Hacker Newsnew | past | comments | ask | show | jobs | submit | fromlogin
Fun with low-rank vocab matrices (and a bonus test loss reduction?) (gilesthomas.com)
3 points by gpjt 3 days ago | past | discuss
Why do OpenAI's GPT-2 weights beat mine? Part five: data quality (gilesthomas.com)
2 points by gpjt 11 days ago | past | discuss
Extending Raschka's GPT-2: an MoE trained from scratch on an RTX 3090 (gilesthomas.com)
2 points by ibobev 32 days ago | past
Extending Raschka's GPT-2: an MoE trained from scratch on an RTX 3090 (gilesthomas.com)
1 point by gpjt 32 days ago | past
Putting my Jax-trained models on the Hugging Face Hub (gilesthomas.com)
2 points by gpjt 39 days ago | past
Why do OpenAI's GPT-2 weights beat mine? Part four: digging into dropout (gilesthomas.com)
1 point by gpjt 46 days ago | past
Adding diagrams to my static site generator with D2 (gilesthomas.com)
2 points by ibobev 48 days ago | past
Adding diagrams to my static site generator with D2 (gilesthomas.com)
3 points by gpjt 48 days ago | past
Use the built-in GELU, don't roll your own (gilesthomas.com)
2 points by ibobev 54 days ago | past
Use the built-in GELU, don't roll your own (gilesthomas.com)
1 point by gpjt 54 days ago | past
A Quick(ish) Chinchilla Check (gilesthomas.com)
1 point by gpjt 66 days ago | past
Why do OpenAI's GPT-2 weights beat mine? (gilesthomas.com)
3 points by gmays 68 days ago | past
I use AI on this blog (gilesthomas.com)
1 point by gpjt 73 days ago | past | 1 comment
Why do OpenAI's GPT-2 weights beat mine? Part three: testing overtraining (gilesthomas.com)
2 points by gpjt 74 days ago | past
Why do OpenAI's GPT-2 weights beat mine? Part two: the bugfix (gilesthomas.com)
8 points by gpjt 74 days ago | past
Why do OpenAI's GPT-2 weights beat mine? (gilesthomas.com)
4 points by gpjt 75 days ago | past
Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090 (gilesthomas.com)
1 point by ibobev 77 days ago | past
Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090 (gilesthomas.com)
16 points by gpjt 80 days ago | past
Poppy the training box, part 1: The beginnings (gilesthomas.com)
3 points by ibobev 3 months ago | past
Building intuition about LLM parameter counts (gilesthomas.com)
3 points by ibobev 3 months ago | past
Building intuition about LLM parameter counts (gilesthomas.com)
2 points by gpjt 3 months ago | past
Poppy the training box, part 1: the beginnings (gilesthomas.com)
3 points by gpjt 3 months ago | past
From bigrams to GPT-2, one component at a time (in Jax) (gilesthomas.com)
1 point by gpjt 3 months ago | past
Building a Jax training loop for an LLM training run (gilesthomas.com)
2 points by gpjt 3 months ago | past
Thoughts on Role Confusion (gilesthomas.com)
2 points by ibobev 3 months ago | past
Thoughts on Role Confusion (gilesthomas.com)
3 points by gpjt 3 months ago | past
Flax debugging: making a hash of things (gilesthomas.com)
3 points by ibobev 3 months ago | past
Flax debugging: making a hash of things (gilesthomas.com)
2 points by gpjt 3 months ago | past
10Gb/s Ethernet: switching to a Broadcom SFP+ module (gilesthomas.com)
195 points by gpjt 3 months ago | past | 170 comments
Jax: Commitment Issues (gilesthomas.com)
3 points by ibobev 3 months ago | past

Consider applying for YC's Winter 2027 batch! Applications are open till November 2.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: