🧠 两小时从 0 训出你的 64M 大模型。Train your own 64M LLM from scratch in 2 hours.
[update] readme
jingyaogong20a43d7db013
diff --git a/README.md b/README.mdindex d9415a6..8ecc4c5 100644--- a/README.md+++ b/README.md@@ -1665,13 +1665,13 @@ MiniMind模型本身预训练数据集小的可怜,也没有针对性的对测# 📌 Others-## 模型转换+## 🔧 模型转换* [./scripts/convert_model.py](./scripts/convert_model.py)可以实现`torch / transformers`模型的互相转换* 如无特别说明,`MiniMind2`模型均默认为`Transformers`格式的模型,需提前`t2t`转换!-## 基于MiniMind-API服务接口+## 🖥️ 基于MiniMind-API服务接口* [./scripts/serve_openai_api.py](./scripts/serve_openai_api.py)完成了兼容openai-api的最简聊天接口,方便将自己的模型接入第三方UI例如FastGPT、OpenWebUI、Dify等等。@@ -1712,6 +1712,14 @@ MiniMind模型本身预训练数据集小的可怜,也没有针对性的对测}'```+## 👨💻 更多++* <a href="https://github.com/jingyaogong/minimind/discussions/618">🔗从MiniMind-LLM微调扩散语言模型</a>++* <a href="https://github.com/jingyaogong/minimind/discussions/611">🔗模型的generate方法说明</a>++---+## <img src="https://avatars.githubusercontent.com/u/136984999" height="28" style="vertical-align: middle;"/> [vllm](https://github.com/vllm-project/vllm)vLLM是极其流行的高效推理框架,支持大模型快速部署,优化显存利用与吞吐量。diff --git a/README_en.md b/README_en.mdindex 296cfda..b495a53 100644--- a/README_en.md+++ b/README_en.md@@ -1599,14 +1599,14 @@ Models generally achieve baseline performance due to small parameter scales and# 📌 Others-## Model Conversion+## 🔧 Model Conversion* [./scripts/convert_model.py](./scripts/convert_model.py) enables mutual conversion of `torch / transformers` models* Unless otherwise specified, `MiniMind2` models are by default in `Transformers` format and require `t2t` conversion beforehand!-## OpenAI-API Based MiniMind Service Interface+## 🖥️ OpenAI-API Based MiniMind Service Interface* [./scripts/serve_openai_api.py](./scripts/serve_openai_api.py) provides extremely simple OpenAI-API compatible chat interface, convenient for integration with third-party UIs like FastGPT, Open-WebUI, Dify, etc.@@ -1646,6 +1646,13 @@ Models generally achieve baseline performance due to small parameter scales and}'```+## 👨💻 More++* <a href="https://github.com/jingyaogong/minimind/discussions/618">🔗Fine-tuning Diffusion Language Models from MiniMind-LLM</a>+* <a href="https://github.com/jingyaogong/minimind/discussions/611">🔗Model generate method explanation</a>++---+## <img src="https://avatars.githubusercontent.com/u/136984999" height="28" style="vertical-align: middle;"/> [vllm](https://github.com/vllm-project/vllm)vLLM is an extremely popular efficient inference framework supporting fast deployment of large models, optimizing GPU memory usage and throughput.