Search results

Low perf for backward pass of dropout fused with normalization

nvfuser

performance

rematerialization

Status: Open.

#2231 In Lightning-AI/lightning-thunder;

· liqiangxl opened on Jun 13, 2025

<code>thunder.jit</code> has a relatively high CPU overhead when processing small graphs with small inputs.

performance

Status: Open.

#1657 In Lightning-AI/lightning-thunder;

· kiya00 opened on Jan 17, 2025

nvFuser has a faster RMSNorm fusion definition than thunder's RMSNorm decomposition

operators

performance

Status: Open.

#1582 In Lightning-AI/lightning-thunder;

· mruberry opened on Dec 23, 2024

Investigate Memory and Performance difference using <code>nvfuser</code> vs <code>torch.compile</code> executor on Qwen2

high priority

memory use

nemoIssues needed to support NVIDIA NeMo models.Issues needed to support NVIDIA NeMo models.

performance

thunderfxfor things that could be applicable to the dynamo+thunder frontendfor things that could be applicable to the dynamo+thunder frontend

Status: Open.

#1552 In Lightning-AI/lightning-thunder;

· kshitij12345 opened on Dec 13, 2024

HF Transformers ViT slower than <code>torch.compile</code> and raw pytorch

performance

Status: Open.

#1502 In Lightning-AI/lightning-thunder;

· 2catycm opened on Dec 2, 2024

Investigate the difference in speedup between Llama3-8b and Mistral-7B-v0.1

enhancementNew feature or requestNew feature or request

performance

Status: Open.

#1341 In Lightning-AI/lightning-thunder;

· mpatel31415 opened on Oct 22, 2024

Strength reduction: fold transpose into a subsequent GEMM call

enhancementNew feature or requestNew feature or request

nemoIssues needed to support NVIDIA NeMo models.Issues needed to support NVIDIA NeMo models.

neva

performance

thunderfxfor things that could be applicable to the dynamo+thunder frontendfor things that could be applicable to the dynamo+thunder frontend

Status: Open.

#1252 In Lightning-AI/lightning-thunder;

· tfogal opened on Oct 2, 2024

Implement GroupNorm to invoke APEX GroupNorm for NeMo Stable Diffusion AutoEncoder performance

bugSomething isn't workingSomething isn't working

nemoIssues needed to support NVIDIA NeMo models.Issues needed to support NVIDIA NeMo models.

performance

Status: Open.

#468 In Lightning-AI/lightning-thunder;

· athitten opened on May 29, 2024

Distributed and Bucketing Performance Improvements

bugSomething isn't workingSomething isn't working

distributed

enhancementNew feature or requestNew feature or request

performance

Status: Open.

#348 In Lightning-AI/lightning-thunder;

· parthmannan opened on May 2, 2024