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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00659 | 128 | 512 | 559,100 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00688 | 128 | 512 | 559,360 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 92,160 | residual | 0.00362 | 128 | 256 | 1,060,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00874 | 512 | 528 | 1,060,000 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00666 | 128 | 64 | 100,860 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00646 | 128 | 64 | 100,860 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | residual | 0.00339 | 128 | 32 | 137,980 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00682 | 128 | 64 | 100,100 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00666 | 128 | 64 | 101,120 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | residual | 0.00346 | 128 | 32 | 137,730 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00634 | 512 | 528 | 138,240 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00576 | 512 | 528 | 138,500 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00592 | 512 | 528 | 138,500 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | cast | 0.0057 | 128 | 1,024 | 154,110 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | cast | 0.00675 | 128 | 1,024 | 154,370 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::AUnaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00442 | 128 | 1,024 | 2,100,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | residual | 0.0073 | 128 | 2,048 | 2,160,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.01341 | 512 | 528 | 2,100,000 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<c10::BFloat16, at::native::MaxOps<c10::BFloat16>, unsigned int, c10::BFloat16, (int)4, (int)4>>(T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.01034 | 512 | 512 | 3,230,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | residual | 0.0073 | 128 | 2,064 | 2,160,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void <unnamed>::softmax_warp_forward<c10::BFloat16, c10::BFloat16, float, (int)8, (bool)0, (bool)0>(T2 *, const T1 *, int, int, int, const bool *, int, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00803 | 128 | 2,048 | 2,120,000 | 29 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | cast | 0.0065 | 128 | 1,024 | 1,070,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | residual | 0.00381 | 128 | 360 | 1,480,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | cast | 0.00685 | 128 | 720 | 770,050 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, void at::native::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | rmsnorm | 0.0041 | 128 | 360 | 1,480,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<float, at::native::MeanOps<float, float, float, float>, unsigned int, float, (int)4, (int)4>>(T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | rmsnorm | 0.0152 | 512 | 8 | 1,540,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::CUDAFunctorOnSelf_add<float>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.00349 | 128 | 1 | 5,630 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.00336 | 128 | 1 | 7,170 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00515 | 128 | 1,440 | 1,490,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00816 | 128 | 720 | 1,550,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::vectorized_elementwise_kernel<(int)8, at::native::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | cast | 0.0039 | 128 | 360 | 1,480,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void <unnamed>::softmax_warp_forward<c10::BFloat16, c10::BFloat16, float, (int)2, (bool)0, (bool)0>(T2 *, const T1 *, int, int, int, const bool *, int, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,024 | other | 0.00355 | 128 | 2 | 7,680 | 21 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 4)]::operator ()() const::[lambda(long) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[la... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,024 | cast | 0.00432 | 128 | 2 | 13,570 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, void at::native::compare_scalar_kernel<long>(at::TensorIteratorBase &, at::native::<unnamed>::OpType, T1)::[lambda(long) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.00349 | 128 | 1 | 13,570 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)2, at::native::<unnamed>::launch_clamp_scalar(at::TensorIteratorBase &, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 4)]::operator ()() const::[lambda(long) (instance 1)], std::array<char... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.00355 | 128 | 1 | 13,820 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_put_kernel_impl<at::native::OpaqueType<(int)8>>(at::TensorIterator &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.0081 | 128 | 1 | 56,060 | 36 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)8>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.00723 | 128 | 1 | 56,060 | 36 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)2>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.00698 | 128 | 1 | 52,740 | 36 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)8>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.00768 | 128 | 1 | 56,060 | 36 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::vectorized_gather_kernel<(int)16, long>(char *, char *, T2 *, int, long, long, long, long, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00458 | 256 | 1,024 | 745,980 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 3)]::operator ()() const::[lambda(int) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>,... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | cast | 0.00381 | 128 | 1 | 36,860 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::FillFunctor<int>, std::array<char *, (unsigned long)1>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | fill | 0.00272 | 128 | 1 | 4,100 | 16 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::vectorized_gather_kernel<(int)16, long>(char *, char *, T2 *, int, long, long, long, long, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,572,864 | other | 0.00566 | 256 | 1,536 | 284,420 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,474,560 | residual | 0.00781 | 128 | 2,880 | 11,800,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::<unnamed>::launch_clamp_scalar(at::TensorIteratorBase &, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lamb... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,474,560 | other | 0.00883 | 128 | 2,880 | 5,920,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::<unnamed>::launch_clamp_scalar(at::TensorIteratorBase &, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lamb... