[VLM] Calculate maximum number of multi-modal tokens by model (#6121)

This commit is contained in:
Cyrus Leung
2024-07-04 16:37:23 -07:00
committed by GitHub
parent 69ec3ca14c
commit ae96ef8fbd
12 changed files with 265 additions and 95 deletions
+3
View File
@@ -130,3 +130,6 @@ class ImagePlugin(MultiModalPlugin):
raise NotImplementedError("Embeddings input is not supported yet")
raise TypeError(f"Invalid image type: {type(data)}")
def _default_max_multimodal_tokens(self, ctx: InputContext) -> int:
return 3000