The torch-checkpoints for the pruned models are listed here. These model checkpoints are in huggingface format and can be seamlessly utilized with the Hugging Face Transformers library, without any intermediary steps or complications.
| Model | Sparsity | Sparsity type | Perplexity | Checkpoint |
|---|---|---|---|---|
| LLaMA-2-7B | 0.5 | Unstructured | 6.86 | MBZUAI-LLM/GBLM-Pruner-LLaMA-2-7B |
| LLaMA-2-13B | 0.5 | Unstructured | 5.88 | MBZUAI-LLM/GBLM-Pruner-LLaMA-2-13B |
| LLaMA-2-70B | 0.5 | Unstructured | 4.17 | MBZUAI-LLM/GBLM-Pruner-LLaMA-2-70B |
| LLaMA-2-7B | 0.5 | 4:8 | 8.63 | Coming Soon |
| LLaMA-2-13B | 0.5 | 4:8 | 6.90 | Coming Soon |
| LLaMA-2-70B | 0.5 | 4:8 | 4.72 | Coming Soon |
| Model | Sparsity | Sparsity type | Perplexity | Checkpoint |
|---|---|---|---|---|
| LLaMA-2-7B-Chat | 0.5 | Unstructured | 8.40 | MBZUAI-LLM/GBLM-Pruner-LLaMA-2-7B-chat |
| LLaMA-2-13B-Chat | 0.5 | Unstructured | 7.10 | MBZUAI-LLM/GBLM-Pruner-LLaMA-2-13B-chat |
| Model | Sparsity | Sparsity type | Perplexity | Checkpoint |
|---|---|---|---|---|
| LLaMA-1-7B | 0.5 | Unstructured | 7.15 | MBZUAI-LLM/GBLM-Pruner-LLaMA-1-7B |
| LLaMA-1-13B | 0.5 | Unstructured | 6.11 | MBZUAI-LLM/GBLM-Pruner-LLaMA-1-13B |
| LLaMA-1-30B | 0.5 | Unstructured | 5.18 | MBZUAI-LLM/GBLM-Pruner-LLaMA-1-30B |
| LLaMA-1-7B | 0.5 | 4:8 | 8.48 | Coming Soon |
| LLaMA-1-13B | 0.5 | 4:8 | 7.26 | Coming Soon |
| LLaMA-1-30B | 0.5 | 4:8 | 5.89 | Coming Soon |