source stringclasses 13
values | gpu stringclasses 4
values | model stringclasses 13
values | kernel_family stringclasses 7
values | kernel_name stringclasses 328
values | dtype stringclasses 2
values | held_out bool 2
classes | M float64 1 16.4k ⌀ | N float64 8 248k ⌀ | K float64 2.05k 29.6k ⌀ | bs float64 1 1 ⌀ | seq float64 1 16.4k ⌀ | n_heads float64 8 64 ⌀ | head_dim float64 64 256 ⌀ | kv_heads float64 1 64 ⌀ | numel float64 64 531M ⌀ | op_type stringclasses 177
values | gpu_time_duration_ms float64 0 703 | launch_block_size float64 16 512 | launch_grid_size float64 1 1.04M | dram_bytes_sum float64 1.27 94.3B ⌀ | launch_registers_per_thread float64 16 255 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctorOnSelf_add<float>, std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.00323 | 128 | 1 | 5,250 | 30 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.00362 | 128 | 1 | 6,780 | 22 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 2, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<float, float, float, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.0104 | 128 | 4,096 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::vectorized_elementwise_kernel<4, at::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | cast | 0.00714 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01117 | 128 | 2,048 | 2,130,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01181 | 128 | 2,048 | 2,150,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | neg | 0.00784 | 128 | 1,024 | 1,070,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.01805 | 512 | 432 | 2,100,000 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01104 | 128 | 2,048 | 2,150,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | residual | 0.01219 | 128 | 2,048 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 131,072 | mul | 0.00723 | 128 | 256 | 314,620 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 65,536 | neg | 0.00688 | 128 | 128 | 148,860 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.00979 | 512 | 432 | 269,060 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 131,072 | mul | 0.00723 | 128 | 256 | 314,620 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 131,072 | residual | 0.00752 | 128 | 256 | 544,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.0087 | 512 | 432 | 269,180 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.00858 | 512 | 432 | 269,180 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctor_add<c10::BFloat16>, std::array<char *, 3>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | residual | 0.0072 | 128 | 1,024 | 4,200,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::unrolled_elementwise_kernel<at::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>, 4, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned in... | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | cast | 0.01277 | 128 | 2,048 | 2,130,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, void at::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | rmsnorm | 0.00803 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | reduce | void at::reduce_kernel<512, 1, at::ReduceOp<float, at::MeanOps<float, float, float, float>, unsigned int, float, 4, 4>>(T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 262,144 | rmsnorm | 0.01018 | 512 | 128 | 4,250,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctorOnSelf_add<float>, std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.00339 | 128 | 1 | 5,250 | 30 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.00358 | 128 | 1 | 6,780 | 22 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 2, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<float, float, float, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.0105 | 128 | 4,096 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::vectorized_elementwise_kernel<4, at::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | cast | 0.00714 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01123 | 128 | 2,048 | 2,130,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::<unnamed>::silu_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | silu | 0.01446 | 128 | 3,696 | 7,580,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>, std::array<char *, 3>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.01434 | 128 | 3,696 | 15,150,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctor_add<c10::BFloat16>, std::array<char *, 3>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | residual | 0.00733 | 128 | 1,024 | 4,200,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::unrolled_elementwise_kernel<at::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>, 4, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned in... | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | cast | 0.0129 | 128 | 2,048 | 2,130,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, void at::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | rmsnorm | 0.00797 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | reduce | void at::reduce_kernel<512, 1, at::ReduceOp<float, at::MeanOps<float, float, float, float>, unsigned int, float, 4, 4>>(T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 262,144 | rmsnorm | 0.01034 | 512 | 128 | 4,250,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctorOnSelf_add<float>, std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.00323 | 128 | 1 | 5,250 | 30 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.00358 | 128 | 1 | 6,780 | 22 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 2, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<float, float, float, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.01034 | 128 | 4,096 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::vectorized_elementwise_kernel<4, at::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | cast | 0.00717 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01123 | 128 | 2,048 | 2,130,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01187 | 128 | 2,048 | 2,150,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | neg | 0.0079 | 128 | 1,024 | 1,070,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.01798 | 512 | 432 | 2,100,000 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01104 | 128 | 2,048 | 2,150,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | residual | 0.0121 | 128 | 2,048 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 131,072 | mul | 0.00739 | 128 | 256 | 314,750 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 65,536 | neg | 0.00698 | 128 | 128 | 148,860 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.0097 | 512 | 432 | 269,060 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 131,072 | mul | 0.00717 | 128 | 256 | 314,620 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 131,072 | residual | 0.00739 | 128 | 256 | 544,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.00864 | 512 | 432 | 269,180 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | copy | void at::<unnamed>::CatArrayBatchedCopy<at::<unnamed>::OpaqueType<2>, unsigned int, 4, 64, 64>(T1 *, at::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::<unnamed>::TensorSizeStride<T2, 4>, int, T2) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | other | 0.00858 | 512 | 432 | 274,820 | 23 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctor_add<c10::BFloat16>, std::array<char *, 3>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | residual | 0.00726 | 128 | 1,024 | 4,200,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::unrolled_elementwise_kernel<at::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>, 4, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned in... | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | cast | 0.01258 | 128 | 2,048 | 2,130,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, void