Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
COMMITS
November 6, 2025
R
Use consistent title case
rasbt committed
S
Fix empty device issue (#904)
Sebastian Raschka committed
S
n_heads × d_head -> d_head × d_head in DeltaNet (#903)
Sebastian Raschka committed
November 3, 2025
S
Image resizing
Sebastian Raschka committed
S
Gated DeltaNet write-up (#901)
Sebastian Raschka committed
November 1, 2025
S
Training on MPS in PyTorch 2.9 (#900)
Sebastian Raschka committed
A
Fix MHAEinsum weight dimension bug when d_in != d_out (#857) (#893)
Aviral Garg committed
S
simplify uv command (#898)
Sebastian Raschka committed
October 29, 2025
S
Add bonus dependencies to pyproject (#897)
Sebastian Raschka committed
October 22, 2025
S
Fix ffn link (#892)
Sebastian Raschka committed
S
Make quote style consistent (#891)
Sebastian Raschka committed
October 21, 2025
C
- docs(moe): correct arXiv link for DeepSeekMoE (#890)
casinca committed
October 20, 2025
S
Mixture-of-Experts intro (#888)
Sebastian Raschka committed
October 17, 2025
S
Make it easier to toggle between thinking and instruct variants (#887)
Sebastian Raschka committed
October 14, 2025
S
Update the compression rate comment in MLA (#883)
Sebastian Raschka committed
October 13, 2025
S
Use figure numbers in ch05-7 (#881)
Sebastian Raschka committed
S
Add alternative attention structure (#880)
Sebastian Raschka committed
S
sliding window attention (#879)
Sebastian Raschka committed
S
Add other appendices for completeness (#878)
Sebastian Raschka committed
October 12, 2025
S
Multi-Head Latent Attention (#876)
Sebastian Raschka committed
October 11, 2025
S
Use GB instead of GiB consistently (#875)
Sebastian Raschka committed
S
Grouped-Query Attention memory (#874)
Sebastian Raschka committed
October 9, 2025
R
Use inference_device
rasbt committed
S
Add simpler BPE, and make previous BPE better (#870)
Sebastian Raschka committed
October 8, 2025
S
Qwen3 and evaluation bonus materials (#869)
Sebastian Raschka committed
October 7, 2025
S
Switch from urllib to requests to improve reliability (#867)
Sebastian Raschka committed
October 6, 2025
S
Add missing comma in imports in README (#865)
Sebastian Raschka committed
October 1, 2025
S
Note about output dimensions (#862)
Sebastian Raschka committed
S
Update ollama address (#861)
Sebastian Raschka committed