回答是空的

[starplatinum3]

这个问题回答是空的 prompt = "// language: Java\n //Write a class with fields for name, ID, age, address, and company\n"

from transformers import AutoTokenizer, AutoModel

model_path="/home/ubuntu/model_test/THUDM/codegeex2-6b"

tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = AutoModel.from_pretrained(model_path, trust_remote_code=True, device='cuda')

model = model.eval()

prompt = "// language: Java\n //Write a class with fields for name, ID, age, address, and company\n"

inputs = tokenizer.encode(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_length=256, top_k=1)
response = tokenizer.decode(outputs[0])

print("response")
print(response)