Megatron-MoE-ModelZoo
Best practices for training DeepSeek, Mixtral, Qwen and other MoE models using Megatron Core.
yanring/Megatron-MoE-ModelZoo
View on GitHub ↗Best practices for training DeepSeek, Mixtral, Qwen and other MoE models using Megatron Core.
Megatron-MoE-ModelZoo grew +2.0% in the last 30 days, ranks 67th of 103 in Pretraining & Training, last commit 149 days ago.
Star trajectory
115 days tracked2026-07-02206 stars today2026-10-25
Momentum
+2.0% in 30d, above the Pretraining & Training median of +0.0%.
Maintenance
Last commit 149 days ago, fresher than the median of 163d.
How it compares in Pretraining & Training
Megatron-MoE-ModelZoo
206
stars
+2.0% 30d
commit 149 days ago
Pretraining & Training median
316
stars
+0.0% 30d
commit ~163d ago
Leader · maxtext
2.4k
stars
+0.9% 30d
commit today
Ranked 67th of 103 by stars.
huggingface/optimum-habana212-6wadeKeith/autoresearch-qwen212-6
yanring/Megatron-MoE-ModelZoo206
kamalkraj/ALBERT-TF2.0204+2MirroS-Lab/AgentGarten202+4- Language
- Python
- License
- MIT License
- Contributors
- 10
- Forks
- 34
- Watchers
- 7
- Open issues
- 8
- Releases
- 0
- Created
- 2025-03-24
- Stars this week
- +0
You've seen where Megatron-MoE-ModelZoo stands today.
Get the moves before everyone else: the 5 Pretraining & Training repos climbing fastest, every Tuesday.
Free weekly digest. Double opt-in, unsubscribe anytime.