Feature Extraction
Transformers
Safetensors
sam3_video
segmentation
sam
sam3.1
nvfp4
quantization
food-segmentation
instance-segmentation
video-segmentation
Instructions to use Reza2kn/sam3.1-nvfp4-detector-no-language with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Reza2kn/sam3.1-nvfp4-detector-no-language with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="Reza2kn/sam3.1-nvfp4-detector-no-language")# Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("Reza2kn/sam3.1-nvfp4-detector-no-language") model = AutoModel.from_pretrained("Reza2kn/sam3.1-nvfp4-detector-no-language") - Notebooks
- Google Colab
- Kaggle
| { | |
| "quant_method": "nvfp4", | |
| "format": "custom_nvfp4_e2m1_e4m3_scales", | |
| "base_model": "facebook/sam3.1", | |
| "base_filename": "sam3.1_multiplex.pt", | |
| "base_model_relation": "quantized", | |
| "checkpoint_path": "/home/rezo/.cache/huggingface/hub/models--facebook--sam3.1/snapshots/daa63191845a41281374e725f4c9e51c7a824460/sam3.1_multiplex.pt", | |
| "block_size": 16, | |
| "scale_rule": "static_6", | |
| "total_parameters": 874365676, | |
| "selected_parameters": 501011510, | |
| "quantized_parameters": 472113408, | |
| "kept_parameters": 402252268, | |
| "quantized_parameter_fraction": 0.5399496125691924, | |
| "selected_quantized_fraction": 0.9423204828168519, | |
| "max_measured_tensor_rel_l2": 0.0, | |
| "quantize_patterns": [ | |
| "^detector\\." | |
| ], | |
| "exclude_patterns": [ | |
| "^detector\\.backbone\\.language_backbone\\." | |
| ], | |
| "tensors": { | |
| "detector.backbone.vision_backbone.trunk.blocks.0.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.0.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.0.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.0.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.0.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.0.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.0.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.0.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.0.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.0.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.0.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.0.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.1.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.1.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.1.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.1.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.1.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.1.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.1.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.1.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.1.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.1.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.1.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.1.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.10.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.10.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.10.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.10.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.10.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.10.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.10.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.10.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.10.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.10.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.10.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.10.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.11.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.11.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.11.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.11.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.11.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.11.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.11.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.11.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.11.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.11.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.11.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.11.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.12.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.12.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.12.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.12.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.12.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.12.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.12.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.12.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.12.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.12.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.12.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.12.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.13.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.13.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.13.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.13.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.13.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.13.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.13.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.13.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.13.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.13.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.13.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.13.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.14.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.14.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.14.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.14.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.14.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.14.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.14.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.14.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.14.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.14.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.14.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.14.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.15.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.15.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.15.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.15.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.15.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.15.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.15.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.15.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.15.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.15.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.15.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.15.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.16.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.16.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.16.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.16.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.16.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.16.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.16.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.16.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.16.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.16.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.16.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.16.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.17.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.17.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.17.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.17.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.17.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.17.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.17.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.17.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.17.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.17.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.17.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.17.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.18.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.18.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.18.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.18.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.18.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.18.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.18.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.18.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.18.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.18.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.18.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.18.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.19.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.19.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.19.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.19.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.19.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.19.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.19.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.19.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.19.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.19.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.19.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.19.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.2.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.2.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.2.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.2.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.2.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.2.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.2.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.2.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.2.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.2.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.2.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.2.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.20.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.20.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.20.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.20.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.20.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.20.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.20.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.20.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.20.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.20.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.20.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.20.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.21.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.21.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.21.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.21.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.21.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.21.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.21.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.21.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.21.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.21.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.21.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.21.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.22.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.22.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.22.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.22.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.22.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.22.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.22.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.22.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.22.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.22.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.22.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.22.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.23.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.23.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.23.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.23.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.23.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.23.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.23.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.23.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.23.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.23.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.23.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.23.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.24.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.24.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.24.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.24.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.24.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.24.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.24.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.24.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.24.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.24.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.24.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.24.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.25.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.25.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.25.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.25.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.25.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.25.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.25.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.25.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.25.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.25.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.25.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.25.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.26.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.26.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.26.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.26.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.26.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.26.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.26.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.26.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.26.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.26.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.26.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.26.