BREAKING
NeMo AutoModel Adds Diffusers Support
Fine-Tuning for Image and Video
Launch Model Recipes
Text-to-Image
FLUX.1-dev
●
12B parameters
●
Full and LoRA recipes
Text-to-Video
Wan 2.1
●
1.3B parameters
●
HunyuanVideo 1.5 too
The Fine-Tuning Workflow
1
Load HF model
↓
2
Encode VAE latents
↓
3
FSDP2 training
↓
4
Generate & infer
0
x
min throughput
0
x
max throughput
0
%
memory cut
A Standardized Fine-Tuning Path
AI NEWS BLITZ
NVIDIA expands its open-source NeMo AutoModel library to Hugging Face Diffusers.