Add A100 PyTorch/CUDA D-FINE benchmark submissions (5 variants)#4
Merged
Conversation
5 new submissions for D-FINE on NVIDIA A100-PCIE-40GB (MIG 7g.40gb slice)
with torch 2.6.0+cu124, libreyolo at commit
3383a8f142a5decc735f362258f6851d3f026fa3 (the 106-add-d-fine-model-family
feature branch). Variants: n, s, m, l, x.
Metadata changes:
- support-matrix.json: bump supported_libreyolo_identifier from a single
string to a list, adding the D-FINE branch SHA alongside the existing
pinned 1c70efb commit. Append the 5 dfine-{n,s,m,l,x} model ids.
- website/src/data/metadata/families.json: add dfine family (USTC, ICLR
2025 paper).
- website/src/data/metadata/models.json: add 5 D-FINE variant entries.
Measured COCO val2017 mAP@50-95 (paper reference in parens):
dfine-n 0.4279 (~0.428)
dfine-s 0.5070 (~0.487)
dfine-m 0.5509 (~0.523)
dfine-l 0.5725 (~0.540)
dfine-x 0.5931 (~0.558)
LibreYOLO weights score above the original paper for s/m/l/x — this is
likely because the LibreYOLO checkpoints use the Objects365+COCO
two-stage pretraining that the original D-FINE repo also publishes
("L_obj365" weights), which beats the COCO-only baseline reported in the
paper's main table. Worth flagging if a reviewer wants the strict
COCO-only numbers instead.
Local: scripts/validate_submission.py and scripts/build_verified_results.py
both exit 0. generated/verified-results.v1.json regenerated and committed.
Note: the D-FINE branch is not yet merged into LibreYOLO main. Once it
lands and a new pinned commit replaces these two SHAs, this PR's
submissions can either be left as-is (the SHA stays in support-matrix
indefinitely as a historical entry) or backfilled to the new pin.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
5 new submissions for D-FINE on NVIDIA A100-PCIE-40GB (MIG 7g.40gb slice) with torch 2.6.0+cu124, libreyolo at commit
3383a8f142a5decc735f362258f6851d3f026fa3 (the 106-add-d-fine-model-family feature branch). Variants: n, s, m, l, x.
Metadata changes:
Measured COCO val2017 mAP@50-95 (paper reference in parens):
dfine-n 0.4279 (~0.428)
dfine-s 0.5070 (~0.487)
dfine-m 0.5509 (~0.523)
dfine-l 0.5725 (~0.540)
dfine-x 0.5931 (~0.558)
LibreYOLO weights score above the original paper for s/m/l/x — this is likely because the LibreYOLO checkpoints use the Objects365+COCO two-stage pretraining that the original D-FINE repo also publishes ("L_obj365" weights), which beats the COCO-only baseline reported in the paper's main table. Worth flagging if a reviewer wants the strict COCO-only numbers instead.
Local: scripts/validate_submission.py and scripts/build_verified_results.py both exit 0. generated/verified-results.v1.json regenerated and committed.
Note: the D-FINE branch is not yet merged into LibreYOLO main. Once it lands and a new pinned commit replaces these two SHAs, this PR's submissions can either be left as-is (the SHA stays in support-matrix indefinitely as a historical entry) or backfilled to the new pin.