Created by: wangchaochaohu
code generator: graph
I0420 10:18:34.893874 66342 fusion_group_pass.cc:56] subgraph: {
Node(sigmoid_1.tmp_0{32x128}), inputs:{sigmoid}, outputs:{elementwise_mul, elementwise_mul_grad}
Node(tanh_0.tmp_0{32x128}), inputs:{tanh}, outputs:{elementwise_mul, elementwise_mul_grad}
Node(tmp_3{32x128}), inputs:{elementwise_mul}, outputs:{elementwise_add, elementwise_add_grad}
Node(tanh_1.tmp_0{32x128}), inputs:{tanh}, outputs:{elementwise_add, elementwise_add_grad, tanh_grad}
Node(sigmoid_2.tmp_0{32x128}), inputs:{sigmoid}, outputs:{elementwise_add, elementwise_add_grad}
Node(assign_0.tmp_0{32x128}), inputs:{assign}, outputs:{elementwise_mul, elementwise_mul_grad}
Node(sigmoid_0.tmp_0{32x128}), inputs:{sigmoid}, outputs:{elementwise_mul, elementwise_mul_grad}
Node(tmp_2{32x128}), inputs:{elementwise_mul}, outputs:{elementwise_add, elementwise_add_grad}
Node(tmp_5@GRAD{32x128}), inputs:{mean_grad}, outputs:{elementwise_add_grad}
Node(Op(elementwise_add_grad), inputs:{Out@GRAD[tmp_5@GRAD], X[tanh_1.tmp_0], Y[sigmoid_2.tmp_0]}, outputs:{X@GRAD[tanh_1.tmp_0@GRAD], Y@GRAD[sigmoid_2.tmp_0@GRAD]}), inputs:{tmp_5@GRAD, tanh_1.tmp_0, sigmoid_2.tmp_0}, outputs:{tanh_1.tmp_0@GRAD, sigmoid_2.tmp_0@GRAD}.
Node(tanh_1.tmp_0@GRAD{32x128}), inputs:{elementwise_add_grad}, outputs:{tanh_grad}
Node(sigmoid_2.tmp_0@GRAD{32x128}), inputs:{elementwise_add_grad}, outputs:{}
Node(Op(tanh_grad), inputs:{Out[tanh_1.tmp_0], Out@GRAD[tanh_1.tmp_0@GRAD]}, outputs:{X@GRAD[tmp_4@GRAD]}), inputs:{tanh_1.tmp_0, tanh_1.tmp_0@GRAD}, outputs:{tmp_4@GRAD}.
Node(tmp_4@GRAD{32x128}), inputs:{tanh_grad}, outputs:{elementwise_add_grad}
Node(Op(elementwise_add_grad), inputs:{Out@GRAD[tmp_4@GRAD], X[tmp_2], Y[tmp_3]}, outputs:{X@GRAD[tmp_2@GRAD], Y@GRAD[tmp_3@GRAD]}), inputs:{tmp_4@GRAD, tmp_2, tmp_3}, outputs:{tmp_2@GRAD, tmp_3@GRAD}.
Node(tmp_2@GRAD{32x128}), inputs:{elementwise_add_grad}, outputs:{elementwise_mul_grad}
Node(tmp_3@GRAD{32x128}), inputs:{elementwise_add_grad}, outputs:{elementwise_mul_grad}
Node(Op(elementwise_mul_grad), inputs:{Out@GRAD[tmp_3@GRAD], X[sigmoid_1.tmp_0], Y[tanh_0.tmp_0]}, outputs:{X@GRAD[sigmoid_1.tmp_0@GRAD], Y@GRAD[tanh_0.tmp_0@GRAD]}), inputs:{tmp_3@GRAD, sigmoid_1.tmp_0, tanh_0.tmp_0}, outputs:{sigmoid_1.tmp_0@GRAD, tanh_0.tmp_0@GRAD}.
