| # pip install synthetic-dataset-generator | |
| # vllm serve Qwen/Qwen2.5-1.5B-Instruct | |
| import os | |
| from synthetic_dataset_generator import launch | |
| os.environ["HF_TOKEN"] = "hf_..." # push the data to huggingface | |
| os.environ["VLLM_BASE_URL"] = "http://127.0.0.1:8000/" # vllm base url | |
| os.environ["MODEL"] = "Qwen/Qwen2.5-1.5B-Instruct" # model id | |
| os.environ["TOKENIZER_ID"] = "Qwen/Qwen2.5-1.5B-Instruct" # tokenizer id | |
| os.environ["MAGPIE_PRE_QUERY_TEMPLATE"] = "qwen2" | |
| os.environ["MAX_NUM_ROWS"] = "10000" | |
| os.environ["DEFAULT_BATCH_SIZE"] = "2" | |
| os.environ["MAX_NUM_TOKENS"] = "1024" | |
| launch() | |