| { |
| "bundle": { |
| "digest": { |
| "algorithm": "sha256", |
| "scope": "structure", |
| "value": "9d469d0ec5c4a2198dc88598ee0104ef38aa48c38a225864aae2cf008f8738cf" |
| }, |
| "files": [ |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/added_tokens.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/chat_template.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/config.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/generation_config.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/merges.txt" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/model.safetensors" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/preprocessor_config.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/processor_config.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/special_tokens_map.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/tokenizer.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/tokenizer_config.json" |
| }, |
| { |
| "path": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct/vocab.json" |
| }, |
| { |
| "path": "config.json" |
| }, |
| { |
| "path": "model.safetensors" |
| }, |
| { |
| "path": "policy_postprocessor.json" |
| }, |
| { |
| "path": "policy_postprocessor_step_0_unnormalizer_processor.safetensors" |
| }, |
| { |
| "path": "policy_preprocessor.json" |
| }, |
| { |
| "path": "policy_preprocessor_step_5_normalizer_processor.safetensors" |
| } |
| ], |
| "name": "smolvla", |
| "revision": 1, |
| "uuid": "7131ec7f-8cad-46a8-862c-9c6abbd32335" |
| }, |
| "deployments": { |
| "rknn_rk3588": { |
| "artifacts": { |
| "action": { |
| "format": "rknn", |
| "path": "artifacts/rknn/rknn_rk3588/generations/cc1c69ec-884f-4bc7-9873-598f74e3dfad/action.rknn", |
| "sha256": "021fbce1e1497881ec3321ed37e12e58fa7ad79e1d6d506336c8758803cd5538" |
| }, |
| "embedding": { |
| "format": "pt", |
| "path": "artifacts/rknn/rknn_rk3588/generations/2602144b-79cb-46cd-ace1-c8edae07dd95/embedding.pt", |
| "sha256": "e05cfb3682483ca4e9d8f3711d83845c4f20b749e759f5f74281b73b33ce54d0" |
| }, |
| "prefill": { |
| "format": "rknn", |
| "path": "artifacts/rknn/rknn_rk3588/generations/b7eac25d-6039-4874-9070-ee3fff065a23/prefill.rknn", |
| "sha256": "b0aa47da27ebbc2f1f7d9c6d1bce0d882a442242b0f5398a390c6e6b7d0794d8" |
| }, |
| "state_projection": { |
| "format": "pt", |
| "path": "artifacts/rknn/rknn_rk3588/generations/b1e5816f-3a51-4749-8f26-a1288a1de15b/state_projection.pt", |
| "sha256": "26bdae373faf23dca73c13582ad64b54d154f377563ff16f648f5954fb482149" |
| }, |
| "vision_top": { |
| "format": "rknn", |
| "path": "artifacts/rknn/rknn_rk3588/generations/f144bf1e-1c9e-433d-bf5e-79fdcbb9a40f/vision.rknn", |
| "sha256": "237b1e99e8c167dbffc72736ff2b8e3d7efbb16780be948d2991d6599cf81089", |
| "share_group": "shared-vision" |
| }, |
| "vision_wrist": { |
| "format": "rknn", |
| "path": "artifacts/rknn/rknn_rk3588/generations/f144bf1e-1c9e-433d-bf5e-79fdcbb9a40f/vision.rknn", |
| "sha256": "237b1e99e8c167dbffc72736ff2b8e3d7efbb16780be948d2991d6599cf81089", |
| "share_group": "shared-vision" |
| } |
| }, |
| "bindings": { |
| "action": { |
| "inputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "x_t", |
| "semantic": "noise", |
| "shape": [ |
| 1, |
| 50, |
| 32 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 1, |
| "runtime_name": "timestep", |
| "semantic": "time", |
| "shape": [ |
| 1 |
| ] |
| }, |
| { |
| "dtype": "bool", |
| "index": 2, |
| "runtime_name": "prefix_pad_masks", |
| "semantic": "internal.prefix_pad_masks", |
| "shape": [ |
| 1, |
| 177 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 3, |
| "runtime_name": "past_key_0", |
| "semantic": "internal.past_key.0", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 4, |
| "runtime_name": "past_value_0", |
| "semantic": "internal.past_value.0", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 5, |
| "runtime_name": "past_key_1", |
| "semantic": "internal.past_key.1", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 6, |
| "runtime_name": "past_value_1", |
| "semantic": "internal.past_value.1", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 7, |
| "runtime_name": "past_key_2", |
| "semantic": "internal.past_key.2", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 8, |
| "runtime_name": "past_value_2", |
| "semantic": "internal.past_value.2", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 9, |
