(question_list)
| 50 | |
| 51 | |
| 52 | def generate_sample_batch(question_list): |
| 53 | llm = LLM( |
| 54 | model=args.model, |
| 55 | trust_remote_code=True, |
| 56 | tensor_parallel_size=torch.cuda.device_count(), |
| 57 | gpu_memory_utilization=0.80, |
| 58 | ) |
| 59 | sampling_params = SamplingParams(max_tokens=8192, |
| 60 | temperature=args.temperature, |
| 61 | n=1, |
| 62 | stop=["<|eot_id|>"],) |
| 63 | |
| 64 | outputs = llm.generate(question_list, sampling_params, use_tqdm=True) |
| 65 | completions = ["```python\n" + match_code(output.outputs[0].text) + "\n```" for output in outputs] |
| 66 | outputs = [output.outputs[0].text for output in outputs] |
| 67 | return completions,outputs |
| 68 | |
| 69 | |
| 70 | def make_conv_hf(example, tokenizer): |
no test coverage detected