NVIDIA-NeMo /Megatron-Bridge
Training library for Megatron-based models with bidirectional Hugging Face conversion capability
★ 997⑂ 583👁 9⑃ 5,947● 985▤ 20
+1
日增长
997
stars
583
forks
9
watchers
5,947
prs
985
issues
20
releases
README
NeMo Megatron Bridge Documentation | Supported Models | Examples | Contributing 📣 News - [11/13/2026] Inkling 2 Small from Thinking Machines is now supported! Megatron Bridge provides HF↔Megatron checkpoint conversion, including the release's quantized experts, image-and-text inference, and GB200 recipes for pretraining, full SFT, 32K long-context SFT with context parallelism, LoRA, and image-text SFT and LoRA. See the examples README and verification card for verified configurations and current limitations. - [11/05/2026] ERNIE-5.5-Flash-Base is now merged on main! See the examples README for sft training details. - [10/16/2026] Reflection AI Beam-1.0-501B-A23B is now supported! Megatron Bridge provides HF↔Megatron checkpoint conversion, inference, and BF16 H200 recipes for pretraining, SFT, long-context SFT, and LoRA PEFT. See the examples README and verification card for verified configurations and current limitations. - [10/15/2026] NVIDIA Nemotron 3.5 Super VL is released! Megatron Bridge supports the 120B-A12B vision-language model with HF↔Megatron checkpoint conversion, image inference, and pretraining, SFT, and LoRA recipes, plus text-only conversion and training workflows. See the model documentation and verification card for supported workflows, verified configurations, and current limitations. - [10/15/2026] Muse-Spark-Open-816B-A42B is now supported! Megatron Bridge adds checkpoint conversion, inference, and GB200 training recipes for the 816B-total / 42B-active vision-and-audio-language MoE model: pretraining, full SFT, long-context SFT at 32K and 64K with context parallelism, LoRA, and image-text SFT and LoRA, with HybridEP MoE dispatch. See the examples README and the verification card for details. - [08/19/2026] Megatron Bridge 0.6.0 released! Highlights include GLM-5.2 with cuDNN fused DSA for 128K THD-packed context-parallel training, new LLM/VLM/diffusion models, more reliable VLM data and sequence packing, improved checkpoint and LoRA workflows, unified training and conversion launchers, and refreshed performance recipes. Huge thanks to our community contributors! See the full release notes. - [08/11/2026] NVIDIA Nemotron 3.5 Lightning is released! Read the technical blog here. See the Nemotron 3.5 Lightning README for details. - [07/31/2026] K-EXAONE-2 is now supported! Day-0 support for LG AI Research's K-EXAONE-2 (750B-A37B MoE) is now available on the k-exaone-2 branch. Users can convert checkpoints, run inference, fine-tune with SFT or PEFT (LoRA), and use the NVFP4 inference-optimized checkpoint released alongside the model. See the branch documentation for setup instructions and examples. - [07/17/2026] LongStraw (MinT-2M) is a long-context RL research system from MindLab Research that uses Megatron Bridge's modeling and LoRA capabilities. LongStraw explores resident-prefix, response-only GRPO by capturing a shared prompt once and replaying only the trainable response branches, and reports a 2.1M-token G
Top issues
#4214Compare Axolotl Qwen vs MBridge Qwen vs MBridge Nemotron Super LoRA SFT performance14 comments · 2026-06-08#1255Nan grad norm in MXFP8 training14 comments · 2025-11-10#3555[bug] There seems to be a consistent increase in CPU memory while using MBridge11 comments · 2026-04-28#5067[bug] [DeepEP][B200+B300][26.06.01] Timeout in internode_dispatch (CPU) during MoE token dispatch10 comments · 2026-07-24#4609[feature] intra-microbatch reordering for MegatronMIMO (+ sequence packing, scalable DP)10 comments · 2026-07-01#4351[feature] Support offline tokenized + index-only packed mode for SFT datasets10 comments · 2026-06-14#2785[bug] Torch FSDP2 does not work10 comments · 2026-03-13#1986Support and switching mechanism for multiple LoRA adapters10 comments · 2026-01-19
Discussions
#4957Accelerating GLM-5.2 Long-Context Training with cuDNN Fused DSA Kernels and THD-Packed Context Parallelism using Megatron Bridge0 comments · 2026-07-19#3460Low-Precision Training with Megatron-Bridge: FP8 (Current Scaling), MXFP8, NVFP4 vs BF160 comments · 2026-04-21#2935Dapt traing with 8xb2000 comments · 2026-03-21#1609TildeOpen30B distilation3 comments · 2025-12-05#1320What is the status of DeepEP/HybridEP Support in Megatron Bridge and what's their hardware requirement?1 comments · 2025-11-12
Latest release
NVIDIA Megatron-Bridge 0.7.0 v0.7.0 2026-10-30Highlights Model Collection Support LLM DeepSeek V4 — enhanced support with MCore HybridModel on main branch; library configs modernized; multi-node inference launcher (PR\5860, PR\5554, PR\6163) GLM-5.2 — enhanced support with MCore HybridModel, with shared GLM-5.2/5.3 bridge compatibility (PR\6158, PR\6117) Kimi K3 — language backbone (PR\5130) Muse Glimmer 30B (PR\5479, PR\6255) K-EXAONE 2.0 (PR\5191, community @eagle705) Muse Spark Open 816B-A42B and Beam 1.0 501B-A23B are on main, not in 0.7.0. Verified with Megatron Bridge main in the NeMo 26.10 container