FlashMLA

Latest news

1. 20 Oct
   DeepSeek open-sources DeepInfer high-performance inference engine

   DeepSeek released DeepInfer, an open-source inference framework used in its large-scale production services. The release includes optimizations for offline inference, parallelism strategies, specialized kernels, and features for deterministic, batch-invariant, and asynchronous scheduling.

   Tools · 1 source

   

Articles

No articles about FlashMLA yet.