Instructions to use Haongchen/MemoryVLA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Piper
How to use Haongchen/MemoryVLA with Piper:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Add runtime aliases and CogACT base metadata
Browse files
base_models/CogACT-Large/config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"data_root_dir": "/mnt/blob/open_x",
|
| 3 |
+
"diffusion_model_type": "DiT-L",
|
| 4 |
+
"future_action_window_size": 15,
|
| 5 |
+
"hf_token": "HF_TOKEN",
|
| 6 |
+
"image_aug": true,
|
| 7 |
+
"is_resume": false,
|
| 8 |
+
"load_all_data_for_training": true,
|
| 9 |
+
"past_action_window_size": 0,
|
| 10 |
+
"pretrained_checkpoint": "/mnt/blob/vla_model/openvla-7b-prismatic/checkpoints/step-295000-epoch-40-loss=0.2200.pt",
|
| 11 |
+
"repeated_diffusion_steps": 8,
|
| 12 |
+
"resume_epoch": null,
|
| 13 |
+
"resume_step": null,
|
| 14 |
+
"run_id": "prism-dinosiglip-224px+oxe+diffusion+n2+b16+x42--image_aug",
|
| 15 |
+
"run_id_note": null,
|
| 16 |
+
"run_root_dir": "/mnt/blob/cogact/ditl_8_lr2e-5_b16_fa15_pa0_shuffle_oxe",
|
| 17 |
+
"save_interval": 2500,
|
| 18 |
+
"seed": 42,
|
| 19 |
+
"trackers": [
|
| 20 |
+
"jsonl",
|
| 21 |
+
"wandb"
|
| 22 |
+
],
|
| 23 |
+
"use_ema": true,
|
| 24 |
+
"vla": {
|
| 25 |
+
"base_vlm": "prism-dinosiglip-224px+7b",
|
| 26 |
+
"data_mix": "oxe_magic_soup_plus_minus",
|
| 27 |
+
"enable_gradient_checkpointing": true,
|
| 28 |
+
"enable_mixed_precision_training": true,
|
| 29 |
+
"epochs": 100,
|
| 30 |
+
"expected_world_size": 16,
|
| 31 |
+
"freeze_llm_backbone": false,
|
| 32 |
+
"freeze_vision_backbone": false,
|
| 33 |
+
"global_batch_size": 256,
|
| 34 |
+
"learning_rate": 2e-05,
|
| 35 |
+
"lr_scheduler_type": "constant",
|
| 36 |
+
"max_grad_norm": 1.0,
|
| 37 |
+
"max_steps": null,
|
| 38 |
+
"per_device_batch_size": 16,
|
| 39 |
+
"reduce_in_full_precision": true,
|
| 40 |
+
"shuffle_buffer_size": 250000,
|
| 41 |
+
"train_strategy": "fsdp-full-shard",
|
| 42 |
+
"type": "prism-dinosiglip-224px+oxe+diffusion",
|
| 43 |
+
"unfreeze_last_llm_layer": false,
|
| 44 |
+
"vla_id": "prism-dinosiglip-224px+oxe+diffusion",
|
| 45 |
+
"warmup_ratio": 0.0,
|
| 46 |
+
"weight_decay": 0.0
|
| 47 |
+
},
|
| 48 |
+
"wandb_entity": null,
|
| 49 |
+
"wandb_project": null
|
| 50 |
+
}
|