哪里玩原版 / LLaMA 2 Demo

• HuggingChat/Llama-2-70b-chat-hf
• Hugging Face Space 7B | 13B | 70B
• Perplexity 7B/13B
• Replicate 7B | 13B | 70B
• sdk.vercel.ai
• nat.dev

怎么玩 / How to Prompt

<s>[INST] <<SYS>>
{{ system_prompt }}
<</SYS>>

{{ user_message }} [/INST]

<s>[INST] <<SYS>>
{{ system_prompt }}
<</SYS>>

{{ user_msg_1 }} [/INST] {{ model_answer_1 }} </s>\
<s>[INST] {{ user_msg_2 }} [/INST] {{ model_answer_2 }} </s>\
<s>[INST] {{ user_msg_3 }} [/INST]

跑分 / Benchmark

• Hugging Face Open LLM Leaderboard
• https://ai.meta.com/llama/

细节 / Detail

• Llama 2: Open Foundation and Fine-Tuned Chat Models
• Llama 2: an incredible open LLM
• Llama 2: Full Breakdown
• https://twitter.com/DrJimFan/status/1681372700881854465

微调 / Fine-tuning

已经有哪些了？ / Where

• https://huggingface.co/spaces/LinkSoul/Chinese-Llama-2-7b 中文
• https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b Hermes
• https://huggingface.co/stabilityai/FreeWilly2 Orca-like

怎么调？/ How

• Fine-tune LLaMA 2 (7-70B) on Amazon SageMaker
• HuggingFace, Fine-tuning LLaMA 2 with PEFT
• Meta Examples and recipes for Llama model
• The EASIEST way to finetune LLAMA-v2 on local machine!
• 知乎，LLaMA2 RLHF 技术细节
• Ollama

其它 / Misc.

美西时间这周六早上九点，美东时间早上十二点，聊一下怎么把 LLaMA 2 搞到超过 GPT 3.5 的效果，有兴趣一起来讨论
Topic: LLaMA-2 搞到 match / 超过 GPT 3.5
Time: Jul 22, 2023 09:00 AM Pacific Time (US and Canada)
https://ed-ac-uk.zoom.us/j/87003419834
Meeting ID: 870 0341 9834
Passcode: abcd1234

批评

• https://twitter.com/MarcosGorgojo/status/1681365374606901255
• https://twitter.com/dotey/status/1681925505644101632

参考资料 / Reference

• LLaMA 2 – Every Resource you need
• HuggingFace, Llama 2 is here – get it on Hugging Face
• Replicate, What happened with Llama 2 in the last 24 hours?
• Replicate, Fine-tune Llama 2 on Replicate

Leave a Comment