze_llm": false,
+  "freeze_vit": false,
+  "freeze_aligner": false,
+  "target_modules": [
+    "all-linear"
+  ],
+  "target_regex": null,
+  "target_parameters": null,
+  "modules_to_save": [],
+  "lora_rank": 8,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "lora_bias": "none",
+  "lora_dtype": null,
+  "lorap_lr_ratio": null,
+  "use_rslora": false,
+  "use_dora": false,
+  "lora_ga_batch_size": 2,
+  "lora_ga_iters": 2,
+  "lora_ga_max_length": 1024,
+  "lora_ga_direction": "ArB2r",
+  "lora_ga_scale": "stable",
+  "lora_ga_stable_gamma": 16,
+  "init_weights": true,
+  "fourier_n_frequency": 2000,
+  "fourier_scaling": 300.0,
+  "boft_block_size": 4,
+  "boft_block_num": 0,
+  "boft_n_butterfly_factor": 1,
+  "boft_dropout": 0.0,
+  "vera_rank": 256,
+  "vera_projection_prng_key": 0,
+  "vera_dropout": 0.0,
+  "vera_d_initial": 0.1,
+  "adapter_act": "gelu",
+  "adapter_length": 128,
+  "adalora_target_r": 8,
+  "adalora_init_r": 12,
+  "adalora_tinit": 0,
+  "adalora_tfinal": 0,
+  "adalora_deltaT": 1,
+  "adalora_beta1": 0.85,
+  "adalora_beta2": 0.85,
+  "adalora_orth_reg_weight": 0.5,
+  "llamapro_num_new_blocks": 4,
+  "llamapro_num_groups": null,
+  "reft_layer_key": null,
+  "reft_layers": null,
+  "reft_rank": 4,
+  "reft_intervention_type": "LoreftIntervention",
+  "reft_args": null,
+  "swanlab_token": null,
+  "swanlab_project": "ms-swift",
+  "swanlab_workspace": null,
+  "swanlab_exp_name": null,
+  "swanlab_notification_method": null,
+  "swanlab_webhook_url": null,
+  "swanlab_secret": null,
+  "swanlab_sender_email": null,
+  "swanlab_receiver_email": null,
+  "swanlab_smtp_server": null,
+  "swanlab_smtp_port": null,
+  "swanlab_email_language": "zh",
+  "swanlab_mode": "cloud",
+  "add_version": true,
+  "create_checkpoint_symlink": false,
+  "zero_hpz_partition_size": null,
+  "deepspeed_autotp_size": null,
+  "swift_version": "4.1.0.dev0",
+  "ckpt_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "rank": 0,
+  "global_world_size": 8,
+  "local_world_size": 8,
+  "model_suffix": "checkpoint-2181",
+  "model_info": "ModelInfo(model_type='llava_siglip2_qwen3', model_dir='/opt/tiger/model_cache/checkpoint-2181', torch_dtype=torch.bfloat16, max_model_len=40960, quant_method=None, quant_bits=None, rope_scaling=None, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
+  "model_meta": "ModelMeta(model_type='llava_siglip2_qwen3', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-0.6B', hf_model_id='Qwen/Qwen3-0.6B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-1.7B', hf_model_id='Qwen/Qwen3-1.7B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-4B', hf_model_id='Qwen/Qwen3-4B', model_path=None, ms_revision=None, hf_revision=None)], template=None, ignore_patterns=None, requires=None, tags=[])], loader=<class 'model_plugin.SigLIP2Qwen3Loader'>, template='llava_siglip2_qwen3', model_arch=MultiModelKeys(arch_name='llava_hf', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['model.language_model', 'lm_head'], aligner=['model.multi_modal_projector'], vision_tower=['model.vision_tower'], generator=[]), architectures=['LlavaOnevisionForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=[], tags=['vision', 'video'])",
+  "model_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "template_meta": "QwenTemplateMeta(template_type='llava_siglip2_qwen3', prefix=[], prompt=['<|im_start|>user\\n{{QUERY}}<|im_end|>\\n<|im_start|>assistant\\n'], chat_sep=['<|im_end|>\\n'], suffix=['<|im_end|>\\n'], template_cls=<class 'model_plugin.SigLIP2LlavaTemplate'>, system_prefix=['<|im_start|>system\\n{{SYSTEM}}<|im_end|>\\n'], default_system=None, auto_add_bos=False, stop_words=['<|endoftext|>'], agent_template='hermes', is_thinking=False, thinking_prefix='', non_thinking_prefix='', history_thinking_prefix='')",
+  "_val_dataset_exists": false,
+  "hub": "<class 'swift.hub.hub.HFHub'>",
+  "evaluation_strategy": "steps",
+  "training_args": "Seq2SeqTrainingArguments(output_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215', per_device_train_batch_size=1, num_train_epochs=3.0, max_steps=500, learning_rate=1e-05, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_steps=0.05, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, optim_target_modules=None, gradient_accumulation_steps=8, average_tokens_across_devices=None, max_grad_norm=1.0, label_smoothing_factor=0.0, bf16=True, fp16=False, bf16_full_eval=False, fp16_full_eval=False, tf32=None, gradient_checkpointing=True, gradient_checkpointing_kwargs={'use_reentrant': False}, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, use_liger_kernel=False, liger_kernel_config=None, use_cache=False, neftune_noise_alpha=None, torch_empty_cache_steps=None, auto_find_batch_size=False, logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_steps=1, logging_first_step=True, log_on_each_node=True, logging_nan_inf_filter=True, include_num_input_tokens_seen=None, log_level='passive', log_level_replica='warning', disable_tqdm=False, report_to=[], run_name='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215', project='huggingface', trackio_space_id='trackio', eval_strategy=<IntervalStrategy.NO: 'no'>, eval_steps=100.0, eval_delay=0, per_device_eval_batch_size=1, prediction_loss_only=False, eval_on_start=False, eval_do_concat_batches=True, eval_use_gather_object=False, eval_accumulation_steps=None, include_for_metrics=[], batch_eval_metrics=False, save_only_model=False, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=100, save_on_each_node=False, save_total_limit=2, enable_jit_checkpoint=False, push_to_hub=False, hub_token=None, hub_private_repo=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_always_push=False, hub_revision=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, restore_callback_states_from_checkpoint=False, full_determinism=False, seed=42, data_seed=42, use_cpu=False, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), parallelism_config=None, dataloader_drop_last=False, dataloader_num_workers=4, dataloader_pin_memory=True, dataloader_persistent_workers=False, dataloader_prefetch_factor=4, remove_unused_columns=False, label_names=None, train_sampling_strategy='random', length_column_name='length', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, ddp_backend=None, ddp_timeout=7200, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 2, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'allgather_partitions': True, 'allgather_bucket_size': 200000000.0, 'overlap_comm': False, 'reduce_scatter': True, 'reduce_bucket_size': 200000000.0, 'contiguous_gradients': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, debug=[], skip_memory_metrics=True, do_train=False, do_eval=False, do_predict=False, resume_from_checkpoint=None, warmup_ratio=0.05, logging_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/runs', local_rank=0, sortish_sampler=False, predict_with_generate=False, generation_max_length=None, generation_num_beams=None, generation_config=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, safe_serialization=True, max_shard_size='5GB', check_model=True, acc_strategy='token', train_dataloader_shuffle=True, group_by_length=False, max_epochs=None, aligner_lr=None, vit_lr=1e-06, use_logits_to_keep=None, ds3_gather_for_generation=True, resume_only_model=False, optimizer='multimodal', loss_type=None, eval_metric=None, callbacks=[], early_stop_interval=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, tuner_type='full', use_galore=False, galore_target_modules=None, galore_rank=128, galore_update_proj_gap=50, galore_scale=1.0, galore_proj_type='std', galore_optim_per_parameter=False, galore_with_embedding=False, galore_quantization=False, galore_proj_quant=False, galore_proj_bits=4, galore_proj_group_size=256, galore_cos_threshold=0.4, galore_gamma_proj=2, galore_queue_size=5, lisa_activated_layers=0, lisa_step_interval=20, use_flash_ckpt=False)"
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/args.json

@@ -0,0 +1,376 @@
+{
+  "output_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215",
+  "per_device_train_batch_size": 1,
+  "num_train_epochs": 3.0,
+  "max_steps": 500,
+  "learning_rate": 1e-05,
+  "lr_scheduler_type": "cosine",
+  "lr_scheduler_kwargs": null,
+  "warmup_steps": 0,
+  "optim": "adamw_torch_fused",
+  "optim_args": null,
+  "weight_decay": 0.1,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_epsilon": 1e-08,
+  "optim_target_modules": null,
+  "gradient_accumulation_steps": 8,
+  "average_tokens_across_devices": true,
+  "max_grad_norm": 1.0,
+  "label_smoothing_factor": 0.0,
+  "bf16": true,
+  "fp16": false,
+  "bf16_full_eval": false,
+  "fp16_full_eval": false,
+  "tf32": null,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": "{\"use_reentrant\": false}",
+  "torch_compile": false,
+  "torch_compile_backend": null,
+  "torch_compile_mode": null,
+  "use_liger_kernel": false,
+  "liger_kernel_config": null,
+  "use_cache": false,
+  "neftune_noise_alpha": null,
+  "torch_empty_cache_steps": null,
+  "auto_find_batch_size": false,
+  "logging_strategy": "steps",
+  "logging_steps": 1,
+  "logging_first_step": true,
+  "log_on_each_node": true,
+  "logging_nan_inf_filter": true,
+  "include_num_input_tokens_seen": false,
+  "log_level": "passive",
+  "log_level_replica": "warning",
+  "disable_tqdm": null,
+  "report_to": [
+    "none"
+  ],
+  "run_name": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215",
+  "project": "huggingface",
+  "trackio_space_id": "trackio",
+  "eval_strategy": "no",
+  "eval_steps": 100.0,
+  "eval_delay": 0,
+  "per_device_eval_batch_size": 1,
+  "prediction_loss_only": false,
+  "eval_on_start": false,
+  "eval_do_concat_batches": true,
+  "eval_use_gather_object": false,
+  "eval_accumulation_steps": null,
+  "include_for_metrics": [],
+  "batch_eval_metrics": false,
+  "save_only_model": false,
+  "save_strategy": "steps",
+  "save_steps": 100.0,
+  "save_on_each_node": false,
+  "save_total_limit": 2,
+  "enable_jit_checkpoint": false,
+  "push_to_hub": false,
+  "hub_token": null,
+  "hub_private_repo": null,
+  "hub_model_id": null,
+  "hub_strategy": "every_save",
+  "hub_always_push": false,
+  "hub_revision": null,
+  "load_best_model_at_end": false,
+  "metric_for_best_model": "loss",
+  "greater_is_better": false,
+  "ignore_data_skip": false,
+  "restore_callback_states_from_checkpoint": false,
+  "full_determinism": false,
+  "seed": 42,
+  "data_seed": 42,
+  "use_cpu": false,
+  "accelerator_config": "{\"dispatch_batches\": false}",
+  "parallelism_config": null,
+  "dataloader_drop_last": false,
+  "dataloader_num_workers": 4,
+  "dataloader_pin_memory": true,
+  "dataloader_persistent_workers": false,
+  "dataloader_prefetch_factor": 4,
+  "remove_unused_columns": true,
+  "label_names": null,
+  "train_sampling_strategy": "random",
+  "length_column_name": "length",
+  "ddp_find_unused_parameters": null,
+  "ddp_bucket_cap_mb": null,
+  "ddp_broadcast_buffers": null,
+  "ddp_backend": null,
+  "ddp_timeout": 7200,
+  "fsdp": [],
+  "fsdp_config": null,
+  "deepspeed": {
+    "fp16": {
+      "enabled": "auto",
+      "loss_scale": 0,
+      "loss_scale_window": 1000,
+      "initial_scale_power": 16,
+      "hysteresis": 2,
+      "min_loss_scale": 1
+    },
+    "bf16": {
+      "enabled": "auto"
+    },
+    "zero_optimization": {
+      "stage": 2,
+      "offload_optimizer": {
+        "device": "none",
+        "pin_memory": true
+      },
+      "allgather_partitions": true,
+      "allgather_bucket_size": 200000000.0,
+      "overlap_comm": false,
+      "reduce_scatter": true,
+      "reduce_bucket_size": 200000000.0,
+      "contiguous_gradients": true
+    },
+    "gradient_accumulation_steps": "auto",
+    "gradient_clipping": "auto",
+    "steps_per_print": 2000,
+    "train_batch_size": "auto",
+    "train_micro_batch_size_per_gpu": "auto",
+    "wall_clock_breakdown": false
+  },
+  "debug": null,
+  "skip_memory_metrics": true,
+  "do_train": false,
+  "do_eval": false,
+  "do_predict": false,
+  "resume_from_checkpoint": null,
+  "warmup_ratio": 0.05,
+  "logging_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/runs",
+  "local_rank": 0,
+  "sortish_sampler": false,
+  "predict_with_generate": false,
+  "generation_max_length": null,
+  "generation_num_beams": null,
+  "generation_config": null,
+  "tuner_backend": "peft",
+  "vit_gradient_checkpointing": null,
+  "router_aux_loss_coef": 0.0,
+  "enable_dft_loss": false,
+  "enable_channel_loss": false,
+  "safe_serialization": true,
+  "max_shard_size": "5GB",
+  "check_model": true,
+  "acc_strategy": "token",
+  "train_dataloader_shuffle": true,
+  "group_by_length": false,
+  "max_epochs": null,
+  "aligner_lr": null,
+  "vit_lr": 1e-06,
+  "use_logits_to_keep": null,
+  "ds3_gather_for_generation": true,
+  "resume_only_model": false,
+  "optimizer": null,
+  "loss_type": null,
+  "eval_metric": null,
+  "callbacks": [],
+  "early_stop_interval": null,
+  "eval_use_evalscope": false,
+  "eval_dataset": [],
+  "eval_dataset_args": null,
+  "eval_limit": null,
+  "eval_generation_config": null,
+  "extra_eval_args": null,
+  "tuner_type": "full",
+  "use_galore": false,
+  "galore_target_modules": null,
+  "galore_rank": 128,
+  "galore_update_proj_gap": 50,
+  "galore_scale": 1.0,
+  "galore_proj_type": "std",
+  "galore_optim_per_parameter": false,
+  "galore_with_embedding": false,
+  "galore_quantization": false,
+  "galore_proj_quant": false,
+  "galore_proj_bits": 4,
+  "galore_proj_group_size": 256,
+  "galore_cos_threshold": 0.4,
+  "galore_gamma_proj": 2,
+  "galore_queue_size": 5,
+  "lisa_activated_layers": 0,
+  "lisa_step_interval": 20,
+  "use_flash_ckpt": false,
+  "use_ray": false,
+  "ray_exp_name": null,
+  "device_groups": null,
+  "model": "/opt/tiger/model_cache/checkpoint-2181",
+  "model_type": "llava_siglip2_qwen3",
+  "model_revision": null,
+  "task_type": "causal_lm",
+  "torch_dtype": "bfloat16",
+  "attn_impl": "flash_attn",
+  "experts_impl": null,
+  "new_special_tokens": [],
+  "num_labels": null,
+  "problem_type": null,
+  "rope_scaling": null,
+  "device_map": null,
+  "max_memory": {},
+  "max_model_len": null,
+  "local_repo_path": null,
+  "init_strategy": null,
+  "template": "llava_siglip2_qwen3",
+  "system": null,
+  "max_length": 16384,
+  "truncation_strategy": "delete",
+  "max_pixels": null,
+  "agent_template": null,
+  "norm_bbox": null,
+  "use_chat_template": true,
+  "padding_side": "right",
+  "padding_free": true,
+  "loss_scale": "default",
+  "sequence_parallel_size": 1,
+  "template_backend": "swift",
+  "response_prefix": null,
+  "enable_thinking": null,
+  "add_non_thinking_prefix": true,
+  "dataset": [],
+  "val_dataset": [],
+  "cached_dataset": [
+    "/mnt/bn/strategy-mllm-train/common/datasets/vmllm_cached/image_10pct/train"
+  ],
+  "cached_val_dataset": [],
+  "split_dataset_ratio": 0.0,
+  "dataset_num_proc": 1,
+  "load_from_cache_file": false,
+  "dataset_shuffle": true,
+  "val_dataset_shuffle": false,
+  "streaming": false,
+  "interleave_prob": null,
+  "stopping_strategy": "first_exhausted",
+  "shuffle_buffer_size": 1000,
+  "download_mode": "reuse_dataset_if_exists",
+  "columns": {},
+  "strict": false,
+  "model_name": null,
+  "model_author": null,
+  "custom_dataset_info": [],
+  "quant_method": null,
+  "quant_bits": null,
+  "hqq_axis": null,
+  "bnb_4bit_compute_dtype": "bfloat16",
+  "bnb_4bit_quant_type": "nf4",
+  "bnb_4bit_use_double_quant": true,
+  "bnb_4bit_quant_storage": null,
+  "max_new_tokens": 64,
+  "temperature": 0.0,
+  "top_k": null,
+  "top_p": null,
+  "repetition_penalty": null,
+  "num_beams": 1,
+  "stream": false,
+  "stop_words": [],
+  "logprobs": false,
+  "top_logprobs": null,
+  "structured_outputs_regex": null,
+  "train_type": null,
+  "adapters": [],
+  "external_plugins": [
+    "video_mllm/model_plugin.py",
+    "video_mllm/dataset_plugin.py"
+  ],
+  "custom_register_path": [],
+  "model_kwargs": {},
+  "load_args": false,
+  "load_data_args": false,
+  "packing": true,
+  "packing_length": 16384,
+  "packing_num_proc": 1,
+  "lazy_tokenize": false,
+  "use_hf": true,
+  "ignore_args_error": false,
+  "use_swift_lora": false,
+  "freeze_parameters": [],
+  "freeze_parameters_regex": null,
+  "freeze_parameters_ratio": 0.0,
+  "trainable_parameters": [
+    "model.multi_modal_projector"
+  ],
+  "trainable_parameters_regex": null,
+  "freeze_llm": false,
+  "freeze_vit": false,
+  "freeze_aligner": false,
+  "target_modules": [
+    "all-linear"
+  ],
+  "target_regex": null,
+  "target_parameters": null,
+  "modules_to_save": [],
+  "lora_rank": 8,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "lora_bias": "none",
+  "lora_dtype": null,
+  "lorap_lr_ratio": null,
+  "use_rslora": false,
+  "use_dora": false,
+  "lora_ga_batch_size": 2,
+  "lora_ga_iters": 2,
+  "lora_ga_max_length": 1024,
+  "lora_ga_direction": "ArB2r",
+  "lora_ga_scale": "stable",
+  "lora_ga_stable_gamma": 16,
+  "init_weights": true,
+  "fourier_n_frequency": 2000,
+  "fourier_scaling": 300.0,
+  "boft_block_size": 4,
+  "boft_block_num": 0,
+  "boft_n_butterfly_factor": 1,
+  "boft_dropout": 0.0,
+  "vera_rank": 256,
+  "vera_projection_prng_key": 0,
+  "vera_dropout": 0.0,
+  "vera_d_initial": 0.1,
+  "adapter_act": "gelu",
+  "adapter_length": 128,
+  "adalora_target_r": 8,
+  "adalora_init_r": 12,
+  "adalora_tinit": 0,
+  "adalora_tfinal": 0,
+  "adalora_deltaT": 1,
+  "adalora_beta1": 0.85,
+  "adalora_beta2": 0.85,
+  "adalora_orth_reg_weight": 0.5,
+  "llamapro_num_new_blocks": 4,
+  "llamapro_num_groups": null,
+  "reft_layer_key": null,
+  "reft_layers": null,
+  "reft_rank": 4,
+  "reft_intervention_type": "LoreftIntervention",
+  "reft_args": null,
+  "swanlab_token": null,
+  "swanlab_project": "ms-swift",
+  "swanlab_workspace": null,
+  "swanlab_exp_name": null,
+  "swanlab_notification_method": null,
+  "swanlab_webhook_url": null,
+  "swanlab_secret": null,
+  "swanlab_sender_email": null,
+  "swanlab_receiver_email": null,
+  "swanlab_smtp_server": null,
+  "swanlab_smtp_port": null,
+  "swanlab_email_language": "zh",
+  "swanlab_mode": "cloud",
+  "add_version": true,
+  "create_checkpoint_symlink": false,
+  "zero_hpz_partition_size": null,
+  "deepspeed_autotp_size": null,
+  "swift_version": "4.1.0.dev0",
+  "ckpt_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "rank": 0,
+  "global_world_size": 8,
+  "local_world_size": 8,
+  "model_suffix": "checkpoint-2181",
+  "model_info": "ModelInfo(model_type='llava_siglip2_qwen3', model_dir='/opt/tiger/model_cache/checkpoint-2181', torch_dtype=torch.bfloat16, max_model_len=40960, quant_method=None, quant_bits=None, rope_scaling=None, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
+  "model_meta": "ModelMeta(model_type='llava_siglip2_qwen3', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-0.6B', hf_model_id='Qwen/Qwen3-0.6B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-1.7B', hf_model_id='Qwen/Qwen3-1.7B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-4B', hf_model_id='Qwen/Qwen3-4B', model_path=None, ms_revision=None, hf_revision=None)], template=None, ignore_patterns=None, requires=None, tags=[])], loader=<class 'model_plugin.SigLIP2Qwen3Loader'>, template='llava_siglip2_qwen3', model_arch=MultiModelKeys(arch_name='llava_hf', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['model.language_model', 'lm_head'], aligner=['model.multi_modal_projector'], vision_tower=['model.vision_tower'], generator=[]), architectures=['LlavaOnevisionForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=[], tags=['vision', 'video'])",
+  "model_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "template_meta": "QwenTemplateMeta(template_type='llava_siglip2_qwen3', prefix=[], prompt=['<|im_start|>user\\n{{QUERY}}<|im_end|>\\n<|im_start|>assistant\\n'], chat_sep=['<|im_end|>\\n'], suffix=['<|im_end|>\\n'], template_cls=<class 'model_plugin.SigLIP2LlavaTemplate'>, system_prefix=['<|im_start|>system\\n{{SYSTEM}}<|im_end|>\\n'], default_system=None, auto_add_bos=False, stop_words=['<|endoftext|>'], agent_template='hermes', is_thinking=False, thinking_prefix='', non_thinking_prefix='', history_thinking_prefix='')",
+  "_val_dataset_exists": false,
+  "hub": "<class 'swift.hub.hub.HFHub'>",
+  "evaluation_strategy": "steps",
+  "training_args": "Seq2SeqTrainingArguments(output_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215', per_device_train_batch_size=1, num_train_epochs=3.0, max_steps=500, learning_rate=1e-05, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_steps=0.05, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, optim_target_modules=None, gradient_accumulation_steps=8, average_tokens_across_devices=None, max_grad_norm=1.0, label_smoothing_factor=0.0, bf16=True, fp16=False, bf16_full_eval=False, fp16_full_eval=False, tf32=None, gradient_checkpointing=True, gradient_checkpointing_kwargs={'use_reentrant': False}, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, use_liger_kernel=False, liger_kernel_config=None, use_cache=False, neftune_noise_alpha=None, torch_empty_cache_steps=None, auto_find_batch_size=False, logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_steps=1, logging_first_step=True, log_on_each_node=True, logging_nan_inf_filter=True, include_num_input_tokens_seen=None, log_level='passive', log_level_replica='warning', disable_tqdm=False, report_to=[], run_name='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215', project='huggingface', trackio_space_id='trackio', eval_strategy=<IntervalStrategy.NO: 'no'>, eval_steps=100.0, eval_delay=0, per_device_eval_batch_size=1, prediction_loss_only=False, eval_on_start=False, eval_do_concat_batches=True, eval_use_gather_object=False, eval_accumulation_steps=None, include_for_metrics=[], batch_eval_metrics=False, save_only_model=False, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=100, save_on_each_node=False, save_total_limit=2, enable_jit_checkpoint=False, push_to_hub=False, hub_token=None, hub_private_repo=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_always_push=False, hub_revision=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, restore_callback_states_from_checkpoint=False, full_determinism=False, seed=42, data_seed=42, use_cpu=False, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), parallelism_config=None, dataloader_drop_last=False, dataloader_num_workers=4, dataloader_pin_memory=True, dataloader_persistent_workers=False, dataloader_prefetch_factor=4, remove_unused_columns=False, label_names=None, train_sampling_strategy='random', length_column_name='length', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, ddp_backend=None, ddp_timeout=7200, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 2, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'allgather_partitions': True, 'allgather_bucket_size': 200000000.0, 'overlap_comm': False, 'reduce_scatter': True, 'reduce_bucket_size': 200000000.0, 'contiguous_gradients': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, debug=[], skip_memory_metrics=True, do_train=False, do_eval=False, do_predict=False, resume_from_checkpoint=None, warmup_ratio=0.05, logging_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/runs', local_rank=0, sortish_sampler=False, predict_with_generate=False, generation_max_length=None, generation_num_beams=None, generation_config=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, safe_serialization=True, max_shard_size='5GB', check_model=True, acc_strategy='token', train_dataloader_shuffle=True, group_by_length=False, max_epochs=None, aligner_lr=None, vit_lr=1e-06, use_logits_to_keep=None, ds3_gather_for_generation=True, resume_only_model=False, optimizer='multimodal', loss_type=None, eval_metric=None, callbacks=[], early_stop_interval=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, tuner_type='full', use_galore=False, galore_target_modules=None, galore_rank=128, galore_update_proj_gap=50, galore_scale=1.0, galore_proj_type='std', galore_optim_per_parameter=False, galore_with_embedding=False, galore_quantization=False, galore_proj_quant=False, galore_proj_bits=4, galore_proj_group_size=256, galore_cos_threshold=0.4, galore_gamma_proj=2, galore_queue_size=5, lisa_activated_layers=0, lisa_step_interval=20, use_flash_ckpt=False)"
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/config.json

