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
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_4x2_v_bz_splitK_TNT
bf16
false
128
1,024
4,096
1
128
null
null
null
null
v_proj
0.00998
384
64
9,450,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_2x1_v_bz_TNT
bf16
false
128
4,096
4,096
1
128
null
null
null
null
o_proj
0.01709
384
128
34,680,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_112x128_64x7_2x1_v_bz_TNN
bf16
false
128
14,336
4,096
1
128
null
null
null
null
gate
0.04506
384
128
118,750,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_112x128_64x7_2x1_v_bz_TNN
bf16
false
128
14,336
4,096
1
128
null
null
null
null
up
0.04454
384
128
118,540,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_96x128_64x7_2x1_v_bz_splitK_TNN
bf16
false
128
4,096
14,336
1
128
null
null
null
null
down_proj
0.0481
384
132
121,170,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_256x128_64x4_2x1_v_bz_coopA_TNT
bf16
false
128
4,096
4,096
1
128
null
null
null
null
q_proj
0.37354
384
132
1,090,000,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
void cutlass::Kernel2<cutlass_80_simt_sgemm_128x32_8x5_nn_align1>(T1::Params)
bf16
false
128
1,024
4,096
1
128
null
null
null
null
k_proj
0.00531
128
2
30,980
80
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_2x1_v_bz_TNT
bf16
false
128
1,024
4,096
1
128
null
null
null
null
v_proj
0.01702
384
128
34,690,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_4x2_v_bz_splitK_TNT
bf16
false
128
1,024
4,096
1
128
null
null
null
null
k_proj
0.01072
384
64
9,450,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_4x2_v_bz_splitK_TNT
bf16
false
128
1,024
4,096
1
128
null
null
null
null
k_proj
0.00982
384
64
9,450,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_2x1_v_bz_TNT
bf16
false
128
4,096
4,096
1
128
null
null
null
null
o_proj
0.01683
384
128
34,650,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_112x128_64x7_2x1_v_bz_TNN
bf16
false
128
14,336
4,096
1
128
null
null
null
null
gate
0.04714
384
128
118,560,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_112x128_64x7_2x1_v_bz_TNN
bf16
false
128
14,336
4,096
1
128
null
null
null
null
up
0.04461
384
128
118,590,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_96x128_64x7_2x1_v_bz_splitK_TNN
bf16
false
128
4,096
14,336
1
128
null
null
null
null
down_proj
0.0463
384
132
121,250,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_2x1_v_bz_TNT
bf16
false
128
4,096
4,096
1
128
null
null
null
null
q_proj
0.0175
384
128
34,640,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_4x2_v_bz_splitK_TNT
bf16
false
128
1,024
4,096
1
128
null
null
null
null
k_proj
0.01062
384
64
9,450,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_4x2_v_bz_splitK_TNT
bf16
false
128
1,024
4,096
1
128
null
null
null
null
v_proj
0.00989
384
64
9,450,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_64x64_64x13_2x1_v_bz_TNT
bf16
false
128
4,096
4,096
1
128
null
null
null
null
o_proj
0.01699
384
128
34,660,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_112x128_64x7_2x1_v_bz_TNN
bf16
false
128
14,336
4,096
1
128
null
null
null
null
gate
0.04454
384
128
118,550,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_112x128_64x7_2x1_v_bz_TNN
bf16
false
128
14,336
4,096
1
128
null
null
null
null
up
0.04486
384
128
118,590,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_96x128_64x7_2x1_v_bz_splitK_TNN
bf16
false
128
4,096
14,336
1
128
null
null
null
null
down_proj
0.04762
384
132
121,200,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
gemm
nvjet_tst_256x128_64x4_2x1_v_bz_coopA_TNT
bf16
false
128
4,096
4,096
1
128
null
null
null
null
q_proj
0.37293
384
132
1,090,000,000
168
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00451
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00467
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
4,096
14,336
null
null
null
null
null
null
null
0.00723
512
512
6,300,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00467
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00445
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
4,096
14,336
null
null
null
null
null
null
null
0.00675
512
512
6,300,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00442
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00461
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
4,096
14,336
null
null
null
null
null
null
null
0.00675
512
512
6,310,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00483
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
1,024
4,096
null
null
null
null
null
null
null
0.00445
512
256
1,060,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)0, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
4,096
14,336
null
null
null
null
null
null
null
0.00678
512
512
6,300,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
flash_attn
void pytorch_flash::flash_fwd_kernel<Flash_fwd_kernel_traits<(int)128, (int)128, (int)64, (int)4, (bool)0, (bool)0, cutlass::bfloat16_t, Flash_kernel_traits<(int)128, (int)128, (int)64, (int)4, cutlass::bfloat16_t>>, (bool)0, (bool)1, (bool)0, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0>(pytorch_flash::Flash_fwd_params...
bf16
false
null
null
null
1
128
32
128
8
null
flash_attn
0.01456
128
32
1,650,000
255
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
flash_attn
void pytorch_flash::flash_fwd_kernel<Flash_fwd_kernel_traits<(int)128, (int)128, (int)64, (int)4, (bool)0, (bool)0, cutlass::bfloat16_t, Flash_kernel_traits<(int)128, (int)128, (int)64, (int)4, cutlass::bfloat16_t>>, (bool)0, (bool)1, (bool)0, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0>(pytorch_flash::Flash_fwd_params...
