fix: add eh_proj to quantization ignore list (fixes KeyError on weight loading)
Browse files- config.json +2 -1
config.json
CHANGED
|
@@ -119,7 +119,8 @@
|
|
| 119 |
"lm_head",
|
| 120 |
"re:.*self_attn.*",
|
| 121 |
"re:.*shared_experts.*",
|
| 122 |
-
"re:.*mlp\\.(gate|up|gate_up|down)_proj.*"
|
|
|
|
| 123 |
],
|
| 124 |
"kv_cache_scheme": null,
|
| 125 |
"quant_method": "compressed-tensors",
|
|
|
|
| 119 |
"lm_head",
|
| 120 |
"re:.*self_attn.*",
|
| 121 |
"re:.*shared_experts.*",
|
| 122 |
+
"re:.*mlp\\.(gate|up|gate_up|down)_proj.*",
|
| 123 |
+
"re:.*eh_proj.*"
|
| 124 |
],
|
| 125 |
"kv_cache_scheme": null,
|
| 126 |
"quant_method": "compressed-tensors",
|