@@ -0,0 +1,248 @@
+{
+  "architectures": [
+    "LlavaOnevisionForConditionalGeneration"
+  ],
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "hidden_size": 2048,
+  "image_grid_pinpoints": [
+    [
+      384,
+      384
+    ],
+    [
+      384,
+      768
+    ],
+    [
+      384,
+      1152
+    ],
+    [
+      384,
+      1536
+    ],
+    [
+      384,
+      1920
+    ],
+    [
+      384,
+      2304
+    ],
+    [
+      768,
+      384
+    ],
+    [
+      768,
+      768
+    ],
+    [
+      768,
+      1152
+    ],
+    [
+      768,
+      1536
+    ],
+    [
+      768,
+      1920
+    ],
+    [
+      768,
+      2304
+    ],
+    [
+      1152,
+      384
+    ],
+    [
+      1152,
+      768
+    ],
+    [
+      1152,
+      1152
+    ],
+    [
+      1152,
+      1536
+    ],
+    [
+      1152,
+      1920
+    ],
+    [
+      1152,
+      2304
+    ],
+    [
+      1536,
+      384
+    ],
+    [
+      1536,
+      768
+    ],
+    [
+      1536,
+      1152
+    ],
+    [
+      1536,
+      1536
+    ],
+    [
+      1536,
+      1920
+    ],
+    [
+      1536,
+      2304
+    ],
+    [
+      1920,
+      384
+    ],
+    [
+      1920,
+      768
+    ],
+    [
+      1920,
+      1152
+    ],
+    [
+      1920,
+      1536
+    ],
+    [
+      1920,
+      1920
+    ],
+    [
+      1920,
+      2304
+    ],
+    [
+      2304,
+      384
+    ],
+    [
+      2304,
+      768
+    ],
+    [
+      2304,
+      1152
+    ],
+    [
+      2304,
+      1536
+    ],
+    [
+      2304,
+      1920
+    ],
+    [
+      2304,
+      2304
+    ]
+  ],
+  "image_token_index": 151669,
+  "keys_to_ignore_at_inference": [
+    "past_key_values"
+  ],
+  "model_type": "llava_onevision",
+  "multimodal_projector_bias": true,
+  "pad_token_id": 151643,
+  "projector_hidden_act": "gelu",
+  "text_config": {
+    "_name_or_path": "/home/tiger/.cache/huggingface/hub/models--Qwen--Qwen3-1.7B/snapshots/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e",
+    "architectures": [
+      "Qwen3ForCausalLM"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 151643,
+    "dtype": "bfloat16",
+    "eos_token_id": 151645,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 2048,
+    "initializer_range": 0.02,
+    "intermediate_size": 6144,
+    "layer_types": [
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 40960,
+    "max_window_layers": 28,
+    "model_type": "qwen3",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 28,
+    "num_key_value_heads": 8,
+    "pad_token_id": 151643,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "rope_theta": 1000000,
+      "rope_type": "default"
+    },
+    "sliding_window": null,
+    "tie_word_embeddings": true,
+    "use_cache": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.2.0",
+  "use_cache": false,
+  "video_token_index": 151670,
+  "vision_aspect_ratio": "anyres_max_9",
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 26,
+    "patch_size": 14,
+    "vision_use_head": false
+  },
+  "vision_feature_layer": -1,
+  "vision_feature_select_strategy": "full"
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "transformers_version": "5.2.0",
+  "use_cache": true
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c691944f9d68071f5078527c05c2c9a342c9eae53ed3aca4960cf54d0270aea
+size 3654094853

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6b400665444084f9a470fd55a4dd4665f88a6195396d13675b85f059618c5d3
+size 3654085701

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f4cb0e6fa0c778897380621cc1b4eca3ee2e0656be8456a912d0f53f47570a3
+size 3654083653

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:178511313175542200cb5f7ef503cb00a4554cace8c5ebf003a303b32f73a38a
+size 3654085317

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19335ebba0a2ac8db8ae180d897a98170d173867e8bc60e174f6cca18d75aa2a
+size 3654085253

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cea43eea1c10652a2dc1e1490a1baad5f205b6ceaa19c77b4f217fa6bd191e7
+size 3654085253

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:479b2d00d5e545d488adaaa9be7d1c80fb487b29efe4392141ab56b3079c88f3
+size 3654085829

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26833aa47b40d64d4c895a305befff1741f31a12a8f127197da1100a93e6acb3
+size 3654081925

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/global_step300/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eeec3962da3dd0f0f61fba2019d51c3dbc764c665843ff710eff7e312f0b3146
+size 4872342181

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/latest

@@ -0,0 +1 @@
+global_step300

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f24efcae4e9726d5380f1a8d964e9d7dee73d00929c92dd46b6cf7c5c5120e8
+size 4872193968

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/processor_config.json

@@ -0,0 +1,206 @@
+{
+  "image_processor": {
+    "data_format": "channels_first",
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_pad": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_grid_pinpoints": [
+      [
+        384,
+        384
+      ],
+      [
+        384,
+        768
+      ],
+      [
+        384,
+        1152
+      ],
+      [
+        384,
+        1536
+      ],
+      [
+        384,
+        1920
+      ],
+      [
+        384,
+        2304
+      ],
+      [
+        768,
+        384
+      ],
+      [
+        768,
+        768
+      ],
+      [
+        768,
+        1152
+      ],
+      [
+        768,
+        1536
+      ],
+      [
+        768,
+        1920
+      ],
+      [
+        768,
+        2304
+      ],
+      [
+        1152,
+        384
+      ],
+      [
+        1152,
+        768
+      ],
+      [
+        1152,
+        1152
+      ],
+      [
+        1152,
+        1536
+      ],
+      [
+        1152,
+        1920
+      ],
+      [
+        1152,
+        2304
+      ],
+      [
+        1536,
+        384
+      ],
+      [
+        1536,
+        768
+      ],
+      [
+        1536,
+        1152
+      ],
+      [
+        1536,
+        1536
+      ],
+      [
+        1536,
+        1920
+      ],
+      [
+        1536,
+        2304
+      ],
+      [
+        1920,
+        384
+      ],
+      [
+        1920,
+        768
+      ],
+      [
+        1920,
+        1152
+      ],
+      [
+        1920,
+        1536
+      ],
+      [
+        1920,
+        1920
+      ],
+      [
+        1920,
+        2304
+      ],
+      [
+        2304,
+        384
+      ],
+      [
+        2304,
+        768
+      ],
+      [
+        2304,
+        1152
+      ],
+      [
+        2304,
+        1536
+      ],
+      [
+        2304,
+        1920
+      ],
+      [
+        2304,
+        2304
+      ]
+    ],
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_processor_type": "LlavaOnevisionImageProcessorFast",
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "height": 384,
+      "width": 384
+    }
+  },
+  "image_token": "<image>",
+  "num_image_tokens": 729,
+  "processor_class": "LlavaOnevisionProcessor",
+  "video_processor": {
+    "data_format": "channels_first",
+    "default_to_square": false,
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "do_sample_frames": false,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_metadata": false,
+    "size": {
+      "height": 384,
+      "width": 384
+    },
+    "video_processor_type": "LlavaOnevisionVideoProcessor"
+  },
+  "video_token": "<video>",
+  "vision_aspect_ratio": "anyres_max_9",
+  "vision_feature_select_strategy": null
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0418786c564bb001fc5f4215375b1a99ed3ea3f30771057e1d8682214a0108f
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35deaabe10396919e53b2274736981b26bc690165c8839845f178b355a6a7588
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c5deee061a76a9e05b4634f5abad786b4042e744cc4c156a2ab76dc3c94fcac
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13579f7ca365ea6f9cef83929c34833cd97f9b5764a434989b4e3e2db87dce56
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:073d0b1d7e2e3930e5c29fc418e6246462ab2a00396c3a86e1448d373503f1b2
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c60731015889939e665a308a0933ac6c327734a4699cf5615217db0cfaf03ca
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:416648e7a35618f28d73625751a8fc2ad066533414d19f6fc0c39932429023fc
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:619b5c49446c94dca2b233ad17098e42f7d58a945c2eced1aa23e50469a75431
+size 16389

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edf547cbeb33ab19d02e6aa7d657d5b0c3c5d9fe3409b74784c83fa0b853d5be
+size 1465

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9024318c850eaadf26be79389d21b07a7afd8f1af749b89f9109b06c0d12173c
+size 11423018

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/tokenizer_config.json

@@ -0,0 +1,19 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<image>",
+    "<video>"
+  ],
+  "is_local": true,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "processor_class": "LlavaOnevisionProcessor",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/trainer_state.json

