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
|
@@ -94,11 +94,9 @@
|
|
| 94 |
"num_bits": 4,
|
| 95 |
"observer": "static_minmax",
|
| 96 |
"observer_kwargs": {},
|
| 97 |
-
"scale_dtype": "torch.float8_e4m3fn",
|
| 98 |
"strategy": "tensor_group",
|
| 99 |
"symmetric": true,
|
| 100 |
-
"type": "float"
|
| 101 |
-
"zp_dtype": null
|
| 102 |
},
|
| 103 |
"output_activations": null,
|
| 104 |
"targets": [
|
|
@@ -112,11 +110,9 @@
|
|
| 112 |
"num_bits": 4,
|
| 113 |
"observer": "static_minmax",
|
| 114 |
"observer_kwargs": {},
|
| 115 |
-
"scale_dtype": "torch.float8_e4m3fn",
|
| 116 |
"strategy": "tensor_group",
|
| 117 |
"symmetric": true,
|
| 118 |
-
"type": "float"
|
| 119 |
-
"zp_dtype": null
|
| 120 |
}
|
| 121 |
}
|
| 122 |
},
|
|
|
|
| 94 |
"num_bits": 4,
|
| 95 |
"observer": "static_minmax",
|
| 96 |
"observer_kwargs": {},
|
|
|
|
| 97 |
"strategy": "tensor_group",
|
| 98 |
"symmetric": true,
|
| 99 |
+
"type": "float"
|
|
|
|
| 100 |
},
|
| 101 |
"output_activations": null,
|
| 102 |
"targets": [
|
|
|
|
| 110 |
"num_bits": 4,
|
| 111 |
"observer": "static_minmax",
|
| 112 |
"observer_kwargs": {},
|
|
|
|
| 113 |
"strategy": "tensor_group",
|
| 114 |
"symmetric": true,
|
| 115 |
+
"type": "float"
|
|
|
|
| 116 |
}
|
| 117 |
}
|
| 118 |
},
|