wzh

hg2wzh

·

AI & ML interests

None yet

Recent Activity

liked a Space 14 days ago

HuggingFaceM4/encoder-free-vlm

liked a model about 1 month ago

reacted to erikkaum's post with ❤️ about 2 months ago

Releasing my first kernel 🔥 MaxSim Late-interaction retrieval (ColBERT / PyLate) bottlenecks on materializing the full similarity matrix. This kernel avoids it by using tiled scoring with simdgroup_matrix (Metal) and WMMA. The result is 3–5× speedup compared to naive PyTorch baseline 🔥 Benchmarks: - SmallRerank (B=32, C=10): up to 3.2× (M3 Pro) / 2.8× (A100) - HeavyRerank (B=32, C=100): up to 3.8× (M3 Pro) / 5.3× (A100) - LongDocStress (Ld=1024): up to 6.2× (L4) Try it out 👇 https://huggingface.co/kernels/erikkaum/maxsim

View all activity

Organizations

None yet

liked a Space 14 days ago

Encoder-Free VLM

Train Your Own Encoder-Free VLM in $100

liked a model about 1 month ago

srpone/gr-lite

Image Feature Extraction • 0.3B • Updated May 26 • 2.77k • 5

reacted to erikkaum's post with ❤️ about 2 months ago

Post

3178

Releasing my first kernel 🔥 MaxSim

Late-interaction retrieval (ColBERT / PyLate) bottlenecks on materializing the full similarity matrix. This kernel avoids it by using tiled scoring with simdgroup_matrix (Metal) and WMMA.

The result is 3–5× speedup compared to naive PyTorch baseline 🔥

Benchmarks:
- SmallRerank (B=32, C=10): up to 3.2× (M3 Pro) / 2.8× (A100)
- HeavyRerank (B=32, C=100): up to 3.8× (M3 Pro) / 5.3× (A100)
- LongDocStress (Ld=1024): up to 6.2× (L4)

Try it out 👇
https://huggingface.co/kernels/erikkaum/maxsim

liked a dataset about 2 months ago

nvidia/Nemotron-Image-Training-v3

Viewer • Updated Apr 28 • 6.92M • 3.96k • 75

liked a model about 2 months ago

jinaai/jina-embeddings-v5-omni-nano-retrieval

Feature Extraction • 0.9B • Updated about 1 month ago • 2.84k • 8

upvoted a collection about 2 months ago

jina-embeddings-v5-omni

Multimodal (text + image + video + audio) embedding models aligned with jina-embeddings-v5-text-*. Two sizes, four task variants each. • 27 items • Updated May 12 • 36

liked 2 models about 2 months ago

Qwen/Qwen3.6-27B

Image-Text-to-Text • 28B • Updated Apr 24 • 4.93M • • 1.88k

Zyphra/ZAYA1-8B

9B • Updated 8 days ago • 61.2k • 580

upvoted a collection about 2 months ago

GenLIP

Model weights of paper "Let ViT Speak: Generative Language-Image Pre-training" • 6 items • Updated May 5 • 8

liked 2 models about 2 months ago

nvidia/llama-nemotron-rerank-vl-1b-v2

Text Ranking • 2B • Updated May 20 • 101k • 53

HuggingFaceTB/nanowhale-100m

Text Generation • 0.1B • Updated May 4 • 873 • 64

liked a dataset 2 months ago

qihoo360/FineHARD

Preview • Updated Oct 9, 2025 • 135 • 11

liked a Space 3 months ago

QED-Nano: Teaching a Tiny Model to Prove Hard Theorems

Who needs 1T parameters? Olympiad proofs with a 4B model

upvoted a collection 3 months ago

TIPSv2

TIPSv2 foundational vision-language models. Webpage: https://gdm-tipsv2.github.io/ • 9 items • Updated Apr 14 • 38

liked 2 models 3 months ago

z-lab/Qwen3.5-27B-DFlash

Text Generation • 2B • Updated 15 days ago • 16.7k • 110

Qwen/Qwen-Image-Edit-2511

Image-to-Image • Updated Dec 23, 2025 • 162k • • 1.1k

upvoted a collection 3 months ago

Qwen-Image

14 items • Updated Dec 31, 2025 • 115

liked a model 3 months ago

zai-org/GLM-5.1

Text Generation • 754B • Updated May 13 • 96.9k • • 1.83k

liked a dataset 3 months ago

TIGER-Lab/MMEB-V2

Updated Nov 11, 2025 • 3.53k • 16

liked a model 3 months ago

ATH-MaaS/Marco-Mini-Instruct

Text Generation • 17B • Updated Apr 10 • 907 • 47