Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
Repository Index
47,133 repositories indexed
1–13 of 13 · Most stars
Open-source AI gateway. Self-hosted LLM gateway with OpenAI-, Claude- and Gemini-compatible APIs, model routing, access control, usage tracking and cost accounting.
TranslatedChinese LLaMA & Alpaca LLMs + local CPU/GPU training and deployment
- llm
- plm
- pre-trained-language-models
- alpaca
QLoRA: Efficient Finetuning of Quantized LLMs
A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.
- avx2
- c99
- cpu-inference
- deep-learning
Serve large Qwen models fast on the GPUs you actually own. Qwen3.8-27B on a single 24 GB card with vLLM: 127 tok/s single-user (381 when the answer quotes the prompt), ~1,035 tok/s at 64 concurrent, 150k-262k context. vLLM patches, requant pipeline, benchmarks.
- kv-cache
- llm-inference
- local-llm
- quantization
Project Maya - run GLM-5.3-Flash (321B MoE), DeepSeek-V4.1-Flash-1017 (748B MoE) and MiniMax-M3.1-Flash (428B MoE) privately on your own GPU(s), from one card to 16, NVIDIA or AMD, Linux or Windows. Engine, server, dashboard, OpenAI- and Anthropic-compatible API, quant tools.
Oracle3: open-source trading engine and MCP server for prediction markets. Finds fee-adjusted no-arbitrage violations across related event contracts and trades them live on Kalshi, Polymarket and Solana, or on paper, under pre-trade risk limits. 13 MCP tools, JSON CLI, agent skills, 600+ tests.
- algorithmic-trading
- kalshi
- polymarket
- python
Official library for RaBitQ vector quantization and approximate nearest-neighbor search, in C++ and Python.
- nearest-neighbor-search
- quantization
- vector-database
- vector-search
Fast, exact inference for StartLux-Decision models on NVIDIA GPUs: the released GGUF files on int8 and 4-bit tensor cores, behind the /v1/systemone API, with up to 5.3× lower latency than vLLM in FP8 and bit-identical answers.
- calibration
- cuda
- cuda-graphs
- decision-making
Legend of Elya — N64 game with a real 6.36M-parameter ternary transformer on the VR4300 MIPS III CPU. Zelda-style dungeon, AI NPCs, byte-level inference at 1.23 tok/s scalar / 2.19 tok/s on the RSP overlay (measured under ares, never on silicon). Built with libdragon.
Research experiments exploring uncommon quant techniques.
- analysis
- jupyter-notebook
- python
- quant
Per-repo memory, outcome telemetry, and a calibrated-confidence gate for Claude Code, with MCP and AGENTS.md projections so other AI coding tools can read its context. Notes survive sessions; success claims need test evidence; your reverts are remembered. Local-only, stdlib runtime.