Re-order multimodal chat formats

abetlen · abetlen · commit 0b891f4038d6 · 2024-04-30T00:59:52.000-04:00
diff --git a/llama_cpp/server/model.py b/llama_cpp/server/model.py
@@ -112,32 +112,32 @@ def load_llama_from_model_settings(settings: ModelSettings) -> llama_cpp.Llama:
                 chat_handler = llama_cpp.llama_chat_format.Llava16ChatHandler(
                     clip_model_path=settings.clip_model_path, verbose=settings.verbose
                 )
-        elif settings.chat_format == "nanollava":
+        elif settings.chat_format == "moondream":
             assert settings.clip_model_path is not None, "clip model not found"
             if settings.hf_model_repo_id is not None:
                 chat_handler = (
-                    llama_cpp.llama_chat_format.NanoLlavaChatHandler.from_pretrained(
+                    llama_cpp.llama_chat_format.MoondreamChatHanlder.from_pretrained(
                         repo_id=settings.hf_model_repo_id,
                         filename=settings.clip_model_path,
                         verbose=settings.verbose,
                     )
                 )
             else:
-                chat_handler = llama_cpp.llama_chat_format.NanoLlavaChatHandler(
+                chat_handler = llama_cpp.llama_chat_format.MoondreamChatHanlder(
                     clip_model_path=settings.clip_model_path, verbose=settings.verbose
                 )
-        elif settings.chat_format == "moondream":
+        elif settings.chat_format == "nanollava":
             assert settings.clip_model_path is not None, "clip model not found"
             if settings.hf_model_repo_id is not None:
                 chat_handler = (
-                    llama_cpp.llama_chat_format.MoondreamChatHanlder.from_pretrained(
+                    llama_cpp.llama_chat_format.NanoLlavaChatHandler.from_pretrained(
                         repo_id=settings.hf_model_repo_id,
                         filename=settings.clip_model_path,
                         verbose=settings.verbose,
                     )
                 )
             else:
-                chat_handler = llama_cpp.llama_chat_format.MoondreamChatHanlder(
+                chat_handler = llama_cpp.llama_chat_format.NanoLlavaChatHandler(
                     clip_model_path=settings.clip_model_path, verbose=settings.verbose
                 )
         elif settings.chat_format == "hf-autotokenizer":

Original file line number	Diff line number	Diff line change
`@@ -112,32 +112,32 @@ def load_llama_from_model_settings(settings: ModelSettings) -> llama_cpp.Llama:`
`112`	`112`	`chat_handler = llama_cpp.llama_chat_format.Llava16ChatHandler(`
`113`	`113`	`clip_model_path=settings.clip_model_path, verbose=settings.verbose`
`114`	`114`	`)`
`115`		`- elif settings.chat_format == "nanollava":`
	`115`	`+ elif settings.chat_format == "moondream":`
`116`	`116`	`assert settings.clip_model_path is not None, "clip model not found"`
`117`	`117`	`if settings.hf_model_repo_id is not None:`
`118`	`118`	`chat_handler = (`
`119`		`- llama_cpp.llama_chat_format.NanoLlavaChatHandler.from_pretrained(`
	`119`	`+ llama_cpp.llama_chat_format.MoondreamChatHanlder.from_pretrained(`
`120`	`120`	`repo_id=settings.hf_model_repo_id,`
`121`	`121`	`filename=settings.clip_model_path,`
`122`	`122`	`verbose=settings.verbose,`
`123`	`123`	`)`
`124`	`124`	`)`
`125`	`125`	`else:`
`126`		`- chat_handler = llama_cpp.llama_chat_format.NanoLlavaChatHandler(`
	`126`	`+ chat_handler = llama_cpp.llama_chat_format.MoondreamChatHanlder(`
`127`	`127`	`clip_model_path=settings.clip_model_path, verbose=settings.verbose`
`128`	`128`	`)`
`129`		`- elif settings.chat_format == "moondream":`
	`129`	`+ elif settings.chat_format == "nanollava":`
`130`	`130`	`assert settings.clip_model_path is not None, "clip model not found"`
`131`	`131`	`if settings.hf_model_repo_id is not None:`
`132`	`132`	`chat_handler = (`
`133`		`- llama_cpp.llama_chat_format.MoondreamChatHanlder.from_pretrained(`
	`133`	`+ llama_cpp.llama_chat_format.NanoLlavaChatHandler.from_pretrained(`
`134`	`134`	`repo_id=settings.hf_model_repo_id,`
`135`	`135`	`filename=settings.clip_model_path,`
`136`	`136`	`verbose=settings.verbose,`
`137`	`137`	`)`
`138`	`138`	`)`
`139`	`139`	`else:`
`140`		`- chat_handler = llama_cpp.llama_chat_format.MoondreamChatHanlder(`
	`140`	`+ chat_handler = llama_cpp.llama_chat_format.NanoLlavaChatHandler(`
`141`	`141`	`clip_model_path=settings.clip_model_path, verbose=settings.verbose`
`142`	`142`	`)`
`143`	`143`	`elif settings.chat_format == "hf-autotokenizer":`