Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# SQFT: Low-cost Model Adaptation in Low-precision Sparse Foundation Models
|
| 7 |
+
|
| 8 |
+
- Base Model: [IntelLabs/sqft-mistral-7b-v0.3-50-base-gptq](https://huggingface.co/IntelLabs/sqft-mistral-7b-v0.3-50-base-gptq)
|
| 9 |
+
- Sparsity: 50%
|
| 10 |
+
- Quantization: INT4 (GPTQ)
|
| 11 |
+
- Finetune Method: SQFT
|
| 12 |
+
- Finetune data: [GSM8K](https://huggingface.co/datasets/openai/gsm8k)
|
| 13 |
+
- Sub-Adapter: Heuristic
|
| 14 |
+
|
| 15 |
+
### Evaluation
|
| 16 |
+
|
| 17 |
+
```bash
|
| 18 |
+
BASE_MODEL_PATH=IntelLabs/sqft-mistral-7b-v0.3-50-base-gptq
|
| 19 |
+
ADAPTER_MODEL_PATH=IntelLabs/sqft-mistral-7b-v0.3-50-gptq-math-heu-adapter
|
| 20 |
+
lm_eval --model hf --model_args pretrained=${BASE_MODEL_PATH},peft=${ADAPTER_MODEL_PATH},add_bos_token=True,trust_remote_code=True --tasks gsm8k --batch_size auto:4
|
| 21 |
+
```
|
| 22 |
+
|
| 23 |
+
Refer to our [repo](https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning/tree/main/SQFT) for the environment information to run this command.
|
| 24 |
+
|