Open Models Almanac
Methodssoftware implementationkernel & quantization library

specific method · filed under software implementation

NVIDIA Model-Optimizer

Model-Optimizer is a software toolkit for post-training quantization of large models, supporting both HuggingFace and Megatron-LM workflows.1 Its HuggingFace PTQ script relies on transformers native inference, which runs in a single process and therefore does not support multi-node quantization.2

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. Model-Optimizer supports both HuggingFace and Megatron-LM for quantization — Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
  2. Model-Optimizer’s HuggingFace PTQ script uses transformers native inference, which does not support multi-node due to being limited to single process execution — Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning

Also called Model-Optimizer.

sources
2
model
1
lab adopt it
1
strongest
used

Documented in

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
mentioned 1used 1

NVIDIA-Nemotron-3-Ultra-550B-A55B

NVIDIAadopted2 sources

Model-Optimizer supports both HuggingFace and Megatron-LM for quantization

usedsoftware implementationin Nemotron 3 UltraNVIDIA
1 more from this model

Model-Optimizer's HuggingFace PTQ script uses transformers native inference

mentionedsoftware implementationin Nemotron 3 UltraNVIDIA
↑