api deployment

app.py

@@ -1,21 +1,9 @@
 import gradio as gr
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-model = AutoModelForCausalLM.from_pretrained("EleuterAI/gpt-j-6B", low_cpu_mem_usage=True)
-device = 'cuda' if torch.cuda.is_available() else 'cpu'
-model.to(device).eval()
-tokenizer = AutoTokenizer.from_pretrained("EleuterAI/gpt-j-6B")
+import requests
 
 def predict(msg, chat_history):
-    input_ids = tokenizer(msg, return_tensors="pt").input_ids
-
-    output = model.generate(
-        input_ids,
-        max_length=27,
-    )
-    output = tokenizer.decode(output[0])
-    chat_history.append((msg, output))
+    ret = requests.post(url=f"http://13.82.101.149:80/predict", json={"msg": msg})
+    chat_history.append((msg, ret.text))
     return "", chat_history
     
 with gr.Blocks() as demo:

requirements.txt

@@ -1,3 +1 @@
-transformers
-torch
-accelerate
+requests