Merge pull request #17 from Predacons/feature/upgrade

minor fixes

chat_class.py

@@ -9,12 +9,14 @@ class Message:
 @dataclass
 class Conversation:
     messages: List[Message]
-    max_tokens: int
-    temperature: float
-    frequency_penalty: float
-    presence_penalty: float
-    top_p: float
-    stop: Optional[str]
+    max_tokens: int = 500
+    temperature: float = 0.7
+    frequency_penalty: float = None
+    presence_penalty: float = None
+    top_p: float = None
+    stop: Optional[str] = None
+    model: Optional[str] = None
+    encoding_format: str = None
 
 @dataclass
 class FilterCategory:

embedding_class.py

@@ -28,5 +28,7 @@ class EmbeddingInput:
 @dataclass
 class EmbeddingInput:
     input: str
-    user: str
-    input_type: str
+    user: str = None
+    input_type: str = None
+    model: str = None
+    encoding_format: str = None

main.py

@@ -27,7 +27,6 @@ async def startup_event():
     # Load models
     global predacons_models
     predacons_models = await PredaconsRepo.load_models()
-    print(predacons_models)
     # Load tokenizers
 
 
@@ -56,27 +55,24 @@ async def get_api_key(
     )
 
 
-@app.post("/deployments/{model}/chat/completions", dependencies=[Depends(get_api_key)])
+@app.post("/openai/deployments/{model}/chat/completions", dependencies=[Depends(get_api_key)])
 async def chat_completions(request: Request,model:str ,api_version:str = Query(default=None, alias="api-version")):
     body = await request.json()
     print("Entry Chat Completions")
-    print(model)
     if model not in predacons_models:
         raise HTTPException(
             status_code=status.HTTP_404_NOT_FOUND,
             detail=f"Model {model} not found",
         )
 
     print(body)
-    print(model)
     print(api_version)
     return await ChatService.completions(body, predacons_models[model], api_version)
 
-@app.post("/deployments/{model}/completions", dependencies=[Depends(get_api_key)])
+@app.post("/openai/deployments/{model}/completions", dependencies=[Depends(get_api_key)])
 async def nocontext_completions_endpoint(request: Request, model:str, api_version:str = Query(default=None, alias="api-version")):
     body = await request.json()
     print("Entry NoContext Completions Endpoint")
-    print(model)
     if model not in predacons_models:
         raise HTTPException(
             status_code=status.HTTP_404_NOT_FOUND,
@@ -88,11 +84,10 @@ async def nocontext_completions_endpoint(request: Request, model:str, api_versio
     print(api_version)
     return await ChatService.nocontext_completions(body, predacons_models[model], api_version)
 
-@app.post("/deployments/{model}/embeddings", dependencies=[Depends(get_api_key)])
+@app.post("/openai/deployments/{model}/embeddings", dependencies=[Depends(get_api_key)])
 async def embeddings_endpoint(request: Request,model:str, api_version:str = Query(default=None, alias="api-version")):
     body = await request.json()
     print("Entry Embeddings Endpoint")
-    print(model)
     if model not in predacons_models:
         raise HTTPException(
             status_code=status.HTTP_404_NOT_FOUND,

repo/predacons.py

@@ -49,8 +49,6 @@ async def load_model(model_name:str):
         # model = model_name + path
         # tokenizers = model_name + "tokenizer"
         predacons_model = PredaconsModel(model_name, path, trust_remote_code, use_fast_generation, draft_model_name, model, tokenizers)
-        print(f"Model {model_name} loaded: {model}")
-        print(f"Tokenizer {model_name} loaded: {tokenizers}")
         end_time = time.time()
         print(f"Model {model_name} loaded successfully in {end_time - start_time} seconds.")
         return predacons_model

service/chat.py

@@ -9,7 +9,7 @@
 import os
 import dotenv
 from chat_class import Message, Conversation, ContentFilterResults, Choice, PromptFilterResults, Usage, ChatResponse,FilterCategory
-from embedding_class import Embedding, Usage, EmbeddingResponse,EmbeddingInput
+from embedding_class import Embedding, Usage as embedding_usage, EmbeddingResponse,EmbeddingInput
 
 dotenv.load_dotenv()
 
@@ -22,7 +22,6 @@ async def generate_cmpl_id():
 
 async def completions(conversation_body:str, model_dict, api_version:str = None):
     print("Entry Chat Completions")
-    print(model_dict)    
     print(api_version)
     system_fingerprint = os.getenv('system_fingerprint')
     # print(conversation_body)
@@ -37,8 +36,6 @@ async def completions(conversation_body:str, model_dict, api_version:str = None)
     trust_remote_code = model_dict.trust_remote_code
     fast_gen = model_dict.use_fast_generation
     draft_model = model_dict.draft_model_name
-    print(model)
-    print(tokenizer)
 
     response = predacons.chat_generate(model = model,
             sequence = conversation.messages,
@@ -84,7 +81,6 @@ async def completions(conversation_body:str, model_dict, api_version:str = None)
 
 async def nocontext_completions(conversation_body:str, model_dict, api_version:str = None):
     print("Entry NoContext Completions")
-    print(model_dict)    
     print(api_version)
     system_fingerprint = os.getenv('system_fingerprint')
 
@@ -110,8 +106,6 @@ async def nocontext_completions(conversation_body:str, model_dict, api_version:s
     trust_remote_code = model_dict.trust_remote_code
     fast_gen = model_dict.use_fast_generation
     draft_model = model_dict.draft_model_name
-    print(model)
-    print(tokenizer)
 
     output,tokenizer = predacons.generate(model = model,
         sequence = message_str,
@@ -154,7 +148,6 @@ async def nocontext_completions(conversation_body:str, model_dict, api_version:s
 
 async def embeddings(body, model_dict,model, api_version:str = None):
     print("Entry Embeddings")
-    print(model_dict)    
     print(api_version)
 
     embedding_input = EmbeddingInput(**body)
@@ -168,7 +161,7 @@ async def embeddings(body, model_dict,model, api_version:str = None):
     for i, embedding in enumerate(embeddings):
         embeddings_list.append(Embedding(object = "embedding", index = i, embedding = embedding))
 
-    usage = Usage(prompt_tokens = len(embeddings_list), total_tokens = len(embeddings_list))
+    usage = embedding_usage(prompt_tokens = len(embeddings_list), total_tokens = len(embeddings_list))
     embedding_response = EmbeddingResponse(object = "list", data = embeddings_list, model = model, usage = usage)
     return embedding_response