xvxuopop 31078aa10a [v1] Support multimodal Ulysses CP and memory-efficient chunk loss for SFT (#10762)

2026-09-09 19:22:25 +08:00

..

train_full_chunk_loss.yaml

[v1] Support multimodal Ulysses CP and memory-efficient chunk loss for SFT (#10762)

2026-09-09 19:22:25 +08:00

train_full_deepspeed.yaml

[v1] refactor registry plugin structure and params (#10641)

2026-07-24 15:23:21 +08:00

train_full_fsdp2.yaml

[v1] refactor registry plugin structure and params (#10641)

2026-07-24 15:23:21 +08:00

train_full_liger_kernel.yaml

[v1] refactor registry plugin structure and params (#10641)

2026-07-24 15:23:21 +08:00

train_full_multimodal_ulysses_cp.yaml

[v1] Support multimodal Ulysses CP and memory-efficient chunk loss for SFT (#10762)

2026-09-09 19:22:25 +08:00

train_full_muon.yaml

[v1] add muon optimizer (#10618)

2026-07-17 21:44:28 +08:00

train_full_qwen3_moe_fsdpturbo_ep_fsdp.yaml

[v1] add FSDPTurbo EP/EFSDP plugin for MoE training (#10676)

2026-08-13 20:45:55 +08:00

train_full_ulysses_cp.yaml

[v1] refactor registry plugin structure and params (#10641)

2026-07-24 15:23:21 +08:00

train_multimodal.yaml

[v1] support GDN Ulysses cp (#10727)

2026-08-20 18:53:39 +08:00