at::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | rmsnorm | 0.00806 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | reduce | void at::reduce_kernel<512, 1, at::ReduceOp<float, at::MeanOps<float, float, float, float>, unsigned int, float, 4, 4>>(T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 262,144 | rmsnorm | 0.01021 | 512 | 128 | 4,250,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctorOnSelf_add<float>, std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.00317 | 128 | 1 | 5,250 | 30 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.00358 | 128 | 1 | 6,780 | 22 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 2, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<float, float, float, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.0104 | 128 | 4,096 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::vectorized_elementwise_kernel<4, at::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | cast | 0.00714 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01098 | 128 | 2,048 | 2,130,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::<unnamed>::silu_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | silu | 0.01427 | 128 | 3,696 | 7,580,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>, std::array<char *, 3>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.01443 | 128 | 3,696 | 15,150,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctor_add<c10::BFloat16>, std::array<char *, 3>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | residual | 0.00726 | 128 | 1,024 | 4,200,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::unrolled_elementwise_kernel<at::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>, 4, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned in... | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | cast | 0.0128 | 128 | 2,048 | 2,130,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, void at::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | rmsnorm | 0.00778 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | reduce | void at::reduce_kernel<512, 1, at::ReduceOp<float, at::MeanOps<float, float, float, float>, unsigned int, float, 4, 4>>(T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 262,144 | rmsnorm | 0.01053 | 512 | 128 | 4,250,000 | 32 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::CUDAFunctorOnSelf_add<float>, std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.0031 | 128 | 1 | 5,250 | 30 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::vectorized_elementwise_kernel<4, at::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.0033 | 128 | 1 | 6,780 | 22 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 2, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<float, float, float, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 2,097,152 | mul | 0.0104 | 128 | 4,096 | 4,210,000 | 16 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | cast | void at::vectorized_elementwise_kernel<4, at::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, 2>>(int, T2, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 524,288 | cast | 0.00707 | 128 | 1,024 | 4,200,000 | 28 |
Qwen2.5-72B-Instruct_prefill | A100 | Qwen-72B | elementwise | void at::elementwise_kernel<128, 4, void at::gpu_kernel_impl_nocast<at::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | true | null | null | null | 1 | 128 | null | null | null | 1,048,576 | mul | 0.01117 | 128 | 2,048 | 2,130,000 | 16 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.0065 | 64 | 6 | 30,590 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_128x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.0999 | 128 | 80 | 106,190,000 | 246 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_128x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | v_proj | 0.06717 | 128 | 96 | 64,250,000 | 246 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void cutlass::Kernel2<cutlass_80_tensorop_s16816gemm_bf16_64x64_64x6_tn_align8>(T1::Params) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.00803 | 128 | 54 | 1,820,000 | 91 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void cutlass::Kernel2<cutlass_80_tensorop_s16816gemm_bf16_64x64_64x6_tn_align8>(T1::Params) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.00829 | 128 | 54 | 1,820,000 | 91 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.04 | 256 | 96 | 6,310,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.02195 | 64 | 384 | 4,750,000 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.02243 | 64 | 384 | 4,750,000 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03942 | 256 | 48 | 3,170,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03936 | 256 | 48 | 4,740,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03936 | 256 | 48 | 4,740,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.01962 | 64 | 192 | 2,390,000 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nt | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.02432 | 256 | 48 | 3,170,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03942 | 256 | 48 | 3,170,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03939 | 256 | 48 | 4,740,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03946 | 256 | 48 | 4,740,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.01942 | 64 | 192 | 2,390,000 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nt | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.02432 | 256 | 48 | 3,170,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_128x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.06947 | 128 | 80 | 64,510,000 | 246 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_256x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.16694 | 256 | 68 | 179,590,000 | 238 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_256x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.1671 | 256 | 68 | 179,590,000 | 238 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_256x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.1751 | 256 | 80 | 182,740,000 | 238 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_128x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.10138 | 128 | 80 | 106,190,000 | 246 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_bf16_s16816gemm_bf16_128x128_ldg8_f2f_stages_64x3_tn | bf16 | false | 128 | 5,112 | 5,120 | 1 | 128 | null | null | null | null | q_proj | 0.06563 | 128 | 96 | 64,250,000 | 246 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void cutlass::Kernel2<cutlass_80_tensorop_s16816gemm_bf16_64x64_64x6_tn_align8>(T1::Params) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.00829 | 128 | 54 | 1,820,000 | 91 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void cutlass::Kernel2<cutlass_80_tensorop_s16816gemm_bf16_64x64_64x6_tn_align8>(T1::Params) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.0081 | 128 | 54 | 1,820,000 | 91 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03994 | 256 | 96 | 6,310,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.02234 | 64 | 384 | 4,750,000 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | void magma_sgemmEx_kernel<float, float, float, 0, 0, 6, 3, 5, 3, 3>(int, int, int, Tensor, int, Tensor, int, Tensor, int, Tensor, int, int, int, const T1 *, const T1 *, T1, T1, int, cublasLtEpilogue_t, int, const void *, long) | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.02205 | 64 | 384 | 4,750,000 | 80 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_tn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.03933 | 256 | 48 | 3,170,000 | 118 |
Qwen3.5-27B_prefill | A100 | Qwen3.5-27B | gemm | ampere_sgemm_128x128_nn | bf16 | false | 128 | 852 | 5,120 | 1 | 128 | null | null | null | null | k_proj | 0.0392 | 256 | 48 | 4,740,000 | 118 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.