Add GPT-OSS from OpenAI - closed in favor of 689#683

Closed

ikawrakow wants to merge 21 commits intomainfrom

Commits on Aug 9, 2025

gmp-oss: common
Iwan Kawrakow
committed
gpt-oss: attnetion sinks, swiglu_oai
Iwan Kawrakow
committed

Commits on Aug 10, 2025

gpt-oss: WIP llama
Iwan Kawrakow
committed
gpt-oss: CPU seems to be working
Iwan Kawrakow
committed
CUDA: ADD_ID
Iwan Kawrakow
committed
gpt-oss: Seems to be working on CUDA
Iwan Kawrakow
committed
gpt-oss: add sinks to the attn-vec kernels
Iwan Kawrakow
committed

Commits on Aug 11, 2025

CUDA: add head size of 64 to new mma
Iwan Kawrakow
committed

Commits on Aug 12, 2025

gpt-oss: add ability to use -fmoe (only CUDA for now)
Iwan Kawrakow
committed
Move row sums to the write place
Iwan Kawrakow
committed
Add sinks to iqk flash attention
Iwan Kawrakow
committed
gpt_oss: Implement -fmoe on the CPU
Iwan Kawrakow
committed
Simdify swiglu_oai
Iwan Kawrakow
committed

Commits on Aug 13, 2025

llama: factor out model loader
Iwan Kawrakow
committed
Builds successfully
Iwan Kawrakow
committed
It runs, but mmap does not work
Iwan Kawrakow
committed
Fix llama_mmap so mmap works
Iwan Kawrakow
committed
Minor
Iwan Kawrakow
committed
Fix CUDA after latest changes
Iwan Kawrakow
committed

Commits on Aug 14, 2025

Fix llama_mmap on non-Linux platforms
Iwan Kawrakow
committed
Iterating on Windows build failures
Iwan Kawrakow
committed