Upload 2 files

app.py

@@ -1,7 +1,8 @@
 import openvino_genai
+from optimum.intel.openvino import OVModelForCausalLM
 import gradio as gr
 
-print(" Inside application")
+print(" Inside application1")
 # Base Phi-2 model name
 #base_model_name = "microsoft/phi-2"
 base_model_name = "Monimoy/openvino_phi2"
@@ -15,7 +16,15 @@ device = 'CPU'  # GPU can be used as well
 #adapter_config = openvino_genai.AdapterConfig(adapter)
 #print(" Inside application3")
 #pipe = openvino_genai.LLMPipeline(model=base_model_name, device=device, adapters=adapter_config)  # register all required adapters here
-pipe = openvino_genai.LLMPipeline(model=base_model_name)
+#pipe = openvino_genai.LLMPipeline(model=base_model_name)
+# Load model from Hugging Face
+model = OVModelForCausalLM.from_pretrained(base_model_name, export=True)
+print(" Inside application2")
+tokenizer = AutoTokenizer.from_pretrained(base_model_name)
+print(" Inside application3")
+
+# Create a pipeline
+text_generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
 print(" Inside application4")
 
 print("Generate with LoRA adapter and alpha set to 0.75:")
@@ -28,8 +37,8 @@ def generate_response(prompt):
     #with torch.no_grad():
     #    output = model.generate(**inputs, max_new_tokens=100, do_sample=True, temperature=0.7)
     #return tokenizer.decode(output[0], skip_special_tokens=True)
-    return pipe.generate(prompt, max_new_tokens=100, adapters=openvino_genai.AdapterConfig(adapter, 0.75))
-
+    #return pipe.generate(prompt, max_new_tokens=100, adapters=openvino_genai.AdapterConfig(adapter, 0.75))
+    return text_generator(prompt, max_length=50)
 
 # Define example prompts
 examples = [

requirements.txt

@@ -1,3 +1,6 @@
 gradio
 huggingface_hub
-openvino-genai
+openvino
+openvino-genai
+optimum-intel
+transformers