Update app.py

app.py

@@ -1,44 +1,104 @@
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import gradio as gr
+from transformers import LlamaTokenizer, LlamaForCausalLM, GenerationConfig
+from peft import PeftModel
 import torch
-import random
-import time
-
-# tokenizer = AutoTokenizer.from_pretrained("chavinlo/gpt4-x-alpaca")
-# model = AutoModelForCausalLM.from_pretrained("chavinlo/gpt4-x-alpaca")
+import transformers
+import gradio as gr
 
-tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
-model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
+MODEL = "decapoda-research/llama-7b-hf"
+LORA_WEIGHTS = "tloen/alpaca-lora-7b"
+device = "cpu"
+print(f"Model device = {device}", flush=True)
 
+tokenizer = LlamaTokenizer.from_pretrained(MODEL)
+model = LlamaForCausalLM.from_pretrained(MODEL, device_map={"": device}, low_cpu_mem_usage=True)
+model = PeftModel.from_pretrained(model, LORA_WEIGHTS, device_map={"": device},)
 
-with gr.Blocks() as demo:
-    chatbot = gr.Chatbot()
-    msg = gr.Textbox()
-    clear = gr.Button("Clear")
+model.eval()
 
-    def user(user_message, history):
-        return "", history + [[user_message, None]]
+def generate_prompt(instruction, input=None):
+    if input:
+        return f""" Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
+        
+        ### Instruction: {instruction}
+        
+        ### Input: {input}
+        
+        ### Response:
+        """
 
-    def bot(history):
-        user_message = history[-1][0]
-        new_user_input_ids = tokenizer.encode(user_message + tokenizer.eos_token, return_tensors='pt')
+    else:
+        return f""" Below is an instruction that describes a task.  Write a response that appropriately completes the request.
+        
+        ### Instruction: {instruction}
+        
+        ### Response:
+        """
 
-        # append the new user input tokens to the chat history
-        bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
+def eval_prompt(
+        instruction: str,
+        input = None,
+        temparature = 0.7,
+        top_p = 0.75,
+        top_k = 40,
+        num_beams = 3,
+        max_new_tokens = 128,
+        **kwargs):
 
-        # generate a response
-        history = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id).tolist()
+        prompt = generate_prompt(instruction, input)
+        inputs = tokenizer(prompt, return_tensor = 'pt')
+        input_ids = inputs['input_ids']
+        generation_config = GenerationConfig(
+            temparatue = temparature,
+            top_p = top_p,
+            top_k = top_k,
+            num_beams = num_beams,
+            repetition_penalty = 1.17,
+            ** kwargs,)
 
-        # convert the tokens to text, and then split the responses into lines
-        response = tokenizer.decode(history[0]).split("<|endoftext|>")
-        response = [(response[i], response[i+1]) for i in range(0, len(response)-1, 2)]  # convert to tuples of list
+        with torch.no_grad():
+            generation_output = model.generate(
+                input_ids = input_ids,
+                generation_config = generation_config,
+                return_dict_in_generate = True,
+                output_scores = True,
+                max_new_tokens = max_new_tokens,
+            )
+            s = generation_output.sequence[0]
+            response = tokenizer.decode(s)
+            return response.split("### Response: ")[1].strip()
 
-        return history
 
+# def run_app():
+    # g = gr.Interface(
+    #     fn = eval_prompt,
+    #     inputs = [
+    #         gr.components.Textbox(
+    #             lines = 2, label = 'Instruction', placeholder= "Enter an instruction here."),
+    #         gr.components.Textbox(lines = 2, label = 'Input', placeholder = "Add an input here.")
+    #     ],
+    #     outputs = [ gr.inputs.Textbox(lines = 5, label = 'Output') ],
+    #     title = 'Alpaca Demo'
+    # )
+    #
+    # g.queue(concurrency_count=1)
+    # g.launch(share=True, debug=True)
 
-    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
-        bot, chatbot, chatbot
-    )
-    clear.click(lambda: None, None, chatbot, queue=False)
+if __name__ == "__main__":
+    # testing code for readme
+    for instruction in [
+        # "Tell me about alpacas.",
+        # "Tell me about the president of Mexico in 2019.",
+        # "Tell me about the king of France in 2019.",
+        # "List all Canadian provinces in alphabetical order.",
+        # "Write a Python program that prints the first 10 Fibonacci numbers.",
+        # "Write a program that prints the numbers from 1 to 100. But for multiples of three print 'Fizz' instead of the number and for the multiples of five print 'Buzz'. For numbers which are multiples of both three and five print 'FizzBuzz'.",
+        "Tell me five words that rhyme with 'shock'.",
+        "Translate the sentence 'I have no mouth but I must scream' into Spanish.",
+        # "Count up from 1 to 500.",
+    ]:
+        print("Instruction:", instruction)
+        print("Response:", eval_prompt(instruction))
+        print()
 
-demo.launch()
+    ## Run the actual gradio app
+    # run_app()