rf-detr-segmentation-NVFP4 / quantization_config.json
Reza2kn's picture
Upload experimental NVFP4 RF-DETR segmentation quant
007dfa6 verified
Raw
History Blame Contribute Delete
67.1 kB
{
"quant_method": "nvfp4",
"format": "custom_nvfp4_e2m1_e4m3_scales",
"base_model": "Roboflow/rf-detr-segmentation",
"base_model_relation": "quantized",
"block_size": 16,
"total_parameters": 34153555,
"quantized_parameters": 31841152,
"kept_parameters": 2312403,
"quantized_parameter_fraction": 0.9322939295777555,
"max_measured_tensor_rel_l2": 0.1563214510679245,
"tensors": {
"backbone.0.encoder.encoder.embeddings.mask_token": {
"shape": [
1,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.embeddings.mask_token.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.embeddings.mask_token.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.0.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.0.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.0.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.0.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.0.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.0.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.0.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.0.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.0.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.0.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.0.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.0.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.0.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.0.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.0.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.0.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.0.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.0.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.1.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.1.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.1.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.1.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.1.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.1.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.1.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.1.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.1.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.1.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.1.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.1.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.1.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.1.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.1.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.1.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.1.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.1.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.10.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.10.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.10.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.10.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.10.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.10.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.10.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.10.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.10.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.10.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.10.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.10.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.10.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.10.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.10.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.10.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.10.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.10.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.11.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.11.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.11.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.11.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.11.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.11.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.11.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.11.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.11.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.11.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.11.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.11.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.11.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.11.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.11.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.11.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.11.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.11.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.2.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.2.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.2.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.2.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.2.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.2.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.2.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.2.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.2.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.2.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.2.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.2.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.2.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.2.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.2.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.2.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.2.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.2.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.3.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.3.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.3.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.3.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.3.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.3.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.3.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.3.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.3.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.3.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.3.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.3.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.3.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.3.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.3.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.3.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.3.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.3.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.4.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.4.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.4.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.4.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.4.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.4.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.4.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.4.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.4.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.4.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.4.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.4.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.4.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.4.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.4.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.4.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.4.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.4.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.5.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.5.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.5.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.5.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.5.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.5.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.5.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.5.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.5.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.5.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.5.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.5.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.5.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.5.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.5.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.5.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.5.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.5.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.6.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.6.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.6.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.6.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.6.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.6.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.6.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.6.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.6.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.6.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.6.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.6.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.6.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.6.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.6.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.6.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.6.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.6.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.7.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.7.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.7.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.7.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.7.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.7.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.7.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.7.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.7.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.7.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.7.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.7.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.7.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.7.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.7.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.7.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.7.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.7.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.8.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.8.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.8.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.8.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.8.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.8.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.8.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.8.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.8.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.8.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.8.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.8.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.8.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.8.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.8.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.8.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.8.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.8.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.9.attention.attention.key.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.9.attention.attention.key.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.9.attention.attention.key.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.9.attention.attention.query.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.9.attention.attention.query.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.9.attention.attention.query.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.9.attention.attention.value.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.9.attention.attention.value.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.9.attention.attention.value.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.9.attention.output.dense.weight": {
