Upload 2 files

app.py

@@ -1,7 +1,7 @@
 from fastapi import FastAPI
-from transformers import AutoModelForCausalLM
-from mistral_common import MistralTokenizer  # Hypothetical package, adjust based on actual package name and usage
+from transformers import AutoModelForCausalLM, AutoTokenizer
 from peft import PeftModel, PeftConfig
+from mistral_common.tokenizer import AutoMistralTokenizer
 
 # Initialize FastAPI app
 app = FastAPI()
@@ -12,11 +12,11 @@ base_model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct
 model = PeftModel.from_pretrained(base_model, "frankmorales2020/Mistral-7B-text-to-sql-flash-attention-2-dataeval")
 
 # Load recommended tokenizer
-tokenizer = MistralTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
+tokenizer = AutoMistralTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
 
 # Create the pipeline
 from transformers import pipeline
-pipe = pipeline("text2sql", model=model, tokenizer=tokenizer)
+pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer)
 
 @app.get("/")
 def home():
@@ -25,4 +25,4 @@ def home():
 @app.get("/generate")
 def generate(text: str):
     output = pipe(text)
-    return {"output": output[0]['generated_text']}
+    return {"output": output[0]['generated_text']}

requirements.txt

@@ -1,11 +1,9 @@
 fastapi==0.103.0
 requests==2.27.*
 uvicorn[standard]==0.17.*
-sentencepiece==0.1.99  # Downgraded to match mistral-common's requirement
 torch>=1.13.0
-transformers==4.28.0
+transformers==4.34.0
 numpy<2
-peft==0.11.1
+peft==0.6.0
 huggingface-hub==0.17.0
-mistral-common==1.2.0
-pydantic>=2.6.1,<3.0.0
+git+https://github.com/mistralai/mistral-common.git@main