Flash attention is an option for accelerating training and inference. Only NVIDIA GPUs of Turing, Ampere, Ada, and Hopper architecture, e.g., H100, A100, RTX 3090, T4, RTX 2080, can support flash attention. You can use our models without installing it.
4.31.0 is preferred.
Please check if you have updated the code to the latest, and correctly downloaded all the sharded checkpoint files.
This is the merge file of the tokenizer. You have to download it. Note that if you just git clone the repo without git-lfs, you cannot download this file.
Run the command pip install -r requirements.txt. You can find the file at https://github.com/QwenLM/Qwen-7B/blob/main/requirements.txt.
Yes, see web_demo.py for web demo and cli_demo.py for CLI demo. See README for more information.
Yes, run python cli_demo.py --cpu-only will load the model and inference on CPU only.
Yes. See the function chat_stream in modeling_qwen.py.
This is because tokens represent bytes and a single token may be a meaningless string. We have updated the default setting of our tokenizer to avoid such decoding results. Please update the code to the latest version.
Please check if you are loading Qwen-7B-Chat instead of Qwen-7B. Qwen-7B is the base model without alignment, which behaves differently from the SFT/Chat model.
Yes, the quantization is supported by bitsandbytes. We are working on an improved version and will release the quantized model checkpoints.
Errors in running quantized models: importlib.metadata.PackageNotFoundError: No package metadata was found for bitsandbytes
For Linux users，running pip install bitsandbytes directly can solve the problem. For Windows users, you can run python -m pip install bitsandbytes --prefer-binary --extra-index-url=https://jllllll.github.io/bitsandbytes-windows-webui·
We solved this problem. Updating the code to the latest version can help.
Please ensure that NTK is applied. use_dynamc_ntk and use_logn_attn in config.json should be set to true (true by default).
We do not provide finetuning or RLHF codes for now. However, some projects have supported finetuning, see FastChat, Firefly, LLaMA Efficient Tuning, etc. We will soon update the relevant codes.
In our training, we only use <|endoftext|> as the separator and padding token. You can set bos_id, eos_id, and pad_id to tokenizer.eod_id. Learn more about our tokenizer from our documents about the tokenizer.