optimum-neuron-cache / neuronxcc-2.19.8089.0+8ab9f450 /0_REGISTRY /0.3.1.dev1 /cvt /hf-internal-testing /tiny-random-CvTModel /e5c893a9ead6645e9326.json
| { | |
| "_entry_class": "SingleModelCacheEntry", | |
| "_model_id": "hf-internal-testing/tiny-random-CvTModel", | |
| "_task": "image-classification", | |
| "architectures": [ | |
| "CvtModel" | |
| ], | |
| "attention_drop_rate": [ | |
| 0.0, | |
| 0.0, | |
| 0.0 | |
| ], | |
| "cls_token": [ | |
| false, | |
| false, | |
| true | |
| ], | |
| "depth": [ | |
| 1, | |
| 2, | |
| 10 | |
| ], | |
| "drop_path_rate": [ | |
| 0.0, | |
| 0.0, | |
| 0.1 | |
| ], | |
| "drop_rate": [ | |
| 0.0, | |
| 0.0, | |
| 0.0 | |
| ], | |
| "embed_dim": [ | |
| 16, | |
| 48, | |
| 96 | |
| ], | |
| "image_size": 64, | |
| "initializer_range": 0.02, | |
| "kernel_qkv": [ | |
| 3, | |
| 3, | |
| 3 | |
| ], | |
| "layer_norm_eps": 1e-12, | |
| "mlp_ratio": [ | |
| 4.0, | |
| 4.0, | |
| 4.0 | |
| ], | |
| "model_type": "cvt", | |
| "neuron": { | |
| "auto_cast": null, | |
| "auto_cast_type": null, | |
| "compiler_type": "neuronx-cc", | |
| "compiler_version": "2.19.8089.0+8ab9f450", | |
| "disable_fallback": false, | |
| "disable_fast_relayout": false, | |
| "dynamic_batch_size": false, | |
| "float_dtype": "fp32", | |
| "inline_weights_to_neff": true, | |
| "input_names": [ | |
| "pixel_values" | |
| ], | |
| "int_dtype": "int64", | |
| "model_type": "cvt", | |
| "optlevel": "2", | |
| "output_attentions": false, | |
| "output_hidden_states": false, | |
| "output_names": [ | |
| "logits" | |
| ], | |
| "static_batch_size": 1, | |
| "static_image_size": 64, | |
| "static_num_channels": 3, | |
| "task": "image-classification", | |
| "tensor_parallel_size": 1 | |
| }, | |
| "num_channels": 3, | |
| "num_heads": [ | |
| 1, | |
| 3, | |
| 6 | |
| ], | |
| "padding_kv": [ | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "padding_q": [ | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "patch_padding": [ | |
| 2, | |
| 1, | |
| 1 | |
| ], | |
| "patch_sizes": [ | |
| 7, | |
| 3, | |
| 3 | |
| ], | |
| "patch_stride": [ | |
| 4, | |
| 2, | |
| 2 | |
| ], | |
| "qkv_bias": [ | |
| true, | |
| true, | |
| true | |
| ], | |
| "qkv_projection_method": [ | |
| "dw_bn", | |
| "dw_bn", | |
| "dw_bn" | |
| ], | |
| "stride_kv": [ | |
| 2, | |
| 2, | |
| 2 | |
| ], | |
| "stride_q": [ | |
| 1, | |
| 1, | |
| 1 | |
| ] | |
| } |