@@ -0,0 +1,2434 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.2886375875067313,
+  "eval_steps": 100.0,
+  "global_step": 300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.004308023694130318,
+      "grad_norm": 34.22294235229492,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 2.5344715118408203,
+      "step": 1,
+      "token_acc": 0.5299950245219988
+    },
+    {
+      "epoch": 0.008616047388260635,
+      "grad_norm": 34.8410758972168,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 2.4708309173583984,
+      "step": 2,
+      "token_acc": 0.5363605898123325
+    },
+    {
+      "epoch": 0.012924071082390954,
+      "grad_norm": 39.80426788330078,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.775846481323242,
+      "step": 3,
+      "token_acc": 0.5183054170395942
+    },
+    {
+      "epoch": 0.01723209477652127,
+      "grad_norm": 34.9159049987793,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 2.3811163902282715,
+      "step": 4,
+      "token_acc": 0.5468338887914401
+    },
+    {
+      "epoch": 0.021540118470651588,
+      "grad_norm": 35.46262741088867,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 2.525376558303833,
+      "step": 5,
+      "token_acc": 0.5244299674267101
+    },
+    {
+      "epoch": 0.025848142164781908,
+      "grad_norm": 39.962608337402344,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 2.5210909843444824,
+      "step": 6,
+      "token_acc": 0.5377253761053478
+    },
+    {
+      "epoch": 0.030156165858912225,
+      "grad_norm": 33.2032356262207,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 2.580561399459839,
+      "step": 7,
+      "token_acc": 0.5228668179957578
+    },
+    {
+      "epoch": 0.03446418955304254,
+      "grad_norm": 23.4835205078125,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 2.2210943698883057,
+      "step": 8,
+      "token_acc": 0.5748507275912618
+    },
+    {
+      "epoch": 0.03877221324717286,
+      "grad_norm": 21.224645614624023,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 2.182731866836548,
+      "step": 9,
+      "token_acc": 0.578366892643934
+    },
+    {
+      "epoch": 0.043080236941303175,
+      "grad_norm": 19.219499588012695,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.9947047233581543,
+      "step": 10,
+      "token_acc": 0.5828964891898785
+    },
+    {
+      "epoch": 0.04738826063543349,
+      "grad_norm": 10.247937202453613,
+      "learning_rate": 4.4e-06,
+      "loss": 1.8489328622817993,
+      "step": 11,
+      "token_acc": 0.5984906381352694
+    },
+    {
+      "epoch": 0.051696284329563816,
+      "grad_norm": 10.421255111694336,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 1.7730863094329834,
+      "step": 12,
+      "token_acc": 0.6098880314722348
+    },
+    {
+      "epoch": 0.05600430802369413,
+      "grad_norm": 9.2549467086792,
+      "learning_rate": 5.2e-06,
+      "loss": 1.705881118774414,
+      "step": 13,
+      "token_acc": 0.6173091377644153
+    },
+    {
+      "epoch": 0.06031233171782445,
+      "grad_norm": 6.33651876449585,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 1.6188805103302002,
+      "step": 14,
+      "token_acc": 0.6224566257421177
+    },
+    {
+      "epoch": 0.06462035541195477,
+      "grad_norm": 7.8872880935668945,
+      "learning_rate": 6e-06,
+      "loss": 1.5559440851211548,
+      "step": 15,
+      "token_acc": 0.6367019934393137
+    },
+    {
+      "epoch": 0.06892837910608508,
+      "grad_norm": 6.443421363830566,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 1.6304481029510498,
+      "step": 16,
+      "token_acc": 0.6337283187360467
+    },
+    {
+      "epoch": 0.0732364028002154,
+      "grad_norm": 4.364102840423584,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 1.5763095617294312,
+      "step": 17,
+      "token_acc": 0.6282391722369747
+    },
+    {
+      "epoch": 0.07754442649434572,
+      "grad_norm": 4.073349952697754,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 1.4519922733306885,
+      "step": 18,
+      "token_acc": 0.6620760163579504
+    },
+    {
+      "epoch": 0.08185245018847603,
+      "grad_norm": 3.76297926902771,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 1.520553708076477,
+      "step": 19,
+      "token_acc": 0.6476499528943788
+    },
+    {
+      "epoch": 0.08616047388260635,
+      "grad_norm": 3.565013885498047,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.443807601928711,
+      "step": 20,
+      "token_acc": 0.6564110824742269
+    },
+    {
+      "epoch": 0.09046849757673667,
+      "grad_norm": 3.0926177501678467,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 1.305448293685913,
+      "step": 21,
+      "token_acc": 0.6728254783782893
+    },
+    {
+      "epoch": 0.09477652127086698,
+      "grad_norm": 2.9559481143951416,
+      "learning_rate": 8.8e-06,
+      "loss": 1.3177809715270996,
+      "step": 22,
+      "token_acc": 0.6766286820038702
+    },
+    {
+      "epoch": 0.0990845449649973,
+      "grad_norm": 3.516502618789673,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 1.412851095199585,
+      "step": 23,
+      "token_acc": 0.6545797141412749
+    },
+    {
+      "epoch": 0.10339256865912763,
+      "grad_norm": 2.693331003189087,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 1.3329224586486816,
+      "step": 24,
+      "token_acc": 0.6691273897406776
+    },
+    {
+      "epoch": 0.10770059235325795,
+      "grad_norm": 3.09365177154541,
+      "learning_rate": 1e-05,
+      "loss": 1.3302955627441406,
+      "step": 25,
+      "token_acc": 0.6788545041009029
+    },
+    {
+      "epoch": 0.11200861604738827,
+      "grad_norm": 3.117713689804077,
+      "learning_rate": 9.999890641901124e-06,
+      "loss": 1.3734593391418457,
+      "step": 26,
+      "token_acc": 0.6594324976051267
+    },
+    {
+      "epoch": 0.11631663974151858,
+      "grad_norm": 2.4900543689727783,
+      "learning_rate": 9.99956257238817e-06,
+      "loss": 1.256556510925293,
+      "step": 27,
+      "token_acc": 0.6908205642274095
+    },
+    {
+      "epoch": 0.1206246634356489,
+      "grad_norm": 2.8557262420654297,
+      "learning_rate": 9.999015805811965e-06,
+      "loss": 1.3106865882873535,
+      "step": 28,
+      "token_acc": 0.6742931211189486
+    },
+    {
+      "epoch": 0.12493268712977922,
+      "grad_norm": 2.0467495918273926,
+      "learning_rate": 9.998250366089848e-06,
+      "loss": 1.3423490524291992,
+      "step": 29,
+      "token_acc": 0.6723827275551414
+    },
+    {
+      "epoch": 0.12924071082390953,
+      "grad_norm": 2.235506057739258,
+      "learning_rate": 9.99726628670463e-06,
+      "loss": 1.3720948696136475,
+      "step": 30,
+      "token_acc": 0.668766154668885
+    },
+    {
+      "epoch": 0.13354873451803984,
+      "grad_norm": 2.093294143676758,
+      "learning_rate": 9.996063610703138e-06,
+      "loss": 1.3393146991729736,
+      "step": 31,
+      "token_acc": 0.6776454554304442
+    },
+    {
+      "epoch": 0.13785675821217017,
+      "grad_norm": 2.331446647644043,
+      "learning_rate": 9.994642390694308e-06,
+      "loss": 1.2755086421966553,
+      "step": 32,
+      "token_acc": 0.685778748701787
+    },
+    {
+      "epoch": 0.1421647819063005,
+      "grad_norm": 4.931087493896484,
+      "learning_rate": 9.993002688846913e-06,
+      "loss": 1.2852654457092285,
+      "step": 33,
+      "token_acc": 0.7003776905870706
+    },
+    {
+      "epoch": 0.1464728056004308,
+      "grad_norm": 2.2281808853149414,
+      "learning_rate": 9.991144576886824e-06,
+      "loss": 1.234990119934082,
+      "step": 34,
+      "token_acc": 0.6911842151458765
+    },
+    {
+      "epoch": 0.15078082929456113,
+      "grad_norm": 2.216933488845825,
+      "learning_rate": 9.989068136093873e-06,
+      "loss": 1.2935469150543213,
+      "step": 35,
+      "token_acc": 0.6736291459523633
+    },
+    {
+      "epoch": 0.15508885298869143,
+      "grad_norm": 2.199500560760498,
+      "learning_rate": 9.986773457298311e-06,
+      "loss": 1.2178046703338623,
+      "step": 36,
+      "token_acc": 0.6948421052631579
+    },
+    {
+      "epoch": 0.15939687668282176,
+      "grad_norm": 3.288256883621216,
+      "learning_rate": 9.984260640876821e-06,
+      "loss": 1.2350926399230957,
+      "step": 37,
+      "token_acc": 0.6875022046633038
+    },
+    {
+      "epoch": 0.16370490037695207,
+      "grad_norm": 1.8213781118392944,
+      "learning_rate": 9.981529796748135e-06,
+      "loss": 1.2279441356658936,
+      "step": 38,
+      "token_acc": 0.6980759578383805
+    },
+    {
+      "epoch": 0.1680129240710824,
+      "grad_norm": 2.273272752761841,
+      "learning_rate": 9.97858104436822e-06,
+      "loss": 1.2051079273223877,
+      "step": 39,
+      "token_acc": 0.6949230265312807
+    },
+    {
+      "epoch": 0.1723209477652127,
+      "grad_norm": 2.2953481674194336,
+      "learning_rate": 9.975414512725058e-06,
+      "loss": 1.2342824935913086,
+      "step": 40,
+      "token_acc": 0.6916462800402936
+    },
+    {
+      "epoch": 0.17662897145934303,
+      "grad_norm": 2.2890288829803467,
+      "learning_rate": 9.972030340333e-06,
+      "loss": 1.207423210144043,
+      "step": 41,
+      "token_acc": 0.6919735017716838
+    },
+    {
+      "epoch": 0.18093699515347333,
+      "grad_norm": 1.9035778045654297,
+      "learning_rate": 9.968428675226714e-06,
+      "loss": 1.1688880920410156,
+      "step": 42,
+      "token_acc": 0.6906148450822046
+    },
+    {
+      "epoch": 0.18524501884760367,
+      "grad_norm": 2.176882743835449,
+      "learning_rate": 9.964609674954696e-06,
+      "loss": 1.238387107849121,
+      "step": 43,
+      "token_acc": 0.6940492431624764
+    },
+    {
+      "epoch": 0.18955304254173397,
+      "grad_norm": 1.740558385848999,
+      "learning_rate": 9.960573506572391e-06,
+      "loss": 1.1984151601791382,
+      "step": 44,
+      "token_acc": 0.6910384253093839
+    },
+    {
+      "epoch": 0.1938610662358643,
+      "grad_norm": 1.9107651710510254,
+      "learning_rate": 9.956320346634877e-06,
+      "loss": 1.16075599193573,
+      "step": 45,
+      "token_acc": 0.6984731549229896
+    },
+    {
+      "epoch": 0.1981690899299946,
+      "grad_norm": 2.126544713973999,
+      "learning_rate": 9.951850381189152e-06,
+      "loss": 1.075717806816101,
+      "step": 46,
+      "token_acc": 0.7196697774587222
+    },
+    {
+      "epoch": 0.20247711362412493,
+      "grad_norm": 1.8616559505462646,
+      "learning_rate": 9.94716380576598e-06,
+      "loss": 1.2275675535202026,
+      "step": 47,
+      "token_acc": 0.6868371180636367
+    },
+    {
+      "epoch": 0.20678513731825526,
+      "grad_norm": 2.2004618644714355,
+      "learning_rate": 9.942260825371359e-06,
+      "loss": 1.1721410751342773,
+      "step": 48,
+      "token_acc": 0.7035488476442994
+    },
+    {
+      "epoch": 0.21109316101238557,
+      "grad_norm": 1.926447868347168,
+      "learning_rate": 9.937141654477529e-06,
+      "loss": 1.205888032913208,
+      "step": 49,
+      "token_acc": 0.6890798786653185
+    },
+    {
+      "epoch": 0.2154011847065159,
+      "grad_norm": 1.9273842573165894,
+      "learning_rate": 9.931806517013612e-06,
+      "loss": 1.1814758777618408,
+      "step": 50,
+      "token_acc": 0.7045194700383348
+    },
+    {
+      "epoch": 0.2197092084006462,
+      "grad_norm": 1.667378544807434,
+      "learning_rate": 9.926255646355804e-06,
+      "loss": 1.1939655542373657,
+      "step": 51,
+      "token_acc": 0.6967874821434504
+    },
+    {
+      "epoch": 0.22401723209477653,
+      "grad_norm": 1.9569302797317505,
+      "learning_rate": 9.920489285317169e-06,
+      "loss": 1.129996418952942,
+      "step": 52,
+      "token_acc": 0.7048677927446368
+    },
+    {
+      "epoch": 0.22832525578890683,
+      "grad_norm": 2.068121910095215,
+      "learning_rate": 9.91450768613702e-06,
+      "loss": 1.2010860443115234,
+      "step": 53,
+      "token_acc": 0.6904986522911051
+    },
+    {
+      "epoch": 0.23263327948303716,
+      "grad_norm": 1.5467214584350586,
+      "learning_rate": 9.908311110469881e-06,
+      "loss": 1.1379708051681519,
+      "step": 54,
+      "token_acc": 0.7055634134842056
+    },
+    {
+      "epoch": 0.23694130317716747,
+      "grad_norm": 2.0830445289611816,
+      "learning_rate": 9.901899829374048e-06,
+      "loss": 1.0704392194747925,
+      "step": 55,
+      "token_acc": 0.7152232895798196
+    },
+    {
+      "epoch": 0.2412493268712978,
+      "grad_norm": 2.495779037475586,
+      "learning_rate": 9.895274123299724e-06,
+      "loss": 1.0942933559417725,
+      "step": 56,
+      "token_acc": 0.7167745978692291
+    },
+    {
+      "epoch": 0.2455573505654281,
+      "grad_norm": 2.6332848072052,
+      "learning_rate": 9.888434282076759e-06,
+      "loss": 0.9827362895011902,
+      "step": 57,
+      "token_acc": 0.7357487741887188
+    },
+    {
+      "epoch": 0.24986537425955843,
+      "grad_norm": 2.415322780609131,
+      "learning_rate": 9.881380604901964e-06,
+      "loss": 1.0783662796020508,
+      "step": 58,
+      "token_acc": 0.7155577080694434
+    },
+    {
+      "epoch": 0.25417339795368876,
+      "grad_norm": 2.1927847862243652,
+      "learning_rate": 9.874113400326031e-06,
+      "loss": 1.2449455261230469,
+      "step": 59,
+      "token_acc": 0.6808658974448356
+    },
+    {
+      "epoch": 0.25848142164781907,
+      "grad_norm": 2.241107225418091,
+      "learning_rate": 9.86663298624003e-06,
+      "loss": 1.112494707107544,
+      "step": 60,
+      "token_acc": 0.7220329651422869
+    },
+    {
+      "epoch": 0.26278944534194937,
+      "grad_norm": 1.7743316888809204,
+      "learning_rate": 9.858939689861506e-06,
+      "loss": 1.1850183010101318,
+      "step": 61,
+      "token_acc": 0.692633709713129
+    },
+    {
+      "epoch": 0.26709746903607967,
+      "grad_norm": 1.6615829467773438,
+      "learning_rate": 9.851033847720167e-06,
+      "loss": 1.1249656677246094,
+      "step": 62,
+      "token_acc": 0.7059144015981597
+    },
+    {
+      "epoch": 0.27140549273021003,
+      "grad_norm": 1.5314289331436157,
+      "learning_rate": 9.842915805643156e-06,
+      "loss": 1.12161386013031,
+      "step": 63,
+      "token_acc": 0.7094002053086963
+    },
+    {
+      "epoch": 0.27571351642434033,
+      "grad_norm": 1.8113492727279663,
+      "learning_rate": 9.834585918739936e-06,
+      "loss": 1.0911247730255127,
+      "step": 64,
+      "token_acc": 0.7131517228937398
+    },
+    {
+      "epoch": 0.28002154011847064,
+      "grad_norm": 1.5588164329528809,
+      "learning_rate": 9.826044551386743e-06,
+      "loss": 1.1018824577331543,
+      "step": 65,
+      "token_acc": 0.7167374220937585
+    },
+    {
+      "epoch": 0.284329563812601,
+      "grad_norm": 1.6779085397720337,
+      "learning_rate": 9.817292077210658e-06,
+      "loss": 1.1914907693862915,
+      "step": 66,
+      "token_acc": 0.6931636612645514
+    },
+    {
+      "epoch": 0.2886375875067313,
+      "grad_norm": 1.8709357976913452,
+      "learning_rate": 9.808328879073251e-06,
+      "loss": 1.1766862869262695,
+      "step": 67,
+      "token_acc": 0.7000365586156471
+    },
+    {
+      "epoch": 0.2929456112008616,
+      "grad_norm": 1.8896335363388062,
+      "learning_rate": 9.79915534905385e-06,
+      "loss": 1.2150582075119019,
+      "step": 68,
+      "token_acc": 0.690418950540116
+    },
+    {
+      "epoch": 0.2972536348949919,
+      "grad_norm": 1.773423433303833,
+      "learning_rate": 9.789771888432375e-06,
+      "loss": 1.1559756994247437,
+      "step": 69,
+      "token_acc": 0.7010006235339529
+    },
+    {
+      "epoch": 0.30156165858912226,
+      "grad_norm": 1.5840319395065308,
+      "learning_rate": 9.780178907671788e-06,
+      "loss": 1.1300228834152222,
+      "step": 70,
+      "token_acc": 0.7048673543421343
+    },
+    {
+      "epoch": 0.30586968228325256,
+      "grad_norm": 1.7504026889801025,
+      "learning_rate": 9.77037682640015e-06,
+      "loss": 1.1307711601257324,
+      "step": 71,
+      "token_acc": 0.7024363156798779
+    },
+    {
+      "epoch": 0.31017770597738287,
+      "grad_norm": 1.7551653385162354,
+      "learning_rate": 9.760366073392246e-06,
+      "loss": 1.1609346866607666,
+      "step": 72,
+      "token_acc": 0.6984935064935065
+    },
+    {
+      "epoch": 0.31448572967151317,
+      "grad_norm": 1.6342157125473022,
+      "learning_rate": 9.750147086550843e-06,
+      "loss": 1.0880463123321533,
+      "step": 73,
+      "token_acc": 0.7135369347962821
+    },
+    {
+      "epoch": 0.31879375336564353,
+      "grad_norm": 1.7132184505462646,
+      "learning_rate": 9.739720312887536e-06,
+      "loss": 1.1172926425933838,
+      "step": 74,
+      "token_acc": 0.7106431450435189
+    },
+    {
+      "epoch": 0.32310177705977383,
+      "grad_norm": 1.726162075996399,
+      "learning_rate": 9.729086208503174e-06,
+      "loss": 1.1162161827087402,
+      "step": 75,
+      "token_acc": 0.7107878787878787
+    },
+    {
+      "epoch": 0.32740980075390413,
+      "grad_norm": 1.6666592359542847,
+      "learning_rate": 9.718245238567939e-06,
+      "loss": 1.076697826385498,
+      "step": 76,
+      "token_acc": 0.7241597747668485
+    },
+    {
+      "epoch": 0.33171782444803444,
+      "grad_norm": 1.4859079122543335,
+      "learning_rate": 9.707197877300974e-06,
+      "loss": 1.09866201877594,
+      "step": 77,
+      "token_acc": 0.7109936635126304
+    },
+    {
+      "epoch": 0.3360258481421648,
+      "grad_norm": 1.6729086637496948,
+      "learning_rate": 9.69594460794965e-06,
+      "loss": 1.1241893768310547,
+      "step": 78,
+      "token_acc": 0.7088581024349286
+    },
+    {
+      "epoch": 0.3403338718362951,
+      "grad_norm": 1.5671727657318115,
+      "learning_rate": 9.684485922768422e-06,
+      "loss": 1.0672322511672974,
+      "step": 79,
+      "token_acc": 0.7204722938385084
+    },
+    {
+      "epoch": 0.3446418955304254,
+      "grad_norm": 1.6157093048095703,
+      "learning_rate": 9.672822322997305e-06,
+      "loss": 1.127738118171692,
+      "step": 80,
+      "token_acc": 0.6991704875573134
+    },
+    {
+      "epoch": 0.34894991922455576,
+      "grad_norm": 1.7361069917678833,
+      "learning_rate": 9.660954318839934e-06,
+      "loss": 1.115564227104187,
+      "step": 81,
+      "token_acc": 0.7092018810659374
+    },
+    {
+      "epoch": 0.35325794291868606,
+      "grad_norm": 1.5783766508102417,
+      "learning_rate": 9.648882429441258e-06,
+      "loss": 1.0751323699951172,
+      "step": 82,
+      "token_acc": 0.7211808283757998
+    },
+    {
+      "epoch": 0.35756596661281637,
+      "grad_norm": 1.9924604892730713,
+      "learning_rate": 9.636607182864828e-06,
+      "loss": 1.063920259475708,
+      "step": 83,
+      "token_acc": 0.7266466605109264
+    },
+    {
+      "epoch": 0.36187399030694667,
+      "grad_norm": 1.3612253665924072,
+      "learning_rate": 9.624129116069695e-06,
+      "loss": 1.133460283279419,
+      "step": 84,
+      "token_acc": 0.7086913565462112
+    },
+    {
+      "epoch": 0.366182014001077,
+      "grad_norm": 1.69752037525177,
+      "learning_rate": 9.611448774886925e-06,
+      "loss": 1.0716493129730225,
+      "step": 85,
+      "token_acc": 0.7221262950172669
+    },
+    {
+      "epoch": 0.37049003769520733,
+      "grad_norm": 1.6512473821640015,
+      "learning_rate": 9.598566713995718e-06,
+      "loss": 1.0333918333053589,
+      "step": 86,
+      "token_acc": 0.7264647063180505
+    },
+    {
+      "epoch": 0.37479806138933763,
+      "grad_norm": 1.6064995527267456,
+      "learning_rate": 9.585483496899151e-06,
+      "loss": 1.123110294342041,
+      "step": 87,
+      "token_acc": 0.7135224274406332
+    },
+    {
+      "epoch": 0.37910608508346794,
+      "grad_norm": 1.8027215003967285,
+      "learning_rate": 9.572199695899522e-06,
+      "loss": 1.030998945236206,
+      "step": 88,
+      "token_acc": 0.7300635712561359
+    },
+    {
+      "epoch": 0.3834141087775983,
+      "grad_norm": 1.6361185312271118,
+      "learning_rate": 9.558715892073324e-06,
+      "loss": 1.1275122165679932,
+      "step": 89,
+      "token_acc": 0.7020269608643144
+    },
+    {
+      "epoch": 0.3877221324717286,
+      "grad_norm": 1.5621289014816284,
+      "learning_rate": 9.545032675245814e-06,
+      "loss": 1.0969953536987305,
+      "step": 90,
+      "token_acc": 0.7070858623746846
+    },
+    {
+      "epoch": 0.3920301561658589,
+      "grad_norm": 1.790278434753418,
+      "learning_rate": 9.531150643965224e-06,
+      "loss": 1.073486566543579,
+      "step": 91,
+      "token_acc": 0.7183880749698586
+    },
+    {
+      "epoch": 0.3963381798599892,
+      "grad_norm": 1.7555965185165405,
+      "learning_rate": 9.517070405476575e-06,
+      "loss": 1.071641445159912,
+      "step": 92,
+      "token_acc": 0.7204091394091026
+    },
+    {
+      "epoch": 0.40064620355411956,
+      "grad_norm": 1.7374863624572754,
+      "learning_rate": 9.502792575695112e-06,
+      "loss": 1.1522419452667236,
+      "step": 93,
+      "token_acc": 0.7053788169158043
+    },
+    {
+      "epoch": 0.40495422724824987,
+      "grad_norm": 1.9136552810668945,
+      "learning_rate": 9.48831777917936e-06,
+      "loss": 1.1674339771270752,
+      "step": 94,
+      "token_acc": 0.6964470088065594
+    },
+    {
+      "epoch": 0.40926225094238017,
+      "grad_norm": 2.252842426300049,
+      "learning_rate": 9.473646649103819e-06,
+      "loss": 1.0893977880477905,
+      "step": 95,
+      "token_acc": 0.7158215010141988
+    },
+    {
+      "epoch": 0.4135702746365105,
+      "grad_norm": 1.7565231323242188,
+      "learning_rate": 9.458779827231237e-06,
+      "loss": 1.0599641799926758,
+      "step": 96,
+      "token_acc": 0.722913735252733
+    },
+    {
+      "epoch": 0.41787829833064083,
+      "grad_norm": 1.720835566520691,
+      "learning_rate": 9.443717963884568e-06,
+      "loss": 1.0994898080825806,
+      "step": 97,
+      "token_acc": 0.7109813161650018
+    },
+    {
+      "epoch": 0.42218632202477113,
+      "grad_norm": 1.5080041885375977,
+      "learning_rate": 9.428461717918512e-06,
+      "loss": 1.0535576343536377,
+      "step": 98,
+      "token_acc": 0.7211945850986876
+    },
+    {
+      "epoch": 0.42649434571890144,
+      "grad_norm": 1.422196626663208,
+      "learning_rate": 9.413011756690686e-06,
+      "loss": 1.0241769552230835,
+      "step": 99,
+      "token_acc": 0.7236528301886792
+    },
+    {
+      "epoch": 0.4308023694130318,
+      "grad_norm": 1.7745169401168823,
+      "learning_rate": 9.397368756032445e-06,
+      "loss": 1.0839262008666992,
+      "step": 100,
+      "token_acc": 0.7179929498237456
+    },
+    {
+      "epoch": 0.4351103931071621,
+      "grad_norm": 1.9213250875473022,
+      "learning_rate": 9.381533400219319e-06,
+      "loss": 1.0448743104934692,
