44 lines
2.4 KiB
Markdown
44 lines
2.4 KiB
Markdown
---
|
|
datasets:
|
|
- HuggingFaceH4/ultrachat_200k
|
|
language:
|
|
- en
|
|
pipeline_tag: text-generation
|
|
---
|
|
|
|
# SparseLlama-2-7b-ultrachat_200k-pruned_50.2of4
|
|
|
|
## Model Overview
|
|
- **Model Architecture:** Llama-2
|
|
- **Input:** Text
|
|
- **Output:** Text
|
|
- **Model Optimizations:**
|
|
- **Pruned:** 50% 2:4
|
|
- **Release Date:** 6/28/2024
|
|
- **Version:** 1.0
|
|
- **Model Developers:** Neural Magic
|
|
|
|
Compressed version of [Llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b-hf) specialized for text-generation.
|
|
This model was obtained by fine-tuning the Sparse Foundational model [Sparse-Llama-2-7b-pruned_50.2of4](https://huggingface.co/nm-testing/SparseLlama-2-7b-pruned_50.2of4) on the [ultrachat_200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) dataset.
|
|
It achieves a win rate of 62.1% on the [AlpacaEval](https://github.com/tatsu-lab/alpaca_eval) benchmark (version 1.0) when using [Llama-2-70b-chat](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) as evaluator, whereas the dense [Llama-2-7b-ultrachat200k](https://huggingface.co/neuralmagic/Llama-2-7b-ultrachat200k) model achieves 57.6% win rate.
|
|
|
|
This model was produced as part if Neural Magic's Sparse Foundational Models initiative, and demostrates the capability of Sparse Foundational Models to transfer to the text-generation domain.
|
|
|
|
**Note:** This model uses the chat template from [zephyr-7b-beta](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta).
|
|
|
|
## Model Optimizations
|
|
|
|
This model is derived from the Sparse Foundational model [Sparse-Llama-2-7b-pruned_50.2of4](https://huggingface.co/nm-testing/SparseLlama-2-7b-pruned_50.2of4), which was obtained by applying the [SparseGPT](https://arxiv.org/abs/2301.00774) algorithm to prune [Llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b-hf) to 50% sparsity with a 2:4 mask.
|
|
This optimization reduces the number of parameters by 50%, reducing the disk size and FLOPs by the same level.
|
|
|
|
## Evaluation
|
|
|
|
This model was evaluated in the [AlpacaEval](https://github.com/tatsu-lab/alpaca_eval) benchmark using [Llama-2-70b-chat](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) as evaluator.
|
|
|
|
## Accuracy
|
|
|
|
| Model | Win rate | Recovery |
|
|
| :----- | :--------: | :--------: |
|
|
| [Llama-2-7b](https://huggingface.co/meta-llama/Llama-2-7b-hf) | 3.7% | -- |
|
|
| [Llama-2-7b-ultrachat200k](https://huggingface.co/neuralmagic/Llama-2-7b-ultrachat200k) | 57.6% | -- |
|
|
| SparseLlama-2-7b-ultrachat_200k-pruned_50.2of4 | 62.1% | 108% | |