Downloads · 30 days
17
24% of all-time downloads
levos06/Bolt-17M
Bolt-17M is a text generation model from levos06. Use it when you need the model to write or continue text. The card lists the license as mit.
Transformer модель с архитектурой DeepSeek (RMSNorm, RoPE, SwiGLU) для генерации текста.
Downloads · 30 days
17
24% of all-time downloads
All-time downloads
70
Public
Parameters
30.2M
121 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors121 MB · 100%
From the Hugging Face model README
Transformer модель с архитектурой DeepSeek (RMSNorm, RoPE, SwiGLU) для генерации текста.
from load_model import load_model, generate
model, tokenizer, device = load_model()
text = generate(model, tokenizer, "The meaning of life is", max_new_tokens=100)
print(text)
import torch
from transformers import GPT2Tokenizer
from safetensors.torch import load_file
# Загрузите модель
state_dict = load_file("model.safetensors")
tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
# Инициализируйте архитектуру (см. training/scripts/training/train_sft.py)
# и загрузите state_dict
Модель обучена методом Supervised Fine-Tuning на инструкциях.
Подробности обучения см. в training/README.md
temperature (0.1-2.0): Контроль случайностиtop_p (0.0-1.0): Nucleus samplingmax_new_tokens: Максимальное количество токеновMIT License