+      "step": 101,
+      "token_acc": 0.7268769599704852
+    },
+    {
+      "epoch": 0.4394184168012924,
+      "grad_norm": 1.8812602758407593,
+      "learning_rate": 9.365506381941066e-06,
+      "loss": 1.0698344707489014,
+      "step": 102,
+      "token_acc": 0.7161785216178521
+    },
+    {
+      "epoch": 0.4437264404954227,
+      "grad_norm": 1.6234725713729858,
+      "learning_rate": 9.349288402271387e-06,
+      "loss": 1.0575649738311768,
+      "step": 103,
+      "token_acc": 0.7228134641383853
+    },
+    {
+      "epoch": 0.44803446418955306,
+      "grad_norm": 1.7460858821868896,
+      "learning_rate": 9.332880170637252e-06,
+      "loss": 1.1082675457000732,
+      "step": 104,
+      "token_acc": 0.7142947461591957
+    },
+    {
+      "epoch": 0.45234248788368336,
+      "grad_norm": 1.6380345821380615,
+      "learning_rate": 9.31628240478787e-06,
+      "loss": 1.1685254573822021,
+      "step": 105,
+      "token_acc": 0.6993712668972022
+    },
+    {
+      "epoch": 0.45665051157781367,
+      "grad_norm": 1.7230206727981567,
+      "learning_rate": 9.299495830763285e-06,
+      "loss": 1.0857363939285278,
+      "step": 106,
+      "token_acc": 0.7167548635169605
+    },
+    {
+      "epoch": 0.46095853527194397,
+      "grad_norm": 1.7054288387298584,
+      "learning_rate": 9.28252118286263e-06,
+      "loss": 1.1256463527679443,
+      "step": 107,
+      "token_acc": 0.7026264236145698
+    },
+    {
+      "epoch": 0.46526655896607433,
+      "grad_norm": 1.6527998447418213,
+      "learning_rate": 9.265359203611988e-06,
+      "loss": 1.0812056064605713,
+      "step": 108,
+      "token_acc": 0.7152996010027894
+    },
+    {
+      "epoch": 0.46957458266020463,
+      "grad_norm": 1.670569658279419,
+      "learning_rate": 9.248010643731936e-06,
+      "loss": 1.0796825885772705,
+      "step": 109,
+      "token_acc": 0.7202207281491818
+    },
+    {
+      "epoch": 0.47388260635433493,
+      "grad_norm": 1.9298903942108154,
+      "learning_rate": 9.230476262104678e-06,
+      "loss": 1.0428919792175293,
+      "step": 110,
+      "token_acc": 0.7195813030138964
+    },
+    {
+      "epoch": 0.4781906300484653,
+      "grad_norm": 1.65512216091156,
+      "learning_rate": 9.212756825740874e-06,
+      "loss": 1.147737741470337,
+      "step": 111,
+      "token_acc": 0.7054659725749112
+    },
+    {
+      "epoch": 0.4824986537425956,
+      "grad_norm": 1.524289846420288,
+      "learning_rate": 9.194853109746073e-06,
+      "loss": 1.1040431261062622,
+      "step": 112,
+      "token_acc": 0.7155831310133981
+    },
+    {
+      "epoch": 0.4868066774367259,
+      "grad_norm": 1.6400195360183716,
+      "learning_rate": 9.176765897286812e-06,
+      "loss": 1.0260937213897705,
+      "step": 113,
+      "token_acc": 0.7253460824889396
+    },
+    {
+      "epoch": 0.4911147011308562,
+      "grad_norm": 1.8784524202346802,
+      "learning_rate": 9.15849597955636e-06,
+      "loss": 1.0686829090118408,
+      "step": 114,
+      "token_acc": 0.7213137813412104
+    },
+    {
+      "epoch": 0.49542272482498656,
+      "grad_norm": 2.1859333515167236,
+      "learning_rate": 9.140044155740102e-06,
+      "loss": 1.010416865348816,
+      "step": 115,
+      "token_acc": 0.7280323345381058
+    },
+    {
+      "epoch": 0.49973074851911686,
+      "grad_norm": 1.671910047531128,
+      "learning_rate": 9.121411232980589e-06,
+      "loss": 1.0519911050796509,
+      "step": 116,
+      "token_acc": 0.7266268924371486
+    },
+    {
+      "epoch": 0.5040387722132472,
+      "grad_norm": 1.5711668729782104,
+      "learning_rate": 9.102598026342223e-06,
+      "loss": 1.0841341018676758,
+      "step": 117,
+      "token_acc": 0.715380567067956
+    },
+    {
+      "epoch": 0.5083467959073775,
+      "grad_norm": 1.5971369743347168,
+      "learning_rate": 9.083605358775612e-06,
+      "loss": 1.1038738489151,
+      "step": 118,
+      "token_acc": 0.709660139277424
+    },
+    {
+      "epoch": 0.5126548196015078,
+      "grad_norm": 1.7582615613937378,
+      "learning_rate": 9.064434061081562e-06,
+      "loss": 1.0597245693206787,
+      "step": 119,
+      "token_acc": 0.716953316953317
+    },
+    {
+      "epoch": 0.5169628432956381,
+      "grad_norm": 1.6436163187026978,
+      "learning_rate": 9.045084971874738e-06,
+      "loss": 1.1113476753234863,
+      "step": 120,
+      "token_acc": 0.7140453152232706
+    },
+    {
+      "epoch": 0.5212708669897684,
+      "grad_norm": 1.723655104637146,
+      "learning_rate": 9.025558937546987e-06,
+      "loss": 1.0558507442474365,
+      "step": 121,
+      "token_acc": 0.7235675306957708
+    },
+    {
+      "epoch": 0.5255788906838987,
+      "grad_norm": 1.7649812698364258,
+      "learning_rate": 9.005856812230304e-06,
+      "loss": 1.1403634548187256,
+      "step": 122,
+      "token_acc": 0.7090575479566306
+    },
+    {
+      "epoch": 0.529886914378029,
+      "grad_norm": 1.9195817708969116,
+      "learning_rate": 8.98597945775948e-06,
+      "loss": 1.110230565071106,
+      "step": 123,
+      "token_acc": 0.7115247201333832
+    },
+    {
+      "epoch": 0.5341949380721593,
+      "grad_norm": 1.826794981956482,
+      "learning_rate": 8.96592774363439e-06,
+      "loss": 1.011481761932373,
+      "step": 124,
+      "token_acc": 0.7357743141683019
+    },
+    {
+      "epoch": 0.5385029617662898,
+      "grad_norm": 1.6854645013809204,
+      "learning_rate": 8.94570254698197e-06,
+      "loss": 1.0935232639312744,
+      "step": 125,
+      "token_acc": 0.7150796340223653
+    },
+    {
+      "epoch": 0.5428109854604201,
+      "grad_norm": 2.66402530670166,
+      "learning_rate": 8.92530475251784e-06,
+      "loss": 1.1172863245010376,
+      "step": 126,
+      "token_acc": 0.7189485136700242
+    },
+    {
+      "epoch": 0.5471190091545504,
+      "grad_norm": 2.063110113143921,
+      "learning_rate": 8.90473525250761e-06,
+      "loss": 1.0691171884536743,
+      "step": 127,
+      "token_acc": 0.7180058651026393
+    },
+    {
+      "epoch": 0.5514270328486807,
+      "grad_norm": 1.7156977653503418,
+      "learning_rate": 8.883994946727848e-06,
+      "loss": 1.0431095361709595,
+      "step": 128,
+      "token_acc": 0.7283595856392792
+    },
+    {
+      "epoch": 0.555735056542811,
+      "grad_norm": 1.8289692401885986,
+      "learning_rate": 8.863084742426719e-06,
+      "loss": 1.132885217666626,
+      "step": 129,
+      "token_acc": 0.706735040999213
+    },
+    {
+      "epoch": 0.5600430802369413,
+      "grad_norm": 2.87642240524292,
+      "learning_rate": 8.842005554284296e-06,
+      "loss": 1.0355212688446045,
+      "step": 130,
+      "token_acc": 0.7289206171661677
+    },
+    {
+      "epoch": 0.5643511039310716,
+      "grad_norm": 2.5777523517608643,
+      "learning_rate": 8.820758304372557e-06,
+      "loss": 1.006464958190918,
+      "step": 131,
+      "token_acc": 0.7324875466998755
+    },
+    {
+      "epoch": 0.568659127625202,
+      "grad_norm": 1.910340666770935,
+      "learning_rate": 8.799343922115045e-06,
+      "loss": 1.047837257385254,
+      "step": 132,
+      "token_acc": 0.7368516767294683
+    },
+    {
+      "epoch": 0.5729671513193323,
+      "grad_norm": 1.7361208200454712,
+      "learning_rate": 8.777763344246209e-06,
+      "loss": 1.011834740638733,
+      "step": 133,
+      "token_acc": 0.7284243277279798
+    },
+    {
+      "epoch": 0.5772751750134626,
+      "grad_norm": 1.8098233938217163,
+      "learning_rate": 8.756017514770444e-06,
+      "loss": 1.044621467590332,
+      "step": 134,
+      "token_acc": 0.7287564494050753
+    },
+    {
+      "epoch": 0.5815831987075929,
+      "grad_norm": 1.5448087453842163,
+      "learning_rate": 8.734107384920771e-06,
+      "loss": 1.073019027709961,
+      "step": 135,
+      "token_acc": 0.7119461765671152
+    },
+    {
+      "epoch": 0.5858912224017232,
+      "grad_norm": 1.7159926891326904,
+      "learning_rate": 8.71203391311725e-06,
+      "loss": 1.1081541776657104,
+      "step": 136,
+      "token_acc": 0.7037540013580367
+    },
+    {
+      "epoch": 0.5901992460958535,
+      "grad_norm": 1.6637221574783325,
+      "learning_rate": 8.689798064925049e-06,
+      "loss": 1.0081043243408203,
+      "step": 137,
+      "token_acc": 0.7252172754820019
+    },
+    {
+      "epoch": 0.5945072697899838,
+      "grad_norm": 1.9624783992767334,
+      "learning_rate": 8.6674008130122e-06,
+      "loss": 0.9942106008529663,
+      "step": 138,
+      "token_acc": 0.7384822910902048
+    },
+    {
+      "epoch": 0.5988152934841141,
+      "grad_norm": 1.822143793106079,
+      "learning_rate": 8.644843137107058e-06,
+      "loss": 1.0651887655258179,
+      "step": 139,
+      "token_acc": 0.7213238913924338
+    },
+    {
+      "epoch": 0.6031233171782445,
+      "grad_norm": 1.6583638191223145,
+      "learning_rate": 8.622126023955446e-06,
+      "loss": 1.1587040424346924,
+      "step": 140,
+      "token_acc": 0.700965954077593
+    },
+    {
+      "epoch": 0.6074313408723748,
+      "grad_norm": 1.8652868270874023,
+      "learning_rate": 8.599250467277483e-06,
+      "loss": 1.0659656524658203,
+      "step": 141,
+      "token_acc": 0.7252379815073868
+    },
+    {
+      "epoch": 0.6117393645665051,
+      "grad_norm": 1.6176389455795288,
+      "learning_rate": 8.576217467724129e-06,
+      "loss": 1.0570114850997925,
+      "step": 142,
+      "token_acc": 0.725997805513647
+    },
+    {
+      "epoch": 0.6160473882606354,
+      "grad_norm": 1.7835546731948853,
+      "learning_rate": 8.553028032833397e-06,
+      "loss": 1.0544447898864746,
+      "step": 143,
+      "token_acc": 0.7251621838326409
+    },
+    {
+      "epoch": 0.6203554119547657,
+      "grad_norm": 2.2224531173706055,
+      "learning_rate": 8.529683176986295e-06,
+      "loss": 1.090584397315979,
+      "step": 144,
+      "token_acc": 0.7182019410863273
+    },
+    {
+      "epoch": 0.624663435648896,
+      "grad_norm": 1.5986168384552002,
+      "learning_rate": 8.506183921362443e-06,
+      "loss": 1.1154588460922241,
+      "step": 145,
+      "token_acc": 0.7075557113233447
+    },
+    {
+      "epoch": 0.6289714593430263,
+      "grad_norm": 2.1650876998901367,
+      "learning_rate": 8.482531293895412e-06,
+      "loss": 0.9825166463851929,
+      "step": 146,
+      "token_acc": 0.7433130484007925
+    },
+    {
+      "epoch": 0.6332794830371568,
+      "grad_norm": 1.9352518320083618,
+      "learning_rate": 8.458726329227748e-06,
+      "loss": 1.0409623384475708,
+      "step": 147,
+      "token_acc": 0.725833571269947
+    },
+    {
+      "epoch": 0.6375875067312871,
+      "grad_norm": 1.8794916868209839,
+      "learning_rate": 8.434770068665723e-06,
+      "loss": 1.037879467010498,
+      "step": 148,
+      "token_acc": 0.7275396795896368
+    },
+    {
+      "epoch": 0.6418955304254174,
+      "grad_norm": 1.9519153833389282,
+      "learning_rate": 8.410663560133784e-06,
+      "loss": 1.0211390256881714,
+      "step": 149,
+      "token_acc": 0.7312037659717552
+    },
+    {
+      "epoch": 0.6462035541195477,
+      "grad_norm": 1.6926748752593994,
+      "learning_rate": 8.386407858128707e-06,
+      "loss": 1.0971746444702148,
+      "step": 150,
+      "token_acc": 0.7090979504289799
+    },
+    {
+      "epoch": 0.650511577813678,
+      "grad_norm": 1.642252802848816,
+      "learning_rate": 8.362004023673473e-06,
+      "loss": 1.0516653060913086,
+      "step": 151,
+      "token_acc": 0.7258933381217454
+    },
+    {
+      "epoch": 0.6548196015078083,
+      "grad_norm": 2.1485331058502197,
+      "learning_rate": 8.337453124270864e-06,
+      "loss": 1.060987949371338,
+      "step": 152,
+      "token_acc": 0.7258541834813022
+    },
+    {
+      "epoch": 0.6591276252019386,
+      "grad_norm": 2.170975923538208,
+      "learning_rate": 8.31275623385675e-06,
+      "loss": 1.0732593536376953,
+      "step": 153,
+      "token_acc": 0.7258163683563328
+    },
+    {
+      "epoch": 0.6634356488960689,
+      "grad_norm": 1.811652421951294,
+      "learning_rate": 8.287914432753123e-06,
+      "loss": 1.0401476621627808,
+      "step": 154,
+      "token_acc": 0.7282481532239196
+    },
+    {
+      "epoch": 0.6677436725901993,
+      "grad_norm": 1.6851332187652588,
+      "learning_rate": 8.262928807620843e-06,
+      "loss": 1.0859062671661377,
+      "step": 155,
+      "token_acc": 0.7122624537991038
+    },
+    {
+      "epoch": 0.6720516962843296,
+      "grad_norm": 1.7066915035247803,
+      "learning_rate": 8.237800451412095e-06,
+      "loss": 1.094170331954956,
+      "step": 156,
+      "token_acc": 0.7179701838554915
+    },
+    {
+      "epoch": 0.6763597199784599,
+      "grad_norm": 2.079630136489868,
+      "learning_rate": 8.212530463322584e-06,
+      "loss": 1.0371711254119873,
+      "step": 157,
+      "token_acc": 0.7281948615281949
+    },
+    {
+      "epoch": 0.6806677436725902,
+      "grad_norm": 1.9780957698822021,
+      "learning_rate": 8.18711994874345e-06,
+      "loss": 1.1454193592071533,
+      "step": 158,
+      "token_acc": 0.7068293139487724
+    },
+    {
+      "epoch": 0.6849757673667205,
+      "grad_norm": 2.0461599826812744,
+      "learning_rate": 8.161570019212921e-06,
+      "loss": 1.0632997751235962,
+      "step": 159,
+      "token_acc": 0.7188578486576815
+    },
+    {
+      "epoch": 0.6892837910608508,
+      "grad_norm": 1.8037797212600708,
+      "learning_rate": 8.135881792367686e-06,
+      "loss": 1.058543086051941,
+      "step": 160,
+      "token_acc": 0.7186254980079682
+    },
+    {
+      "epoch": 0.6935918147549811,
+      "grad_norm": 1.8418726921081543,
+      "learning_rate": 8.110056391894005e-06,
+      "loss": 1.0211877822875977,
+      "step": 161,
+      "token_acc": 0.7313764967680407
+    },
+    {
+      "epoch": 0.6978998384491115,
+      "grad_norm": 2.1125383377075195,
+      "learning_rate": 8.084094947478556e-06,
+      "loss": 1.1065531969070435,
+      "step": 162,
+      "token_acc": 0.7097420151520208
+    },
+    {
+      "epoch": 0.7022078621432418,
+      "grad_norm": 1.665132999420166,
+      "learning_rate": 8.057998594759022e-06,
+      "loss": 1.0586419105529785,
+      "step": 163,
+      "token_acc": 0.7197357402065693
+    },
+    {
+      "epoch": 0.7065158858373721,
+      "grad_norm": 1.541516900062561,
+      "learning_rate": 8.031768475274412e-06,
+      "loss": 1.0898858308792114,
+      "step": 164,
+      "token_acc": 0.7077255377696245
+    },
+    {
+      "epoch": 0.7108239095315024,
+      "grad_norm": 1.8763457536697388,
+      "learning_rate": 8.005405736415127e-06,
+      "loss": 1.1182177066802979,
+      "step": 165,
+      "token_acc": 0.7080204455101912
+    },
+    {
+      "epoch": 0.7151319332256327,
+      "grad_norm": 1.6244707107543945,
+      "learning_rate": 7.978911531372764e-06,
+      "loss": 1.052537441253662,
+      "step": 166,
+      "token_acc": 0.72174548690567
+    },
+    {
+      "epoch": 0.719439956919763,
+      "grad_norm": 1.7351884841918945,
+      "learning_rate": 7.952287019089686e-06,
+      "loss": 1.0168427228927612,
+      "step": 167,
+      "token_acc": 0.7289631977233038
+    },
+    {
+      "epoch": 0.7237479806138933,
+      "grad_norm": 1.883098840713501,
+      "learning_rate": 7.925533364208308e-06,
+      "loss": 1.0756115913391113,
+      "step": 168,
+      "token_acc": 0.7224274583557228
+    },
+    {
+      "epoch": 0.7280560043080236,
+      "grad_norm": 1.633523941040039,
+      "learning_rate": 7.898651737020166e-06,
+      "loss": 1.0162701606750488,
+      "step": 169,
+      "token_acc": 0.7267775147591479
+    },
+    {
+      "epoch": 0.732364028002154,
+      "grad_norm": 1.8218071460723877,
+      "learning_rate": 7.871643313414718e-06,
+      "loss": 1.0270060300827026,
+      "step": 170,
+      "token_acc": 0.7299303379673288
+    },
+    {
+      "epoch": 0.7366720516962844,
+      "grad_norm": 1.9670600891113281,
+      "learning_rate": 7.844509274827907e-06,
+      "loss": 1.0233960151672363,
+      "step": 171,
+      "token_acc": 0.7328961244397575
+    },
+    {
+      "epoch": 0.7409800753904147,
+      "grad_norm": 2.1067371368408203,
+      "learning_rate": 7.817250808190483e-06,
+      "loss": 1.1208573579788208,
+      "step": 172,
+      "token_acc": 0.7087912087912088
+    },
+    {
+      "epoch": 0.745288099084545,
+      "grad_norm": 1.7854297161102295,
+      "learning_rate": 7.789869105876083e-06,
+      "loss": 0.9967523217201233,
+      "step": 173,
+      "token_acc": 0.7338152553236948
+    },
+    {
+      "epoch": 0.7495961227786753,
+      "grad_norm": 2.0477850437164307,
+      "learning_rate": 7.762365365649068e-06,
+      "loss": 1.0872858762741089,
+      "step": 174,
+      "token_acc": 0.7145077370092574
+    },
+    {
+      "epoch": 0.7539041464728056,
+      "grad_norm": 1.975886344909668,
+      "learning_rate": 7.734740790612137e-06,
+      "loss": 0.9649370908737183,
+      "step": 175,
+      "token_acc": 0.7424413401299661
+    },
+    {
+      "epoch": 0.7582121701669359,
+      "grad_norm": 2.2943193912506104,
+      "learning_rate": 7.70699658915369e-06,
+      "loss": 0.9680527448654175,
+      "step": 176,
+      "token_acc": 0.743770432255721
+    },
+    {
+      "epoch": 0.7625201938610663,
+      "grad_norm": 1.7437077760696411,
+      "learning_rate": 7.679133974894984e-06,
+      "loss": 1.013915777206421,
+      "step": 177,
+      "token_acc": 0.7305739813185326
+    },
+    {
+      "epoch": 0.7668282175551966,
+      "grad_norm": 1.8506817817687988,
+      "learning_rate": 7.651154166637025e-06,
+      "loss": 1.060949683189392,
+      "step": 178,
+      "token_acc": 0.722198551342139
+    },
+    {
+      "epoch": 0.7711362412493269,
+      "grad_norm": 2.2398781776428223,
+      "learning_rate": 7.62305838830727e-06,
+      "loss": 0.9670058488845825,
+      "step": 179,
+      "token_acc": 0.7442888725128961
+    },
+    {
+      "epoch": 0.7754442649434572,
+      "grad_norm": 1.9097397327423096,
+      "learning_rate": 7.594847868906076e-06,
+      "loss": 1.0389031171798706,
+      "step": 180,
+      "token_acc": 0.7278524254156229
+    },
+    {
+      "epoch": 0.7797522886375875,
+      "grad_norm": 2.028334379196167,
+      "learning_rate": 7.566523842452958e-06,
+      "loss": 1.0487357378005981,
+      "step": 181,
+      "token_acc": 0.7233856893542757
+    },
+    {
+      "epoch": 0.7840603123317178,
+      "grad_norm": 1.7536522150039673,
+      "learning_rate": 7.5380875479325855e-06,
+      "loss": 0.8992788791656494,
+      "step": 182,
+      "token_acc": 0.7524805373225462
+    },
+    {
+      "epoch": 0.7883683360258481,
+      "grad_norm": 1.8125722408294678,
+      "learning_rate": 7.509540229240601e-06,
+      "loss": 1.0258413553237915,
+      "step": 183,
+      "token_acc": 0.7229109127111568
+    },
+    {
+      "epoch": 0.7926763597199784,
+      "grad_norm": 1.7677319049835205,
+      "learning_rate": 7.480883135129211e-06,
+      "loss": 1.0057992935180664,
+      "step": 184,
+      "token_acc": 0.7275143376696042
+    },
+    {
+      "epoch": 0.7969843834141088,
+      "grad_norm": 1.6595399379730225,
+      "learning_rate": 7.452117519152542e-06,
+      "loss": 1.0524992942810059,
+      "step": 185,
+      "token_acc": 0.723226874272053
+    },
+    {
+      "epoch": 0.8012924071082391,
+      "grad_norm": 1.6659682989120483,
+      "learning_rate": 7.4232446396118265e-06,
+      "loss": 1.029264211654663,
+      "step": 186,
+      "token_acc": 0.7303199773467365
+    },
+    {
+      "epoch": 0.8056004308023694,
+      "grad_norm": 1.933450698852539,
+      "learning_rate": 7.394265759500348e-06,
+      "loss": 1.0707944631576538,
+      "step": 187,
+      "token_acc": 0.7189290654020944
+    },
+    {
+      "epoch": 0.8099084544964997,
+      "grad_norm": 2.3765671253204346,
+      "learning_rate": 7.365182146448205e-06,
+      "loss": 0.9349066019058228,
+      "step": 188,
+      "token_acc": 0.7443648805954355
+    },
+    {
+      "epoch": 0.81421647819063,
+      "grad_norm": 1.8340600728988647,
+      "learning_rate": 7.335995072666848e-06,
+      "loss": 1.0280508995056152,
+      "step": 189,
+      "token_acc": 0.730319563522993
+    },
+    {
+      "epoch": 0.8185245018847603,
+      "grad_norm": 1.7334873676300049,
+      "learning_rate": 7.30670581489344e-06,
+      "loss": 1.0868083238601685,
+      "step": 190,
+      "token_acc": 0.7095239724855412
+    },
+    {
+      "epoch": 0.8228325255788906,
+      "grad_norm": 1.4876623153686523,
+      "learning_rate": 7.2773156543349965e-06,
+      "loss": 1.0421819686889648,
+      "step": 191,
+      "token_acc": 0.7225835322195704
+    },
+    {
+      "epoch": 0.827140549273021,
+      "grad_norm": 1.4609578847885132,
+      "learning_rate": 7.247825876612353e-06,
+      "loss": 1.034349799156189,
+      "step": 192,
+      "token_acc": 0.7214945373332529
+    },
+    {
+      "epoch": 0.8314485729671514,
+      "grad_norm": 1.4783042669296265,
+      "learning_rate": 7.218237771703921e-06,
+      "loss": 1.0758285522460938,
+      "step": 193,
+      "token_acc": 0.7069245624252828
+    },
+    {
+      "epoch": 0.8357565966612817,
+      "grad_norm": 1.647188663482666,
+      "learning_rate": 7.18855263388926e-06,
+      "loss": 1.0096200704574585,
+      "step": 194,
+      "token_acc": 0.7351176105401117
+    },
+    {
+      "epoch": 0.840064620355412,
+      "grad_norm": 1.7604641914367676,
+      "learning_rate": 7.158771761692464e-06,
+      "loss": 0.9966579079627991,
+      "step": 195,
+      "token_acc": 0.7323502967092249
+    },
+    {
+      "epoch": 0.8443726440495423,
+      "grad_norm": 1.7427420616149902,
+      "learning_rate": 7.128896457825364e-06,
+      "loss": 1.0532734394073486,
+      "step": 196,
+      "token_acc": 0.7239461255903445
+    },
+    {
+      "epoch": 0.8486806677436726,
+      "grad_norm": 1.604024887084961,
+      "learning_rate": 7.098928029130529e-06,
+      "loss": 1.0371193885803223,
+      "step": 197,
+      "token_acc": 0.724610928996531
+    },
+    {
+      "epoch": 0.8529886914378029,
+      "grad_norm": 1.562129020690918,
+      "learning_rate": 7.068867786524116e-06,
+      "loss": 0.9111225008964539,
+      "step": 198,
+      "token_acc": 0.7505460899956313
+    },
+    {
+      "epoch": 0.8572967151319332,
+      "grad_norm": 1.6844677925109863,
+      "learning_rate": 7.038717044938519e-06,
+      "loss": 1.0580235719680786,
+      "step": 199,
+      "token_acc": 0.7201209784631747
+    },
+    {
+      "epoch": 0.8616047388260636,
+      "grad_norm": 1.4355993270874023,
+      "learning_rate": 7.008477123264849e-06,
+      "loss": 1.062284231185913,
+      "step": 200,
+      "token_acc": 0.7225402403083201
+    },
+    {
+      "epoch": 0.8659127625201939,
+      "grad_norm": 1.7106071710586548,
+      "learning_rate": 6.978149344295242e-06,
+      "loss": 1.065187931060791,
