Upload args.json with huggingface_hub
Browse files
args.json
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"exp_id": "vidcab_ret_m",
|
| 3 |
+
"debug": false,
|
| 4 |
+
"llm_model": "gpt2-medium",
|
| 5 |
+
"llm_8bit": false,
|
| 6 |
+
"freeze_lm": false,
|
| 7 |
+
"lora": false,
|
| 8 |
+
"vis_model": "openai/clip-vit-base-patch32",
|
| 9 |
+
"precision": "bf16",
|
| 10 |
+
"n_visual_tokens": 1,
|
| 11 |
+
"rand_init": false,
|
| 12 |
+
"num_layers": -1,
|
| 13 |
+
"hidden_dim": 1152,
|
| 14 |
+
"nheads": 8,
|
| 15 |
+
"dim_feedforward": 2048,
|
| 16 |
+
"dropout": 0.1,
|
| 17 |
+
"droppath": 0.1,
|
| 18 |
+
"vis_pooling": false,
|
| 19 |
+
"vis_query_pooling": false,
|
| 20 |
+
"last_vis_mean": false,
|
| 21 |
+
"vocab_model": "google/siglip-so400m-patch14-384",
|
| 22 |
+
"dataset_dir": "/blob/v-lqinghong/data/Ego_database",
|
| 23 |
+
"log_base_dir": "/blob/v-lqinghong/experiments/VLog",
|
| 24 |
+
"dataset": "ret",
|
| 25 |
+
"metadata": "egoclip_vidcab",
|
| 26 |
+
"fullset": false,
|
| 27 |
+
"val_dataset": "ret",
|
| 28 |
+
"val_metadata": "egoclip_vidcab",
|
| 29 |
+
"workers": 16,
|
| 30 |
+
"visual_input": "feature",
|
| 31 |
+
"image_size": 224,
|
| 32 |
+
"num_frame": 1,
|
| 33 |
+
"add_special_tokens": false,
|
| 34 |
+
"num_history": 0,
|
| 35 |
+
"past_len": 0,
|
| 36 |
+
"train_narrator": "narration_pass_2",
|
| 37 |
+
"add_eos": true,
|
| 38 |
+
"max_len": 128,
|
| 39 |
+
"max_len_eval": 128,
|
| 40 |
+
"max_clip_len": 128,
|
| 41 |
+
"temperature": 0.7,
|
| 42 |
+
"epochs": 10,
|
| 43 |
+
"start_epoch": 0,
|
| 44 |
+
"steps_per_epoch": -1,
|
| 45 |
+
"val_steps_per_epoch": -1,
|
| 46 |
+
"batch_size": 32,
|
| 47 |
+
"val_batch_size": 32,
|
| 48 |
+
"print_freq": 1,
|
| 49 |
+
"resume": "",
|
| 50 |
+
"evaluate": false,
|
| 51 |
+
"only_best": true,
|
| 52 |
+
"do_sample": false,
|
| 53 |
+
"lock_lm": false,
|
| 54 |
+
"lr": 0.0003,
|
| 55 |
+
"beta1": 0.9,
|
| 56 |
+
"beta2": 0.95,
|
| 57 |
+
"weight_decay": 0.0,
|
| 58 |
+
"lr_warmup_steps": 100,
|
| 59 |
+
"lr_schedule_step_size": 10,
|
| 60 |
+
"lr_schedule_gamma": 0.1,
|
| 61 |
+
"grad_accumulation_steps": 1,
|
| 62 |
+
"grad_clip": 1.0,
|
| 63 |
+
"loss": "nce",
|
| 64 |
+
"nce_temperature": 0.05,
|
| 65 |
+
"scale_lm_loss": 1.0,
|
| 66 |
+
"scale_ret_loss": 1.0,
|
| 67 |
+
"train_class": false,
|
| 68 |
+
"main_node": true,
|
| 69 |
+
"world_size": -1,
|
| 70 |
+
"local_rank": -1,
|
| 71 |
+
"dist_url": "tcp://localhost:44122",
|
| 72 |
+
"dist_backend": "nccl",
|
| 73 |
+
"seed": null,
|
| 74 |
+
"gpu": null,
|
| 75 |
+
"multiprocessing_distributed": false,
|
| 76 |
+
"log_dir": "/blob/v-lqinghong/experiments/VLog/vidcab_ret_m/2025-03-08_12-48-28"
|
| 77 |
+
}
|