Update config.json
Browse filesRemoving "scale_dtype" and "zp_dtype"from config.json. These are not currently supported in VLLM and there's an active issue about this open:
https://github.com/vllm-project/llm-compressor/issues/2057
- config.json +2 -6
config.json
CHANGED
|
@@ -62,11 +62,9 @@
|
|
| 62 |
"num_bits": 4,
|
| 63 |
"observer": "static_minmax",
|
| 64 |
"observer_kwargs": {},
|
| 65 |
-
"scale_dtype": "torch.float8_e4m3fn",
|
| 66 |
"strategy": "tensor_group",
|
| 67 |
"symmetric": true,
|
| 68 |
-
"type": "float"
|
| 69 |
-
"zp_dtype": null
|
| 70 |
},
|
| 71 |
"output_activations": null,
|
| 72 |
"targets": [
|
|
@@ -80,11 +78,9 @@
|
|
| 80 |
"num_bits": 4,
|
| 81 |
"observer": "static_minmax",
|
| 82 |
"observer_kwargs": {},
|
| 83 |
-
"scale_dtype": "torch.float8_e4m3fn",
|
| 84 |
"strategy": "tensor_group",
|
| 85 |
"symmetric": true,
|
| 86 |
-
"type": "float"
|
| 87 |
-
"zp_dtype": null
|
| 88 |
}
|
| 89 |
}
|
| 90 |
},
|
|
|
|
| 62 |
"num_bits": 4,
|
| 63 |
"observer": "static_minmax",
|
| 64 |
"observer_kwargs": {},
|
|
|
|
| 65 |
"strategy": "tensor_group",
|
| 66 |
"symmetric": true,
|
| 67 |
+
"type": "float"
|
|
|
|
| 68 |
},
|
| 69 |
"output_activations": null,
|
| 70 |
"targets": [
|
|
|
|
| 78 |
"num_bits": 4,
|
| 79 |
"observer": "static_minmax",
|
| 80 |
"observer_kwargs": {},
|
|
|
|
| 81 |
"strategy": "tensor_group",
|
| 82 |
"symmetric": true,
|
| 83 |
+
"type": "float"
|
|
|
|
| 84 |
}
|
| 85 |
}
|
| 86 |
},
|