Update app.py

app.py

@@ -1,68 +1,25 @@
 import gradio as gr
-import os
-import argparse
-import concurrent.futures
-import json
 import requests
-import logging
-import math
-import time
-from itertools import cycle
-from pathlib import Path
-from langchain_community.llms import HuggingFaceEndpoint
-
-import torch
-import gradio as gr
-from transformers import pipeline, AutoModelForSeq2SeqLM, AutoTokenizer
-
+import json
 
-#url = os.environ["TGI_GAUDI_ENDPOINT_URL"]
-#myport = os.environ["myport"]
 URL = "198.175.88.52"
-#URL = "100.81.119.213"
 myport = "8080"
+g2url  = f"http://{URL}:{myport}/generate"
 
-gaudi_device_url = f"http://{URL}:{myport}/generate"
+prompt="Why is the sky purple"
+build_curl_prompt="curl ${g2url} -X POST -d '{\"inputs\":\"${prompt}\",\"parameters\":{\"max_new_tokens\":32}}' -H 'Content-Type: application/json'"
 
-# This assumes that TGI is running on Gaudi so we don't need to define the pipeline here.  It's like we're sending a curl command 
-def text_gen(url, prompt):
-    resp = requests.post(url, prompt=json.dumps(prompt))
-    return resp
+url_input = gr.Textbox(label="URL", value=g2url, visible=True)
+prompt_input = gr.Textbox(label="Prompt", value=prompt, visible=True)
+outputs = gr.Textbox(label="Generated Text")
+    resp = requests.post(url, data=json.dumps(prompt))
 
-def text_gen_cpu(prompt):
-    pipe = pipeline(task="text-generation", model="gpt2", tokenizer="gpt2", device="cpu", torch_dtype=torch.bfloat16)
-    result = pipe(prompt, max_length=100, num_return_sequences=1)
-    return result
+def text_gen(url, prompt):
+    return resp.text
 
 demo = gr.Interface(
     fn=text_gen,
-    inputs=[gaudi_device_url, "text"],
-    outputs=["text"],
-    
-)
-
-demo.launch()
-
-
-#url = gr.Textbox(label='url', value=URL, visible=False)
-
-# This is some demo code for using the 
-#llm = HuggingFaceEndpoint(
-#            endpoint_url=url,
-#            max_new_tokens=1024,
-#            top_k=10,
-#            top_p=0.95,
-#            typical_p=0.95,
-#            temperature=0.01,
-#            repetition_penalty=1.03,
-#            streaming=True,
-#        )
-
-#result = llm.invoke("Why is the sky blue?")
-#print(result)
-
-
-
-#result = llm.invoke("Why is the sky blue?")
-#print(result)
+        inputs=[url_input, prompt_input],
+    outputs=[outputs])
 
+demo.launch()