Upload 24_10_27_15_11_26_log.txt with huggingface_hub
Browse files- 24_10_27_15_11_26_log.txt +115 -0
24_10_27_15_11_26_log.txt
ADDED
|
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"run": {
|
| 3 |
+
"task": "image_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": 50,
|
| 10 |
+
"iters_per_epoch": 600,
|
| 11 |
+
"batch_size_train": 15,
|
| 12 |
+
"batch_size_eval": 1,
|
| 13 |
+
"num_workers": 4,
|
| 14 |
+
"warmup_steps": 200,
|
| 15 |
+
"seed": 42,
|
| 16 |
+
"output_dir": "output/xraygpt_vqa_finetune",
|
| 17 |
+
"amp": true,
|
| 18 |
+
"resume_ckpt_path": null,
|
| 19 |
+
"evaluate": false,
|
| 20 |
+
"train_splits": [
|
| 21 |
+
"train"
|
| 22 |
+
],
|
| 23 |
+
"device": "cuda",
|
| 24 |
+
"world_size": 2,
|
| 25 |
+
"dist_url": "env://",
|
| 26 |
+
"distributed": true,
|
| 27 |
+
"rank": 0,
|
| 28 |
+
"gpu": 0,
|
| 29 |
+
"dist_backend": "nccl"
|
| 30 |
+
},
|
| 31 |
+
"model": {
|
| 32 |
+
"arch": "mini_gpt4",
|
| 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": "Joycean0301/Llama-3.2-3B-Instruct-Medical-Conversational",
|
| 41 |
+
"prompt": "",
|
| 42 |
+
"lora_r": 16,
|
| 43 |
+
"lora_alpha": 16,
|
| 44 |
+
"lora_target_modules": [
|
| 45 |
+
"q_proj",
|
| 46 |
+
"k_proj",
|
| 47 |
+
"v_proj",
|
| 48 |
+
"o_proj",
|
| 49 |
+
"gate_proj",
|
| 50 |
+
"up_proj",
|
| 51 |
+
"down_proj"
|
| 52 |
+
],
|
| 53 |
+
"lora_dropout": 0.1,
|
| 54 |
+
"lora_bias": "none",
|
| 55 |
+
"model_type": "pretrain_vicuna",
|
| 56 |
+
"max_txt_len": 160,
|
| 57 |
+
"end_sym": "<|eot_id|>",
|
| 58 |
+
"ckpt": "24_10_26_14_03_09_pretrain100_finetune2_lora_openi.pth"
|
| 59 |
+
},
|
| 60 |
+
"preprocess": {
|
| 61 |
+
"vis_processor": {
|
| 62 |
+
"train": {
|
| 63 |
+
"name": "blip2_image_train",
|
| 64 |
+
"image_size": 224
|
| 65 |
+
},
|
| 66 |
+
"eval": {
|
| 67 |
+
"name": "blip2_image_eval",
|
| 68 |
+
"image_size": 224
|
| 69 |
+
}
|
| 70 |
+
},
|
| 71 |
+
"text_processor": {
|
| 72 |
+
"train": {
|
| 73 |
+
"name": "blip_caption"
|
| 74 |
+
},
|
| 75 |
+
"eval": {
|
| 76 |
+
"name": "blip_caption"
|
| 77 |
+
}
|
| 78 |
+
}
|
| 79 |
+
},
|
| 80 |
+
"datasets": {
|
| 81 |
+
"vqa": {
|
| 82 |
+
"data_type": "images",
|
| 83 |
+
"build_info": {
|
| 84 |
+
"storage": "dataset/llava_med_60k_instruct"
|
| 85 |
+
},
|
| 86 |
+
"vis_processor": {
|
| 87 |
+
"train": {
|
| 88 |
+
"name": "blip2_image_train",
|
| 89 |
+
"image_size": 224
|
| 90 |
+
}
|
| 91 |
+
},
|
| 92 |
+
"text_processor": {
|
| 93 |
+
"train": {
|
| 94 |
+
"name": "blip_caption"
|
| 95 |
+
}
|
| 96 |
+
}
|
| 97 |
+
}
|
| 98 |
+
}
|
| 99 |
+
}
|
| 100 |
+
{"train_lr": "0.000025", "train_loss": "1.129822"}
|
| 101 |
+
{"train_lr": "0.000030", "train_loss": "0.771452"}
|
| 102 |
+
{"train_lr": "0.000030", "train_loss": "0.699119"}
|
| 103 |
+
{"train_lr": "0.000030", "train_loss": "0.650587"}
|
| 104 |
+
{"train_lr": "0.000030", "train_loss": "0.610647"}
|
| 105 |
+
{"train_lr": "0.000029", "train_loss": "0.574458"}
|
| 106 |
+
{"train_lr": "0.000029", "train_loss": "0.542380"}
|
| 107 |
+
{"train_lr": "0.000029", "train_loss": "0.509769"}
|
| 108 |
+
{"train_lr": "0.000029", "train_loss": "0.478779"}
|
| 109 |
+
{"train_lr": "0.000028", "train_loss": "0.448832"}
|
| 110 |
+
{"train_lr": "0.000028", "train_loss": "0.420626"}
|
| 111 |
+
{"train_lr": "0.000028", "train_loss": "0.389205"}
|
| 112 |
+
{"train_lr": "0.000027", "train_loss": "0.360322"}
|
| 113 |
+
{"train_lr": "0.000027", "train_loss": "0.331934"}
|
| 114 |
+
{"train_lr": "0.000026", "train_loss": "0.303854"}
|
| 115 |
+
{"train_lr": "0.000026", "train_loss": "0.276897"}
|