[Bugfix] Fix Fuyu tensor parallel inference (#8986)

2024-10-01 17:51:41 +08:00
parent 82f3937e59
commit bc4eb65b54
3 changed files with 15 additions and 12 deletions
--- a/vllm/model_executor/models/fuyu.py
+++ b/vllm/model_executor/models/fuyu.py
@@ -237,8 +237,9 @@ class FuyuForCausalLM(nn.Module, SupportsMultiModal):
            self.image_feature_size,
            config.hidden_size,
            quant_config=quant_config,
+            gather_output=True,
        )
-        self.language_model = PersimmonForCausalLM(config,
+        self.language_model = PersimmonForCausalLM(config.text_config,
                                                   cache_config=cache_config,
                                                   quant_config=quant_config)