"shape": [
384,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.9.attention.output.dense.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.9.attention.output.dense.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.9.mlp.fc1.weight": {
"shape": [
1536,
384
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.9.mlp.fc1.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.9.mlp.fc1.weight.nvfp4_scale_e4m3"
},
"backbone.0.encoder.encoder.encoder.layer.9.mlp.fc2.weight": {
"shape": [
384,
1536
],
"source_dtype": "float32",
"block_size": 16,
"packed": "backbone.0.encoder.encoder.encoder.layer.9.mlp.fc2.weight.nvfp4_packed",
"scale_e4m3": "backbone.0.encoder.encoder.encoder.layer.9.mlp.fc2.weight.nvfp4_scale_e4m3"
},
"bbox_embed.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "bbox_embed.layers.0.weight.nvfp4_packed",
"scale_e4m3": "bbox_embed.layers.0.weight.nvfp4_scale_e4m3"
},
"bbox_embed.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "bbox_embed.layers.1.weight.nvfp4_packed",
"scale_e4m3": "bbox_embed.layers.1.weight.nvfp4_scale_e4m3"
},
"bbox_embed.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "bbox_embed.layers.2.weight.nvfp4_packed",
"scale_e4m3": "bbox_embed.layers.2.weight.nvfp4_scale_e4m3"
},
"class_embed.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "class_embed.weight.nvfp4_packed",
"scale_e4m3": "class_embed.weight.nvfp4_scale_e4m3"
},
"query_feat.weight": {
"shape": [
2600,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "query_feat.weight.nvfp4_packed",
"scale_e4m3": "query_feat.weight.nvfp4_scale_e4m3"
},
"segmentation_head.blocks.0.pwconv1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.blocks.0.pwconv1.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.blocks.0.pwconv1.weight.nvfp4_scale_e4m3"
},
"segmentation_head.blocks.1.pwconv1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.blocks.1.pwconv1.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.blocks.1.pwconv1.weight.nvfp4_scale_e4m3"
},
"segmentation_head.blocks.2.pwconv1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.blocks.2.pwconv1.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.blocks.2.pwconv1.weight.nvfp4_scale_e4m3"
},
"segmentation_head.blocks.3.pwconv1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.blocks.3.pwconv1.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.blocks.3.pwconv1.weight.nvfp4_scale_e4m3"
},
"segmentation_head.query_features_block.layers.0.weight": {
"shape": [
1024,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.query_features_block.layers.0.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.query_features_block.layers.0.weight.nvfp4_scale_e4m3"
},
"segmentation_head.query_features_block.layers.2.weight": {
"shape": [
256,
1024
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.query_features_block.layers.2.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.query_features_block.layers.2.weight.nvfp4_scale_e4m3"
},
"segmentation_head.query_features_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "segmentation_head.query_features_proj.weight.nvfp4_packed",
"scale_e4m3": "segmentation_head.query_features_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.cross_attn.attention_weights.weight": {
"shape": [
32,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.cross_attn.attention_weights.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.cross_attn.attention_weights.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.cross_attn.output_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.cross_attn.output_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.cross_attn.output_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.cross_attn.sampling_offsets.weight": {
"shape": [
64,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.cross_attn.sampling_offsets.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.cross_attn.sampling_offsets.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.cross_attn.value_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.cross_attn.value_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.cross_attn.value_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.linear1.weight": {
"shape": [
2048,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.linear1.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.linear1.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.linear2.weight": {
"shape": [
256,
2048
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.linear2.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.linear2.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.self_attn.in_proj_weight": {
"shape": [
768,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.self_attn.in_proj_weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.self_attn.in_proj_weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.0.self_attn.out_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.0.self_attn.out_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.0.self_attn.out_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.cross_attn.attention_weights.weight": {
"shape": [
32,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.cross_attn.attention_weights.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.cross_attn.attention_weights.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.cross_attn.output_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.cross_attn.output_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.cross_attn.output_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.cross_attn.sampling_offsets.weight": {
"shape": [
64,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.cross_attn.sampling_offsets.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.cross_attn.sampling_offsets.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.cross_attn.value_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.cross_attn.value_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.cross_attn.value_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.linear1.weight": {
"shape": [
2048,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.linear1.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.linear1.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.linear2.weight": {
"shape": [
256,
2048
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.linear2.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.linear2.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.self_attn.in_proj_weight": {
"shape": [
768,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.self_attn.in_proj_weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.self_attn.in_proj_weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.1.self_attn.out_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.1.self_attn.out_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.1.self_attn.out_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.cross_attn.attention_weights.weight": {
"shape": [
32,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.cross_attn.attention_weights.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.cross_attn.attention_weights.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.cross_attn.output_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.cross_attn.output_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.cross_attn.output_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.cross_attn.sampling_offsets.weight": {
"shape": [
64,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.cross_attn.sampling_offsets.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.cross_attn.sampling_offsets.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.cross_attn.value_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.cross_attn.value_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.cross_attn.value_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.linear1.weight": {
"shape": [
2048,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.linear1.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.linear1.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.linear2.weight": {
"shape": [
256,
2048
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.linear2.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.linear2.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.self_attn.in_proj_weight": {
"shape": [
768,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.self_attn.in_proj_weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.self_attn.in_proj_weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.2.self_attn.out_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.2.self_attn.out_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.2.self_attn.out_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.cross_attn.attention_weights.weight": {
"shape": [
32,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.cross_attn.attention_weights.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.cross_attn.attention_weights.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.cross_attn.output_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.cross_attn.output_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.cross_attn.output_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.cross_attn.sampling_offsets.weight": {
"shape": [
64,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.cross_attn.sampling_offsets.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.cross_attn.sampling_offsets.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.cross_attn.value_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.cross_attn.value_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.cross_attn.value_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.linear1.weight": {
"shape": [
2048,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.linear1.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.linear1.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.linear2.weight": {
"shape": [
256,
2048
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.linear2.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.linear2.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.self_attn.in_proj_weight": {
"shape": [
768,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.self_attn.in_proj_weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.self_attn.in_proj_weight.nvfp4_scale_e4m3"
},