Node(Op(elementwise_mul_grad), inputs:{Out@GRAD[tmp_2@GRAD], X[assign_0.tmp_0], Y[sigmoid_0.tmp_0]}, outputs:{X@GRAD[assign_0.tmp_0@GRAD], Y@GRAD[sigmoid_0.tmp_0@GRAD]}), inputs:{tmp_2@GRAD, assign_0.tmp_0, sigmoid_0.tmp_0}, outputs:{assign_0.tmp_0@GRAD, sigmoid_0.tmp_0@GRAD}.
Node(sigmoid_1.tmp_0@GRAD{32x128}), inputs:{elementwise_mul_grad}, outputs:{}
Node(tanh_0.tmp_0@GRAD{32x128}), inputs:{elementwise_mul_grad}, outputs:{}
Node(assign_0.tmp_0@GRAD{32x128}), inputs:{elementwise_mul_grad}, outputs:{}
Node(sigmoid_0.tmp_0@GRAD{32x128}), inputs:{elementwise_mul_grad}, outputs:{}
}code
extern "C" __global__ void FusedElementwise9(int N, const float* __restrict__ arg0, const float* __restrict__ arg1, const float* __restrict__ arg2, const float* __restrict__ arg3, const float* __restrict__ arg4, const float* __restrict__ arg5, const float* __restrict__ arg6, const float* __restrict__ arg7, const float* __restrict__ arg8, float* arg10, float* arg14, float* arg15, float* arg16, float* arg17) {
for(int idx = blockIdx.x * blockDim.x + threadIdx.x;
idx < N;
idx += gridDim.x * blockDim.x) {
float tmp0 = __ldg(&arg0[idx]);
float tmp1 = __ldg(&arg1[idx]);
float tmp3 = __ldg(&arg3[idx]);
float tmp5 = __ldg(&arg5[idx]);
float tmp6 = __ldg(&arg6[idx]);
float tmp8 = __ldg(&arg8[idx]);
float tmp9 = tmp8;
float tmp10 = tmp8;
float tmp11 = tmp9 * (1.0 - tmp3 * tmp3);
float tmp12 = tmp11;
float tmp13 = tmp11;
float tmp14 = tmp13 * tmp1;
float tmp15 = tmp13 * tmp0;
float tmp16 = tmp12 * tmp6;
float tmp17 = tmp12 * tmp5;
arg10[idx] = tmp10;
arg14[idx] = tmp14;
arg15[idx] = tmp15;
arg16[idx] = tmp16;
arg17[idx] = tmp17;
}
}develop
extern "C" __global__ void FusedElementwise2(int N, float* arg0, float* arg1, float* arg2, float* arg3, float* arg4, float* arg5, float* arg6, float* arg7, float* arg8, float* arg9, float* arg10, float* arg11, float* arg12, float* arg13, float* arg14, float* arg15, float* arg16, float* arg17) {
for(int idx = blockIdx.x * blockDim.x + threadIdx.x;
idx < N;
idx += gridDim.x * blockDim.x) {
float tmp1 = arg1[idx];
float tmp3 = arg3[idx];
float tmp4 = arg4[idx];
float tmp5 = arg5[idx];
float tmp6 = arg6[idx];
float tmp8 = arg8[idx];
float tmp9 = tmp4;
float tmp10 = tmp4;
float tmp11 = tmp9 * (1.0 - tmp1 * tmp1);
float tmp12 = tmp11;
float tmp13 = tmp11;
float tmp16 = tmp13 * tmp5;
float tmp17 = tmp13 * tmp8;
float tmp14 = tmp12 * tmp6;
float tmp15 = tmp12 * tmp3;
arg9[idx] = tmp9;
arg10[idx] = tmp10;
arg11[idx] = tmp11;
arg12[idx] = tmp12;
arg13[idx] = tmp13;
arg14[idx] = tmp14;
arg15[idx] = tmp15;
arg16[idx] = tmp16;
arg17[idx] = tmp17;
}
}ResNet50 Memory:(BS=128) without fusion_group
I0403 11:39:16.828281 76940 parallel_executor.cc:481] The Program will be executed on CUDA using ParallelExecutor, 1 cards are used, so 1 programs are executed in parallel.
