update
Browse files- VideoLLaMA-7B-ActivityNet-VTune.pth +3 -0
- log.txt +105 -0
VideoLLaMA-7B-ActivityNet-VTune.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7ff8913891b40a17e7368bf28cc949972bdb30d696faedce625e7d8ff07e8524
|
3 |
+
size 265436196
|
log.txt
ADDED
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"run": {
|
3 |
+
"task": "video_text_pretrain",
|
4 |
+
"lr_sched": "linear_warmup_cosine_lr",
|
5 |
+
"init_lr": 3e-05,
|
6 |
+
"min_lr": 1e-05,
|
7 |
+
"warmup_lr": 1e-06,
|
8 |
+
"weight_decay": 0.05,
|
9 |
+
"max_epoch": 3,
|
10 |
+
"iters_per_epoch": 51377,
|
11 |
+
"batch_size_train": 1,
|
12 |
+
"batch_size_eval": 4,
|
13 |
+
"num_workers": 4,
|
14 |
+
"warmup_steps": 25688,
|
15 |
+
"seed": 42,
|
16 |
+
"output_dir": "/data/mjjung/Video-LLaMA/videollama_activitynet_vtune",
|
17 |
+
"amp": true,
|
18 |
+
"resume_ckpt_path": null,
|
19 |
+
"evaluate": false,
|
20 |
+
"train_splits": [
|
21 |
+
"train"
|
22 |
+
],
|
23 |
+
"device": "cuda",
|
24 |
+
"world_size": 4,
|
25 |
+
"dist_url": "env://",
|
26 |
+
"distributed": true,
|
27 |
+
"rank": 0,
|
28 |
+
"gpu": 0,
|
29 |
+
"dist_backend": "nccl"
|
30 |
+
},
|
31 |
+
"model": {
|
32 |
+
"arch": "video_llama",
|
33 |
+
"image_size": 224,
|
34 |
+
"drop_path_rate": 0,
|
35 |
+
"use_grad_checkpoint": false,
|
36 |
+
"vit_precision": "fp16",
|
37 |
+
"freeze_vit": true,
|
38 |
+
"freeze_qformer": true,
|
39 |
+
"num_query_token": 32,
|
40 |
+
"llama_model": "/data/mjjung/vicuna-7b-v1.5",
|
41 |
+
"prompt": "",
|
42 |
+
"model_type": "pretrain_vicuna",
|
43 |
+
"imagebind_ckpt_path": "/data/mjjung/Video-LLaMA/Video-LLaMA-2-7B-Finetuned/",
|
44 |
+
"ckpt": "/data/mjjung/Video-LLaMA/Video-LLaMA-2-7B-Finetuned/VL_LLaMA_2_7B_Finetuned.pth",
|
45 |
+
"equip_audio_branch": false,
|
46 |
+
"frozen_llama_proj": false,
|
47 |
+
"frozen_video_Qformer": false,
|
48 |
+
"frozen_audio_Qformer": true,
|
49 |
+
"fusion_head_layers": 2,
|
50 |
+
"max_frame_pos": 32,
|
51 |
+
"fusion_header_type": "seqTransf",
|
52 |
+
"max_txt_len": 320,
|
53 |
+
"end_sym": "###",
|
54 |
+
"prompt_path": "",
|
55 |
+
"prompt_template": "###Human: {} ###Assistant: "
|
56 |
+
},
|
57 |
+
"preprocess": {
|
58 |
+
"vis_processor": {
|
59 |
+
"train": {
|
60 |
+
"name": "alpro_video_train",
|
61 |
+
"image_size": 224,
|
62 |
+
"n_frms": 8
|
63 |
+
},
|
64 |
+
"eval": {
|
65 |
+
"name": "alpro_video_eval",
|
66 |
+
"image_size": 224,
|
67 |
+
"n_frms": 8
|
68 |
+
}
|
69 |
+
},
|
70 |
+
"text_processor": {
|
71 |
+
"train": {
|
72 |
+
"name": "blip_caption"
|
73 |
+
},
|
74 |
+
"eval": {
|
75 |
+
"name": "blip_caption"
|
76 |
+
}
|
77 |
+
}
|
78 |
+
},
|
79 |
+
"datasets": {
|
80 |
+
"webvid_instruct": {
|
81 |
+
"data_type": "video",
|
82 |
+
"build_info": {
|
83 |
+
"anno_dir": "data/activitynet_filtered_qa_grounding.json",
|
84 |
+
"videos_dir": "/data/video_datasets/"
|
85 |
+
},
|
86 |
+
"vis_processor": {
|
87 |
+
"train": {
|
88 |
+
"name": "alpro_video_train",
|
89 |
+
"n_frms": 8,
|
90 |
+
"image_size": 224
|
91 |
+
}
|
92 |
+
},
|
93 |
+
"text_processor": {
|
94 |
+
"train": {
|
95 |
+
"name": "blip_caption"
|
96 |
+
}
|
97 |
+
},
|
98 |
+
"num_video_query_token": 32,
|
99 |
+
"tokenizer_name": "/data/mjjung/vicuna-7b-v1.5",
|
100 |
+
"model_type": "vicuna"
|
101 |
+
}
|
102 |
+
}
|
103 |
+
}
|
104 |
+
{"train_lr": "0.000", "train_loss": "0.230"}
|
105 |
+
{"train_lr": "0.000", "train_loss": "0.204"}
|