Kimi-K2.6 · original · l12_default

L1 P1 1_Square_matrix_multiplication_.

correct 7.39x 0.4μs

L1 P2 2_Standard_matrix_multiplication

error —

L1 P3 3_Batched_matrix_multiplication.

error —

L1 P4 4_Matrix_vector_multiplication_.

error —

L1 P5 5_Matrix_scalar_multiplication.p

error —

L1 P6 6_Matmul_with_large_K_dimension_

error —

L1 P7 7_Matmul_with_small_K_dimension_

error —

L1 P8 8_Matmul_with_irregular_shapes_.

error —

L1 P9 9_Tall_skinny_matrix_multiplicat

correct 0.43x 6.1μs

L1 P10 10_3D_tensor_matrix_multiplicati

correct 1.00x 1.0μs

L1 P11 11_4D_tensor_matrix_multiplicati

error —

L1 P12 12_Matmul_with_diagonal_matrices

correct 1.27x 0.1μs

L1 P13 13_Matmul_for_symmetric_matrices

error —

L1 P14 14_Matmul_for_upper_triangular_m

correct 1.02x 2.6μs

L1 P15 15_Matmul_for_lower_triangular_m

correct 1.02x 2.6μs

L1 P16 16_Matmul_with_transposed_A.py

correct 5.05x 0.5μs

L1 P17 17_Matmul_with_transposed_B.py

correct 1.00x 2.7μs

L1 P18 18_Matmul_with_transposed_both.p

error —

L1 P19 19_ReLU.py

correct 0.93x 4.6μs

L1 P20 20_LeakyReLU.py

correct 0.92x 4.7μs

L1 P21 21_Sigmoid.py

correct 0.92x 4.7μs

L1 P22 22_Tanh.py

correct 0.93x 4.6μs

L1 P23 23_Softmax.py

error —

L1 P24 24_LogSoftmax.py

error —

L1 P25 25_Swish.py

error —

L1 P26 26_GELU_.py

correct 0.59x 7.3μs

L1 P27 27_SELU_.py

correct 0.99x 4.3μs

L1 P28 28_HardSigmoid.py

correct 0.98x 4.4μs

L1 P29 29_Softplus.py

error —

L1 P30 30_Softsign.py

correct 3.22x 4.6μs

L1 P31 31_ELU.py

correct 0.93x 4.6μs

L1 P32 32_HardTanh.py

correct 0.93x 4.6μs

L1 P33 33_BatchNorm.py

correct 0.30x 29.2μs

L1 P34 34_InstanceNorm.py

error —

L1 P35 35_GroupNorm_.py

error —

L1 P36 36_RMSNorm_.py

correct 1.29x 11.0μs

L1 P37 37_FrobeniusNorm_.py

error —

L1 P38 38_L1Norm_.py

incorrect —

L1 P39 39_L2Norm_.py

incorrect —

L1 P40 40_LayerNorm.py

correct 19.64x 0.4μs

L1 P41 41_Max_Pooling_1D.py

correct 4.01x 3.9μs

L1 P42 42_Max_Pooling_2D.py

correct 1.55x 7.0μs

L1 P43 43_Max_Pooling_3D.py

correct 1.49x 2.6μs

L1 P44 44_Average_Pooling_1D.py

correct 3.64x 1.5μs

L1 P45 45_Average_Pooling_2D.py

error —

L1 P46 46_Average_Pooling_3D.py

correct 1.51x 5.7μs

L1 P47 47_Sum_reduction_over_a_dimensio

correct 0.86x 3.5μs

L1 P48 48_Mean_reduction_over_a_dimensi

correct 0.90x 3.4μs

L1 P49 49_Max_reduction_over_a_dimensio

error —

L1 P50 50_conv_standard_2D__square_inpu

error —

L1 P51 51_Argmax_over_a_dimension.py

correct 0.86x 3.8μs

L1 P52 52_Argmin_over_a_dimension.py

correct 0.91x 3.6μs

L1 P53 53_Min_reduction_over_a_dimensio

correct 1.62x 3.6μs

L1 P54 54_conv_standard_3D__square_inpu

correct 0.24x 7.8μs

L1 P55 55_conv_standard_2D__asymmetric_

error —

