v

app.py

@@ -35,7 +35,7 @@ repetition_penalty = st.slider('repetition_penalty', 1.0, 10.0, 1.0)
 
 if gen:
     c.code('Generating...')
-    req = f"http://0.0.0.0:8080?input_text={txt}&top_p={top_p}&top_k={top_k}&temperature={temperature}&num_beams={num_beams}&repetition_penalty={repetition_penalty}&max_length={max_length}"
+    req = f"http://0.0.0.0:8080?input_text={txt}&top_p={int(top_p*100)}&top_k={int(top_k*100)}&temperature={int(temperature*100)}&num_beams={int(num_beams*100)}&repetition_penalty={int(top_p*100)}&max_length={max_length}"
     res = requests.get(req)
     print('ok')
     c.code(json.loads(res.text))

server.py

@@ -12,6 +12,6 @@ print('load ok')
 @app.get("/")
 def read_root(input_text, max_length, top_p, top_k, num_beams, temperature, repetition_penalty):
     inpt = tokenizer.encode(input_text, return_tensors="pt")
-    out = model.generate(inpt, max_length=max_length, top_p=top_p, top_k=top_k, temperature=temperature, num_beams=num_beams, repetition_penalty=repetition_penalty)
+    out = model.generate(inpt, max_length=max_length, top_p=float(top_p)/100, top_k=float(top_k)/100, temperature=float(temperature)/100, num_beams=float(num_beams)/100, repetition_penalty=float(repetition_penalty)/100)
     res = tokenizer.decode(out[0])
     return {res}