|
--- |
|
language: |
|
- zh |
|
license: bigscience-bloom-rail-1.0 |
|
library_name: peft |
|
tags: |
|
- bloom |
|
thumbnail: https://s3.amazonaws.com/moonup/production/uploads/1677459920577-63b8e3432adad59f41dc65f4.jpeg?w=200&h=200&f=face |
|
pipeline_tag: text-generation |
|
widget: |
|
- {} |
|
base_model: bigscience/bloomz-7b1-mt |
|
--- |
|
|
|
# Bloom 7B1 LightNovel ZH_CN LoRa Finetuned |
|
|
|
|
|
BigScience Large Open-science Open-access Multilingual Language Model with 7,1 billion parameters finetuned on Chinese Translation of Japanese LightNovel using LoRa from PEFT (?) |
|
|
|
## Model Details |
|
|
|
I just downloaded 50 LightNovels then finetuned the model on raw text. |
|
|
|
Trained by Rorical |
|
|
|
## Use |
|
|
|
```python |
|
import torch |
|
from peft import PeftModel, PeftConfig |
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
peft_model_id = "Rorical/bloom-7b1-lightnovel-lora" |
|
config = PeftConfig.from_pretrained(peft_model_id) |
|
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=True, device_map='auto', cache_dir="cache") |
|
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path, cache_dir="cache") |
|
model = PeftModel.from_pretrained(model, peft_model_id, cache_dir="cache") |
|
|
|
prompt = "你是谁?\n" |
|
batch = tokenizer(prompt, return_tensors='pt').to("cuda") |
|
|
|
with torch.cuda.amp.autocast(): |
|
output_tokens = model.generate(**batch, max_new_tokens=150, do_sample=True, top_k=50, top_p=0.95) |
|
|
|
print(tokenizer.decode(output_tokens[0], skip_special_tokens=True)) |
|
``` |