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.27.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.27.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.27.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.27.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.27.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.27.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.27.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.27.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.27.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.27.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.27.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.27.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.28.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.28.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.28.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.28.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.28.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.28.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.28.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.28.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.28.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.28.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.28.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.28.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.29.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.29.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.29.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.29.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.29.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.29.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.29.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.29.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.29.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.29.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.29.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.29.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.3.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.3.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.3.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.3.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.3.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.3.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.3.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.3.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.3.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.3.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.3.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.3.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.30.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.30.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.30.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.30.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.30.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.30.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.30.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.30.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.30.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.30.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.30.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.30.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.31.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.31.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.31.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.31.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.31.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.31.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.31.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.31.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.31.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.31.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.31.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.31.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.4.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.4.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.4.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.4.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.4.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.4.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.4.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.4.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.4.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.4.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.4.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.4.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.5.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.5.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.5.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.5.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.5.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.5.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.5.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.5.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.5.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.5.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.5.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.5.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.6.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.6.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.6.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.6.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.6.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.6.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.6.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.6.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.6.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.6.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.6.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.6.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.7.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.7.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.7.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.7.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.7.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.7.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.7.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.7.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.7.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.7.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.7.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.7.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.8.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.8.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.8.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.8.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.8.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.8.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.8.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.8.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.8.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.8.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.8.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.8.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.9.attn.proj.weight": { | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.9.attn.proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.9.attn.proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.9.attn.qkv.weight": { | |
| "shape": [ | |
| 3072, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.9.attn.qkv.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.9.attn.qkv.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.9.mlp.fc1.weight": { | |
| "shape": [ | |
| 4736, | |
| 1024 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.9.mlp.fc1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.9.mlp.fc1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.backbone.vision_backbone.trunk.blocks.9.mlp.fc2.weight": { | |
| "shape": [ | |
| 1024, | |
| 4736 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.backbone.vision_backbone.trunk.blocks.9.mlp.fc2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.backbone.vision_backbone.trunk.blocks.9.mlp.fc2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.dot_prod_scoring.hs_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.dot_prod_scoring.hs_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.dot_prod_scoring.hs_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.dot_prod_scoring.prompt_mlp.layers.0.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.dot_prod_scoring.prompt_mlp.layers.0.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.dot_prod_scoring.prompt_mlp.layers.0.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.dot_prod_scoring.prompt_mlp.layers.1.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.dot_prod_scoring.prompt_mlp.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.dot_prod_scoring.prompt_mlp.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.dot_prod_scoring.prompt_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.dot_prod_scoring.prompt_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.dot_prod_scoring.prompt_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.cls_embed.weight": { | |
| "shape": [ | |
| 1, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.cls_embed.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.cls_embed.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.0.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.0.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.0.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.0.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.0.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.0.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.0.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.0.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.0.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.0.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.0.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.0.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.0.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.0.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.0.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.0.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.0.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.0.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.1.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.1.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.1.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.1.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.1.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.1.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.1.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.1.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.1.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.1.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.1.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.1.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.1.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.1.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.1.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.1.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.1.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.1.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.2.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.2.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.2.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.2.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.2.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.2.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.2.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.2.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.2.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.2.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.2.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.2.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.2.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.2.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.2.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.encode.2.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.encode.2.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.encode.2.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.final_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.final_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.final_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.label_embed.weight": { | |
| "shape": [ | |
| 2, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.label_embed.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.label_embed.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.points_pool_project.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.points_pool_project.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.points_pool_project.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.geometry_encoder.points_pos_enc_project.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.geometry_encoder.points_pos_enc_project.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.geometry_encoder.points_pos_enc_project.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.segmentation_head.cross_attend_prompt.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.segmentation_head.cross_attend_prompt.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.segmentation_head.cross_attend_prompt.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.segmentation_head.cross_attend_prompt.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.segmentation_head.cross_attend_prompt.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.segmentation_head.cross_attend_prompt.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.segmentation_head.mask_predictor.mask_embed.layers.0.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.segmentation_head.mask_predictor.mask_embed.layers.0.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.segmentation_head.mask_predictor.mask_embed.layers.0.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.segmentation_head.mask_predictor.mask_embed.layers.1.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.segmentation_head.mask_predictor.mask_embed.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.segmentation_head.mask_predictor.mask_embed.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.segmentation_head.mask_predictor.mask_embed.layers.2.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.segmentation_head.mask_predictor.mask_embed.layers.2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.segmentation_head.mask_predictor.mask_embed.layers.2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.bbox_embed.layers.0.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.bbox_embed.layers.0.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.bbox_embed.layers.0.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.bbox_embed.layers.1.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.bbox_embed.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.bbox_embed.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.bbox_embed.layers.2.weight": { | |
