File size: 1,481 Bytes
950ff1f
 
 
d38f83b
950ff1f
 
f2f02d8
950ff1f
 
 
 
 
 
 
 
 
 
 
f2f02d8
 
950ff1f
 
 
 
f2f02d8
 
 
6579fe3
f2f02d8
 
 
 
 
 
 
6579fe3
f2f02d8
 
 
 
 
 
d38f83b
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
---
language: en
license: apache-2.0
library_name: transformers
---

# SQFT Fine-tuned Model: sqft-sparsepeft-mistral-7b-v0.3-50-gsm8k-heu

- Base Model: [IntelLabs/sqft-mistral-7b-v0.3-50-base](https://huggingface.co/IntelLabs/sqft-mistral-7b-v0.3-50-base)
- Sparsity: 50%
- Quantization: No
- Finetune Method: SQFT + SparsePEFT
- Finetune data: [GSM8K](https://huggingface.co/datasets/openai/gsm8k)
- Sub-Adapter: Heuristic

### Evaluation

```bash
MODEL_NAME=IntelLabs/sqft-sparsepeft-mistral-7b-v0.3-50-gsm8k-heu
lm_eval --model hf --model_args pretrained=${MODEL_NAME},add_bos_token=True,trust_remote_code=True --tasks gsm8k --batch_size auto:4
```

Refer to our [repo](https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning/tree/main/SQFT) for the environment information to run this command.

## Model Sources

- **Repository:** [https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning/tree/main/SQFT](https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning/tree/main/SQFT)
- **Paper:** [SQFT: Low-cost Model Adaptation in Low-precision Sparse Foundation Models](https://arxiv.org/abs/2410.03750)

## Citation

```bash
@article{munoz2024sqft,
  title = {SQFT: Low-cost Model Adaptation in Low-precision Sparse Foundation Models},
  author={J. Pablo Munoz and Jinjie Yuan and Nilesh Jain},
  journal={The 2024 Conference on Empirical Methods in Natural Language Processing (Findings)},
  year={2024}
}
```

## License

Apache-2.0