Repository Index

Jev-like family of decision models built on top of Qwen3.5/3.8 you can train and run on your own

• decision-model
• jev
• qwen3

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

Implementing LLM architectures from scratch in PyTorch

• gemma3
• llama4
• llm
• moe