EADST

Save Hugging Face Model with One Bin

max_shard_size (int or str, optional, defaults to "10GB") — Only applicable for models. The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size lower than this size. If expressed as a string, needs to be digits followed by a unit (like "5MB").

Based on the introduction, one bin model can be saved by changing the "max_shard_size".

LlamaForCausalLM.save_pretrained(base_model, output_dir, max_shard_size="100GB") # save one bin if the model is less than 100GB

Reference

PreTrainedModel

About Me
XD
Goals determine what you are going to be.
Category
标签云
API Vim Ptyhon TSV Augmentation MD5 Hilton Jetson OpenCV OCR Land VSCode 报税 Food 继承 递归学习法 阿里云 FlashAttention CUDA 图标 Markdown Anaconda Base64 NLP Quantize 图形思考法 SQLite Diagram Clash Quantization Zip 音频 Distillation WebCrawler tqdm Llama GoogLeNet hf 飞书 关于博主 CLAP ONNX Tensor Bert SVR Rebuttal torchinfo Permission InvalidArgumentError FastAPI Freesound Mixtral GIT Ubuntu Domain Conda tar Michelin Windows Plotly Animate Website TTS CSV Bipartite SPIE 顶会 Math 论文 EXCEL XGBoost mmap git FP32 Color FP64 Streamlit CC Algorithm Pickle 财报 XML ResNet-50 PyCharm Claude Git Data GGML DeepStream Qwen2.5 Heatmap printf scipy SAM Bitcoin ChatGPT Vmess PIP llama.cpp Paddle DeepSeek VGG-16 腾讯云 HaggingFace Knowledge 搞笑 Logo 多线程 RL UI Jupyter 多进程 v0.dev Tiktoken Pillow Python News LLAMA v2ray Qwen UNIX PDF 签证 Transformers IndexTTS2 FP16 VPN QWEN CV PyTorch Firewall SQL Numpy COCO 证件照 CAM Pandas Github WAN BeautifulSoup 域名 LaTeX Breakpoint JSON LoRA uwsgi CEIR Dataset Card LLM Review 公式 BF16 ModelScope Translation Agent Django Statistics Plate C++ Cloudreve Tracking OpenAI Proxy BTC Image2Text Magnet Baidu Use Password 算法题 版权 云服务器 Random YOLO Datetime Attention LeetCode Interview Search AI Excel Linux PDB Bin Miniforge Video Paper Shortcut NameSilo diffusers Template 强化学习 Input RGB Google Nginx GPTQ uWSGI TensorFlow Hotel CTC RAR Sklearn 净利润 logger Safetensors ms-swift Web Gemma Disk HuggingFace NLTK 第一性原理 Pytorch transformers Crawler Docker Qwen2 Hungarian git-lfs 论文速读 TensorRT icon FP8 GPT4
站点统计

本站现有博文332篇,共被浏览896795

本站已经建立2597天!

热门文章
文章归档
回到顶部