About us
Statistics, AI, and ML meetups (lectures, working groups, demos, talks)
GPU time available on AMD for free with accepted proposal. Free $100 to anyone to get started.
https://www.amd.com/en/developer/ai-dev-program.html
We understand the frustrating nature of learning and try to be patient by providing a free service and access to GPU time and mentors. Everybody is a volunteer and the ease of manipulation via social media has led to the growth of unreasonable expectations no longer bounded by common manners. This group is designed to help individuals with sufficient programming skill the ability to grow by collaboration.
Upcoming events
58
LLM programming workshop.
Location not specified yetReview programming fundamentals. Pytorch, HuggingFace apis, The popularity of LLMs has caused extensive copy and paste in open source projects such as vllm. We are going to cover some basic programming idioms needed for distilling models from big to smaller models. These skills are transferrable and are different than the classic RL exercises.
pre distillation exercisesYou will need a colab account. We will review each problem and leave time for everyone to complete their solutions without the aid of a LLM.
This event is about repetition and practice.- HF apis to tokenizer and model
- how to speed up model downloads
- basic idioms, batching, padding, attention masks
- logits and probability distributions with softmax
- minimizing KL divergence
- top-k and softening teacher distribution by raising T. Dark Knowledge
- token level knowledge distillation. (distillation event we will stop here)
- Sequence level distillation
- DAgger Ross 2011
- Off policy Distillation
- On policy Distillation Agarwal 2024
- Self Distilled Reasoning Zhao 2016
- Self Distillation Fine Tuning Shenfeld 2026
Generated by Codex.
https://colab.research.google.com/drive/1EokTq_U-_JcPHtPTTjfVUcSjTW7dyDiB?usp=drive_link
Zoom link 10m before 630pm time on Thu.16 attendeesKubernetes and vllm Working Group
·OnlineOnlineWe have had consistent zoom bombing problems. Link up 10m before start time
Many jobs require an opensource contribution to get in the door. LLMs can do github tickets.
vllm ticket review: https://github.com/vllm-project/vllm/issues/50128.
This ticket starts with comparing warm start vs. cold start. Someone found a path where there was a bug causing longer restarts because of tracing.
Speedgraph###
3 attendees
Past events
1196
