Yueming Yuan 6d0cd93f10 DeepSeek-V4 support-v2: native megatron implementation as option (#2706)

2026-08-31 13:21:05 -07:00

..

deepseek_v4

DeepSeek-V4 support-v2: native megatron implementation as option (#2706)

2026-08-31 13:21:05 -07:00

glm5

Bump Megatron-LM to miles-main-20260819 (latest NVIDIA dev) (#2673)

2026-08-24 12:37:44 -07:00

inkling

Bump Megatron-LM to miles-main-20260819 (latest NVIDIA dev) (#2673)

2026-08-24 12:37:44 -07:00

__init__.py

a new start

2025-10-08 23:09:54 -07:00

cp_utils.py

feat: real cp support with relayout fix for qwen3.5 train/rollout mismatch (#885)

2026-04-10 20:06:54 -07:00

dsa_topk.py

[SGLang v0.5.13] Add configurable Miles DSA top-k backend (#1058)

2026-06-22 19:48:20 -07:00

glm4.py

Bump Megatron-LM to miles-main-20260819 (latest NVIDIA dev) (#2673)

2026-08-24 12:37:44 -07:00

hf_attention.py

Bump Megatron-LM to miles-main-20260819 (latest NVIDIA dev) (#2673)

2026-08-24 12:37:44 -07:00

qwen3_5.py

Bump Megatron-LM to miles-main-20260819 (latest NVIDIA dev) (#2673)

2026-08-24 12:37:44 -07:00

qwen3_next.py

Bump Megatron-LM to miles-main-20260819 (latest NVIDIA dev) (#2673)

2026-08-24 12:37:44 -07:00

qwen3_vl.py

feat: Fix Qwen3-VL THD packed mRoPE positions (#1272)

2026-06-19 10:46:45 -07:00

qwen_gdn_backend.py

feat: add FlashQLA backend for Qwen GDN linear-attention layers (#1318)

2026-06-15 19:56:44 -07:00