|
|
--- |
|
|
library_name: peft |
|
|
license: gpl-3.0 |
|
|
--- |
|
|
## Training procedure |
|
|
|
|
|
|
|
|
The following `bitsandbytes` quantization config was used during training: |
|
|
- quant_method: bitsandbytes |
|
|
- load_in_8bit: True |
|
|
- load_in_4bit: False |
|
|
- llm_int8_threshold: 6.0 |
|
|
- llm_int8_skip_modules: None |
|
|
- llm_int8_enable_fp32_cpu_offload: False |
|
|
- llm_int8_has_fp16_weight: False |
|
|
- bnb_4bit_quant_type: fp4 |
|
|
- bnb_4bit_use_double_quant: False |
|
|
- bnb_4bit_compute_dtype: float32 |
|
|
### Framework versions |
|
|
|
|
|
|
|
|
- PEFT 0.6.0.dev0 |
|
|
|
|
|
## Citation |
|
|
```bash |
|
|
@misc{christ2024mathwell, |
|
|
title={MATHWELL: Generating Educational Math Word Problems at Scale}, |
|
|
author={Bryan R Christ and Jonathan Kropko and Thomas Hartvigsen}, |
|
|
year={2024}, |
|
|
eprint={2402.15861}, |
|
|
archivePrefix={arXiv}, |
|
|
primaryClass={cs.CL} |
|
|
} |
|
|
``` |
