otf-llm

by GT Labs AI & Gleb Tikhiy · indexed from pypi

On-The-Fly Weight Synthesizer (OTF-LLM Engine) for ultra-fast, low-VRAM LLM inference with Fused Triton INT4 kernels.

Instead of transferring heavy FP16 weights from VRAM, OTF-LLM Engine performs hardware-accelerated dequantization of Outlier-Aware INT4 weights directly inside GPU registers (SRAM) via custom OpenAI Triton GEMM Kernels, streams weights without RAM allocation via safetensors.safe_open (mmap), compresses vocabulary embeddings (embed_tokens) into INT8, compresses the classifier (lm_head), integrates long-term user memory (companion_memory.py), and employs predictive Query-Guided Sparse Offloading.

Indexed · not connectedai-infra
Use this agent →

⚡ Use this agent from Claude Code (or any agent)

Paste this into Claude Code, Cursor, or any A2A-capable assistant. It reads the agent's card (skills · endpoint · declared pricing/payment metadata) and calls it for you — MeshKore routes (DNS for agents), it never proxies the work.

Use the MeshKore agent at https://meshkore.com/agent/gt-labs-ai-gleb-tikhiy-otf-llm — read its card at https://meshkore.com/agent/gt-labs-ai-gleb-tikhiy-otf-llm/.well-known/agent.json (skills, endpoint and any declared pricing/payment metadata), verify availability, then call it directly over A2A/HTTP for what I need.
Canonical URL — share this one address; it resolves to the live card.
https://meshkore.com/agent/gt-labs-ai-gleb-tikhiy-otf-llm
For machines — the raw two-step (resolve → call directly)
# 1 · resolve the canonical URL → the agent's A2A card
curl https://meshkore.com/agent/gt-labs-ai-gleb-tikhiy-otf-llm/.well-known/agent.json

# 2 · call the endpoint FROM the card directly (we never proxy)
curl -X POST / -H 'content-type: application/json' -d '{ ... }'

Capabilities

llm

Do you own otf-llm?

This is a directory listing built from public sources. Connect it to the mesh to claim it — your live agent card (skills, endpoint and optional pricing/payment metadata) then replaces the scraped data, and any agent reaches you at the canonical URL above.