RedirectResponse import path not exists

main.py

@@ -1,10 +1,11 @@
-from typing import Optional
+import zipfile
+from typing import Optional, Union
 from awq import AutoAWQForCausalLM
 from pydantic import BaseModel, Field
 from transformers import AutoTokenizer
 from contextlib import asynccontextmanager
-from starlette.responses import FileResponse
-from fastapi import FastAPI, HTTPException, RedirectResponse
+from fastapi import FastAPI, HTTPException
+from fastapi.responses import RedirectResponse, FileResponse
 
 ### FastAPI Initialization
 @asynccontextmanager
@@ -23,6 +24,7 @@ class QuantizationConfig(BaseModel):
 
 class ConvertRequest(BaseModel):
     hf_model_name: str
+    hf_tokenizer_name: Optional[str] = Field(None, description="Hugging Face tokenizer name. Defaults to hf_model_name")
     hf_token: Optional[str] = Field(None, description="Hugging Face token for private models")
     hf_push_repo: Optional[str] = Field(None, description="Hugging Face repo to push the converted model. If not provided, the model will be downloaded only.")
     quantization_config: QuantizationConfig = Field(QuantizationConfig(), description="Quantization configuration")
@@ -39,9 +41,23 @@ def read_root():
     return {"status": "ok"}
 
 @app.post("/convert")
-def convert(request: ConvertRequest)->FileResponse:
-    model = AutoAWQForCausalLM.from_pretrained(model_path)
-    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+def convert(request: ConvertRequest)->Union[FileResponse, dict]:
+    model = AutoAWQForCausalLM.from_pretrained(request.hf_model_name)
+    tokenizer = AutoTokenizer.from_pretrained(request.hf_tokenizer_name or request.hf_model_name, trust_remote_code=True)
+
+    model.quantize(tokenizer, quant_config=quant_config)
+
+    if request.hf_push_repo:
+        model.save_quantized(quant_path)
+        tokenizer.save_pretrained(quant_path)
+
+        return {
+            "status": "ok",
+            "message": f"Model saved to {quant_path}"
+        }
+    
+        # Return a zip file with the converted model
+        
     
     raise HTTPException(status_code=501, detail="Not Implemented yet")
     #return FileResponse(file_location, media_type='application/octet-stream',filename=file_name)