公开知识图谱记录
qwenGPU
记录摘要
Compiles a SIMT GPU into Qwen3.5-2B's weights, running real CUDA inside its forw
来自项目知识图谱的有来源实体；具体工具属性仍需补齐。
已记录信息
- 基准测试: reference interpreter matched NVIDIA Tesla T4 on 49 of 49 kernels
- 基准测试: Qwen-written CUDA passed 4 of 5
- 基准测试: reference interpreter matched an NVIDIA Tesla T4 on 49 of 49 kernels
- 基准测试: answered 20 of 24 sampled chat questions with GPU-computed numbers
- 基准测试: wrote a correct kernel first try on 4 of 5 questions
- 能力: supports integer arithmetic, predicates, divergent branches, loops
- 能力: supports shared memory, atomicAdd, __syncthreads(), warp shuffles
- 能力: supports integer arithmetic, predicates, divergent branches and loops
资料覆盖
知识图谱当前记录了 10 个已知资料维度，另有 7 个维度尚未记录，关联 1 篇公开资料。这些数字只描述资料覆盖，不代表模型性能或适配性已经验证。
相关公开记录
- Percepta — 竞品
- Qwen3.5-2B — 基于
- Percepta — 竞品
- Qwenputer — 竞品
- NVIDIA Tesla T4 — 评测于