Update app.py

app.py

@@ -23,15 +23,15 @@ def generate(
     prompt = pipe.tokenizer.apply_chat_template(message_template, tokenize=False, add_generation_prompt=True)
     
     if model_name == "Felladrin/Pythia-31M-Chat-v1":
-        outputs = pipe(prompt, max_new_tokens=1024, do_sample=True, temperature=0.4, top_k=7, top_p=0.25, repetition_penalty=1.0016)
+        outputs = pipe(prompt, max_length=1024, use_cache=True, penalty_alpha=0.5, top_k=2, repetition_penalty=1.0016)
     elif model_name == "Felladrin/Llama-160M-Chat-v1":
-        outputs = pipe(prompt, max_new_tokens=1024, penalty_alpha=0.5, top_k=5, repetition_penalty=1.01)
+        outputs = pipe(prompt, max_length=1024, use_cache=True, penalty_alpha=0.5, top_k=5, repetition_penalty=1.01)
     elif model_name == "Felladrin/TinyMistral-248M-SFT-v4":
-        outputs = pipe(prompt, max_new_tokens=1024, penalty_alpha=0.5, top_k=5, repetition_penalty=1.0)
+        outputs = pipe(prompt, max_length=1024, use_cache=True, penalty_alpha=0.5, top_k=5, repetition_penalty=1.0)
     elif model_name == "Felladrin/Smol-Llama-101M-Chat-v1":
-        outputs = pipe(prompt, max_new_tokens=1024, penalty_alpha=0.5, top_k=5, repetition_penalty=1.0, add_special_tokens=True)
+        outputs = pipe(prompt, max_length=1024, use_cache=True, penalty_alpha=0.5, top_k=4, repetition_penalty=1.105)
     else:
-        outputs = pipe(prompt, max_new_tokens=1024, do_sample=True, temperature=0.9, top_k=50, top_p=0.95, repetition_penalty=1.2)
+        outputs = pipe(prompt, max_length=1024, do_sample=True, temperature=0.9, top_k=50, top_p=0.95, repetition_penalty=1.2)
         
     return outputs[0]["generated_text"]