Add files using upload-large-folder tool
Browse files- checkpoint.pth +3 -0
- config.json +63 -0
checkpoint.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ffda05fe89da03029cbf35fd793c5419d16de69bffe6ba6d161db6b81d943ee0
|
| 3 |
+
size 10881688184
|
config.json
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": {
|
| 3 |
+
"pose_dir": "data/kinetics_processed",
|
| 4 |
+
"desc_dir": "data/kinetics_full_output/descriptions",
|
| 5 |
+
"video_dir": "data/kinetics-dataset/k700-2020",
|
| 6 |
+
"val_split": 0.02,
|
| 7 |
+
"max_samples_per_class": null,
|
| 8 |
+
"num_workers": 8,
|
| 9 |
+
"image_size": 224,
|
| 10 |
+
"normalize_pose": true,
|
| 11 |
+
"use_joint_angles": true,
|
| 12 |
+
"sample_stride": 16,
|
| 13 |
+
"include_temporal_context": false,
|
| 14 |
+
"action_focus_prompt": false,
|
| 15 |
+
"video_fps": 10.0,
|
| 16 |
+
"augment_flip": true,
|
| 17 |
+
"seed": 42
|
| 18 |
+
},
|
| 19 |
+
"model_config": {
|
| 20 |
+
"qwen_model_name": "/root/.cache/huggingface/hub/models--Qwen--Qwen3-VL-4B-Instruct/snapshots/ebb281ec70b05090aa6165b016eac8ec08e71b17",
|
| 21 |
+
"qwen_hidden_size": 2560,
|
| 22 |
+
"use_intermediate_hidden": true,
|
| 23 |
+
"hidden_layer_fraction": 0.5,
|
| 24 |
+
"hidden_layer_index": 18,
|
| 25 |
+
"use_early_exit": true,
|
| 26 |
+
"use_deepstack_features": true,
|
| 27 |
+
"use_flash_attention": true,
|
| 28 |
+
"projection_dim": 1024,
|
| 29 |
+
"action_dim": 44,
|
| 30 |
+
"diffusion_hidden_dim": 1536,
|
| 31 |
+
"num_diffusion_layers": 24,
|
| 32 |
+
"num_diffusion_heads": 24,
|
| 33 |
+
"num_future_tokens": 4,
|
| 34 |
+
"action_horizon": 16,
|
| 35 |
+
"num_frames": 4,
|
| 36 |
+
"use_lora": true,
|
| 37 |
+
"lora_rank": 128,
|
| 38 |
+
"lora_alpha": 128,
|
| 39 |
+
"lora_dropout": 0.05,
|
| 40 |
+
"freeze_vision_encoder": true,
|
| 41 |
+
"freeze_qwen_layers": 0,
|
| 42 |
+
"use_thinking_mode": false,
|
| 43 |
+
"diffusion_steps": 2,
|
| 44 |
+
"init_from_current_pose": false
|
| 45 |
+
},
|
| 46 |
+
"learning_rate": 1e-05,
|
| 47 |
+
"llm_learning_rate": 1e-06,
|
| 48 |
+
"weight_decay": 0.01,
|
| 49 |
+
"batch_size": 8,
|
| 50 |
+
"num_epochs": 4,
|
| 51 |
+
"unfreeze_pct": 0.5,
|
| 52 |
+
"gradient_accumulation_steps": 16,
|
| 53 |
+
"gradient_clip": 1.0,
|
| 54 |
+
"use_amp": true,
|
| 55 |
+
"gradient_checkpointing": false,
|
| 56 |
+
"log_every_n_steps": 1,
|
| 57 |
+
"save_every_n_steps": 750,
|
| 58 |
+
"val_max_batches": 100,
|
| 59 |
+
"perf_log_every": 100,
|
| 60 |
+
"max_checkpoints": 5,
|
| 61 |
+
"log_dir": "logs/kinetics_vla",
|
| 62 |
+
"checkpoint_dir": "checkpoints/kinetics_vla"
|
| 63 |
+
}
|