+      "step": 201,
+      "token_acc": 0.718659797581205
+    },
+    {
+      "epoch": 0.8702207862143242,
+      "grad_norm": 1.8265399932861328,
+      "learning_rate": 6.9477350346650016e-06,
+      "loss": 1.0669317245483398,
+      "step": 202,
+      "token_acc": 0.7238228874337387
+    },
+    {
+      "epoch": 0.8745288099084545,
+      "grad_norm": 1.6417402029037476,
+      "learning_rate": 6.9172355247945586e-06,
+      "loss": 1.0068140029907227,
+      "step": 203,
+      "token_acc": 0.7297520946109023
+    },
+    {
+      "epoch": 0.8788368336025848,
+      "grad_norm": 1.6923401355743408,
+      "learning_rate": 6.88665214883128e-06,
+      "loss": 1.0113261938095093,
+      "step": 204,
+      "token_acc": 0.7302916443013061
+    },
+    {
+      "epoch": 0.8831448572967151,
+      "grad_norm": 1.7035597562789917,
+      "learning_rate": 6.855986244591104e-06,
+      "loss": 0.9667624831199646,
+      "step": 205,
+      "token_acc": 0.7364136181010784
+    },
+    {
+      "epoch": 0.8874528809908454,
+      "grad_norm": 1.7706389427185059,
+      "learning_rate": 6.825239153500029e-06,
+      "loss": 1.0399131774902344,
+      "step": 206,
+      "token_acc": 0.7230998509687034
+    },
+    {
+      "epoch": 0.8917609046849758,
+      "grad_norm": 1.7236924171447754,
+      "learning_rate": 6.794412220535426e-06,
+      "loss": 1.0391285419464111,
+      "step": 207,
+      "token_acc": 0.7210826975424606
+    },
+    {
+      "epoch": 0.8960689283791061,
+      "grad_norm": 2.0680770874023438,
+      "learning_rate": 6.763506794167207e-06,
+      "loss": 1.0605911016464233,
+      "step": 208,
+      "token_acc": 0.7222977447802293
+    },
+    {
+      "epoch": 0.9003769520732364,
+      "grad_norm": 2.0485286712646484,
+      "learning_rate": 6.732524226298841e-06,
+      "loss": 0.9372836351394653,
+      "step": 209,
+      "token_acc": 0.7482249475552687
+    },
+    {
+      "epoch": 0.9046849757673667,
+      "grad_norm": 1.5812981128692627,
+      "learning_rate": 6.701465872208216e-06,
+      "loss": 1.0708039999008179,
+      "step": 210,
+      "token_acc": 0.7196462448387457
+    },
+    {
+      "epoch": 0.908992999461497,
+      "grad_norm": 1.6296844482421875,
+      "learning_rate": 6.670333090488357e-06,
+      "loss": 1.0564963817596436,
+      "step": 211,
+      "token_acc": 0.7194455658726955
+    },
+    {
+      "epoch": 0.9133010231556273,
+      "grad_norm": 1.4809433221817017,
+      "learning_rate": 6.6391272429879886e-06,
+      "loss": 1.0199100971221924,
+      "step": 212,
+      "token_acc": 0.7275283419580629
+    },
+    {
+      "epoch": 0.9176090468497576,
+      "grad_norm": 1.912920355796814,
+      "learning_rate": 6.607849694751978e-06,
+      "loss": 1.0309929847717285,
+      "step": 213,
+      "token_acc": 0.7279354587869362
+    },
+    {
+      "epoch": 0.9219170705438879,
+      "grad_norm": 2.052734375,
+      "learning_rate": 6.576501813961609e-06,
+      "loss": 1.0631229877471924,
+      "step": 214,
+      "token_acc": 0.7230589022757697
+    },
+    {
+      "epoch": 0.9262250942380184,
+      "grad_norm": 2.039206027984619,
+      "learning_rate": 6.545084971874738e-06,
+      "loss": 1.022275686264038,
+      "step": 215,
+      "token_acc": 0.7276333345353575
+    },
+    {
+      "epoch": 0.9305331179321487,
+      "grad_norm": 1.7929805517196655,
+      "learning_rate": 6.513600542765816e-06,
+      "loss": 1.0045278072357178,
+      "step": 216,
+      "token_acc": 0.7384766366974801
+    },
+    {
+      "epoch": 0.934841141626279,
+      "grad_norm": 1.3928626775741577,
+      "learning_rate": 6.4820499038657695e-06,
+      "loss": 1.0367810726165771,
+      "step": 217,
+      "token_acc": 0.7207414090051669
+    },
+    {
+      "epoch": 0.9391491653204093,
+      "grad_norm": 1.959754228591919,
+      "learning_rate": 6.450434435301751e-06,
+      "loss": 1.0611419677734375,
+      "step": 218,
+      "token_acc": 0.7265009882383436
+    },
+    {
+      "epoch": 0.9434571890145396,
+      "grad_norm": 2.0485498905181885,
+      "learning_rate": 6.418755520036775e-06,
+      "loss": 1.0431468486785889,
+      "step": 219,
+      "token_acc": 0.7260826115922718
+    },
+    {
+      "epoch": 0.9477652127086699,
+      "grad_norm": 1.8070369958877563,
+      "learning_rate": 6.387014543809224e-06,
+      "loss": 0.9908697605133057,
+      "step": 220,
+      "token_acc": 0.7407475433924143
+    },
+    {
+      "epoch": 0.9520732364028002,
+      "grad_norm": 1.7055253982543945,
+      "learning_rate": 6.355212895072223e-06,
+      "loss": 1.0319141149520874,
+      "step": 221,
+      "token_acc": 0.724164978103144
+    },
+    {
+      "epoch": 0.9563812600969306,
+      "grad_norm": 1.73855459690094,
+      "learning_rate": 6.323351964932909e-06,
+      "loss": 1.029426097869873,
+      "step": 222,
+      "token_acc": 0.7305472017698503
+    },
+    {
+      "epoch": 0.9606892837910609,
+      "grad_norm": 1.6930358409881592,
+      "learning_rate": 6.291433147091583e-06,
+      "loss": 1.0009236335754395,
+      "step": 223,
+      "token_acc": 0.7313872069817001
+    },
+    {
+      "epoch": 0.9649973074851912,
+      "grad_norm": 1.84811270236969,
+      "learning_rate": 6.259457837780741e-06,
+      "loss": 1.0688104629516602,
+      "step": 224,
+      "token_acc": 0.7212521519165233
+    },
+    {
+      "epoch": 0.9693053311793215,
+      "grad_norm": 1.8543241024017334,
+      "learning_rate": 6.227427435703997e-06,
+      "loss": 1.0186160802841187,
+      "step": 225,
+      "token_acc": 0.7319165059244971
+    },
+    {
+      "epoch": 0.9736133548734518,
+      "grad_norm": 1.9619593620300293,
+      "learning_rate": 6.1953433419748995e-06,
+      "loss": 0.985061764717102,
+      "step": 226,
+      "token_acc": 0.7314210677500103
+    },
+    {
+      "epoch": 0.9779213785675821,
+      "grad_norm": 1.9096139669418335,
+      "learning_rate": 6.163206960055652e-06,
+      "loss": 1.0002927780151367,
+      "step": 227,
+      "token_acc": 0.7308709879551626
+    },
+    {
+      "epoch": 0.9822294022617124,
+      "grad_norm": 2.1386733055114746,
+      "learning_rate": 6.131019695695702e-06,
+      "loss": 1.082963466644287,
+      "step": 228,
+      "token_acc": 0.721509163632022
+    },
+    {
+      "epoch": 0.9865374259558427,
+      "grad_norm": 1.8572235107421875,
+      "learning_rate": 6.098782956870266e-06,
+      "loss": 1.0095869302749634,
+      "step": 229,
+      "token_acc": 0.7395705299689104
+    },
+    {
+      "epoch": 0.9908454496499731,
+      "grad_norm": 1.8879094123840332,
+      "learning_rate": 6.066498153718735e-06,
+      "loss": 1.0653495788574219,
+      "step": 230,
+      "token_acc": 0.7210560484743562
+    },
+    {
+      "epoch": 0.9951534733441034,
+      "grad_norm": 1.6340022087097168,
+      "learning_rate": 6.034166698482984e-06,
+      "loss": 1.0494650602340698,
+      "step": 231,
+      "token_acc": 0.7245733575045321
+    },
+    {
+      "epoch": 0.9994614970382337,
+      "grad_norm": 1.8272323608398438,
+      "learning_rate": 6.001790005445607e-06,
+      "loss": 1.0112372636795044,
+      "step": 232,
+      "token_acc": 0.7338219959516653
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 5.849401473999023,
+      "learning_rate": 5.969369490868042e-06,
+      "loss": 1.02152419090271,
+      "step": 233,
+      "token_acc": 0.7415776499589154
+    },
+    {
+      "epoch": 1.0043080236941304,
+      "grad_norm": 1.852284550666809,
+      "learning_rate": 5.936906572928625e-06,
+      "loss": 0.8803678154945374,
+      "step": 234,
+      "token_acc": 0.7592773805043281
+    },
+    {
+      "epoch": 1.0086160473882606,
+      "grad_norm": 1.5524327754974365,
+      "learning_rate": 5.904402671660551e-06,
+      "loss": 0.9862585067749023,
+      "step": 235,
+      "token_acc": 0.7289358264081256
+    },
+    {
+      "epoch": 1.012924071082391,
+      "grad_norm": 1.633418321609497,
+      "learning_rate": 5.871859208889759e-06,
+      "loss": 0.9620030522346497,
+      "step": 236,
+      "token_acc": 0.7423674343867167
+    },
+    {
+      "epoch": 1.0172320947765212,
+      "grad_norm": 1.5976009368896484,
+      "learning_rate": 5.839277608172739e-06,
+      "loss": 0.9446237087249756,
+      "step": 237,
+      "token_acc": 0.7452129084092126
+    },
+    {
+      "epoch": 1.0215401184706516,
+      "grad_norm": 2.123077869415283,
+      "learning_rate": 5.806659294734256e-06,
+      "loss": 0.9748927354812622,
+      "step": 238,
+      "token_acc": 0.737117873625039
+    },
+    {
+      "epoch": 1.0258481421647818,
+      "grad_norm": 2.096855401992798,
+      "learning_rate": 5.774005695405008e-06,
+      "loss": 0.9755522012710571,
+      "step": 239,
+      "token_acc": 0.7440971441421182
+    },
+    {
+      "epoch": 1.0301561658589122,
+      "grad_norm": 1.9100697040557861,
+      "learning_rate": 5.74131823855921e-06,
+      "loss": 0.9641351699829102,
+      "step": 240,
+      "token_acc": 0.7406723844678464
+    },
+    {
+      "epoch": 1.0344641895530426,
+      "grad_norm": 2.227088451385498,
+      "learning_rate": 5.708598354052122e-06,
+      "loss": 0.981192946434021,
+      "step": 241,
+      "token_acc": 0.7360852713178294
+    },
+    {
+      "epoch": 1.0387722132471728,
+      "grad_norm": 1.58879554271698,
+      "learning_rate": 5.675847473157485e-06,
+      "loss": 1.0112268924713135,
+      "step": 242,
+      "token_acc": 0.7281371199799825
+    },
+    {
+      "epoch": 1.0430802369413033,
+      "grad_norm": 2.0497395992279053,
+      "learning_rate": 5.643067028504931e-06,
+      "loss": 1.0331170558929443,
+      "step": 243,
+      "token_acc": 0.7243439772368006
+    },
+    {
+      "epoch": 1.0473882606354334,
+      "grad_norm": 1.6697180271148682,
+      "learning_rate": 5.610258454017301e-06,
+      "loss": 0.9160628318786621,
+      "step": 244,
+      "token_acc": 0.7527061945666597
+    },
+    {
+      "epoch": 1.0516962843295639,
+      "grad_norm": 2.58595609664917,
+      "learning_rate": 5.577423184847932e-06,
+      "loss": 0.8940185308456421,
+      "step": 245,
+      "token_acc": 0.7573819567288066
+    },
+    {
+      "epoch": 1.056004308023694,
+      "grad_norm": 2.416323184967041,
+      "learning_rate": 5.544562657317863e-06,
+      "loss": 0.9627000689506531,
+      "step": 246,
+      "token_acc": 0.7391393024678768
+    },
+    {
+      "epoch": 1.0603123317178245,
+      "grad_norm": 1.7874491214752197,
+      "learning_rate": 5.5116783088530255e-06,
+      "loss": 0.8945722579956055,
+      "step": 247,
+      "token_acc": 0.7548346386548384
+    },
+    {
+      "epoch": 1.0646203554119547,
+      "grad_norm": 1.9529039859771729,
+      "learning_rate": 5.478771577921351e-06,
+      "loss": 0.9477412700653076,
+      "step": 248,
+      "token_acc": 0.7437445495404843
+    },
+    {
+      "epoch": 1.068928379106085,
+      "grad_norm": 1.5508562326431274,
+      "learning_rate": 5.445843903969854e-06,
+      "loss": 0.9583556652069092,
+      "step": 249,
+      "token_acc": 0.735274946669827
+    },
+    {
+      "epoch": 1.0732364028002155,
+      "grad_norm": 1.7498481273651123,
+      "learning_rate": 5.412896727361663e-06,
+      "loss": 0.9596320390701294,
+      "step": 250,
+      "token_acc": 0.7402946059557032
+    },
+    {
+      "epoch": 1.0775444264943457,
+      "grad_norm": 1.8939355611801147,
+      "learning_rate": 5.379931489313016e-06,
+      "loss": 0.9707253575325012,
+      "step": 251,
+      "token_acc": 0.7383231378193594
+    },
+    {
+      "epoch": 1.081852450188476,
+      "grad_norm": 2.6653194427490234,
+      "learning_rate": 5.346949631830221e-06,
+      "loss": 0.8993242979049683,
+      "step": 252,
+      "token_acc": 0.7550643118436062
+    },
+    {
+      "epoch": 1.0861604738826063,
+      "grad_norm": 2.204688310623169,
+      "learning_rate": 5.3139525976465675e-06,
+      "loss": 0.9620447754859924,
+      "step": 253,
+      "token_acc": 0.7355941012148379
+    },
+    {
+      "epoch": 1.0904684975767367,
+      "grad_norm": 1.9297153949737549,
+      "learning_rate": 5.280941830159228e-06,
+      "loss": 0.9538339376449585,
+      "step": 254,
+      "token_acc": 0.739151133409016
+    },
+    {
+      "epoch": 1.0947765212708669,
+      "grad_norm": 1.8989887237548828,
+      "learning_rate": 5.247918773366112e-06,
+      "loss": 0.8881052732467651,
+      "step": 255,
+      "token_acc": 0.7648537005163512
+    },
+    {
+      "epoch": 1.0990845449649973,
+      "grad_norm": 1.740999698638916,
+      "learning_rate": 5.214884871802703e-06,
+      "loss": 0.9053628444671631,
+      "step": 256,
+      "token_acc": 0.7495991360973854
+    },
+    {
+      "epoch": 1.1033925686591277,
+      "grad_norm": 1.664394497871399,
+      "learning_rate": 5.1818415704788725e-06,
+      "loss": 1.0543291568756104,
+      "step": 257,
+      "token_acc": 0.7163526926527187
+    },
+    {
+      "epoch": 1.107700592353258,
+      "grad_norm": 1.8559702634811401,
+      "learning_rate": 5.148790314815662e-06,
+      "loss": 1.0188100337982178,
+      "step": 258,
+      "token_acc": 0.7235763996023176
+    },
+    {
+      "epoch": 1.1120086160473883,
+      "grad_norm": 2.0520901679992676,
+      "learning_rate": 5.11573255058207e-06,
+      "loss": 0.970104455947876,
+      "step": 259,
+      "token_acc": 0.7349447229279681
+    },
+    {
+      "epoch": 1.1163166397415185,
+      "grad_norm": 1.664621353149414,
+      "learning_rate": 5.082669723831793e-06,
+      "loss": 0.8900072574615479,
+      "step": 260,
+      "token_acc": 0.7525534978734805
+    },
+    {
+      "epoch": 1.120624663435649,
+      "grad_norm": 2.323301076889038,
+      "learning_rate": 5.049603280839982e-06,
+      "loss": 1.0096113681793213,
+      "step": 261,
+      "token_acc": 0.7256683273094331
+    },
+    {
+      "epoch": 1.1249326871297791,
+      "grad_norm": 1.8697978258132935,
+      "learning_rate": 5.016534668039976e-06,
+      "loss": 0.92182457447052,
+      "step": 262,
+      "token_acc": 0.7499677751997937
+    },
+    {
+      "epoch": 1.1292407108239095,
+      "grad_norm": 1.6699860095977783,
+      "learning_rate": 4.983465331960025e-06,
+      "loss": 0.9860280156135559,
+      "step": 263,
+      "token_acc": 0.7302170595746006
+    },
+    {
+      "epoch": 1.13354873451804,
+      "grad_norm": 2.074275016784668,
+      "learning_rate": 4.950396719160019e-06,
+      "loss": 0.9644788503646851,
+      "step": 264,
+      "token_acc": 0.7365843842232358
+    },
+    {
+      "epoch": 1.1378567582121701,
+      "grad_norm": 1.9523180723190308,
+      "learning_rate": 4.917330276168208e-06,
+      "loss": 0.9171963930130005,
+      "step": 265,
+      "token_acc": 0.7560794880431122
+    },
+    {
+      "epoch": 1.1421647819063006,
+      "grad_norm": 2.0590438842773438,
+      "learning_rate": 4.884267449417932e-06,
+      "loss": 0.9525893926620483,
+      "step": 266,
+      "token_acc": 0.7446058169959425
+    },
+    {
+      "epoch": 1.1464728056004307,
+      "grad_norm": 1.599365472793579,
+      "learning_rate": 4.851209685184339e-06,
+      "loss": 0.9242900609970093,
+      "step": 267,
+      "token_acc": 0.7477739359260732
+    },
+    {
+      "epoch": 1.1507808292945612,
+      "grad_norm": 1.7137466669082642,
+      "learning_rate": 4.818158429521129e-06,
+      "loss": 0.9517988562583923,
+      "step": 268,
+      "token_acc": 0.7356900661362339
+    },
+    {
+      "epoch": 1.1550888529886914,
+      "grad_norm": 1.980363130569458,
+      "learning_rate": 4.785115128197298e-06,
+      "loss": 0.9219489097595215,
+      "step": 269,
+      "token_acc": 0.7447838315128523
+    },
+    {
+      "epoch": 1.1593968766828218,
+      "grad_norm": 1.784232258796692,
+      "learning_rate": 4.752081226633888e-06,
+      "loss": 0.8524366617202759,
+      "step": 270,
+      "token_acc": 0.7703314616917225
+    },
+    {
+      "epoch": 1.1637049003769522,
+      "grad_norm": 1.9212590456008911,
+      "learning_rate": 4.719058169840773e-06,
+      "loss": 0.9193314909934998,
+      "step": 271,
+      "token_acc": 0.7476786980450417
+    },
+    {
+      "epoch": 1.1680129240710824,
+      "grad_norm": 1.7464808225631714,
+      "learning_rate": 4.686047402353433e-06,
+      "loss": 0.8871129155158997,
+      "step": 272,
+      "token_acc": 0.7609986092737695
+    },
+    {
+      "epoch": 1.1723209477652128,
+      "grad_norm": 1.8662816286087036,
+      "learning_rate": 4.65305036816978e-06,
+      "loss": 0.9364889860153198,
+      "step": 273,
+      "token_acc": 0.7490751506183279
+    },
+    {
+      "epoch": 1.176628971459343,
+      "grad_norm": 1.917938470840454,
+      "learning_rate": 4.620068510686985e-06,
+      "loss": 0.9322699308395386,
+      "step": 274,
+      "token_acc": 0.7451706895931508
+    },
+    {
+      "epoch": 1.1809369951534734,
+      "grad_norm": 1.8484588861465454,
+      "learning_rate": 4.587103272638339e-06,
+      "loss": 0.9433363080024719,
+      "step": 275,
+      "token_acc": 0.7470407575660631
+    },
+    {
+      "epoch": 1.1852450188476036,
+      "grad_norm": 1.7483965158462524,
+      "learning_rate": 4.554156096030149e-06,
+      "loss": 0.8598661422729492,
+      "step": 276,
+      "token_acc": 0.7575344381511783
+    },
+    {
+      "epoch": 1.189553042541734,
+      "grad_norm": 1.9640449285507202,
+      "learning_rate": 4.5212284220786495e-06,
+      "loss": 0.9079041481018066,
+      "step": 277,
+      "token_acc": 0.7517377674798964
+    },
+    {
+      "epoch": 1.1938610662358644,
+      "grad_norm": 1.5985236167907715,
+      "learning_rate": 4.488321691146975e-06,
+      "loss": 0.9669654369354248,
+      "step": 278,
+      "token_acc": 0.7323662000474528
+    },
+    {
+      "epoch": 1.1981690899299946,
+      "grad_norm": 1.9148527383804321,
+      "learning_rate": 4.4554373426821375e-06,
+      "loss": 1.015479564666748,
+      "step": 279,
+      "token_acc": 0.7254908271644673
+    },
+    {
+      "epoch": 1.202477113624125,
+      "grad_norm": 2.022810935974121,
+      "learning_rate": 4.42257681515207e-06,
+      "loss": 0.961472749710083,
+      "step": 280,
+      "token_acc": 0.7417307902702308
+    },
+    {
+      "epoch": 1.2067851373182552,
+      "grad_norm": 1.8741397857666016,
+      "learning_rate": 4.3897415459827e-06,
+      "loss": 0.9811180830001831,
+      "step": 281,
+      "token_acc": 0.7363141301232445
+    },
+    {
+      "epoch": 1.2110931610123856,
+      "grad_norm": 1.9402947425842285,
+      "learning_rate": 4.356932971495071e-06,
+      "loss": 0.9020291566848755,
+      "step": 282,
+      "token_acc": 0.7514281474886861
+    },
+    {
+      "epoch": 1.2154011847065158,
+      "grad_norm": 1.8439514636993408,
+      "learning_rate": 4.324152526842517e-06,
+      "loss": 1.0201821327209473,
+      "step": 283,
+      "token_acc": 0.730798455011714
+    },
+    {
+      "epoch": 1.2197092084006462,
+      "grad_norm": 2.052015542984009,
+      "learning_rate": 4.291401645947879e-06,
+      "loss": 0.9672775864601135,
+      "step": 284,
+      "token_acc": 0.7426946128060136
+    },
+    {
+      "epoch": 1.2240172320947766,
+      "grad_norm": 1.5881297588348389,
+      "learning_rate": 4.25868176144079e-06,
+      "loss": 0.889592170715332,
+      "step": 285,
+      "token_acc": 0.7558629189633332
+    },
+    {
+      "epoch": 1.2283252557889068,
+      "grad_norm": 1.7183021306991577,
+      "learning_rate": 4.225994304594994e-06,
+      "loss": 0.9233750104904175,
+      "step": 286,
+      "token_acc": 0.7486460772833724
+    },
+    {
+      "epoch": 1.2326332794830372,
+      "grad_norm": 1.6355488300323486,
+      "learning_rate": 4.193340705265746e-06,
+      "loss": 0.958794116973877,
+      "step": 287,
+      "token_acc": 0.7398538711856082
+    },
+    {
+      "epoch": 1.2369413031771674,
+      "grad_norm": 1.7922086715698242,
+      "learning_rate": 4.160722391827262e-06,
+      "loss": 0.9110997319221497,
+      "step": 288,
+      "token_acc": 0.754541538019674
+    },
+    {
+      "epoch": 1.2412493268712979,
+      "grad_norm": 2.0297932624816895,
+      "learning_rate": 4.128140791110243e-06,
+      "loss": 0.9580734372138977,
+      "step": 289,
+      "token_acc": 0.7436807779878826
+    },
+    {
+      "epoch": 1.245557350565428,
+      "grad_norm": 1.7132781744003296,
+      "learning_rate": 4.0955973283394525e-06,
+      "loss": 1.0088131427764893,
+      "step": 290,
+      "token_acc": 0.7320859993328683
+    },
+    {
+      "epoch": 1.2498653742595585,
+      "grad_norm": 1.5488715171813965,
+      "learning_rate": 4.063093427071376e-06,
+      "loss": 0.9522176384925842,
+      "step": 291,
+      "token_acc": 0.7438972284729541
+    },
+    {
+      "epoch": 1.2541733979536889,
+      "grad_norm": 1.8754502534866333,
+      "learning_rate": 4.03063050913196e-06,
+      "loss": 0.8881997466087341,
+      "step": 292,
+      "token_acc": 0.7582413769172045
+    },
+    {
+      "epoch": 1.258481421647819,
+      "grad_norm": 1.789201259613037,
+      "learning_rate": 3.998209994554395e-06,
+      "loss": 0.9494738578796387,
+      "step": 293,
+      "token_acc": 0.747545311141873
+    },
+    {
+      "epoch": 1.2627894453419493,
+      "grad_norm": 1.6978973150253296,
+      "learning_rate": 3.965833301517017e-06,
+      "loss": 0.903584361076355,
+      "step": 294,
+      "token_acc": 0.7527433967079239
+    },
+    {
+      "epoch": 1.2670974690360797,
+      "grad_norm": 1.9547784328460693,
+      "learning_rate": 3.9335018462812664e-06,
+      "loss": 1.0251178741455078,
+      "step": 295,
+      "token_acc": 0.7242036482439422
+    },
+    {
+      "epoch": 1.27140549273021,
+      "grad_norm": 1.8591620922088623,
+      "learning_rate": 3.901217043129735e-06,
+      "loss": 0.8809949159622192,
+      "step": 296,
+      "token_acc": 0.7596185737976783
+    },
+    {
+      "epoch": 1.2757135164243403,
+      "grad_norm": 1.9652024507522583,
+      "learning_rate": 3.8689803043043e-06,
+      "loss": 0.9077787399291992,
+      "step": 297,
+      "token_acc": 0.7569960352592479
+    },
+    {
+      "epoch": 1.2800215401184707,
+      "grad_norm": 1.7500567436218262,
+      "learning_rate": 3.8367930399443495e-06,
+      "loss": 1.0176526308059692,
+      "step": 298,
+      "token_acc": 0.7242611551091366
+    },
+    {
+      "epoch": 1.284329563812601,
+      "grad_norm": 1.750616431236267,
+      "learning_rate": 3.8046566580251e-06,
+      "loss": 0.9793616533279419,
+      "step": 299,
+      "token_acc": 0.7369895658894757
+    },
+    {
+      "epoch": 1.2886375875067313,
+      "grad_norm": 1.5951060056686401,
+      "learning_rate": 3.7725725642960047e-06,
+      "loss": 0.9102382659912109,
+      "step": 300,
+      "token_acc": 0.7541037534129412
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.863143068056158e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a86fe59726eb18bce544401943fcd0c83bf5581ebf5afd514f471c24e331e2e
+size 8977

