nvMelissa

Results 13 issues of nvMelissa

**Is your feature request related to a problem? Please describe.** N/A **Describe the solution you'd like** Support DeepSeek FP8 recipe in JAX. Already supported in Pytorch. **Describe alternatives you've considered**...

FP8
Priority = P1

**Is your feature request related to a problem? Please describe.** At the moment, we enumerate the parameters in C APIs like this: https://github.com/NVIDIA/TransformerEngine/blob/5e4e0b2c378d2b1ec2ee65dfa85124e1dd805389/transformer_engine/common/fused_attn/fused_attn.cpp#L835 As we add more features to attention,...

refactor
attention

**Is your feature request related to a problem? Please describe.** This is not related to a problem, it is a feature request to expand model coverage **Describe the solution you'd...

attention

Is your feature request related to a problem? Please describe. To be added Describe the solution you'd like Work on improving performance for FP8 current scaling Describe alternatives you've considered...

performance
attention

**Is your feature request related to a problem? Please describe.** The logic around cuDNN's support matrix for SDPA is getting long and hard to maintain. **Describe the solution you'd like**...

refactor

## 🚀 Feature Dynamic Shapes - Rework Thunder to be shape-agnostic wherever possible. We should only specialize code for specific shapes in small, targeted areas where it's truly necessary. ###...

high priority

## 🚀 Model / language coverage Investigate and fix slowdowns and long compile times for target MoE models: Llama 4, GPT OSS, DeepSeek V3.1, and Qwen3-Next ### Pitch ### Alternatives...

performance

## 🚀 Feature Match and replace computation, in particular sequences of multiple operations. This feature will make it easier and faster for developers to improve model performance of new models...

high priority
developer efficiency
pattern matching