hkust-nlp/drkernel-8b is a Qwen3-8B-based model specialized for GPU kernel generation and optimization (especially Triton) in the DR.Kernel framework. It is trained for iterative optimization with execution feedback from KernelGYM, rather than single-shot code generation only. — Model type: Qwen3ForCausalLM — Parameter count: 8,190,735,360 (from model.safetensors.index.json) — Weight dtype: BF16 — Base model family: Qwen3-8B — Main capability: generate and iteratively refine optimized ModelNew kernel implementations from PyTorch reference tasks 1. Cold-start SFT — Dataset: hkust-nlp/drkernel-coldstart-8k — Multi-turn trajectory warm-up for kernel generation/refinement 2. Multi-turn RL — Train dataset: hkust-nlp/drkernel-rl-data — Validation dataset: hkust-nlp/drkernel-validation-data (KernelBench Level 2 validation split) — Core methods: TRLOO + MRS + PR + PRS — Execution environment: KernelGYM with compilation/correctness/performance/profiling feedback — drkernel/kernel/scripts/sft/8b-coldstart.sh — drkernel/kernel/scripts/rl/8btrloomrsprprs.sh — Kernel generation research and benchmarking — Triton kernel optimization with iterative feedback
Модальности:
Генерация текста
Области применения:
Генерация кода Диалог / чат
Задача: Генерация текста
Автор: hkust-nlp
Теги: qwen3, triton, kernel-generation, reinforcement-learning, code, conversational, text-generation-inference, endpoints_compatible
Лайков: 4 | Загрузок: 1,577
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