"transformer.decoder.layers.3.self_attn.out_proj.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.layers.3.self_attn.out_proj.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.layers.3.self_attn.out_proj.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.ref_point_head.layers.0.weight": {
"shape": [
256,
512
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.ref_point_head.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.ref_point_head.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.decoder.ref_point_head.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.decoder.ref_point_head.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.decoder.ref_point_head.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.0.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.0.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.0.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.0.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.0.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.0.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.0.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.0.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.0.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.1.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.1.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.1.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.1.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.1.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.1.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.1.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.1.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.1.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.10.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.10.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.10.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.10.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.10.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.10.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.10.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.10.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.10.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.11.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.11.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.11.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.11.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.11.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.11.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.11.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.11.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.11.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.12.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.12.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.12.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.12.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.12.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.12.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.12.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.12.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.12.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.2.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.2.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.2.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.2.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.2.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.2.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.2.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.2.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.2.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.3.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.3.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.3.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.3.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.3.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.3.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.3.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.3.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.3.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.4.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.4.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.4.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.4.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.4.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.4.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.4.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.4.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.4.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.5.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.5.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.5.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.5.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.5.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.5.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.5.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.5.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.5.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.6.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.6.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.6.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.6.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.6.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.6.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.6.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.6.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.6.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.7.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.7.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.7.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.7.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.7.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.7.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.7.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.7.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.7.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.8.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.8.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.8.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.8.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.8.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.8.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.8.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.8.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.8.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.9.layers.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.9.layers.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.9.layers.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.9.layers.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.9.layers.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.9.layers.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_bbox_embed.9.layers.2.weight": {
"shape": [
4,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_bbox_embed.9.layers.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_bbox_embed.9.layers.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.0.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.1.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.10.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.10.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.10.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.11.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.11.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.11.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.12.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.12.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.12.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.2.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.3.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.3.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.3.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.4.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.4.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.4.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.5.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.5.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.5.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.6.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.6.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.6.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.7.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.7.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.7.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.8.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.8.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.8.weight.nvfp4_scale_e4m3"
},
"transformer.enc_out_class_embed.9.weight": {
"shape": [
91,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_out_class_embed.9.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_out_class_embed.9.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.0.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.0.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.0.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.1.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.1.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.1.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.10.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.10.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.10.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.11.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.11.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.11.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.12.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.12.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.12.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.2.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.2.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.2.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.3.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.3.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.3.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.4.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.4.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.4.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.5.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.5.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.5.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.6.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.6.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.6.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.7.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.7.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.7.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.8.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.8.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.8.weight.nvfp4_scale_e4m3"
},
"transformer.enc_output.9.weight": {
"shape": [
256,
256
],
"source_dtype": "float32",
"block_size": 16,
"packed": "transformer.enc_output.9.weight.nvfp4_packed",
"scale_e4m3": "transformer.enc_output.9.weight.nvfp4_scale_e4m3"
}
},
"runtime_note": "This artifact is packed NVFP4 storage. Use the companion dequant loader or a native FP4 runtime adapter before expecting Blackwell speedups."
}