sapientinc/HRM-Text-1B
Text Generation • 1B • Updated • 103k • 368
Pedro Cuenca PRO
pcuenq
AI & ML interests
None yet
Recent Activity
liked a model about 7 hours ago
sapientinc/HRM-Text-1B upvoted an article about 8 hours ago
Eight Days in China: What I Learned from the AI Labs, Robotics Startups and Academia liked a model about 24 hours ago
OsaurusAI/DeepSeek-V4-Flash-JANGTQ2Organizations
liked a
model about 7 hours ago
upvoted an article about 8 hours ago
Article
Eight Days in China: What I Learned from the AI Labs, Robotics Startups and Academia
matthew-d-white
• • 3 liked a
model about 24 hours ago
add pi0 to important_models
2
#10 opened 6 days ago
by 
pagezyhf
liked a
Space 3 days ago
liked a
model 3 days ago
updated a
dataset 4 days ago
published a
dataset 4 days ago
published a
bucket 4 days ago
upvoted a changelog 4 days ago
Hugging Face Changelog
Copy Repo Contents to Buckets Instantly
• 40
upvoted a changelog 5 days ago
Hugging Face Changelog
Filter Leaderboards by Model Size
• 98
liked a
Space 5 days ago
liked 2
models 6 days ago
reacted to alvarobartt's post with 🚀❤️ 6 days ago
Post
3229
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
upvoted 2
articles 6 days ago
Article
Why We Built the OpenMDW License: A Comprehensive License for ML Models
linuxfoundation
• • 25 Article
Software Forgets: Agent Traces Are the Memory
huggingface
• • 8 upvoted 2
collections 6 days ago