bf16
false
null
null
null
1
128
32
128
8
null
flash_attn
0.01478
128
32
1,650,000
255
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
flash_attn
void pytorch_flash::flash_fwd_kernel<Flash_fwd_kernel_traits<(int)128, (int)128, (int)64, (int)4, (bool)0, (bool)0, cutlass::bfloat16_t, Flash_kernel_traits<(int)128, (int)128, (int)64, (int)4, cutlass::bfloat16_t>>, (bool)0, (bool)1, (bool)0, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0>(pytorch_flash::Flash_fwd_params...
bf16
false
null
null
null
1
128
32
128
8
null
flash_attn
0.01549
128
32
1,650,000
255
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
flash_attn
void pytorch_flash::flash_fwd_kernel<Flash_fwd_kernel_traits<(int)128, (int)128, (int)64, (int)4, (bool)0, (bool)0, cutlass::bfloat16_t, Flash_kernel_traits<(int)128, (int)128, (int)64, (int)4, cutlass::bfloat16_t>>, (bool)0, (bool)1, (bool)0, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0>(pytorch_flash::Flash_fwd_params...
bf16
false
null
null
null
1
128
32
128
8
null
flash_attn
0.01478
128
32
1,650,000
255
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
other
0.00509
256
256
1,010,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::vectorized_elementwise_kernel<(int)2, at::native::CUDAFunctorOnSelf_add<long>, std::array<char *, (unsigned long)2>>(int, T2, T3)
bf16
false
null
null
null
1
128
null
null
null
512
residual
0.00314
128
1
6,660
38
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
512
cast
0.00365
128
1
34,050
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
copy
void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)4>, unsigned int, (int)3, (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
1,048,576
other
0.00528
512
528
39,680
29
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::cos_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
8,192
rope
0.00413
128
16
109,060
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::AUnaryFunctor<float, float, float, 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
8,192
mul
0.00358
128
16
70,660
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::sin_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
8,192
rope
0.00397
128
16
109,060
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::AUnaryFunctor<float, float, float, 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
8,192
mul
0.00339
128
16
70,660
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
8,192
cast
0.00342
128
16
70,660
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
8,192
cast
0.0033
128
16
70,660
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
cast
0.00605
128
1,024
1,080,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
rmsnorm
0.00413
128
512
2,100,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.0185
512
8
2,160,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.00333
128
1
5,380
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.00333
128
1
7,420
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
mul
0.00573
128
2,048
2,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
cast
0.00438
128
512
2,100,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.0063
128
1,024
1,080,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.0072
128
1,024
1,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[l...
bf16
false
null
null
null
1
128
null
null
null
262,144
neg
0.00608
128
512
547,070
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.00877
512
528
1,060,000
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.0072
128
1,024
1,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
residual
0.00726
128
1,024
2,120,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
131,072
mul
0.00704
128
256
326,400
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[l...
bf16
false
null
null
null
1
128
null
null
null
65,536
neg
0.00586
128
128
178,690
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.0071
512
528
270,340
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
131,072
mul
0.00685
128
256
321,540
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
131,072
residual
0.00675
128
256
550,660
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.0065
512
528
269,310
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.00646
512
528
269,310
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
residual
0.00429
128
512
2,100,000
34
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
cast
0.00678
128
1,024
1,080,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
rmsnorm
0.00442
128
512
2,100,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.01882
512
8
2,160,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.00339
128
1
5,890
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.00352
128
1
6,910
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
mul
0.0055
128
2,048
2,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
cast
0.00406
128
512
2,100,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.00669
128
1,024
1,080,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::vectorized_elementwise_kernel<(int)8, at::native::<unnamed>::silu_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 6)]::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
1,048,576
silu
0.00726
128
1,792
3,680,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
mul
0.00701
128
1,792
7,350,000
40
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
residual
0.0041
128
512
2,100,000
34
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
cast
0.00589
128
1,024
1,080,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
rmsnorm
0.00432
128
512
2,100,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.01827
512
8
2,160,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.00323
128
1
5,890
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
0.00384
128
1
7,170
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
mul
0.0056
128
2,048
2,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
262,144
cast
0.00406
128
512
2,100,000
32
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.00624
128
1,024
1,080,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.00742
128
1,024
1,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[l...
bf16
false
null
null
null
1
128
null
null
null
262,144
neg
0.00662
128
512
547,580
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.00858
512
528
1,060,000
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
mul
0.00694
128
1,024
1,110,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
524,288
residual
0.00669
128
1,024
2,120,000
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
131,072
mul
0.0065
128
256
321,280
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
elementwise
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 9)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[l...
bf16
false
null
null
null
1
128
null
null
null
65,536
neg
0.00627
128
128
154,110
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.00662
512
528
269,570
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
131,072
mul
0.00624
128
256
322,560
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
131,072
residual
0.00688
128
256
550,910
18
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.00621
512
528
269,310
25
Llama-3.1-8B-Instruct_prefill
H100
Llama-8B
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
1,048,576
other
0.00656
512
528
273,920
25