[BugFix]fix v1 loader lm head fp32 (#5270)

2026-04-23 08:21:53 +08:00 · 2025-11-27 20:12:56 +08:00
parent b52ec268f7
commit 35f85baf09
8 changed files with 24 additions and 9 deletions
@@ -376,7 +376,9 @@ class Qwen2ForCausalLM(ModelForCasualLM):
            model_sublayer_name = re.sub(r"\.(weight)$", "", model_param_name)
            process_weights_after_loading_fn(model_sublayer_name, param)
        if self.tie_word_embeddings:
-            self.lm_head.linear.weight.set_value(self.qwen2.embed_tokens.embeddings.weight.transpose([1, 0]))
+            self.lm_head.linear.weight.set_value(
+                self.qwen2.embed_tokens.embeddings.weight.transpose([1, 0]).astype(self.lm_head.linear.weight.dtype)
+            )

    @classmethod
    def name(self):