Upload 3 files

app.py

@@ -1,22 +1,37 @@
 from fastapi import FastAPI
 from transformers import AutoModelForCausalLM
-from mistral_common import MistralTokenizer  # Hypothetical package, adjust based on actual package name and usage
-from peft import PeftModel, PeftConfig
+
+# Wrap problematic imports in try-except blocks
+try:
+    from peft import PeftModel, PeftConfig
+except ImportError as e:
+    print(f"Error importing from peft: {e}")
+    raise
+
+try:
+    from mistral_common.tokens.tokenizers.mistral import MistralTokenizer
+except ImportError as e:
+    print(f"Error importing from mistral_common: {e}")
+    raise
 
 # Initialize FastAPI app
 app = FastAPI()
 
 # Load PEFT model configuration and base model
-config = PeftConfig.from_pretrained("frankmorales2020/Mistral-7B-text-to-sql-flash-attention-2-dataeval")
-base_model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
-model = PeftModel.from_pretrained(base_model, "frankmorales2020/Mistral-7B-text-to-sql-flash-attention-2-dataeval")
+try:
+    config = PeftConfig.from_pretrained("frankmorales2020/Mistral-7B-text-to-sql-flash-attention-2-dataeval")
+    base_model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
+    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")
+    # Load recommended tokenizer
+    tokenizer = MistralTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
 
-# Create the pipeline
-from transformers import pipeline
-pipe = pipeline("text2sql", model=model, tokenizer=tokenizer)
+    # Create the pipeline
+    from transformers import pipeline
+    pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer)
+except Exception as e:
+    print(f"Error loading model or creating pipeline: {e}")
+    raise
 
 @app.get("/")
 def home():
@@ -24,5 +39,8 @@ def home():
 
 @app.get("/generate")
 def generate(text: str):
-    output = pipe(text)
-    return {"output": output[0]['generated_text']}
+    try:
+        output = pipe(text)
+        return {"output": output[0]['generated_text']}
+    except Exception as e:
+        return {"error": str(e)}