prompt list | data_source string | ability string | answer string | raw_problem string | level int64 | type string | reward_model dict | extra_info dict | problem_idx string |
|---|---|---|---|---|---|---|---|---|---|
[
{
"content": "You are an expert CUDA engineer tasked with translating PyTorch code into highly optimized CUDA kernel code.\n\nThe CUDA code you generate will be saved in `cuda_fname` and loaded using \n↪ torch.utils.cpp_extension.load():\n```python\ncuda_fn = load(\n name=task_name,\n sources=[cuda_fname... | deterministic_inference | code | You are an expert CUDA engineer tasked with translating PyTorch code into highly optimized CUDA kernel code.
The CUDA code you generate will be saved in `cuda_fname` and loaded using
↪ torch.utils.cpp_extension.load():
```python
cuda_fn = load(
name=task_name,
sources=[cuda_fname],
extra_cuda_cflags=[""-... | -1 | {
"ground_truth": "n/a",
"style": "cuda-sandbox-v2"
} | {
"problem_id": "matmul_kernel_persistent"
} | matmul_kernel_persistent |
README.md exists but content is empty.
- Downloads last month
- 7