Nemotron Models Ideas Portal

1 Vote

Al-powered USD translation engine in Omniverse that learns application-specific conventions (Houdini primitives + Maya meshes + Unreal actors • Blender objects) and automatically resolves schema conflicts, attribute mappings, and pipeline incompatibilities-enabling true interoperability across the VFX ecosystem.

No description provided

Created 14 Oct 23:21

0
1 Vote

Per-shot learning of quality thresholds that dynamically adjusts LOD transitions, Nanite cluster sizes, and impostor distances based on camera behavior and framerate targets.

No description provided

Created 14 Oct 23:04

0
1 Vote

Maintain full pipeline context across Houdini→USD→Unreal workflows to detect incompatibilities before export and auto-fix common translation issues.

No description provided

Created 14 Oct 23:03

0
1 Vote

Per-shot learning of quality thresholds that dynamically adjusts LOD transitions, Nanite cluster sizes, and impostor distances based on camera behavior and framerate targets.

No description provided

Created 14 Oct 23:02

0
6 Vote

Hardware optimization - A model optimized for B200 GPUs

Develop a model that is most efficient on an B200 GPU

Created 10 Sep 20:45

0
6 Vote

Expanded Model Availability - Bedrock

No description provided

Created 10 Sep 13:27

0
3 Vote

Efficiency needs - balanced throughput and latency

Develop a model that excels in workloads with intermediate batches and has good throughput and good latency, rather the maximizing one at the expense of the other

Created 10 Sep 20:49

0
3 Vote

Hardware optimization - A model optimized for H100 GPUs

Develop a model that is most efficient on an H100 GPU

Created 10 Sep 20:44

0
2 Vote

Efficiency needs - I want to minimize latency for tiny batches

Develop a model that excels in workloads with small batches and low latencies, matching online scenarios

Created 10 Sep 20:49

0
2 Vote

Controlling reasoning length - Low/Medium/High thinking modes

To control the length of reasoning I want to be able to choose between low, medium or high thinking modes. This effectively tells the model how much time to spend on reasoning before returning the final answer.

Created 10 Sep 13:40

0