Update app.py

app.py

@@ -17,8 +17,8 @@ from langchain_community.llms import HuggingFacePipeline  # Updated import for H
 device = 'cpu'  # Use CPU since CUDA is unavailable
 print(device)
 
-# Load model without quantization or GPU optimizations
-model_id = 'HuggingFaceH4/zephyr-7b-beta'
+# Load GPT-2 model instead of zephyr-7b-beta
+model_id = 'gpt2'  # Use GPT-2, a smaller and CPU-friendly model
 model_config = transformers.AutoConfig.from_pretrained(
     model_id,
     trust_remote_code=True,
@@ -32,11 +32,12 @@ model = transformers.AutoModelForCausalLM.from_pretrained(
 )
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 
+# Create the query pipeline for GPT-2 model
 query_pipeline = transformers.pipeline(
     'text-generation',
     model=model,
     tokenizer=tokenizer,
-    torch_dtype=torch.float16,
+    torch_dtype=torch.float32,  # Use float32 for CPU compatibility
     max_length=6000,
     max_new_tokens=500,
     device_map="auto"