66 lines
1.8 KiB
Markdown
66 lines
1.8 KiB
Markdown
|
---
|
||
|
|
license: bigscience-bloom-rail-1.0
|
||
|
|
language:
|
||
|
|
- vi
|
||
|
|
- en
|
||
|
|
library_name: transformers
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
tags:
|
||
|
|
- bloom
|
||
|
|
- causal-lm
|
||
|
|
- pytorch
|
||
|
|
model-index:
|
||
|
|
- name: vlsp-2023-vllm/hoa-1b4
|
||
|
|
results:
|
||
|
|
- task:
|
||
|
|
name: Word prediction
|
||
|
|
type: text-generation
|
||
|
|
dataset:
|
||
|
|
type: vlsp-2023-vllm/vi_lambada
|
||
|
|
name: vi_lambada
|
||
|
|
split: test
|
||
|
|
metrics:
|
||
|
|
- type: Perplexity
|
||
|
|
value: 8.606673731963474
|
||
|
|
- task:
|
||
|
|
name: Fewshot Translation
|
||
|
|
type: translation
|
||
|
|
dataset:
|
||
|
|
type: vlsp-2023-vllm/en-to-vi-formal-informal-tranlations
|
||
|
|
name: English to Vietnamese Formal/Informal translation
|
||
|
|
split: test
|
||
|
|
metrics:
|
||
|
|
- type: SacreBLEU
|
||
|
|
value: 25.5
|
||
|
|
datasets:
|
||
|
|
- vlsp-2023-vllm/vi_lambada
|
||
|
|
metrics:
|
||
|
|
- perplexity
|
||
|
|
---
|
||
|
|
|
||
|
|
# Hoa 1B4 (Bloom architecture)
|
||
|
|
|
||
|
|
Hoa is an autoregressive Large Language Model (LLM), based on Bloom's model architecture.
|
||
|
|
Hoa was trained on part of the Common Crawl dataset in Vietnamese and English.
|
||
|
|
|
||
|
|
Details will be available soon.
|
||
|
|
|
||
|
|
To contact us, mail to: leanhcuong@gmail.com (Lê Anh Cường) | hieunguyen1053@outlook.com (Hiếu) | nv.cuong@int2.vn (Nguyễn Việt Cường)
|
||
|
|
|
||
|
|
### How to use
|
||
|
|
```python
|
||
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||
|
|
|
||
|
|
tokenizer = AutoTokenizer.from_pretrained("vlsp-2023-vllm/hoa-1b4")
|
||
|
|
model = AutoModelForCausalLM.from_pretrained("vlsp-2023-vllm/hoa-1b4", low_cpu_mem_usage=True)
|
||
|
|
|
||
|
|
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
||
|
|
model.to(device)
|
||
|
|
|
||
|
|
prompt = "Địa chỉ trường Đại học Tôn Đức Thắng nằm ở số"
|
||
|
|
input_ids = tokenizer(prompt, return_tensors="pt")['input_ids'].to(device)
|
||
|
|
|
||
|
|
gen_tokens = model.generate(input_ids, max_length=max_length, repetition_penalty=1.1)
|
||
|
|
|
||
|
|
print(tokenizer.batch_decode(gen_tokens)[0])
|
||
|
|
```
|