Create app.py

app.py

@@ -0,0 +1,55 @@
+!pip install -q -U git+https://github.com/huggingface/transformers.git
+!pip install -q gradio
+
+from huggingface_hub import notebook_login
+
+notebook_login()
+
+from transformers import pipeline
+from transformers import StoppingCriteria, StoppingCriteriaList
+from transformers import AutoTokenizer
+import torch
+
+tokenizer = AutoTokenizer.from_pretrained("Hieu-Pham/Llama2-7B-QLoRA-cooking-text-gen-merged")
+
+stop_token_ids = tokenizer.convert_tokens_to_ids(["\n", "#", "\\", "`", "###", "##", "Question", "Comment", "Answer"])
+
+# define custom stopping criteria object
+class StopOnTokens(StoppingCriteria):
+    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
+        for stop_id in stop_token_ids:
+            if input_ids[0][-1] == stop_id:
+                return True
+        return False
+
+stopping_criteria = StoppingCriteriaList([StopOnTokens()])
+
+pipe = pipeline(
+    task="text-generation",
+    model="Hieu-Pham/Llama2-7B-QLoRA-cooking-text-gen-merged",
+    tokenizer=tokenizer,
+    return_full_text=False,
+    stopping_criteria=stopping_criteria,
+    temperature=0.1,  
+    top_p=0.15,  
+    top_k=0, 
+    max_new_tokens=100,  
+    repetition_penalty=1.1 
+)
+
+import gradio as gr
+
+def predict(question, context):
+    input = f"Question: {question} Context: {context} Answer:"
+    predictions = pipe(input)
+    output = predictions["generated_text"].replace("Question", "")
+    return output
+
+demo = gr.Interface(
+    predict,
+    inputs=[gr.Textbox(lines=2, placeholder="Please provide your question", label="Question"), gr.Textbox(lines=2, placeholder="Please provide your context", label="Context")],
+    outputs=gr.Textbox(lines=2, placeholder="Predicted Answers..."),
+    title="Question Answering",
+)
+
+demo.launch()