| "runtime_name": "past_key_3", |
| "semantic": "internal.past_key.3", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 10, |
| "runtime_name": "past_value_3", |
| "semantic": "internal.past_value.3", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 11, |
| "runtime_name": "past_key_4", |
| "semantic": "internal.past_key.4", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 12, |
| "runtime_name": "past_value_4", |
| "semantic": "internal.past_value.4", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 13, |
| "runtime_name": "past_key_5", |
| "semantic": "internal.past_key.5", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 14, |
| "runtime_name": "past_value_5", |
| "semantic": "internal.past_value.5", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 15, |
| "runtime_name": "past_key_6", |
| "semantic": "internal.past_key.6", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 16, |
| "runtime_name": "past_value_6", |
| "semantic": "internal.past_value.6", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 17, |
| "runtime_name": "past_key_7", |
| "semantic": "internal.past_key.7", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 18, |
| "runtime_name": "past_value_7", |
| "semantic": "internal.past_value.7", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 19, |
| "runtime_name": "past_key_8", |
| "semantic": "internal.past_key.8", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 20, |
| "runtime_name": "past_value_8", |
| "semantic": "internal.past_value.8", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 21, |
| "runtime_name": "past_key_9", |
| "semantic": "internal.past_key.9", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 22, |
| "runtime_name": "past_value_9", |
| "semantic": "internal.past_value.9", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 23, |
| "runtime_name": "past_key_10", |
| "semantic": "internal.past_key.10", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 24, |
| "runtime_name": "past_value_10", |
| "semantic": "internal.past_value.10", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 25, |
| "runtime_name": "past_key_11", |
| "semantic": "internal.past_key.11", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 26, |
| "runtime_name": "past_value_11", |
| "semantic": "internal.past_value.11", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 27, |
| "runtime_name": "past_key_12", |
| "semantic": "internal.past_key.12", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 28, |
| "runtime_name": "past_value_12", |
| "semantic": "internal.past_value.12", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 29, |
| "runtime_name": "past_key_13", |
| "semantic": "internal.past_key.13", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 30, |
| "runtime_name": "past_value_13", |
| "semantic": "internal.past_value.13", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 31, |
| "runtime_name": "past_key_14", |
| "semantic": "internal.past_key.14", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 32, |
| "runtime_name": "past_value_14", |
| "semantic": "internal.past_value.14", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 33, |
| "runtime_name": "past_key_15", |
| "semantic": "internal.past_key.15", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 34, |
| "runtime_name": "past_value_15", |
| "semantic": "internal.past_value.15", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| } |
| ], |
| "outputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "v_t", |
| "semantic": "action", |
| "shape": [ |
| 1, |
| 50, |
| 32 |
| ] |
| } |
| ] |
| }, |
| "embedding": { |
| "inputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "image_0", |
| "semantic": "internal.image_embedding.top", |
| "shape": [ |
| 1, |
| 64, |
| 960 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 1, |
| "runtime_name": "image_1", |
| "semantic": "internal.image_embedding.wrist", |
| "shape": [ |
| 1, |
| 64, |
| 960 |
| ] |
| }, |
| { |
| "dtype": "int64", |
| "index": 2, |
| "runtime_name": "tokens", |
| "semantic": "observation.language.tokens", |
| "shape": [ |
| 1, |
| 48 |
| ] |
| }, |
| { |
| "dtype": "bool", |
| "index": 3, |
| "runtime_name": "language_mask", |
| "semantic": "observation.language.attention_mask", |
| "shape": [ |
| 1, |