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,474,560 | other | 0.00896 | 128 | 2,880 | 5,920,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::AUnaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.0048 | 128 | 1,440 | 2,960,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::sigmoid_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 4)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00573 | 128 | 1,440 | 2,970,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00672 | 128 | 1,440 | 5,900,000 | 40 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctorOnSelf_add<c10::BFloat16>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | residual | 0.0048 | 128 | 1,440 | 2,960,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00656 | 128 | 1,440 | 5,900,000 | 40 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::vectorized_gather_kernel<(int)16, long>(char *, char *, T2 *, int, long, long, long, long, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00445 | 256 | 1,024 | 146,180 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | residual | 0.00662 | 128 | 1,440 | 5,910,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,474,560 | mul | 0.0087 | 128 | 2,880 | 2,980,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)1>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | other | 0.00682 | 128 | 1 | 52,480 | 36 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::<unnamed>::masked_fill_kernel(at::TensorIterator &, const c10::Scalar &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16, bool) (instance 1)]>(at:... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 1,474,560 | other | 0.00902 | 128 | 2,880 | 2,980,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::vectorized_gather_kernel<(int)16, long>(char *, char *, T2 *, int, long, long, long, long, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00528 | 256 | 1,024 | 2,960,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void at::native::reduce_kernel<(int)128, (int)4, at::native::ReduceOp<c10::BFloat16, at::native::func_wrapper_t<c10::BFloat16, at::native::sum_functor<c10::BFloat16, float, c10::BFloat16>::operator ()(at::TensorIterator &)::[lambda(float, float) (instance 1)]>, unsigned int, c10::BFloat16, (int)4, (int)8>>(T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | other | 0.00922 | 128 | 720 | 3,020,000 | 54 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | residual | 0.00384 | 128 | 360 | 1,480,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | cast | 0.00582 | 128 | 720 | 770,050 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, void at::native::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | rmsnorm | 0.00394 | 128 | 360 | 1,480,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<float, at::native::MeanOps<float, float, float, float>, unsigned int, float, (int)4, (int)4>>(T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | rmsnorm | 0.01475 | 512 | 8 | 1,540,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::CUDAFunctorOnSelf_add<float>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.00317 | 128 | 1 | 5,630 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.00336 | 128 | 1 | 7,170 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00518 | 128 | 1,440 | 1,490,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00832 | 128 | 720 | 1,550,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::vectorized_elementwise_kernel<(int)8, at::native::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | cast | 0.0039 | 128 | 360 | 1,480,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00669 | 128 | 512 | 558,850 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00672 | 128 | 512 | 559,870 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 92,160 | residual | 0.0039 | 128 | 256 | 1,060,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00669 | 128 | 512 | 558,850 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00662 | 128 | 512 | 558,850 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 92,160 | residual | 0.00362 | 128 | 256 | 1,060,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00877 | 512 | 528 | 1,060,000 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00675 | 128 | 64 | 100,100 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00643 | 128 | 64 | 100,610 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | residual | 0.00358 | 128 | 32 | 137,470 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.0057 | 128 | 64 | 100,860 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 32,768 | mul | 0.00675 | 128 | 64 | 100,860 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | residual | 0.00349 | 128 | 32 | 137,730 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00662 | 512 | 528 | 138,750 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00682 | 512 | 528 | 138,500 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00634 | 512 | 528 | 138,240 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | cast | 0.00576 | 128 | 1,024 | 153,600 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | cast | 0.00586 | 128 | 1,024 | 154,110 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::AUnaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00435 | 128 | 1,024 | 2,100,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | residual | 0.0073 | 128 | 2,048 | 2,160,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | copy | void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)2>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.0136 | 512 | 528 | 2,100,000 | 25 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<c10::BFloat16, at::native::MaxOps<c10::BFloat16>, unsigned int, c10::BFloat16, (int)4, (int)4>>(T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.01027 | 512 | 512 | 3,230,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::CUDAFunctor_add<c10::BFloat16>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | residual | 0.00726 | 128 | 2,064 | 2,160,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void <unnamed>::softmax_warp_forward<c10::BFloat16, c10::BFloat16, float, (int)8, (bool)0, (bool)0>(T2 *, const T1 *, int, int, int, const bool *, int, bool) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | other | 0.00813 | 128 | 2,048 | 2,120,000 | 29 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | cast | 0.00602 | 128 | 1,024 | 1,070,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | residual | 0.00397 | 128 | 360 | 1,480,000 | 34 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | cast | void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int... | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | cast | 0.00589 | 128 | 720 | 769,540 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, void at::native::<unnamed>::pow_tensor_scalar_kernel_impl<float, float>(at::TensorIteratorBase &, T2)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 184,320 | rmsnorm | 0.00406 | 128 | 360 | 1,480,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | reduce | void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<float, at::native::MeanOps<float, float, float, float>, unsigned int, float, (int)4, (int)4>>(T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 16,384 | rmsnorm | 0.01533 | 512 | 8 | 1,540,000 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::CUDAFunctorOnSelf_add<float>, std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | residual | 0.0033 | 128 | 1 | 5,890 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::vectorized_elementwise_kernel<(int)4, at::native::rsqrt_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 512 | rmsnorm | 0.0033 | 128 | 1 | 7,170 | 32 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 737,280 | mul | 0.00534 | 128 | 1,440 | 1,490,000 | 18 |
gpt-oss-20b_prefill | H100 | gpt-oss-20b | elementwise | void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3) | bf16 | false | null | null | null | 1 | 128 | null | null | null | 368,640 | mul | 0.00829 | 128 | 720 | 1,550,000 | 32 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.