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-300/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if ZERO_STAGE not in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/args.json

@@ -0,0 +1,376 @@
+{
+  "output_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215",
+  "per_device_train_batch_size": 1,
+  "num_train_epochs": 3.0,
+  "max_steps": 500,
+  "learning_rate": 1e-05,
+  "lr_scheduler_type": "cosine",
+  "lr_scheduler_kwargs": null,
+  "warmup_steps": 0,
+  "optim": "adamw_torch_fused",
+  "optim_args": null,
+  "weight_decay": 0.1,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_epsilon": 1e-08,
+  "optim_target_modules": null,
+  "gradient_accumulation_steps": 8,
+  "average_tokens_across_devices": true,
+  "max_grad_norm": 1.0,
+  "label_smoothing_factor": 0.0,
+  "bf16": true,
+  "fp16": false,
+  "bf16_full_eval": false,
+  "fp16_full_eval": false,
+  "tf32": null,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": "{\"use_reentrant\": false}",
+  "torch_compile": false,
+  "torch_compile_backend": null,
+  "torch_compile_mode": null,
+  "use_liger_kernel": false,
+  "liger_kernel_config": null,
+  "use_cache": false,
+  "neftune_noise_alpha": null,
+  "torch_empty_cache_steps": null,
+  "auto_find_batch_size": false,
+  "logging_strategy": "steps",
+  "logging_steps": 1,
+  "logging_first_step": true,
+  "log_on_each_node": true,
+  "logging_nan_inf_filter": true,
+  "include_num_input_tokens_seen": false,
+  "log_level": "passive",
+  "log_level_replica": "warning",
+  "disable_tqdm": null,
+  "report_to": [
+    "none"
+  ],
+  "run_name": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215",
+  "project": "huggingface",
+  "trackio_space_id": "trackio",
+  "eval_strategy": "no",
+  "eval_steps": 100.0,
+  "eval_delay": 0,
+  "per_device_eval_batch_size": 1,
+  "prediction_loss_only": false,
+  "eval_on_start": false,
+  "eval_do_concat_batches": true,
+  "eval_use_gather_object": false,
+  "eval_accumulation_steps": null,
+  "include_for_metrics": [],
+  "batch_eval_metrics": false,
+  "save_only_model": false,
+  "save_strategy": "steps",
+  "save_steps": 100.0,
+  "save_on_each_node": false,
+  "save_total_limit": 2,
+  "enable_jit_checkpoint": false,
+  "push_to_hub": false,
+  "hub_token": null,
+  "hub_private_repo": null,
+  "hub_model_id": null,
+  "hub_strategy": "every_save",
+  "hub_always_push": false,
+  "hub_revision": null,
+  "load_best_model_at_end": false,
+  "metric_for_best_model": "loss",
+  "greater_is_better": false,
+  "ignore_data_skip": false,
+  "restore_callback_states_from_checkpoint": false,
+  "full_determinism": false,
+  "seed": 42,
+  "data_seed": 42,
+  "use_cpu": false,
+  "accelerator_config": "{\"dispatch_batches\": false}",
+  "parallelism_config": null,
+  "dataloader_drop_last": false,
+  "dataloader_num_workers": 4,
+  "dataloader_pin_memory": true,
+  "dataloader_persistent_workers": false,
+  "dataloader_prefetch_factor": 4,
+  "remove_unused_columns": true,
+  "label_names": null,
+  "train_sampling_strategy": "random",
+  "length_column_name": "length",
+  "ddp_find_unused_parameters": null,
+  "ddp_bucket_cap_mb": null,
+  "ddp_broadcast_buffers": null,
+  "ddp_backend": null,
+  "ddp_timeout": 7200,
+  "fsdp": [],
+  "fsdp_config": null,
+  "deepspeed": {
+    "fp16": {
+      "enabled": "auto",
+      "loss_scale": 0,
+      "loss_scale_window": 1000,
+      "initial_scale_power": 16,
+      "hysteresis": 2,
+      "min_loss_scale": 1
+    },
+    "bf16": {
+      "enabled": "auto"
+    },
+    "zero_optimization": {
+      "stage": 2,
+      "offload_optimizer": {
+        "device": "none",
+        "pin_memory": true
+      },
+      "allgather_partitions": true,
+      "allgather_bucket_size": 200000000.0,
+      "overlap_comm": false,
+      "reduce_scatter": true,
+      "reduce_bucket_size": 200000000.0,
+      "contiguous_gradients": true
+    },
+    "gradient_accumulation_steps": "auto",
+    "gradient_clipping": "auto",
+    "steps_per_print": 2000,
+    "train_batch_size": "auto",
+    "train_micro_batch_size_per_gpu": "auto",
+    "wall_clock_breakdown": false
+  },
+  "debug": null,
+  "skip_memory_metrics": true,
+  "do_train": false,
+  "do_eval": false,
+  "do_predict": false,
+  "resume_from_checkpoint": null,
+  "warmup_ratio": 0.05,
+  "logging_dir": "/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/runs",
+  "local_rank": 0,
+  "sortish_sampler": false,
+  "predict_with_generate": false,
+  "generation_max_length": null,
+  "generation_num_beams": null,
+  "generation_config": null,
+  "tuner_backend": "peft",
+  "vit_gradient_checkpointing": null,
+  "router_aux_loss_coef": 0.0,
+  "enable_dft_loss": false,
+  "enable_channel_loss": false,
+  "safe_serialization": true,
+  "max_shard_size": "5GB",
+  "check_model": true,
+  "acc_strategy": "token",
+  "train_dataloader_shuffle": true,
+  "group_by_length": false,
+  "max_epochs": null,
+  "aligner_lr": null,
+  "vit_lr": 1e-06,
+  "use_logits_to_keep": null,
+  "ds3_gather_for_generation": true,
+  "resume_only_model": false,
+  "optimizer": null,
+  "loss_type": null,
+  "eval_metric": null,
+  "callbacks": [],
+  "early_stop_interval": null,
+  "eval_use_evalscope": false,
+  "eval_dataset": [],
+  "eval_dataset_args": null,
+  "eval_limit": null,
+  "eval_generation_config": null,
+  "extra_eval_args": null,
+  "tuner_type": "full",
+  "use_galore": false,
+  "galore_target_modules": null,
+  "galore_rank": 128,
+  "galore_update_proj_gap": 50,
+  "galore_scale": 1.0,
+  "galore_proj_type": "std",
+  "galore_optim_per_parameter": false,
+  "galore_with_embedding": false,
+  "galore_quantization": false,
+  "galore_proj_quant": false,
+  "galore_proj_bits": 4,
+  "galore_proj_group_size": 256,
+  "galore_cos_threshold": 0.4,
+  "galore_gamma_proj": 2,
+  "galore_queue_size": 5,
+  "lisa_activated_layers": 0,
+  "lisa_step_interval": 20,
+  "use_flash_ckpt": false,
+  "use_ray": false,
+  "ray_exp_name": null,
+  "device_groups": null,
+  "model": "/opt/tiger/model_cache/checkpoint-2181",
+  "model_type": "llava_siglip2_qwen3",
+  "model_revision": null,
+  "task_type": "causal_lm",
+  "torch_dtype": "bfloat16",
+  "attn_impl": "flash_attn",
+  "experts_impl": null,
+  "new_special_tokens": [],
+  "num_labels": null,
+  "problem_type": null,
+  "rope_scaling": null,
+  "device_map": null,
+  "max_memory": {},
+  "max_model_len": null,
+  "local_repo_path": null,
+  "init_strategy": null,
+  "template": "llava_siglip2_qwen3",
+  "system": null,
+  "max_length": 16384,
+  "truncation_strategy": "delete",
+  "max_pixels": null,
+  "agent_template": null,
+  "norm_bbox": null,
+  "use_chat_template": true,
+  "padding_side": "right",
+  "padding_free": true,
+  "loss_scale": "default",
+  "sequence_parallel_size": 1,
+  "template_backend": "swift",
+  "response_prefix": null,
+  "enable_thinking": null,
+  "add_non_thinking_prefix": true,
+  "dataset": [],
+  "val_dataset": [],
+  "cached_dataset": [
+    "/mnt/bn/strategy-mllm-train/common/datasets/vmllm_cached/image_10pct/train"
+  ],
+  "cached_val_dataset": [],
+  "split_dataset_ratio": 0.0,
+  "dataset_num_proc": 1,
+  "load_from_cache_file": false,
+  "dataset_shuffle": true,
+  "val_dataset_shuffle": false,
+  "streaming": false,
+  "interleave_prob": null,
+  "stopping_strategy": "first_exhausted",
+  "shuffle_buffer_size": 1000,
+  "download_mode": "reuse_dataset_if_exists",
+  "columns": {},
+  "strict": false,
+  "model_name": null,
+  "model_author": null,
+  "custom_dataset_info": [],
+  "quant_method": null,
+  "quant_bits": null,
+  "hqq_axis": null,
+  "bnb_4bit_compute_dtype": "bfloat16",
+  "bnb_4bit_quant_type": "nf4",
+  "bnb_4bit_use_double_quant": true,
+  "bnb_4bit_quant_storage": null,
+  "max_new_tokens": 64,
+  "temperature": 0.0,
+  "top_k": null,
+  "top_p": null,
+  "repetition_penalty": null,
+  "num_beams": 1,
+  "stream": false,
+  "stop_words": [],
+  "logprobs": false,
+  "top_logprobs": null,
+  "structured_outputs_regex": null,
+  "train_type": null,
+  "adapters": [],
+  "external_plugins": [
+    "video_mllm/model_plugin.py",
+    "video_mllm/dataset_plugin.py"
+  ],
+  "custom_register_path": [],
+  "model_kwargs": {},
+  "load_args": false,
+  "load_data_args": false,
+  "packing": true,
+  "packing_length": 16384,
+  "packing_num_proc": 1,
+  "lazy_tokenize": false,
+  "use_hf": true,
+  "ignore_args_error": false,
+  "use_swift_lora": false,
+  "freeze_parameters": [],
+  "freeze_parameters_regex": null,
+  "freeze_parameters_ratio": 0.0,
+  "trainable_parameters": [
+    "model.multi_modal_projector"
+  ],
+  "trainable_parameters_regex": null,
+  "freeze_llm": false,
+  "freeze_vit": false,
+  "freeze_aligner": false,
+  "target_modules": [
+    "all-linear"
+  ],
+  "target_regex": null,
+  "target_parameters": null,
+  "modules_to_save": [],
+  "lora_rank": 8,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "lora_bias": "none",
+  "lora_dtype": null,
+  "lorap_lr_ratio": null,
+  "use_rslora": false,
+  "use_dora": false,
+  "lora_ga_batch_size": 2,
+  "lora_ga_iters": 2,
+  "lora_ga_max_length": 1024,
+  "lora_ga_direction": "ArB2r",
+  "lora_ga_scale": "stable",
+  "lora_ga_stable_gamma": 16,
+  "init_weights": true,
+  "fourier_n_frequency": 2000,
+  "fourier_scaling": 300.0,
+  "boft_block_size": 4,
+  "boft_block_num": 0,
+  "boft_n_butterfly_factor": 1,
+  "boft_dropout": 0.0,
+  "vera_rank": 256,
+  "vera_projection_prng_key": 0,
+  "vera_dropout": 0.0,
+  "vera_d_initial": 0.1,
+  "adapter_act": "gelu",
+  "adapter_length": 128,
+  "adalora_target_r": 8,
+  "adalora_init_r": 12,
+  "adalora_tinit": 0,
+  "adalora_tfinal": 0,
+  "adalora_deltaT": 1,
+  "adalora_beta1": 0.85,
+  "adalora_beta2": 0.85,
+  "adalora_orth_reg_weight": 0.5,
+  "llamapro_num_new_blocks": 4,
+  "llamapro_num_groups": null,
+  "reft_layer_key": null,
+  "reft_layers": null,
+  "reft_rank": 4,
+  "reft_intervention_type": "LoreftIntervention",
+  "reft_args": null,
+  "swanlab_token": null,
+  "swanlab_project": "ms-swift",
+  "swanlab_workspace": null,
+  "swanlab_exp_name": null,
+  "swanlab_notification_method": null,
+  "swanlab_webhook_url": null,
+  "swanlab_secret": null,
+  "swanlab_sender_email": null,
+  "swanlab_receiver_email": null,
+  "swanlab_smtp_server": null,
+  "swanlab_smtp_port": null,
+  "swanlab_email_language": "zh",
+  "swanlab_mode": "cloud",
+  "add_version": true,
+  "create_checkpoint_symlink": false,
+  "zero_hpz_partition_size": null,
+  "deepspeed_autotp_size": null,
+  "swift_version": "4.1.0.dev0",
+  "ckpt_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "rank": 0,
+  "global_world_size": 8,
+  "local_world_size": 8,
+  "model_suffix": "checkpoint-2181",
+  "model_info": "ModelInfo(model_type='llava_siglip2_qwen3', model_dir='/opt/tiger/model_cache/checkpoint-2181', torch_dtype=torch.bfloat16, max_model_len=40960, quant_method=None, quant_bits=None, rope_scaling=None, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
+  "model_meta": "ModelMeta(model_type='llava_siglip2_qwen3', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-0.6B', hf_model_id='Qwen/Qwen3-0.6B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-1.7B', hf_model_id='Qwen/Qwen3-1.7B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-4B', hf_model_id='Qwen/Qwen3-4B', model_path=None, ms_revision=None, hf_revision=None)], template=None, ignore_patterns=None, requires=None, tags=[])], loader=<class 'model_plugin.SigLIP2Qwen3Loader'>, template='llava_siglip2_qwen3', model_arch=MultiModelKeys(arch_name='llava_hf', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['model.language_model', 'lm_head'], aligner=['model.multi_modal_projector'], vision_tower=['model.vision_tower'], generator=[]), architectures=['LlavaOnevisionForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=[], tags=['vision', 'video'])",
+  "model_dir": "/opt/tiger/model_cache/checkpoint-2181",
+  "template_meta": "QwenTemplateMeta(template_type='llava_siglip2_qwen3', prefix=[], prompt=['<|im_start|>user\\n{{QUERY}}<|im_end|>\\n<|im_start|>assistant\\n'], chat_sep=['<|im_end|>\\n'], suffix=['<|im_end|>\\n'], template_cls=<class 'model_plugin.SigLIP2LlavaTemplate'>, system_prefix=['<|im_start|>system\\n{{SYSTEM}}<|im_end|>\\n'], default_system=None, auto_add_bos=False, stop_words=['<|endoftext|>'], agent_template='hermes', is_thinking=False, thinking_prefix='', non_thinking_prefix='', history_thinking_prefix='')",
+  "_val_dataset_exists": false,
+  "hub": "<class 'swift.hub.hub.HFHub'>",
+  "evaluation_strategy": "steps",
+  "training_args": "Seq2SeqTrainingArguments(output_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215', per_device_train_batch_size=1, num_train_epochs=3.0, max_steps=500, learning_rate=1e-05, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_steps=0.05, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, optim_target_modules=None, gradient_accumulation_steps=8, average_tokens_across_devices=None, max_grad_norm=1.0, label_smoothing_factor=0.0, bf16=True, fp16=False, bf16_full_eval=False, fp16_full_eval=False, tf32=None, gradient_checkpointing=True, gradient_checkpointing_kwargs={'use_reentrant': False}, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, use_liger_kernel=False, liger_kernel_config=None, use_cache=False, neftune_noise_alpha=None, torch_empty_cache_steps=None, auto_find_batch_size=False, logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_steps=1, logging_first_step=True, log_on_each_node=True, logging_nan_inf_filter=True, include_num_input_tokens_seen=None, log_level='passive', log_level_replica='warning', disable_tqdm=False, report_to=[], run_name='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215', project='huggingface', trackio_space_id='trackio', eval_strategy=<IntervalStrategy.NO: 'no'>, eval_steps=100.0, eval_delay=0, per_device_eval_batch_size=1, prediction_loss_only=False, eval_on_start=False, eval_do_concat_batches=True, eval_use_gather_object=False, eval_accumulation_steps=None, include_for_metrics=[], batch_eval_metrics=False, save_only_model=False, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=100, save_on_each_node=False, save_total_limit=2, enable_jit_checkpoint=False, push_to_hub=False, hub_token=None, hub_private_repo=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_always_push=False, hub_revision=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, restore_callback_states_from_checkpoint=False, full_determinism=False, seed=42, data_seed=42, use_cpu=False, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), parallelism_config=None, dataloader_drop_last=False, dataloader_num_workers=4, dataloader_pin_memory=True, dataloader_persistent_workers=False, dataloader_prefetch_factor=4, remove_unused_columns=False, label_names=None, train_sampling_strategy='random', length_column_name='length', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, ddp_backend=None, ddp_timeout=7200, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 2, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'allgather_partitions': True, 'allgather_bucket_size': 200000000.0, 'overlap_comm': False, 'reduce_scatter': True, 'reduce_bucket_size': 200000000.0, 'contiguous_gradients': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, debug=[], skip_memory_metrics=True, do_train=False, do_eval=False, do_predict=False, resume_from_checkpoint=None, warmup_ratio=0.05, logging_dir='/mnt/bn/strategy-mllm-train/user/wangjunjie/code/xiaomoguhzz/exps/video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/runs', local_rank=0, sortish_sampler=False, predict_with_generate=False, generation_max_length=None, generation_num_beams=None, generation_config=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, safe_serialization=True, max_shard_size='5GB', check_model=True, acc_strategy='token', train_dataloader_shuffle=True, group_by_length=False, max_epochs=None, aligner_lr=None, vit_lr=1e-06, use_logits_to_keep=None, ds3_gather_for_generation=True, resume_only_model=False, optimizer='multimodal', loss_type=None, eval_metric=None, callbacks=[], early_stop_interval=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, tuner_type='full', use_galore=False, galore_target_modules=None, galore_rank=128, galore_update_proj_gap=50, galore_scale=1.0, galore_proj_type='std', galore_optim_per_parameter=False, galore_with_embedding=False, galore_quantization=False, galore_proj_quant=False, galore_proj_bits=4, galore_proj_group_size=256, galore_cos_threshold=0.4, galore_gamma_proj=2, galore_queue_size=5, lisa_activated_layers=0, lisa_step_interval=20, use_flash_ckpt=False)"
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/config.json

