Robotics
LeRobot
Safetensors
distributional_value_function
reward-model
lilkm's picture
lilkm HF Staff
Upload reward model weights, train config and readme
30f807e verified
Raw
History Blame Contribute Delete
966 Bytes
{
"type": "distributional_value_function",
"input_features": {},
"output_features": {},
"device": "cuda",
"pretrained_path": null,
"pretrained_revision": null,
"push_to_hub": true,
"repo_id": "lilkm/vf_stackblocks_v10_stage1",
"license": null,
"tags": null,
"private": null,
"siglip_path": "google/siglip2-so400m-patch14-224",
"gemma3_path": "google/gemma-3-270m",
"num_value_bins": 201,
"value_support_min": -1.0,
"value_support_max": 0.0,
"hl_gauss_sigma_ratio": 5.0,
"target_method": "dirac_delta",
"use_one_hot_terminal": true,
"image_resolution": [
224,
224
],
"tokenizer_max_length": 200,
"value_dropout": 0.0,
"freeze_vision_encoder": true,
"freeze_language_model": true,
"stop_gradient_to_vlm": false,
"vision_encoder_lr_multiplier": 0.5,
"readout": "mean_pool",
"normalization_mapping": {
"VISUAL": "IDENTITY"
}
}