Tag
Explore by tags
Unified framework for few-shot evaluation of generative language models across 60+ academic benchmarks. Supports multiple model backends (Hugging Face, vLLM, APIs, local servers), configurable prompts/YAML configs, and reproducible exports for leaderboards and research comparisons.
Serves predictive and generative ML models on Kubernetes via a single InferenceService CRD, with scale-to-zero, canary rollouts, and an OpenAI-compatible LLM path on vLLM. One autoscaling abstraction over PyTorch, XGBoost, ONNX, and HuggingFace.
Contains 40,000 teacher-generated reasoning traces distilled from the Qwen3.8-27B model for supervised fine-tuning and analysis. Covers code, math, science and logic; each example pairs a <think> chain-of-thought with a final response and is distributed in JSONL/Parquet for SFT workflows.
Open-source LLM inference and serving engine built around PagedAttention, which manages the KV cache like OS virtual memory to cut waste and raise throughput. Supports continuous batching, KV cache sharing, quantization, and an OpenAI-compatible API.
Ensures LLM outputs match precise, schema-defined structures by enforcing Python types, Pydantic models, JSON schemas or grammars at generation time. Provider-agnostic integrations (OpenAI, transformers, vLLM, Ollama, etc.) and function-call style mapping reduce brittle post-processing and make structured generation reliable for production pipelines.
Puts OpenAI-, Anthropic- and Ollama-compatible endpoints in front of 60+ inference backends, so existing client code runs unchanged against local models for text, vision, audio, image and embeddings. Runs CPU-only or accelerated, data stays local.
Fine-tunes 100+ LLMs and VLMs from one config file or a no-code web UI, unifying LoRA, QLoRA, full tuning, DPO, PPO, KTO and ORPO behind a single interface. Bundles GaLore, Unsloth, FlashAttention-2 and 2-8bit quantization to fit a single 24GB GPU.
Run any open-source LLM, embedding, speech, image, or multimodal model behind one OpenAI-compatible API — swap GPT for an open model in a single line. Routes across vLLM, llama.cpp, GGML, and TensorRT, scaling from a laptop to a multi-node GPU cluster.
Trains LLMs with RLHF at scale by splitting actor, critic, reward, and reference models across separate GPU groups via Ray, with vLLM-accelerated generation and DeepSpeed ZeRO-3. Supports PPO, GRPO, REINFORCE++, DPO, plus async and agentic multi-turn RL.
Builds custom AI inference servers in pure Python on top of FastAPI, keeping full control over request logic while batching, GPU autoscaling, streaming, and OpenAI-spec endpoints come built in. Claims a 2x+ throughput edge over plain FastAPI.
Performs document OCR, layout analysis, reading-order detection and table recognition across 90+ languages using a ~650M-parameter vision–language model; offers per-page and per-block modes and supports GPU (vllm) and CPU/Apple Silicon backends.
Pocket-sized multimodal LLM for efficient image- and video-understanding on mobile and edge devices, featuring mixed 4x/16x visual-token compression (MiniCPM‑V 4.6), compact 1.3B variants, and ready guides for iOS/Android/HarmonyOS deployment.