@@ -0,0 +1,248 @@
+{
+  "architectures": [
+    "LlavaOnevisionForConditionalGeneration"
+  ],
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "hidden_size": 2048,
+  "image_grid_pinpoints": [
+    [
+      384,
+      384
+    ],
+    [
+      384,
+      768
+    ],
+    [
+      384,
+      1152
+    ],
+    [
+      384,
+      1536
+    ],
+    [
+      384,
+      1920
+    ],
+    [
+      384,
+      2304
+    ],
+    [
+      768,
+      384
+    ],
+    [
+      768,
+      768
+    ],
+    [
+      768,
+      1152
+    ],
+    [
+      768,
+      1536
+    ],
+    [
+      768,
+      1920
+    ],
+    [
+      768,
+      2304
+    ],
+    [
+      1152,
+      384
+    ],
+    [
+      1152,
+      768
+    ],
+    [
+      1152,
+      1152
+    ],
+    [
+      1152,
+      1536
+    ],
+    [
+      1152,
+      1920
+    ],
+    [
+      1152,
+      2304
+    ],
+    [
+      1536,
+      384
+    ],
+    [
+      1536,
+      768
+    ],
+    [
+      1536,
+      1152
+    ],
+    [
+      1536,
+      1536
+    ],
+    [
+      1536,
+      1920
+    ],
+    [
+      1536,
+      2304
+    ],
+    [
+      1920,
+      384
+    ],
+    [
+      1920,
+      768
+    ],
+    [
+      1920,
+      1152
+    ],
+    [
+      1920,
+      1536
+    ],
+    [
+      1920,
+      1920
+    ],
+    [
+      1920,
+      2304
+    ],
+    [
+      2304,
+      384
+    ],
+    [
+      2304,
+      768
+    ],
+    [
+      2304,
+      1152
+    ],
+    [
+      2304,
+      1536
+    ],
+    [
+      2304,
+      1920
+    ],
+    [
+      2304,
+      2304
+    ]
+  ],
+  "image_token_index": 151669,
+  "keys_to_ignore_at_inference": [
+    "past_key_values"
+  ],
+  "model_type": "llava_onevision",
+  "multimodal_projector_bias": true,
+  "pad_token_id": 151643,
+  "projector_hidden_act": "gelu",
+  "text_config": {
+    "_name_or_path": "/home/tiger/.cache/huggingface/hub/models--Qwen--Qwen3-1.7B/snapshots/70d244cc86ccca08cf5af4e1e306ecf908b1ad5e",
+    "architectures": [
+      "Qwen3ForCausalLM"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 151643,
+    "dtype": "bfloat16",
+    "eos_token_id": 151645,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 2048,
+    "initializer_range": 0.02,
+    "intermediate_size": 6144,
+    "layer_types": [
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 40960,
+    "max_window_layers": 28,
+    "model_type": "qwen3",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 28,
+    "num_key_value_heads": 8,
+    "pad_token_id": 151643,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "rope_theta": 1000000,
+      "rope_type": "default"
+    },
+    "sliding_window": null,
+    "tie_word_embeddings": true,
+    "use_cache": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.2.0",
+  "use_cache": false,
+  "video_token_index": 151670,
+  "vision_aspect_ratio": "anyres_max_9",
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 26,
+    "patch_size": 14,
+    "vision_use_head": false
+  },
+  "vision_feature_layer": -1,
+  "vision_feature_select_strategy": "full"
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "transformers_version": "5.2.0",
+  "use_cache": true
+}

