Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
COMMITS
June 21, 2025
S
Llama 3 KV Cache (#685)
Sebastian Raschka committed
June 20, 2025
S
Fix formatting in Qwen3 nb (#680)
Sebastian Raschka committed
June 19, 2025
C
minor readability improvements (#668)
casinca committed
D
fixed plot_losses (#677)
Daniel Kleine committed
S
Qwen3 From Scratch (#678)
Sebastian Raschka committed
June 17, 2025
C
removed old args in GQA class (#674)
casinca committed
June 16, 2025
S
Optimize KV cache (#673)
Sebastian Raschka committed
June 15, 2025
S
Optimized KV cache (#672)
Sebastian Raschka committed
S
Add KV cache (#671)
Sebastian Raschka committed
June 13, 2025
S
Remove redundant model = (#663)
Sebastian Raschka committed
S
Update pixi (#661)
Sebastian Raschka committed
D
fixed gqa qkv code comments (#660)
Daniel Kleine committed
G
Update ch07.ipynb (#643)
Greg Gandenberger committed
S
fixed video link (#646)
Shimpei Kojio committed
June 12, 2025
P
Correcting the wrong reference (#649)
Pratyush Subhadarshi committed
S
Reduce Llama 3 RoPE memory requirements (#658)
Sebastian Raschka committed
April 20, 2025
S
DeBERTa-v3 baseline (#630)
Sebastian Raschka committed
April 18, 2025
S
BPE cosmetics (#629)
Sebastian Raschka committed
S
Dpo vocab size clarification (#628)
Sebastian Raschka committed
April 16, 2025
S
Llama3 from scratch improvements (#621)
Sebastian Raschka committed
C
Minor DPO fixes (#617)
casinca committed
D
fixed `<|endoftext|>` token (#620)
Daniel Kleine committed
April 13, 2025
H
updated exercise 5.3 (#615)
Henry Shi committed
S
add special token handling to bpe from scratch code (#616)
Sebastian Raschka committed
April 12, 2025
S
Ch06 and Ch07 videos (#613)
Sebastian Raschka committed
P
fix: typo in ch07.ipynb (#612)
PRASHANTH REDDY NIMMAKAYALA committed
April 6, 2025
S
Improve ModernBERT comments (#606)
Sebastian Raschka committed
S
align formulas in notes with code (#605)
Sebastian Raschka committed
S
Disable mask saving as weight in Llama 3 model (#604)
Sebastian Raschka committed
April 5, 2025
S
reformat nbs (#602)
Sebastian Raschka committed