| 48 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 4, |
| "runtime_name": "state", |
| "semantic": "observation.state", |
| "shape": [ |
| 1, |
| 32 |
| ] |
| } |
| ], |
| "outputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "prefix_embeddings", |
| "semantic": "internal.prefix_embeddings", |
| "shape": [ |
| 1, |
| 177, |
| 960 |
| ] |
| }, |
| { |
| "dtype": "bool", |
| "index": 1, |
| "runtime_name": "prefix_pad_masks", |
| "semantic": "internal.prefix_pad_masks", |
| "shape": [ |
| 1, |
| 177 |
| ] |
| }, |
| { |
| "dtype": "int64", |
| "index": 2, |
| "runtime_name": "attention_mask", |
| "semantic": "internal.attention_mask", |
| "shape": [ |
| 1, |
| 177, |
| 177 |
| ] |
| }, |
| { |
| "dtype": "int64", |
| "index": 3, |
| "runtime_name": "position_ids", |
| "semantic": "internal.position_ids", |
| "shape": [ |
| 1, |
| 177 |
| ] |
| } |
| ] |
| }, |
| "prefill": { |
| "inputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "prefix_embs", |
| "semantic": "internal.prefix_embeddings", |
| "shape": [ |
| 1, |
| 177, |
| 960 |
| ] |
| }, |
| { |
| "dtype": "int64", |
| "index": 1, |
| "runtime_name": "attention_mask", |
| "semantic": "internal.attention_mask", |
| "shape": [ |
| 1, |
| 177, |
| 177 |
| ] |
| }, |
| { |
| "dtype": "int64", |
| "index": 2, |
| "runtime_name": "position_ids", |
| "semantic": "internal.position_ids", |
| "shape": [ |
| 1, |
| 177 |
| ] |
| } |
| ], |
| "outputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "past_key_0", |
| "semantic": "internal.past_key.0", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 1, |
| "runtime_name": "past_value_0", |
| "semantic": "internal.past_value.0", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 2, |
| "runtime_name": "past_key_1", |
| "semantic": "internal.past_key.1", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 3, |
| "runtime_name": "past_value_1", |
| "semantic": "internal.past_value.1", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 4, |
| "runtime_name": "past_key_2", |
| "semantic": "internal.past_key.2", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 5, |
| "runtime_name": "past_value_2", |
| "semantic": "internal.past_value.2", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 6, |
| "runtime_name": "past_key_3", |
| "semantic": "internal.past_key.3", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 7, |
| "runtime_name": "past_value_3", |
| "semantic": "internal.past_value.3", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 8, |
| "runtime_name": "past_key_4", |
| "semantic": "internal.past_key.4", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 9, |
| "runtime_name": "past_value_4", |
| "semantic": "internal.past_value.4", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 10, |
| "runtime_name": "past_key_5", |
| "semantic": "internal.past_key.5", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 11, |
| "runtime_name": "past_value_5", |
| "semantic": "internal.past_value.5", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 12, |
| "runtime_name": "past_key_6", |
| "semantic": "internal.past_key.6", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 13, |
| "runtime_name": "past_value_6", |
| "semantic": "internal.past_value.6", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 14, |
| "runtime_name": "past_key_7", |
| "semantic": "internal.past_key.7", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 15, |
| "runtime_name": "past_value_7", |
| "semantic": "internal.past_value.7", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 16, |
| "runtime_name": "past_key_8", |
| "semantic": "internal.past_key.8", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 17, |
| "runtime_name": "past_value_8", |
| "semantic": "internal.past_value.8", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 18, |
| "runtime_name": "past_key_9", |
| "semantic": "internal.past_key.9", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 19, |
| "runtime_name": "past_value_9", |
| "semantic": "internal.past_value.9", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 20, |
| "runtime_name": "past_key_10", |
| "semantic": "internal.past_key.10", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 21, |
| "runtime_name": "past_value_10", |
| "semantic": "internal.past_value.10", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 22, |
