Create config.json
Browse files- config.json +124 -0
config.json
ADDED
|
@@ -0,0 +1,124 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "vitamin_large",
|
| 3 |
+
"exp_name": "meditok_s2_clipv01",
|
| 4 |
+
"output_dir": "meditok/outputs/ckpts/meditok_s2_clipv01",
|
| 5 |
+
"resume_from": "",
|
| 6 |
+
"resume_net_only": false,
|
| 7 |
+
"lpips_path": "external/lpips_with_vgg.pth",
|
| 8 |
+
"dino_path": "external/dinov2_vits14_pretrain.pth",
|
| 9 |
+
"fid_eval_src": "",
|
| 10 |
+
"fid_eval_dst": "",
|
| 11 |
+
"vis_img_dir": "assets/vis_imgs/",
|
| 12 |
+
"fid_feature_extractor": "external/weights-inception-2015-12-05-6726825d.pth",
|
| 13 |
+
"clip_pretrain_path": "",
|
| 14 |
+
"fp16": false,
|
| 15 |
+
"bf16": true,
|
| 16 |
+
"tf32": true,
|
| 17 |
+
"compile_model": false,
|
| 18 |
+
"ddp_static": false,
|
| 19 |
+
"grad_ckpt": true,
|
| 20 |
+
"grad_accu": 1,
|
| 21 |
+
"train_data": "meditok/datasets/meta/meta2d_train.csv",
|
| 22 |
+
"val_data": null,
|
| 23 |
+
"train_root": "datasets/medical_preprocessed",
|
| 24 |
+
"val_root": null,
|
| 25 |
+
"dataset_type": "csvimg",
|
| 26 |
+
"csv_img_key": "identifier",
|
| 27 |
+
"csv_caption_key": "caption",
|
| 28 |
+
"csv_separator": ",",
|
| 29 |
+
"subset_ratio": 1.0,
|
| 30 |
+
"img_size": 256,
|
| 31 |
+
"resize_ratio": 1.125,
|
| 32 |
+
"workers": 8,
|
| 33 |
+
"train_num_samples": null,
|
| 34 |
+
"train_data_upsampling_factors": null,
|
| 35 |
+
"val_num_samples": null,
|
| 36 |
+
"dataset_resampled": false,
|
| 37 |
+
"use_aug": false,
|
| 38 |
+
"prob_flip": 0.5,
|
| 39 |
+
"prob_rot": 0.5,
|
| 40 |
+
"prob_grayscale": 0.5,
|
| 41 |
+
"vocab_size": 32768,
|
| 42 |
+
"vocab_width": 64,
|
| 43 |
+
"vocab_norm": true,
|
| 44 |
+
"vq_beta": 0.25,
|
| 45 |
+
"num_codebooks": 8,
|
| 46 |
+
"quant_proj": "attn",
|
| 47 |
+
"embed_dim": 768,
|
| 48 |
+
"num_query": 0,
|
| 49 |
+
"use_clip_pretrain": false,
|
| 50 |
+
"patch_size": 16,
|
| 51 |
+
"drop_path": 0.5,
|
| 52 |
+
"text_width": 768,
|
| 53 |
+
"text_heads": 12,
|
| 54 |
+
"text_layers": 12,
|
| 55 |
+
"text_vocab_size": 49408,
|
| 56 |
+
"text_context_length": 77,
|
| 57 |
+
"local_loss": true,
|
| 58 |
+
"gather_with_grad": true,
|
| 59 |
+
"pretrained_clip": null,
|
| 60 |
+
"pretrained_clip_text": null,
|
| 61 |
+
"lock_text": true,
|
| 62 |
+
"force_custom_text": false,
|
| 63 |
+
"force_custom_vision": false,
|
| 64 |
+
"dino_depth": 12,
|
| 65 |
+
"dino_kernel_size": 9,
|
| 66 |
+
"disc_norm": "gn",
|
| 67 |
+
"disc_aug_prob": 1.0,
|
| 68 |
+
"disc_specnorm": false,
|
| 69 |
+
"step_disc_every": 1,
|
| 70 |
+
"vae_init": -0.5,
|
| 71 |
+
"vocab_init": -1,
|
| 72 |
+
"disc_init": -0.5,
|
| 73 |
+
"epoch": 2,
|
| 74 |
+
"local_bs": 512,
|
| 75 |
+
"global_bs": 1,
|
| 76 |
+
"lr": 0.0005,
|
| 77 |
+
"wd": 0.02,
|
| 78 |
+
"disc_lr": 2e-05,
|
| 79 |
+
"disc_wd": 0.2,
|
| 80 |
+
"grad_clip": 10,
|
| 81 |
+
"ema": 0.9999,
|
| 82 |
+
"warmup_iter": null,
|
| 83 |
+
"warmup_ep": 0.01,
|
| 84 |
+
"disc_start_ep": 0.375,
|
| 85 |
+
"disc_warmup_ep": 0.03,
|
| 86 |
+
"schedule": "cos",
|
| 87 |
+
"lr_start_ratio": 0.0,
|
| 88 |
+
"lr_end_ratio": 0.5,
|
| 89 |
+
"disc_lr_end_ratio": 0.5,
|
| 90 |
+
"custom_lr_multiplier": null,
|
| 91 |
+
"optimizer": "adamw",
|
| 92 |
+
"optim_eps": 1e-06,
|
| 93 |
+
"fuse_opt": false,
|
| 94 |
+
"optim_beta": "0.9_0.95",
|
| 95 |
+
"disc_optim_beta": "0.5_0.9",
|
| 96 |
+
"l1": 0.2,
|
| 97 |
+
"l2": 1.0,
|
| 98 |
+
"lp": 1.0,
|
| 99 |
+
"lpr": 48,
|
| 100 |
+
"ld": 0.4,
|
| 101 |
+
"le": 0.0,
|
| 102 |
+
"lq": 1.0,
|
| 103 |
+
"lc": 1.0,
|
| 104 |
+
"lcos": 0.0,
|
| 105 |
+
"e_temp": 0.01,
|
| 106 |
+
"gada": 1,
|
| 107 |
+
"bcr": 4.0,
|
| 108 |
+
"bcr_cut": 0.2,
|
| 109 |
+
"dcrit": "hg",
|
| 110 |
+
"report_wandb": false,
|
| 111 |
+
"wandb_notes": null,
|
| 112 |
+
"run_id": null,
|
| 113 |
+
"eval_per_epoch": 10,
|
| 114 |
+
"dbg_unused_param": false,
|
| 115 |
+
"dbg_nan": false,
|
| 116 |
+
"seed": null,
|
| 117 |
+
"deterministic": false,
|
| 118 |
+
"use_biomedclip": true,
|
| 119 |
+
"vision_as_text": false,
|
| 120 |
+
"ignore_text_params": true,
|
| 121 |
+
"ct_bias": 1024,
|
| 122 |
+
"freeze_logit_scale": false,
|
| 123 |
+
"pretrained_core_path": null
|
| 124 |
+
}
|