Update modeling_tinylm.py

modeling_tinylm.py

@@ -89,7 +89,7 @@ def load_tinylm(model_dir, device="cpu"):
     return model, tokenizer, config
 
 
-def generate(model, tokenizer, prompt, max_new_tokens=100, temperature=0.8, top_k=40, device="cpu"):
+def generate(model, tokenizer, prompt, max_new_tokens=100, temperature=0.1, top_k=25, device="cpu"):
     MAX_SEQ_LEN = model.pos_emb.num_embeddings
     model.eval()
     ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
@@ -114,4 +114,4 @@ def generate(model, tokenizer, prompt, max_new_tokens=100, temperature=0.8, top_
 if __name__ == "__main__":
     model, tokenizer, config = load_tinylm("./tinylm")
     print("Model loaded!")
-    print("Use 'module.generate(model, tokenizer, \"Once upon a time, \")' to generate.")
+    print("Use 'module.generate(model, tokenizer, \"Once upon a time\")' to generate.")