Pruna 0.3.2: More OSS Algos, More Ways to Optimize
β’ 5
AI & ML interests
Efficient machine learning for any model and hardware: pruning, quantization, compilation, and more.
Recent Activity
View all activity
Articles
README: license section and dataset citations
#11 opened about 18 hours ago
by 
begumcig
sdiazlorΒ
updated a
Space about 21 hours ago
studio with gradio server and workflow
#1 opened 3 days ago
by 
akhaliq
nifleischΒ
published 2
models 3 days ago
nifleischΒ
published a
Space 3 days ago
update-for-launch
#1 opened 4 days ago
by sdiazlor
nifleischΒ
updated a
model 3 days ago
update-for-launch
#1 opened 4 days ago
by sdiazlor
nifleischΒ
updated a
Space 3 days ago
minette-kaunismakiΒ
updated a
dataset 3 days ago
nifleischΒ
updated a
model 3 days ago
sdiazlorΒ
updated 2
models 3 days ago
sdiazlorΒ
updated a
collection 3 days ago
Post
102
Hey! We published a set of 6-step LoRAs, one without the upsampling step: PrunaAI/P-Q-Image-2.1
- Generate or edit images in just 6 steps instead of 40. Use 6-steps for preserving inputs when editing and 6-steps-upsampled for improved visual quality and shorter end-to-end generation.
- 6-step distillation provides improved quality with few generation steps. 6-steps-upsampled is the first distillation that removes the need for a separate prompt-upsampling step. This means 41 to 80 seconds saved, depending on your prompt style.
- The LoRA adapters load directly on top of Qwen-Image-2.1, so the pipeline stays unchanged. Few-step distillation cuts the diffusion cost, and the upsampled adapter also removes the prompt-upsampling step before image generation.
- Generate or edit images in just 6 steps instead of 40. Use 6-steps for preserving inputs when editing and 6-steps-upsampled for improved visual quality and shorter end-to-end generation.
- 6-step distillation provides improved quality with few generation steps. 6-steps-upsampled is the first distillation that removes the need for a separate prompt-upsampling step. This means 41 to 80 seconds saved, depending on your prompt style.
- The LoRA adapters load directly on top of Qwen-Image-2.1, so the pipeline stays unchanged. Few-step distillation cuts the diffusion cost, and the upsampled adapter also removes the prompt-upsampling step before image generation.
Post
135
Hi! We open-source PrunaSuperPoint in collaboration with Eclipse Aidge for the DeepGreen project
It's an optimized version of SuperPoint built for faster keypoint detection on edge devices.
- Up to 2.1Γ faster on Jetson Orin Nano, with optimizations applicable to other runtimes.
- The distilled model retains strong keypoint coverage across indoor and outdoor data, with low descriptor differences from the original model.
- We structurally prune the most expensive convolutional layers, recover performance through distillation, and accelerate keypoint selection with hierarchical top-k, all while preserving the original architectureβs core behavior.
Check it here: https://huggingface.co/PrunaAI/PrunaSuperPoint
It's an optimized version of SuperPoint built for faster keypoint detection on edge devices.
- Up to 2.1Γ faster on Jetson Orin Nano, with optimizations applicable to other runtimes.
- The distilled model retains strong keypoint coverage across indoor and outdoor data, with low descriptor differences from the original model.
- We structurally prune the most expensive convolutional layers, recover performance through distillation, and accelerate keypoint selection with hierarchical top-k, all while preserving the original architectureβs core behavior.
Check it here: https://huggingface.co/PrunaAI/PrunaSuperPoint
Post
1,897
Today, we open-source Pruna-Qwen-Image-2.1, a set of a few-step LoRA adapters that make Qwen-Image-2. up to 6.3Γ faster for image generation and editing.
Try it here: https://huggingface.co/PrunaAI/Pruna-Qwen-Image-2.1
Try it here: https://huggingface.co/PrunaAI/Pruna-Qwen-Image-2.1
Post
156
P-Bench: new public benchmark for T2I models
- Compare optimized or accelerated endpoints on quality, speed, and price.
- Analyze the paretos
- Explore the samples
Check it here π PrunaAI/P-Bench
- Compare optimized or accelerated endpoints on quality, speed, and price.
- Analyze the paretos
- Explore the samples
Check it here π PrunaAI/P-Bench
Post
169
Hey! We've released PrunaVAED, a drop-in replacement decoder for LTX-2.3
- ~1.7-2.1x faster decoder & ~50% lower peak VRAM
- Near-original visual quality & no changes to the latent encoding
Check it here: https://huggingface.co/PrunaAI/PrunaVAED
- ~1.7-2.1x faster decoder & ~50% lower peak VRAM
- Near-original visual quality & no changes to the latent encoding
Check it here: https://huggingface.co/PrunaAI/PrunaVAED