| "shape": [ | |
| 4, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.bbox_embed.layers.2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.bbox_embed.layers.2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.boxRPB_embed_x.layers.1.weight": { | |
| "shape": [ | |
| 8, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.boxRPB_embed_x.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.boxRPB_embed_x.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.boxRPB_embed_y.layers.1.weight": { | |
| "shape": [ | |
| 8, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.boxRPB_embed_y.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.boxRPB_embed_y.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.ca_text.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.ca_text.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.ca_text.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.ca_text.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.ca_text.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.ca_text.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.cross_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.cross_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.cross_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.cross_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.cross_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.cross_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.0.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.0.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.0.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.ca_text.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.ca_text.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.ca_text.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.ca_text.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.ca_text.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.ca_text.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.cross_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.cross_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.cross_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.cross_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.cross_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.cross_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.1.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.1.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.1.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.ca_text.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.ca_text.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.ca_text.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.ca_text.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.ca_text.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.ca_text.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.cross_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.cross_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.cross_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.cross_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.cross_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.cross_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.2.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.2.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.2.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.ca_text.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.ca_text.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.ca_text.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.ca_text.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.ca_text.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.ca_text.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.cross_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.cross_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.cross_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.cross_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.cross_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.cross_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.3.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.3.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.3.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.ca_text.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.ca_text.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.ca_text.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.ca_text.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.ca_text.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.ca_text.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.cross_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.cross_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.cross_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.cross_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.cross_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.cross_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.4.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.4.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.4.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.ca_text.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.ca_text.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.ca_text.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.ca_text.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.ca_text.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.ca_text.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.cross_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.cross_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.cross_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.cross_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.cross_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.cross_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.layers.5.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.layers.5.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.layers.5.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.presence_token.weight": { | |
| "shape": [ | |
| 1, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.presence_token.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.presence_token.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.presence_token_head.layers.0.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.presence_token_head.layers.0.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.presence_token_head.layers.0.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.presence_token_head.layers.1.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.presence_token_head.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.presence_token_head.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.presence_token_head.layers.2.weight": { | |
| "shape": [ | |
| 1, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.presence_token_head.layers.2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.presence_token_head.layers.2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.query_embed.weight": { | |
| "shape": [ | |
| 200, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.query_embed.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.query_embed.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.ref_point_head.layers.0.weight": { | |
| "shape": [ | |
| 256, | |
| 512 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.ref_point_head.layers.0.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.ref_point_head.layers.0.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.decoder.ref_point_head.layers.1.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.decoder.ref_point_head.layers.1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.decoder.ref_point_head.layers.1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.0.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.0.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.0.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.0.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.0.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.0.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.0.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.0.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.0.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.0.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.0.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.0.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.0.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.0.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.0.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.0.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.0.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.0.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.1.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.1.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.1.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.1.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.1.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.1.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.1.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.1.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.1.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.1.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.1.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.1.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.1.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.1.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.1.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.1.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.1.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.1.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.2.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.2.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.2.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.2.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.2.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.2.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.2.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.2.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.2.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.2.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.2.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.2.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.2.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.2.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.2.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.2.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.2.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.2.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.3.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.3.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.3.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.3.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.3.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.3.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.3.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.3.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.3.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.3.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.3.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.3.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.3.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.3.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.3.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.3.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.3.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.3.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.4.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.4.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.4.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.4.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.4.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.4.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.4.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.4.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.4.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.4.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.4.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.4.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.4.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.4.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.4.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.4.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.4.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.4.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.5.cross_attn_image.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.5.cross_attn_image.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.5.cross_attn_image.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.5.cross_attn_image.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.5.cross_attn_image.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.5.cross_attn_image.out_proj.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.5.linear1.weight": { | |
| "shape": [ | |
| 2048, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.5.linear1.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.5.linear1.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.5.linear2.weight": { | |
| "shape": [ | |
| 256, | |
| 2048 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.5.linear2.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.5.linear2.weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.5.self_attn.in_proj_weight": { | |
| "shape": [ | |
| 768, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.5.self_attn.in_proj_weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.5.self_attn.in_proj_weight.nvfp4_scale_e4m3" | |
| }, | |
| "detector.transformer.encoder.layers.5.self_attn.out_proj.weight": { | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "source_dtype": "float32", | |
| "block_size": 16, | |
| "packed": "detector.transformer.encoder.layers.5.self_attn.out_proj.weight.nvfp4_packed", | |
| "scale_e4m3": "detector.transformer.encoder.layers.5.self_attn.out_proj.weight.nvfp4_scale_e4m3" | |
| } | |
| }, | |
| "runtime_note": "This is an NVFP4 storage artifact. Validation dequantizes weights before running the official SAM3 image model; native FP4 speed needs a runtime that consumes packed E2M1 values directly." | |
| } | |