Ruihan11 481d7dbb9e perf(rlhf): vectorize padding-free sequence reductions (#9958)

2026-08-24 15:39:07 +08:00

..

app

…

deploy

[NPU] update select_device (#9559)

2026-06-15 14:36:28 +08:00

eval

[NPU] update select_device (#9559)

2026-06-15 14:36:28 +08:00

export

[NPU] update select_device (#9559)

2026-06-15 14:36:28 +08:00

general

fix(dataset): honor explicit hub prefixes (#9971)

2026-08-24 15:04:53 +08:00

infer

fix(infer): reject prompts without generation space (#9851)

2026-08-04 20:39:17 +08:00

llm

test(ci): drop legacy structbert tests and dependency pins (#9938)

2026-08-18 19:45:03 +08:00

loss_scale

fix: handle tool messages in last round loss scaling (#9923)

2026-08-16 21:53:23 +08:00

megatron

fix:Compatibility with Python 3.14 (#9961)

2026-08-21 15:23:47 +08:00

models

[NPU] update select_device (#9559)

2026-06-15 14:36:28 +08:00

sample

…

test_align

perf(rlhf): vectorize padding-free sequence reductions (#9958)

2026-08-24 15:39:07 +08:00

train

Add SDAR confidence-gated teacher distillation loss to GRPO (#9768)

2026-07-20 15:34:47 +08:00

utils

fix(metrics): count empty NLG predictions in score averages (#9962)

2026-08-24 14:59:22 +08:00

__init__.py

…

model_tag.py

…

run_config.yaml

…

run.py

…

test_utils.py

…