L1 P56 56_conv_standard_2D__asymmetric_

error —

L1 P57 57_conv_transposed_2D__square_in

error —

L1 P58 58_conv_transposed_3D__asymmetri

error —

L1 P59 59_conv_standard_3D__asymmetric_

error —

L1 P60 60_conv_standard_3D__square_inpu

correct 0.18x 55.0μs

L1 P61 61_conv_transposed_3D__square_in

error —

L1 P62 62_conv_standard_2D__square_inpu

error —

L1 P63 63_conv_standard_2D__square_inpu

incorrect —

L1 P64 64_conv_transposed_1D.py

error —

L1 P65 65_conv_transposed_2D__square_in

error —

L1 P66 66_conv_standard_3D__asymmetric_

error —

L1 P67 67_conv_standard_1D.py

error —

L1 P68 68_conv_transposed_3D__square_in

error —

L1 P69 69_conv_transposed_2D__asymmetri

correct 0.97x 3.0μs

L1 P70 70_conv_transposed_3D__asymmetri

error —

L1 P71 71_conv_transposed_2D__asymmetri

error —

L1 P72 72_conv_transposed_3D_asymmetric

error —

L1 P73 73_conv_transposed_3D_asymmetric

correct 0.99x 2.1μs

L1 P74 74_conv_transposed_1D_dilated.py

error —

L1 P75 75_conv_transposed_2D_asymmetric

error —

L1 P76 76_conv_standard_1D_dilated_stri

error —

L1 P77 77_conv_transposed_3D_square_inp

error —

L1 P78 78_conv_transposed_2D_asymmetric

error —

L1 P79 79_conv_transposed_1D_asymmetric

error —

L1 P80 80_conv_standard_2D_square_input

error —

L1 P81 81_conv_transposed_2D_asymmetric

error —

L1 P82 82_conv_depthwise_2D_square_inpu

correct 2.58x 1.0μs

L1 P83 83_conv_depthwise_2D_square_inpu

correct 3.89x 0.4μs

L1 P84 84_conv_depthwise_2D_asymmetric_

error —

L1 P85 85_conv_depthwise_2D_asymmetric_

correct 1.93x 1.2μs

L1 P86 86_conv_depthwise_separable_2D.p

error —

L1 P87 87_conv_pointwise_2D.py

error —

L1 P88 88_MinGPTNewGelu.py

error —

L1 P89 89_cumsum.py

correct 0.06x 78.8μs

L1 P90 90_cumprod.py

correct 0.20x 23.0μs

L1 P91 91_cumsum_reverse.py

correct 1.07x 10.7μs

L1 P92 92_cumsum_exclusive.py

error —

L1 P93 93_masked_cumsum.py

error —

L1 P94 94_MSELoss.py

correct 2.96x 2.9μs

L1 P95 95_CrossEntropyLoss.py

correct 2.03x 0.2μs

L1 P96 96_HuberLoss.py

error —

L1 P97 97_ScaledDotProductAttention.py

error —

L1 P98 98_KLDivLoss.py

correct 5.20x 0.8μs

L1 P99 99_TripletMarginLoss.py

error —

L1 P100 100_HingeLoss.py

correct 7.16x 1.5μs

L2 P1 1_Conv2D_ReLU_BiasAdd.py

correct 1.12x 3.9μs

L2 P2 2_ConvTranspose2d_BiasAdd_Clamp_

correct 0.95x 13.6μs

L2 P3 3_ConvTranspose3d_Sum_LayerNorm_

correct 2.85x 3.5μs

L2 P4 4_Conv2d_Mish_Mish.py

error —

L2 P5 5_ConvTranspose2d_Subtract_Tanh.

error —

L2 P6 6_Conv3d_Softmax_MaxPool_MaxPool

correct 0.86x 1.0μs

L2 P7 7_Conv3d_ReLU_LeakyReLU_GELU_Sig

correct 1.44x 4.5μs

L2 P8 8_Conv3d_Divide_Max_GlobalAvgPoo

correct 1.17x 2.2μs

L2 P9 9_Matmul_Subtract_Multiply_ReLU.

