update

TimeChat-7B-ActivityNet-VTune.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1bf7dfc879b35a0afad7f54bd39a415323a78a9c0cba52f77369df64ca5e7cf
+size 2897697291

log.txt

@@ -0,0 +1,113 @@
+{
+    "run": {
+        "task": "video_text_pretrain",
+        "lr_sched": "linear_warmup_cosine_lr",
+        "init_lr": 3e-05,
+        "min_lr": 1e-05,
+        "warmup_lr": 1e-06,
+        "weight_decay": 0.05,
+        "max_epoch": 3,
+        "iters_per_epoch": 51377,
+        "batch_size_train": 2,
+        "batch_size_eval": 4,
+        "num_workers": 4,
+        "warmup_steps": 25688,
+        "accum_grad_iters": 8,
+        "seed": 42,
+        "output_dir": "/data/mjjung/timechat/activitynet_vtune",
+        "amp": true,
+        "resume_ckpt_path": null,
+        "evaluate": false,
+        "train_splits": [
+            "train"
+        ],
+        "device": "cuda",
+        "world_size": 4,
+        "dist_url": "env://",
+        "distributed": true,
+        "rank": 0,
+        "gpu": 0,
+        "dist_backend": "nccl"
+    },
+    "model": {
+        "arch": "timechat",
+        "image_size": 224,
+        "drop_path_rate": 0,
+        "use_grad_checkpoint": true,
+        "vit_precision": "fp16",
+        "freeze_vit": true,
+        "freeze_qformer": false,
+        "num_query_token": 32,
+        "llama_model": "/data/mjjung/Video-LLaMA/Video-LLaMA-2-7B-Finetuned/llama-2-7b-chat-hf/",
+        "prompt": "",
+        "model_type": "pretrain_llama_v2",
+        "vit_model": "/data/mjjung/timechat/eva_vit_g.pth",
+        "q_former_model": "/data/mjjung/timechat/instruct_blip_vicuna7b_trimmed.pth",
+        "ckpt": "/data/mjjung/timechat/timechat_7b.pth",
+        "frozen_llama_proj": false,
+        "frozen_video_Qformer": false,
+        "fusion_head_layers": 2,
+        "max_frame_pos": 96,
+        "fusion_header_type": "seqTransf",
+        "max_txt_len": 2048,
+        "end_sym": "</s>",
+        "prompt_path": "",
+        "prompt_template": "[INST] <<SYS>>\\n \\n<</SYS>>\\n\\n{} [/INST] ",
+        "lora": true,
+        "lora_inference_mode": false,
+        "qformer_text_input": true,
+        "window_size": 32,
+        "stride": 32
+    },
+    "preprocess": {
+        "vis_processor": {
+            "train": {
+                "name": "alpro_video_train",
+                "image_size": 224,
+                "n_frms": 8
+            },
+            "eval": {
+                "name": "alpro_video_eval",
+                "image_size": 224,
+                "n_frms": 8
+            }
+        },
+        "text_processor": {
+            "train": {
+                "name": "blip_caption"
+            },
+            "eval": {
+                "name": "blip_caption"
+            }
+        }
+    },
+    "datasets": {
+        "charades_instruct": {
+            "data_type": "video",
+            "build_info": {
+                "anno_dir": "data/activitynet_filtered_qa_grounding.json",
+                "videos_dir": "/data/video_datasets/"
+            },
+            "vis_processor": {
+                "train": {
+                    "name": "alpro_video_train",
+                    "n_frms": 96,
+                    "image_size": 224
+                }
+            },
+            "text_processor": {
+                "train": {
+                    "name": "blip_caption"
+                }
+            },
+            "num_video_query_token": 32,
+            "tokenizer_name": "/data/mjjung/Video-LLaMA/Video-LLaMA-2-7B-Finetuned/llama-2-7b-chat-hf/",
+            "model_type": "llama_v2",
+            "num_frm": 96,
+            "sample_type": "rand",
+            "max_txt_len": 2048,
+            "stride": 32
+        }
+    }
+}
+{"train_lr": "0.000", "train_loss": "0.162"}