alpha
Login
Join now
clee.sh
/
qxmx
Star
Starred
Star
Something went wrong. Try again.
Fork
1
Eye filled
Watch
Watching
Watch
A single-model (Bonsai-27B) inference engine for Intel Arc GPUs
Star
Starred
Star
Something went wrong. Try again.
Fork
1
Eye filled
Watch
Watching
Watch
Overview
Overview
Issues
Issues
Pulls
Pulls
Pipelines
Pipelines
qxmx
/
tools
7 files
at
main
Chris Lee
·
prefill: fix DEVICE_LOST on short prompts (< 16 tokens)
·
2mo ago
3ae4eff6
qxmx_batch.cpp
coarsen prefix-cache snapshot stride + real device banner in CLI tools
2mo ago
qxmx_bench.cpp
prefill: batch the tail in one M=16 call (fixes 1.58x serve slowdown)
2mo ago
qxmx_diff.cpp
prefill: fix DEVICE_LOST on short prompts (< 16 tokens)
2mo ago
qxmx_ref_run.cpp
Reorganize source tree: src/ + tools/ + tests/
2mo ago
qxmx_run.cpp
coarsen prefix-cache snapshot stride + real device banner in CLI tools
2mo ago
qxmx_serve.cpp
prefill_cached: fast path when no snapshot group fires + --driver-cpu pin
2mo ago
qxmx_tok.cpp
Reorganize source tree: src/ + tools/ + tests/
2mo ago
