Upload Masked HWM/checkpoint-60000/config.json with huggingface_hub
Browse files
Masked HWM/checkpoint-60000/config.json
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"num_layers": 3,
|
| 3 |
+
"num_heads": 4,
|
| 4 |
+
"d_model": 96,
|
| 5 |
+
"mlp_hidden": 384,
|
| 6 |
+
"mlp_ratio": 4.0,
|
| 7 |
+
"num_past_clips": 2,
|
| 8 |
+
"num_future_clips": 1,
|
| 9 |
+
"num_past_frames": 34,
|
| 10 |
+
"num_future_frames": 17,
|
| 11 |
+
"action_dim": 25,
|
| 12 |
+
"vocab_size": 65536,
|
| 13 |
+
"num_factored_vocabs": 3,
|
| 14 |
+
"spatial_size": 32,
|
| 15 |
+
"temporal_compression": 8,
|
| 16 |
+
"frames_per_clip": 17,
|
| 17 |
+
"shared_layers_start": 1,
|
| 18 |
+
"learning_rate": 0.0005,
|
| 19 |
+
"warmup_steps": 50,
|
| 20 |
+
"max_steps": 100000,
|
| 21 |
+
"batch_size": 2,
|
| 22 |
+
"gradient_accumulation_steps": 8,
|
| 23 |
+
"max_corrupt_rate": 0.2,
|
| 24 |
+
"mask_token_id": 65536,
|
| 25 |
+
"qkv_bias": false,
|
| 26 |
+
"proj_bias": true,
|
| 27 |
+
"attn_drop": 0.0,
|
| 28 |
+
"mlp_drop": 0.0,
|
| 29 |
+
"use_rope": true,
|
| 30 |
+
"use_stream_type_emb": false,
|
| 31 |
+
"init_std": 0.02,
|
| 32 |
+
"use_gradient_checkpointing": true,
|
| 33 |
+
"mixed_precision": "bf16",
|
| 34 |
+
"tokenizer_checkpoint_dir": "/media/skr/storage/robot_world/humanoid_wm/cosmos_tokenizer",
|
| 35 |
+
"train_data_dir": "/media/skr/storage/robot_world/humanoid_wm/1xgpt/data/train_v2.0",
|
| 36 |
+
"val_data_dir": "/media/skr/storage/robot_world/humanoid_wm/1xgpt/data/val_v2.0",
|
| 37 |
+
"test_data_dir": null,
|
| 38 |
+
"seed": 42,
|
| 39 |
+
"save_steps": 1000,
|
| 40 |
+
"eval_steps": 1000,
|
| 41 |
+
"logging_steps": 100
|
| 42 |
+
}
|