SDLLM-EsoLM-1.7B-Base is a 1.7B parameter language model combining autoregressive and masked diffusion language modeling objectives. This model is evaluated as part of the study «Scaling Beyond Masked Diffusion Language Models,» which investigates diffusion-based approaches to language modeling. SDLLM-EsoLM-1.7B-Base is released under the CC BY-NC 4.0 license. By using this model, you are agreeing to the terms and conditions of the license. This model is intended for research purposes only and may not be used for commercial applications. SDLLM-EsoLM-1.7B-Base is intended for use by researchers studying diffusion-based language modeling approaches. Hugging Face 09/02/2026 via https://huggingface.co/nvidia/SDLLM-EsoLM-1.7B-Base Scaling Beyond Masked Diffusion Language Models Nemotron Nano 2 Dataset This repository includes a self-contained SDLLM inference and evaluation package. Use python sample.py —help for sampler overrides. With no prompt the script performs unconditional generation. Release defaults are: AR uses Gumbel-max ancestral sampling (temperature 1); MDLM uses ancestral masked diffusion; EsoLM uses its first-hitting ancestral sampler; and Duo uses its uniform-state…
Модальности:
Генерация текста
Задача: Генерация текста
Автор: nvidia
Теги: diffusion-language-modeling, en
Лайков: 3 | Загрузок: 25
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
