# LLM Models

##### Resources

- [7 Popular LLMs Explained in 7 Minutes - KDnuggets](https://www.kdnuggets.com/7-popular-llms-explained-in-7-minutes)
- [Awesome Free LLM APIs](https://github.com/mnfst/awesome-free-llm-apis)

##### Chinese LLMs

- [Taiwan LLM](https://twllm.com/) - Project TAME (TAiwanese Mixture of Experts) 
    - GitHub: [https://github.com/MiuLab/Taiwan-LLM](https://github.com/MiuLab/Taiwan-LLM)
    - HF: [https://huggingface.co/yentinglin](https://huggingface.co/yentinglin)
    - HF: [https://huggingface.co/audreyt](https://huggingface.co/audreyt)
    - [臺灣繁中LLM另一里程碑！Project TAME以5,000億個Token訓練而成並開源釋出 | iThome](https://www.ithome.com.tw/news/163730)
    - [Project TAME上線！「最台AI」獲輝達算力支援，能懂「很盤」涵義，3大優勢有哪些？|數位時代 BusinessNext](https://www.bnext.com.tw/article/79599/taiwan-project-tame-ai-model-llm-industry)
- [TAIDE](https://taide.tw/index) (台德)- Trustworthy AI Dialogue Engine 
    - GitHub: [https://github.com/taide-taiwan](https://github.com/taide-taiwan)
    - HF: [https://huggingface.co/taide](https://huggingface.co/taide)
    - [TAIDE | iThome](https://www.ithome.com.tw/tags/taide)
- 01.AI - [Yi](https://01.ai/)
    - GitHub: [https://github.com/01-ai/Yi](https://github.com/01-ai/Yi)
    - HF: [https://huggingface.co/01-ai/](https://huggingface.co/01-ai/)
- CKIP-Llama-2-7b 是中央研究院詞庫小組(CKIP)開發的開源可商用繁體中文大型語言模型，以商用開源模型Llama-2-7b以及Atom-7b為基礎，再補強繁體中文的處理能力，並對405個可商用的任務檔案同步進行訓練優化，參數量達70億(7 billion)。 
    - GitHub: [https://github.com/f901107/CKIP-Llama-2-7b](https://github.com/f901107/CKIP-Llama-2-7b)
    - HF: [https://huggingface.co/spaces/ckiplab/CKIP-Llama-2-7b-chat](https://huggingface.co/spaces/ckiplab/CKIP-Llama-2-7b-chat)
- [Qwen](https://qwenlm.github.io/) - 阿里雲通義千問 
    - GitHub: [https://github.com/QwenLM/Qwen](https://github.com/QwenLM/Qwen)
    - GitHub: [https://github.com/QwenLM/Qwen2](https://github.com/QwenLM/Qwen2)
    - HF: [https://huggingface.co/Qwen](https://huggingface.co/Qwen)
    - Doc: [https://help.aliyun.com/zh/dashscope/create-a-chat-foundation-model?spm=a2c4g.11186623.0.0.20ea4937azFCan](https://help.aliyun.com/zh/dashscope/create-a-chat-foundation-model?spm=a2c4g.11186623.0.0.20ea4937azFCan)
- GLM-4 - 智譜 AI 推出的中文多語言模型 
    - GitHub: [https://github.com/THUDM/GLM-4](https://github.com/THUDM/GLM-4)
    - HF: [https://huggingface.co/collections/THUDM/glm-4-665fcf188c414b03c2f7e3b7](https://huggingface.co/collections/THUDM/glm-4-665fcf188c414b03c2f7e3b7)
- [Chinese-Mixtral](https://github.com/ymcui/Chinese-mixtral)
- [DeepSeek](https://www.deepseek.com/) - 深度求索 
    - GitHub: [https://github.com/deepseek-ai/DeepSeek-V2](https://github.com/deepseek-ai/DeepSeek-V2)
    - HF: [https://huggingface.co/deepseek-ai/DeepSeek-V2](https://huggingface.co/deepseek-ai/DeepSeek-V2)

##### Code LLMs

- [Granite](https://www.ibm.com/granite) - Open sourcing IBM’s Granite code models 
    - H F: [https://huggingface.co/ibm-granite](https://huggingface.co/ibm-granite)
    - GitHub: [https://github.com/ibm-granite](https://github.com/ibm-granite)
    - [IBM開源程式開發專用Granite語言模型，效能超越當前多數開源模型 | iThome](https://www.ithome.com.tw/news/163056)
    - [IBM Granite 3.0 models · Ollama Blog](https://ollama.com/blog/ibm-granite)
    - [IBM Granite.Code - Visual Studio Marketplace](https://marketplace.visualstudio.com/items?itemName=IBM.wca-core&utm_source=ibm_developer&utm_content=in_content_link&utm_id=blogs_awb-introducing-ibm-granite-code)
- [Codestral](https://mistral.ai/news/codestral/) - Mistral's first generative AI model for code 
    - HF: [https://huggingface.co/mistralai/Codestral-22B-v0.1](https://huggingface.co/mistralai/Codestral-22B-v0.1)
    - [Mistral AI推出輕量程式撰寫輔助模型 | iThome](https://www.ithome.com.tw/news/163190)
- [Gemini Code Assist](https://codeassist.google/) - Google 推出的程式編寫助理，可以在 VS Code、JetBrains IDE 上使用。

##### Evaluation/Monitor

- [PromptBench](https://github.com/microsoft/promptbench): A Unified Library for Evaluating and Understanding Large Language Models.
- AI產品與系統評測中心: [AI評測模擬測試題庫.xlsx](https://www.aiec.org.tw/Home/DownloadZone)
- [Opik](https://github.com/comet-ml/opik) is an open-source platform for evaluating, testing and monitoring LLM applications.

##### Function Calling LLMs

- [Firefunction-v2](https://fireworks.ai/blog/firefunction-v2-launch-post)
    - HF: [https://huggingface.co/fireworks-ai/firefunction-v2](https://huggingface.co/fireworks-ai/firefunction-v2)

##### Content Safty

- [Google ShieldGemma](https://ai.google.dev/gemma/docs/shieldgemma?hl=zh-tw)  
    ShieldGemma則是個安全分類模型，可額外部署在模型的輸入及輸出端，用以過濾有害內容，它主要篩選4大領域的內容，包括仇恨言論、騷擾、裸露的色情內容，以及危險內容。

##### Hardware Requirements

- [如何計算 Model 需要多少 GPU VRAM](https://substratus.ai/blog/calculating-gpu-memory-for-llm)
- [Calculates how much GPU memory you need and how much token/s you can get for any LLM & GPU/CPU](https://github.com/RahulSChand/gpu_poor)
- [LLM RAM Calculator](https://llm-calc.rayfernando.ai/)
- [llmfit](https://github.com/AlexsJones/llmfit) - A terminal tool that right-sizes LLM models to your system's RAM, CPU, and GPU.
- [CanIRun.ai](https://www.canirun.ai/)