correct 1.02x 2.7μs

L2 P10 10_ConvTranspose2d_MaxPool_Hardt

correct 1.17x 8.3μs

L2 P11 11_ConvTranspose2d_BatchNorm_Tan

correct 1.10x 3.9μs

L2 P12 12_Gemm_Multiply_LeakyReLU.py

error —

L2 P13 13_ConvTranspose3d_Mean_Add_Soft

error —

L2 P14 14_Gemm_Divide_Sum_Scaling.py

error —

L2 P15 15_ConvTranspose3d_BatchNorm_Sub

error —

L2 P16 16_ConvTranspose2d_Mish_Add_Hard

correct 1.51x 7.3μs

L2 P17 17_Conv2d_InstanceNorm_Divide.py

correct 1.36x 3.6μs

L2 P18 18_Matmul_Sum_Max_AvgPool_LogSum

correct 115.83x 0.0μs

L2 P19 19_ConvTranspose2d_GELU_GroupNor

error —

L2 P20 20_ConvTranspose3d_Sum_ResidualA

correct 2.12x 1.7μs

L2 P21 21_Conv2d_Add_Scale_Sigmoid_Grou

error —

L2 P22 22_Matmul_Scale_ResidualAdd_Clam

correct 1.06x 2.7μs

L2 P23 23_Conv3d_GroupNorm_Mean.py

correct 1.39x 0.9μs

L2 P24 24_Conv3d_Min_Softmax.py

correct 0.97x 1.0μs

L2 P25 25_Conv2d_Min_Tanh_Tanh.py

correct 1.04x 6.2μs

L2 P26 26_ConvTranspose3d_Add_HardSwish

error —

L2 P27 27_Conv3d_HardSwish_GroupNorm_Me

error —

L2 P28 28_BMM_InstanceNorm_Sum_Residual

correct 1.03x 2.7μs

L2 P29 29_Matmul_Mish_Mish.py

correct 1.00x 2.7μs

L2 P30 30_Gemm_GroupNorm_Hardtanh.py

correct 1.05x 2.7μs

L2 P31 31_Conv2d_Min_Add_Multiply.py

correct 1.35x 3.8μs

L2 P32 32_Conv2d_Scaling_Min.py

correct 1.23x 6.4μs

L2 P33 33_Gemm_Scale_BatchNorm.py

error —

L2 P34 34_ConvTranspose3d_LayerNorm_GEL

correct 2.00x 5.0μs

L2 P35 35_Conv2d_Subtract_HardSwish_Max

error —

L2 P36 36_ConvTranspose2d_Min_Sum_GELU_

correct 0.28x 5.2μs

L2 P37 37_Matmul_Swish_Sum_GroupNorm.py

correct 1.20x 7.2μs

L2 P38 38_ConvTranspose3d_AvgPool_Clamp

error —

L2 P39 39_Gemm_Scale_BatchNorm.py

correct 0.92x 12.5μs

L2 P40 40_Matmul_Scaling_ResidualAdd.py

correct 1.05x 11.1μs

L2 P41 41_Gemm_BatchNorm_GELU_ReLU.py

correct 1.01x 11.5μs

L2 P42 42_ConvTranspose2d_GlobalAvgPool

error —

L2 P43 43_Conv3d_Max_LogSumExp_ReLU.py

error —

L2 P44 44_ConvTranspose2d_Multiply_Glob

error —

L2 P45 45_Gemm_Sigmoid_LogSumExp.py

correct 0.99x 8.6μs

L2 P46 46_Conv2d_Subtract_Tanh_Subtract

error —

L2 P47 47_Conv3d_Mish_Tanh.py

error —

L2 P48 48_Conv3d_Scaling_Tanh_Multiply_

correct 0.58x 5.8μs

L2 P49 49_ConvTranspose3d_Softmax_Sigmo

error —

L2 P50 50_ConvTranspose3d_Scaling_AvgPo

correct 10.04x 0.6μs

L2 P51 51_Gemm_Subtract_GlobalAvgPool_L

error —

L2 P52 52_Conv2d_Activation_BatchNorm.p

error —

L2 P53 53_Gemm_Scaling_Hardtanh_GELU.py

correct 1.10x 0.7μs

L2 P54 54_Conv2d_Multiply_LeakyReLU_GEL

correct 1.31x 4.7μs

L2 P55 55_Matmul_MaxPool_Sum_Scale.py

error —

