Skip to content

Actions: sgl-project/SpecForge

Actions

PR Test

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
test.yaml
788 workflow runs
788 workflow runs

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
fix: Bump sglang version from 0.5.9 to 0.5.10 PR Test #932: Pull request #529 synchronize by moehanabi
34m 31s moehanabi:bump_sglang0.5.10
moehanabi:bump_sglang0.5.10
34m 31s
fix: Bump sglang version from 0.5.9 to 0.5.10 PR Test #931: Pull request #529 synchronize by moehanabi
33m 58s moehanabi:bump_sglang0.5.10
moehanabi:bump_sglang0.5.10
33m 58s
Fix incorrect LSE gradients in cached FlashAttention for Eagle training PR Test #930: Pull request #536 synchronize by uygnef
24m 34s uygnef:fix/fa
uygnef:fix/fa
24m 34s
support quant PR Test #929: Pull request #540 opened by liusy58
Action required liusy58:quant_support
liusy58:quant_support
Action required
Fix/gemma3 eagle3 hooks PR Test #928: Pull request #538 synchronize by tcligg
Action required tcligg:fix/gemma3-eagle3-hooks
tcligg:fix/gemma3-eagle3-hooks
Action required
Fix/gemma3 eagle3 hooks PR Test #927: Pull request #538 opened by tcligg
Action required tcligg:fix/gemma3-eagle3-hooks
tcligg:fix/gemma3-eagle3-hooks
Action required
Fix: Ensure aux hidden states are correctly set for EAGLE3 in Gemma3.… PR Test #925: Pull request #531 synchronize by tcligg
-1s tcligg:fix/gemma3-eagle3-hooks
tcligg:fix/gemma3-eagle3-hooks
-1s
Fix: Ensure aux hidden states are correctly set for EAGLE3 in Gemma3.… PR Test #924: Pull request #531 synchronize by tcligg
Action required tcligg:fix/gemma3-eagle3-hooks
tcligg:fix/gemma3-eagle3-hooks
Action required
fix: correct vocab_size to 262144 in gemma3-1b-eagle3 configuration PR Test #922: Pull request #537 opened by javierlimt6
Action required javierlimt6:fix/vocab_size_gemma
javierlimt6:fix/vocab_size_gemma
Action required
Fix incorrect LSE gradients in cached FlashAttention for Eagle training PR Test #921: Pull request #536 opened by uygnef
30m 11s uygnef:fix/fa
uygnef:fix/fa
30m 11s
fix: EAGLE-3 training compatibility with multimodal-wrapped targets and large vocabs PR Test #920: Pull request #535 opened by elad-inferize
Action required elad-inferize:fix/mm-wrapper-eagle3-compat
elad-inferize:fix/mm-wrapper-eagle3-compat
Action required
fix: Bump sglang version from 0.5.9 to 0.5.10 PR Test #919: Pull request #529 synchronize by moehanabi
33m 31s moehanabi:bump_sglang0.5.10
moehanabi:bump_sglang0.5.10
33m 31s
Preserve drafter vocab mapping when fine-tuning from a checkpoint PR Test #918: Pull request #534 opened by luv-bansal
Action required luv-bansal:fix-vocab-mapping
luv-bansal:fix-vocab-mapping
Action required
fix: Bump sglang version from 0.5.9 to 0.5.10 PR Test #917: Pull request #529 synchronize by moehanabi
10m 7s moehanabi:bump_sglang0.5.10
moehanabi:bump_sglang0.5.10
10m 7s
fix: Bump sglang version from 0.5.9 to 0.5.10 PR Test #916: Pull request #529 synchronize by moehanabi
3m 35s moehanabi:bump_sglang0.5.10
moehanabi:bump_sglang0.5.10
3m 35s
Fix: support preformatted text datasets in train_eagle3.py (avoid forcing conversations generator) PR Test #915: Pull request #498 synchronize by Seun-Ajayi
Action required Seun-Ajayi:main
Seun-Ajayi:main
Action required
[Fix] preserve image data in preprocess for VLM training on multimodal data PR Test #914: Pull request #532 opened by jamesahou
Action required jamesahou:mm-preprocess-fix
jamesahou:mm-preprocess-fix
Action required
feat: add EAGLE3 support for Step-3.5-Flash PR Test #912: Pull request #530 opened by zijiexia
Action required zijiexia:support_step3p5
zijiexia:support_step3p5
Action required
fix: Bump sglang version from 0.5.9 to 0.5.10 PR Test #911: Pull request #529 opened by moehanabi
3m 22s moehanabi:bump_sglang0.5.10
moehanabi:bump_sglang0.5.10
3m 22s
Reduce peak GPU memory in Eagle3 online target generation by avoiding an extra logits copy PR Test #910: Pull request #528 opened by zijiexia
49m 20s zijiexia:fix_target_out_oom
zijiexia:fix_target_out_oom
49m 20s
support qwen3vl-235b PR Test #909: Pull request #525 synchronize by liusy58
26m 33s liusy58:qwen3-235b
liusy58:qwen3-235b
26m 33s
Fix multimodal hidden-state preparation for Qwen3-VL models PR Test #908: Pull request #526 synchronize by liusy58
38m 31s liusy58:fix_prepare_hidden_states
liusy58:fix_prepare_hidden_states
38m 31s
Fix VLM preprocessing and add mRoPE position handling in target head PR Test #907: Pull request #527 synchronize by liusy58
14m 5s liusy58:fix_train_vlm
liusy58:fix_train_vlm
14m 5s
feat: reduce Eagle3 training memory spike via all-to-all sharding PR Test #906: Pull request #524 synchronize by cih9088
20m 25s laoconeth:feat/shard-target-output
laoconeth:feat/shard-target-output
20m 25s
feat: reduce Eagle3 training memory spike via all-to-all sharding PR Test #905: Pull request #524 synchronize by cih9088
28m 57s laoconeth:feat/shard-target-output
laoconeth:feat/shard-target-output
28m 57s