checkpoint step 2500 epoch 1

0324_1332_checkpoint_step_2500_epoch_1/config.json

@@ -0,0 +1,95 @@
+{
+  "epoch": 1,
+  "global_step": 2500,
+  "loss": 0.03748499974608421,
+  "training_config": {
+    "llm_backbone": "dasheng",
+    "qwen3_name": "Qwen/Qwen3-Reranker-0.6B",
+    "dasheng_name": "mispeech/midashenglm-7b-0804-fp32",
+    "dasheng_path": "/workspace/cache/huggingface/dasheng_lm",
+    "trainable_modules": [
+      "backbone",
+      "dasheng",
+      "dasheng_down",
+      "dasheng_proj",
+      "lm_head"
+    ],
+    "use_lora": true,
+    "lora_r": 16,
+    "lora_alpha": 32,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj"
+    ],
+    "use_dasheng_lora": false,
+    "dasheng_lora_r": 8,
+    "dasheng_lora_alpha": 16,
+    "dasheng_lora_dropout": 0.1,
+    "dasheng_lora_target_modules": null,
+    "train_layer_ratio": 1.0,
+    "train_layer_strategy": "last_n",
+    "output_dim": null,
+    "output_identity": false,
+    "use_logit_scale": true,
+    "use_checkpointing": true,
+    "checkpoint_reentrant": false,
+    "s3_base_path": "https://d2j287p0ytux1o.cloudfront.net",
+    "dataset_config": "/workspace/SpeechRAG_exp/dataset_configs/setting_multi_task_reranker.json",
+    "aws_profile": "test_user",
+    "cache_dir": "/workspace/cache/huggingface",
+    "enable_audio_cache": true,
+    "audio_cache_dir": "/workspace/cache/huggingface/audio_cache",
+    "target_sr": 16000,
+    "mono": true,
+    "max_query_audio_length": 45.0,
+    "max_doc_audio_length": 45.0,
+    "max_query_text_length": 1000,
+    "max_doc_text_length": 1000,
+    "eval_max_query_audio_length": null,
+    "eval_max_doc_audio_length": null,
+    "eval_max_query_text_length": null,
+    "eval_max_doc_text_length": null,
+    "loss_type": "pointwise",
+    "batch_size": 4,
+    "num_epochs": 10,
+    "learning_rate": 0.0001,
+    "gradient_accumulation_steps": 4,
+    "weight_decay": 0.001,
+    "optimizer_bits": "default",
+    "num_workers": 4,
+    "train_batch_task_mode": "single_task",
+    "task_batch_ratio": "{\"semantic\": 0.5, \"cross\": 0.5}",
+    "save_dir": "checkpoints",
+    "save_steps": 500,
+    "keep_checkpoints": 1,
+    "upload_steps": 500,
+    "upload_repo_id": "jdosjcd/reranker_checkpoint",
+    "log_dir": "logs",
+    "log_steps": 5,
+    "mixed_precision": "bf16",
+    "use_deepspeed": false,
+    "deepspeed_config": null,
+    "use_fsdp": false,
+    "fsdp_config": null,
+    "scheduler_type": "warmup_cosine_decay",
+    "warmup_steps": 500,
+    "warmup_ratio": 0.1,
+    "min_lr": 1e-05,
+    "eval_steps": 500,
+    "eval_batch_size": 8,
+    "eval_retrieval_mode": "embedding_topk",
+    "eval_embedding_checkpoint_dir": "/workspace/SpeechRAG_exp/checkpoints_512/checkpoint_step_1700_epoch_1",
+    "eval_embedding_cache_path": null,
+    "eval_embedding_use_ema_model": false,
+    "eval_embedding_top_k": 25,
+    "eval_embedding_batch_size": null,
+    "resume_from": null,
+    "add_speaker_mask": false,
+    "positive_token": "yes",
+    "negative_token": "no",
+    "pointwise_batch_mode": "offline",
+    "pointwise_num_negatives": 0
+  }
+}

0324_1332_checkpoint_step_2500_epoch_1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe190397fb4ba42c297b8ac65467c41cfbc55877cea06da506a9ff21268b1a75
+size 33154937619

0324_1332_checkpoint_step_2500_epoch_1/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f988a445e39e513515ead4ac03a6732d3ea401d049f5d14193cc9ac7f140fbb
+size 9744503428

0324_1332_checkpoint_step_2500_epoch_1/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:328070cbbb97d33f78a16c9002ec8a914ec4325ea6f8aff9a436544dca414432
+size 1401