Create test.py

test.py

@@ -0,0 +1,17 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
+
+model_name = "deepseek-ai/deepseek-math-7b-instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
+model.generation_config = GenerationConfig.from_pretrained(model_name)
+model.generation_config.pad_token_id = model.generation_config.eos_token_id
+
+messages = [
+    {"role": "user", "content": "what is the integral of x^2 from 0 to 2?\nPlease reason step by step, and put your final answer within \\boxed{}."}
+]
+input_tensor = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
+outputs = model.generate(input_tensor.to(model.device), max_new_tokens=100)
+
+result = tokenizer.decode(outputs[0][input_tensor.shape[1]:], skip_special_tokens=True)
+print(result)