| "runtime_name": "past_key_11", |
| "semantic": "internal.past_key.11", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 23, |
| "runtime_name": "past_value_11", |
| "semantic": "internal.past_value.11", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 24, |
| "runtime_name": "past_key_12", |
| "semantic": "internal.past_key.12", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 25, |
| "runtime_name": "past_value_12", |
| "semantic": "internal.past_value.12", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 26, |
| "runtime_name": "past_key_13", |
| "semantic": "internal.past_key.13", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 27, |
| "runtime_name": "past_value_13", |
| "semantic": "internal.past_value.13", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 28, |
| "runtime_name": "past_key_14", |
| "semantic": "internal.past_key.14", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 29, |
| "runtime_name": "past_value_14", |
| "semantic": "internal.past_value.14", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 30, |
| "runtime_name": "past_key_15", |
| "semantic": "internal.past_key.15", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "index": 31, |
| "runtime_name": "past_value_15", |
| "semantic": "internal.past_value.15", |
| "shape": [ |
| 1, |
| 177, |
| 5, |
| 64 |
| ] |
| } |
| ] |
| }, |
| "vision_top": { |
| "inputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "layout": "NCHW", |
| "runtime_name": "pixel_values", |
| "semantic": "observation.images.top", |
| "shape": [ |
| 1, |
| 3, |
| 512, |
| 512 |
| ] |
| } |
| ], |
| "outputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "image_embeddings", |
| "semantic": "internal.image_embedding.top", |
| "shape": [ |
| 1, |
| 64, |
| 960 |
| ] |
| } |
| ] |
| }, |
| "vision_wrist": { |
| "inputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "layout": "NCHW", |
| "runtime_name": "pixel_values", |
| "semantic": "observation.images.wrist", |
| "shape": [ |
| 1, |
| 3, |
| 512, |
| 512 |
| ] |
| } |
| ], |
| "outputs": [ |
| { |
| "dtype": "float32", |
| "index": 0, |
| "runtime_name": "image_embeddings", |
| "semantic": "internal.image_embedding.wrist", |
| "shape": [ |
| 1, |
| 64, |
| 960 |
| ] |
| } |
| ] |
| } |
| }, |
| "execution": [ |
| "vision_top", |
| "vision_wrist", |
| "embedding", |
| "prefill", |
| "action" |
| ], |
| "execution_contract": { |
| "cancellation_granularity": "request_boundary", |
| "execution_structure": "direct", |
| "state_scope": "request" |
| }, |
| "revision": 1, |
| "runtime_profile": { |
| "backend": "rknn", |
| "profile": { |
| "device_id": 0, |
| "target_name": "rk3588" |
| }, |
| "target": { |
| "runtime": "rknn-lite2", |
| "soc": "rk3588" |
| } |
| }, |
| "uuid": "f0697c61-8379-48ce-a12c-ba46afff4ffa" |
| }, |
| "torch-cpu": { |
| "execution_contract": { |
| "cancellation_granularity": "request_boundary", |
| "execution_structure": "direct", |
| "state_scope": "request" |
| }, |
| "revision": 1, |
| "runtime_profile": { |
| "backend": "torch", |
| "profile": { |
| "device": "cpu" |
| }, |
| "target": { |
| "runtime": "torch" |
| } |
| }, |
| "uuid": "063cdce6-c908-4e78-aa39-847d4601eb33" |
| }, |
| "torch-cuda": { |
| "execution_contract": { |
| "cancellation_granularity": "request_boundary", |
| "execution_structure": "direct", |
| "state_scope": "request" |
| }, |
| "revision": 1, |
| "runtime_profile": { |
| "backend": "torch", |
| "profile": { |
| "device": "cuda" |
| }, |
| "target": { |
| "runtime": "torch" |
| } |
| }, |
| "uuid": "adaa0826-0bab-4b83-a861-7acacbbc1b44" |
| } |
| }, |
| "model": { |
| "inputs": [ |
| { |
| "dtype": "float32", |
| "semantic": "observation.state", |
| "shape": [ |
| 6 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "semantic": "observation.current", |
| "shape": [ |
| 6 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "semantic": "observation.images.top", |
| "shape": [ |
| 3, |
| 480, |
| 640 |
| ] |
| }, |
| { |
| "dtype": "float32", |
| "semantic": "observation.images.wrist", |
| "shape": [ |
| 3, |
| 480, |
| 640 |
| ] |
| } |
| ], |
| "interface": "policy", |
| "model_type": "smolvla", |
| "operation": "predict", |
| "outputs": [ |
| { |
| "dtype": "float32", |
| "semantic": "action", |
| "shape": [ |
| 6 |
| ] |
| } |
| ] |
| }, |
| "schema_version": 3 |
| } |
|
|
