Repository navigation

LLM runner: validate and drive hf_static float-cache programs - #24572

Draft

psiddh wants to merge 7 commits into

pytorch:mainfrom

psiddh:stack2/pr07-hf-static-runtime

Draft

LLM runner: validate and drive hf_static float-cache programs#24572
psiddh wants to merge 7 commits into
pytorch:mainfrom
psiddh:stack2/pr07-hf-static-runtime

Commits on Nov 3, 2026
- Add backend-contributed edge transform passes to Partitioner
  psiddh
  committed
  1b88bb9
- Avoid reused names when lifting constant tensors
  psiddh
  committed
  33eee76
- LLM runner: select I/O contracts and accept custom buffer allocators
  psiddh
  committed
  3823faf
- QNN: build partitioners from options and allocate LLM I/O in rpcmem
  psiddh
  committed
  df791aa
- QNN: bind delegate outputs in exported program order
  psiddh
  committed
  109bacd
- Export hf_static float-cache programs with a registered HF runner
  psiddh
  committed
  82bb0a1
- LLM runner: validate and drive hf_static v1 cache buffers
  psiddh
  committed
  0ef7a34