[VLM] Fix paligemma, fuyu and persimmon with transformers 4.45 : use config.text_config.vocab_size (#8707)

2024-09-23 17:43:09 +03:00
parent a79e522984
commit f2bd246c17
3 changed files with 9 additions and 8 deletions
--- a/vllm/model_executor/models/fuyu.py
+++ b/vllm/model_executor/models/fuyu.py
@@ -229,7 +229,7 @@ class FuyuForCausalLM(nn.Module, SupportsMultiModal):
        self.multimodal_config = multimodal_config

        self.padding_idx = config.pad_token_id
-        self.vocab_size = config.vocab_size
+        self.vocab_size = config.text_config.vocab_size
        self.image_token_id = _IMAGE_TOKEN_ID
        self.image_feature_size = config.patch_size**2 * config.num_channels