metadata
datasets:
- truongpdd/vietnamese_poetry
language:
- vi
base_model:
- NlpHUST/gpt2-vietnamese
pipeline_tag: text-generation
To use model, follow the below code:
from transformers import pipeline
from transformers import GPT2Tokenizer, GPT2LMHeadModel
import torch
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = GPT2LMHeadModel.from_pretrained('meomeo163/luc-bat-poet-model')
tokenizer = GPT2Tokenizer.from_pretrained('meomeo163/luc-bat-poet-model')
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=device)
prompt_text = "Ngẫm hay trăm sự tại trời"
print(f"\nPrompt: '{prompt_text}'")
try:
generated_output = pipe(
prompt_text,
max_length=100, # max length of poet
num_return_sequences=1, # max poet generated
do_sample=True,
temperature=0.7, # creative param
top_k=30, # get highest proba of next 30 token
top_p=0.85, # highest proba of token
no_repeat_ngram_size=3,
repetition_penalty=1.15,
pad_token_id=tokenizer.pad_token_id,
eos_token_id=tokenizer.eos_token_id,
)
print(f"\nBài thơ với từ khoá {prompt_text}:")
for i, seq in enumerate(generated_output):
poem_text = seq['generated_text']
print(f"Bài thơ {i+1}:")
print(poem_text)
print("-" * 30)
except Exception as e:
print(f"Lỗi khi sử dụng pipeline: {e}")
print("-------------------------------------")