L2 P56 56_Matmul_Sigmoid_Sum.py

error —

L2 P57 57_Conv2d_ReLU_HardSwish.py

correct 0.41x 7.3μs

L2 P58 58_ConvTranspose3d_LogSumExp_Har

correct 1.74x 4.0μs

L2 P59 59_Matmul_Swish_Scaling.py

correct 1.03x 5.3μs

L2 P60 60_ConvTranspose3d_Swish_GroupNo

incorrect —

L2 P61 61_ConvTranspose3d_ReLU_GroupNor

correct 1.13x 3.2μs

L2 P62 62_Matmul_GroupNorm_LeakyReLU_Su

correct 1.14x 2.7μs

L2 P63 63_Gemm_ReLU_Divide.py

correct 1.01x 2.7μs

L2 P64 64_Gemm_LogSumExp_LeakyReLU_Leak

correct 1.02x 5.2μs

L2 P65 65_Conv2d_AvgPool_Sigmoid_Sum.py

correct 1.20x 10.1μs

L2 P66 66_Matmul_Dropout_Softmax.py

error —

L2 P67 67_Conv2d_GELU_GlobalAvgPool.py

correct 1.27x 5.2μs

L2 P68 68_Matmul_Min_Subtract.py

error —

L2 P69 69_Conv2d_HardSwish_ReLU.py

correct 1.34x 1.3μs

L2 P70 70_Gemm_Sigmoid_Scaling_Residual

correct 1.02x 2.7μs

L2 P71 71_Conv2d_Divide_LeakyReLU.py

correct 0.96x 1.9μs

L2 P72 72_ConvTranspose3d_BatchNorm_Avg

correct 0.86x 14.2μs

L2 P73 73_Conv2d_BatchNorm_Scaling.py

error —

L2 P74 74_ConvTranspose3d_LeakyReLU_Mul

error —

L2 P75 75_Gemm_GroupNorm_Min_BiasAdd.py

correct 1.06x 2.9μs

L2 P76 76_Gemm_Add_ReLU.py

correct 1.01x 2.8μs

L2 P77 77_ConvTranspose3d_Scale_BatchNo

correct 1.00x 6.2μs

L2 P78 78_ConvTranspose3d_Max_Max_Sum.p

correct 1.00x 47.3μs

L2 P79 79_Conv3d_Multiply_InstanceNorm_

correct 1.67x 0.6μs

L2 P80 80_Gemm_Max_Subtract_GELU.py

correct 496.35x 0.0μs

L2 P81 81_Gemm_Swish_Divide_Clamp_Tanh_

correct 1.04x 2.7μs

L2 P82 82_Conv2d_Tanh_Scaling_BiasAdd_M

correct 1.91x 5.2μs

L2 P83 83_Conv3d_GroupNorm_Min_Clamp_Dr

correct 13.70x 0.3μs

L2 P84 84_Gemm_BatchNorm_Scaling_Softma

error —

L2 P85 85_Conv2d_GroupNorm_Scale_MaxPoo

correct 1.67x 1.5μs

L2 P86 86_Matmul_Divide_GELU.py

error —

L2 P87 87_Conv2d_Subtract_Subtract_Mish

correct 1.48x 5.8μs

L2 P88 88_Gemm_GroupNorm_Swish_Multiply

error —

L2 P89 89_ConvTranspose3d_MaxPool_Softm

correct 1.17x 4.9μs

L2 P90 90_Conv3d_LeakyReLU_Sum_Clamp_GE

correct 1.43x 8.0μs

L2 P91 91_ConvTranspose2d_Softmax_BiasA

correct 2.01x 4.6μs

L2 P92 92_Conv2d_GroupNorm_Tanh_HardSwi

correct 0.77x 5.8μs

L2 P93 93_ConvTranspose2d_Add_Min_GELU_

correct 1.34x 4.3μs

L2 P94 94_Gemm_BiasAdd_Hardtanh_Mish_Gr

error —

L2 P95 95_Matmul_Add_Swish_Tanh_GELU_Ha

correct 1.05x 2.7μs

L2 P96 96_ConvTranspose3d_Multiply_Max_

correct 1.20x 4.5μs

L2 P97 97_Matmul_BatchNorm_BiasAdd_Divi

correct 1.03x 2.8μs

L2 P98 98_Matmul_AvgPool_GELU_Scale_Max

error —

L2 P99 99_Matmul_GELU_Softmax.py

error —

L2 P100 100_ConvTranspose3d_Clamp_Min_Di

error —