Repository navigation

feat(benchmark): qualify DINOv3 image features - #1349

Closed

chaofengw-nv wants to merge 33 commits into

NVIDIA:mainfrom

chaofengw-nv:feat/image-features-qualification

Closed

feat(benchmark): qualify DINOv3 image features#1349
chaofengw-nv wants to merge 33 commits into
NVIDIA:mainfrom
chaofengw-nv:feat/image-features-qualification

Commits on Sep 18, 2026
- feat(benchmark): restore text model qualification
  chaofengw-nv
  committed
  aadfd17
- fix(benchmark): restore OPT performance workload
  chaofengw-nv
  committed
  fce8ae6
- feat(benchmark): qualify more text models
  chaofengw-nv
  committed
  36c9b47
- fix(builder): ignore unused alternate weights
  chaofengw-nv
  committed
  e61a12b
- feat(benchmark): restore more text profiles
  chaofengw-nv
  committed
  5e0ba6d
- feat(benchmark): restore additional text profiles
  chaofengw-nv
  committed
  8c32d3c
- fix(benchmark): align MoE accuracy references
  chaofengw-nv
  committed
  a7989d4
- fix(benchmark): reserve Llama BOS cache row
  chaofengw-nv
  committed
  01c1fb3
- feat(benchmark): qualify more generation models
  chaofengw-nv
  committed
  5c54840
- fix(benchmark): propagate trusted HF loading
  chaofengw-nv
  committed
  07516d4
- fix(benchmark): isolate InternLM reference environment
  chaofengw-nv
  committed
  c3e0c95
- fix(benchmark): inherit runtime packages in reference envs
  chaofengw-nv
  committed
  d13b7dd
- fix(benchmark): support pinned Transformers references
  chaofengw-nv
  committed
  70aae84
- feat(benchmark): qualify code generation models
  chaofengw-nv
  committed
  58c1eff
- feat(benchmark): restore translation qualification
  chaofengw-nv
  committed
  1b6ecfd
- feat(benchmark): restore encoder qualification
  chaofengw-nv
  committed
  271e53d
- fix(benchmark): align embedding timing
  chaofengw-nv
  committed
  270d742
- refactor(benchmark): keep model details family-local
  chaofengw-nv
  committed
  4821352
- fix(benchmark): prime compiled measurements
  chaofengw-nv
  committed
  740d6a1
- test(runtime): preserve factory mismatch coverage
  chaofengw-nv
  committed
  eccd88a
- fix(benchmark): align translation qualification
  chaofengw-nv
  committed
  69cdfc9
- feat(benchmark): add more text qualification profiles
  chaofengw-nv
  committed
  a6df9da
- fix(benchmark): align Llama accuracy prompts
  chaofengw-nv
  committed
  90abb22
- fix(llama): sanitize decoded UTF-8
  chaofengw-nv
  committed
  41cc2d7
- fix(benchmark): size TinyLlama MMLU context
  chaofengw-nv
  committed
  d66231f
- feat(benchmark): qualify additional text models
  chaofengw-nv
  committed
  17d4d1e
- feat(benchmark): qualify large text models
  chaofengw-nv
  committed
  6250be0
- feat(benchmark): qualify current text models
  chaofengw-nv
  committed
  7002081
- feat(benchmark): qualify more text models
  chaofengw-nv
  committed
  2edf1ca
- feat(benchmark): qualify time-series models
  chaofengw-nv
  committed
  bc1349f
- feat(benchmark): qualify TIMM classifiers
  chaofengw-nv
  committed
  01f13c8
- feat(benchmark): qualify speech recognition
  chaofengw-nv
  committed
  db48933
- feat(benchmark): qualify DINOv3 image features
  chaofengw-nv
  committed
  1fac312

Commits

Commits on Sep 18, 2026