metadata
license: apache-2.0
language:
- en
Model Card for TurboSparse-Mixtral
The TurboSparse-Mixtral Large Language Model (LLM) is an sparsified version of the Mixtral.
![avatar](/PowerInfer/TurboSparse-Mixtral-Instruct/resolve/main/takeaway.png)
The average performance is evaluated using benchmarks from the OpenLLM Leaderboard.
Inference
Our code for accelerating TurboSparse-Mixtral is currently being refined. Stay tuned! Now you can run this model like dense model.
Chat-Template
During sparsification, we also utilize some SFT datasets. We take ChatML as our chat template:
<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n
Allow Finetuning
As we merged the predictors for FFN neurons in models, you can finetune TurboSparse-Mixtral with any framework and algorithm.
Limitations
- TurboSparse, having just undergone training with 150B tokens, may still exhibit performance gaps in certain tasks.
- The TurboSparse model has only been trained on English-language datasets, hence its capabilities in other languages are still lacking.
- The model may produce unexpected outputs due to its small size, limited training tokens and probabilistic generation paradigm.
License
The model is licensed under Apache-2.0, while model weights are fully open for academic research and also allow free commercial usage.