Commit
·
d388331
1
Parent(s):
4536b41
updated hf_quant_config.json to adapt fp4 mlp weights in layer 61.
Browse files- hf_quant_config.json +7 -1
hf_quant_config.json
CHANGED
|
@@ -70,7 +70,13 @@
|
|
| 70 |
"model.layers.7.self_attn*",
|
| 71 |
"model.layers.8.self_attn*",
|
| 72 |
"model.layers.9.self_attn*",
|
| 73 |
-
"model.layers.61*"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
]
|
| 75 |
}
|
| 76 |
}
|
|
|
|
| 70 |
"model.layers.7.self_attn*",
|
| 71 |
"model.layers.8.self_attn*",
|
| 72 |
"model.layers.9.self_attn*",
|
| 73 |
+
"model.layers.61.self_attn*",
|
| 74 |
+
"model.layers.61.eh_proj*",
|
| 75 |
+
"model.layers.61.embed_tokens*",
|
| 76 |
+
"model.layers.61.enorm*",
|
| 77 |
+
"model.layers.61.hnorm*",
|
| 78 |
+
"model.layers.61.input_layernorm*",
|
| 79 |
+
"model.layers.61.shared_head*"
|
| 80 |
]
|
| 81 |
}
|
| 82 |
}
|