|
--- |
|
library_name: peft |
|
license: mit |
|
language: |
|
- ar |
|
--- |
|
BLOOM-7B Arabic [LAPT + CLP+] |
|
=== |
|
|
|
## How to use |
|
```python |
|
from peft import AutoPeftModelForCausalLM |
|
from transformers import AutoTokenizer |
|
|
|
model = AutoPeftModelForCausalLM.from_pretrained( |
|
"atsuki-yamaguchi/bloom-7b1-clpp-ar" |
|
) |
|
|
|
# w/ GPU |
|
model = AutoPeftModelForCausalLM.from_pretrained( |
|
"atsuki-yamaguchi/bloom-7b1-clpp-ar", |
|
device_map="auto", |
|
load_in_8bit=True, |
|
) |
|
``` |
|
|
|
## Citation |
|
``` |
|
@article{yamaguchi2024empirical, |
|
title={An Empirical Study on Cross-lingual Vocabulary Adaptation for Efficient Generative {LLM} Inference}, |
|
author={Atsuki Yamaguchi and Aline Villavicencio and Nikolaos Aletras}, |
|
journal={ArXiv}, |
|
year={2024}, |
|
volume={abs/2402.10712}, |
|
url={https://arxiv.org/abs/2402.10712} |
|
} |
|
``` |
|
|
|
## Link |
|
For more details, please visit https://github.com/gucci-j/llm-cva |
|
|
|
|
|
## 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 0.5.0 |