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b6f7108aa92babd57bc68d7e3c494107421bcc43a4f753473912dc4313ce92d
+size 3654094853

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6341bf82b810d59c875955c6ef7ea49dff070a6a5226d2fe09313410416dfee
+size 3654085701

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c119a987b2753a67bc5217be2ea83813b05ad6cba0d0702fc6faccb75fe2140b
+size 3654083653

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d6108d8f82c4d80ebcf2e80640ae747d44054cb939e843e46b1c819fdb0468b
+size 3654085317

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bc1da87848a5ce9fe6c3666785d1991bab6d3eba25ee218e74a16431c9a8a35
+size 3654085253

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aaab50a25a3443d520429d546f01e6f3af722ce44442a2876464c20cdffb717f
+size 3654085253

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a3ed640aa1f7493f71a13f2f640da901cef99d43971529f07afa691a38000cc
+size 3654085829

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4239cd7d02795789a3173d35f9e6a23c9fdfbeacd04090904b79b7a36dace193
+size 3654081925

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/global_step400/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9883708414a1d3633701a95e9e776c115e32ff8a19ef96a0b6eec1633882f18
+size 4872342181

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/latest

@@ -0,0 +1 @@
+global_step400

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eeb86520f0dc7205bbebbef63aada47f5effd4f3ed95bfeb0164abdf29959313
+size 4872193968

video_mllm_swift/s2_image_only_10pct/v1-20260316-135215/checkpoint-400/processor_config.json

@@ -0,0 +1,206 @@
+{
+  "image_processor": {
+    "data_format": "channels_first",
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_pad": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_grid_pinpoints": [
+      [
+        384,
+        384
+      ],
+      [
+        384,
+        768
+      ],
+      [
+        384,
+        1152
+      ],
+      [
+        384,
+        1536
+      ],
+      [
+        384,
+        1920
+      ],
+      [
+        384,
+        2304
+      ],
+      [
+        768,
+        384
+      ],
+      [
+        768,
+        768
+      ],
+      [
+        768,
+        1152
+      ],
+      [
+        768,
+        1536
+      ],
+      [
+        768,
+        1920
+      ],
+      [
+        768,
+        2304
+      ],
+      [
+        1152,
+        384
+      ],
+      [
+        1152,
+        768
+      ],
+      [
+        1152,
+        1152
+      ],
+      [
+        1152,
+        1536
+      ],
+      [
+        1152,
+        1920
+      ],
+      [
+        1152,
+        2304
+      ],
+      [
+        1536,
+        384
+      ],
+      [
+        1536,
+        768
+      ],
+      [
+        1536,
+        1152
+      ],
+      [
+        1536,
+        1536
+      ],
+      [
+        1536,
+        1920
+      ],
+      [
+        1536,
+        2304
+      ],
+      [
+        1920,
+        384
+      ],
+      [
+        1920,
+        768
+      ],
+      [
+        1920,
+        1152
+      ],
+      [
+        1920,
+        1536
+      ],
+      [
+        1920,
+        1920
+      ],
+      [
+        1920,
+        2304
+      ],
+      [
+        2304,
+        384
+      ],
+      [
+        2304,
+        768
+      ],
+      [
+        2304,
+        1152
+      ],
+      [
+        2304,
+        1536
+      ],
+      [
+        2304,
+        1920
+      ],
+      [
+        2304,
+        2304
+      ]
+    ],
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_processor_type": "LlavaOnevisionImageProcessorFast",
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "height": 384,
+      "width": 384
+    }
+  },
+  "image_token": "<image>",
+  "num_image_tokens": 729,
+  "processor_class": "LlavaOnevisionProcessor",
+  "video_processor": {
+    "data_format": "channels_first",
+    "default_to_square": false,
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "do_sample_frames": false,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_metadata": false,
+    "size": {
+      "height": 384,
+      "width": 384
+    },
+    "video_processor_type": "LlavaOnevisionVideoProcessor"
+  },
+  "video_token": "<video>",
+  "vision_aspect_ratio": "anyres_max_9",
+  "vision_feature_select_strategy": null
+}