Stuido
🐢
1
Rewarx is an AI-powered product photography tool designed fo
Keble Zhu
keblezhu
AI & ML interests
None yet
Recent Activity
liked a Space 3 days ago
rewarx/stuido commentedon an article 3 days ago
AI Trends 2026: Test-Time Reasoning and the Rise of Reflective Agents upvoted an article 3 days ago
AI Trends 2026: Test-Time Reasoning and the Rise of Reflective AgentsOrganizations
None yet
liked a
Space 3 days ago
commented on AI Trends 2026: Test-Time Reasoning and the Rise of Reflective Agents 3 days ago
upvoted an article 3 days ago
Article
AI Trends 2026: Test-Time Reasoning and the Rise of Reflective Agents
aufklarer
• • 3 liked a
Space 4 days ago
updated a
Space 4 days ago
published a
Space 4 days ago
liked a
Space 4 days ago
updated a
Space 4 days ago
published a
Space 4 days ago
reacted to HannesVonEssen's post with ❤️ 4 days ago
Post
4835
📣 Add architecture visualization to model card!
🌟 For all creators out there: add a model visualization to your model card to capture your audience's attention!
🖱️ When clicked, it opens an interactive view with multiple levels of granularity!
1️⃣ Paste url at https://hfviewer.com/model-card-embed
2️⃣ Paste generated code in your README.md!
3️⃣ ✨
🌟 For all creators out there: add a model visualization to your model card to capture your audience's attention!
🖱️ When clicked, it opens an interactive view with multiple levels of granularity!
1️⃣ Paste url at https://hfviewer.com/model-card-embed
2️⃣ Paste generated code in your README.md!
3️⃣ ✨
reacted to erikkaum's post with ❤️ 4 days ago
Post
3052
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
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
reacted to alvarobartt's post with ❤️ 4 days ago
Post
3190
Latest
TL; DR MoEs can be misleading to reason about from active parameters alone, since each token only activates a subset of experts, while the serving setup still needs to account for the full resident memory footprint.
🧠
🏗️ Dense models usually load and use most weights every forward pass, while MoEs load many experts but only route each token to a few of them
⚡ Active params isn't the same as memory footprint, especially for sparse architectures
📦 Runtime memory is about what is used per request/token, while loading memory also includes the expert weights that need to be resident
📚 KV cache can still dominate depending on context length, batch size, and concurrency
🔀 Expert Parallelism (EP) helps shard experts across accelerators when expert weights dominate
🚀 Data Parallelism (DP) + EP is often a good fit for throughput-oriented MoE serving
Check the repository at https://github.com/alvarobartt/hf-mem
hf-mem release added a breakdown of Mixture-of-Experts (MoE) memory usage!TL; DR MoEs can be misleading to reason about from active parameters alone, since each token only activates a subset of experts, while the serving setup still needs to account for the full resident memory footprint.
🧠
hf-mem now splits MoE memory into base model weights, routed experts, and KV cache🏗️ Dense models usually load and use most weights every forward pass, while MoEs load many experts but only route each token to a few of them
⚡ Active params isn't the same as memory footprint, especially for sparse architectures
📦 Runtime memory is about what is used per request/token, while loading memory also includes the expert weights that need to be resident
📚 KV cache can still dominate depending on context length, batch size, and concurrency
🔀 Expert Parallelism (EP) helps shard experts across accelerators when expert weights dominate
🚀 Data Parallelism (DP) + EP is often a good fit for throughput-oriented MoE serving
Check the repository at https://github.com/alvarobartt/hf-mem
reacted to RiverRider's post with 👍 4 days ago
Post
3267
Natural Language Autoencoders: A Window into Latent Structure
I introduced a concise mathematical formulation of the P versus NP question into the SRT-NLA-AV-v1 demonstration:
P vs NP asks whether every problem whose solution can be verified in polynomial time (NP) can also be solved in polynomial time (P). Integer factorization — given N = p·q where p and q are large primes (p < q) — is in NP but widely believed not to be in P.
The resulting activation verbalization (best-of-N, reranked by AR fidelity) surfaced:
“This article originally appeared in the August 2016 edition of CACM. A new method of proving computational hardness of problems, known as multilinearization, can improve efficiency, reduce complexity and simplify proofs. In this article, I describe multilinearization and its application to several key problems, from the discrete logarithm and factoring to RSA and elliptic-curve discrete logarithms.”
What emerges is not a literal restatement, but a structured articulation of the model’s internal associations: hardness proofs, algebraic techniques, and the cryptographic implications that orbit this foundational question in computational complexity.
The demo offers a compelling interface for exploring these latent representations.
Explore it here:
RiverRider/srt-nla-av-v1-demo
Recommended: Best-of-N sampling with round-trip evaluation for highest fidelity.
I introduced a concise mathematical formulation of the P versus NP question into the SRT-NLA-AV-v1 demonstration:
P vs NP asks whether every problem whose solution can be verified in polynomial time (NP) can also be solved in polynomial time (P). Integer factorization — given N = p·q where p and q are large primes (p < q) — is in NP but widely believed not to be in P.
The resulting activation verbalization (best-of-N, reranked by AR fidelity) surfaced:
“This article originally appeared in the August 2016 edition of CACM. A new method of proving computational hardness of problems, known as multilinearization, can improve efficiency, reduce complexity and simplify proofs. In this article, I describe multilinearization and its application to several key problems, from the discrete logarithm and factoring to RSA and elliptic-curve discrete logarithms.”
What emerges is not a literal restatement, but a structured articulation of the model’s internal associations: hardness proofs, algebraic techniques, and the cryptographic implications that orbit this foundational question in computational complexity.
The demo offers a compelling interface for exploring these latent representations.
Explore it here:
RiverRider/srt-nla-av-v1-demo
Recommended: Best-of-N sampling with round-trip evaluation for highest fidelity.
- 1 reply