I0403 11:39:16.888053 76940 graph_pattern_detector.cc:101] --- detected 33 subgraphs
I0403 11:39:17.092631 76940 graph_pattern_detector.cc:101] --- detected 33 subgraphs
I0403 11:39:17.181783 76940 graph_pattern_detector.cc:101] --- detected 16 subgraphs
I0403 11:39:17.198555 76940 graph_pattern_detector.cc:101] --- detected 16 subgraphs
I0403 11:39:17.279995 76940 build_strategy.cc:376] SeqOnlyAllReduceOps:0, num_trainers:1
I0403 11:39:17.407210 76940 parallel_executor.cc:333] Inplace strategy is enabled, when build_strategy.enable_inplace = True
I0403 11:39:17.442723 76940 parallel_executor.cc:401] Garbage collection strategy is enabled, when FLAGS_eager_delete_tensor_gb = 1
2020-04-03 11:39:17,695-INFO: [Pass 0, train batch 0] loss 58.28773, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.9014 sec
2020-04-03 11:39:19,361-INFO: [Pass 0, train batch 10] loss 37.40405, acc1 0.00781, acc5 0.01562, lr 0.10000, elapse 0.1647 sec
2020-04-03 11:39:21,045-INFO: [Pass 0, train batch 20] loss 35.53636, acc1 0.00000, acc5 0.02344, lr 0.10000, elapse 0.1616 sec
2020-04-03 11:39:22,662-INFO: [Pass 0, train batch 30] loss 30.75803, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1630 sec
2020-04-03 11:39:24,269-INFO: [Pass 0, train batch 40] loss 22.00988, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1612 sec
2020-04-03 11:39:25,881-INFO: [Pass 0, train batch 50] loss 12.28711, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1619 sec
2020-04-03 11:39:27,615-INFO: [Pass 0, train batch 60] loss 8.07491, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1659 sec
2020-04-03 11:39:29,239-INFO: [Pass 0, train batch 70] loss 7.84996, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1597 sec
2020-04-03 11:39:30,850-INFO: [Pass 0, train batch 80] loss 7.88690, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1618 sec
2020-04-03 11:39:32,464-INFO: [Pass 0, train batch 90] loss 7.37603, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1610 sec
2020-04-03 11:39:34,081-INFO: [Pass 0, train batch 100] loss 7.40630, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1616 sec
2020-04-03 11:39:35,704-INFO: [Pass 0, train batch 110] loss 8.60868, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1616 sec
2020-04-03 11:39:37,321-INFO: [Pass 0, train batch 120] loss 7.49577, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1630 sec
2020-04-03 11:39:38,957-INFO: [Pass 0, train batch 130] loss 7.61791, acc1 0.00000, acc5 0.01562, lr 0.10000, elapse 0.1648 sec
2020-04-03 11:39:40,568-INFO: [Pass 0, train batch 140] loss 7.55739, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1604 sec
2020-04-03 11:39:42,205-INFO: [Pass 0, train batch 150] loss 7.34031, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1631 sec
2020-04-03 11:39:43,821-INFO: [Pass 0, train batch 160] loss 7.34924, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1611 sec
2020-04-03 11:39:45,435-INFO: [Pass 0, train batch 170] loss 7.53807, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1627 sec
2020-04-03 11:39:47,072-INFO: [Pass 0, train batch 180] loss 7.39394, acc1 0.00781, acc5 0.02344, lr 0.10000, elapse 0.1611 sec
2020-04-03 11:39:48,696-INFO: [Pass 0, train batch 190] loss 7.38567, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1639 sec
2020-04-03 11:39:50,317-INFO: [Pass 0, train batch 200] loss 8.64147, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1607 sec
2020-04-03 11:39:51,938-INFO: [Pass 0, train batch 210] loss 7.43214, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1617 sec
2020-04-03 11:39:53,554-INFO: [Pass 0, train batch 220] loss 7.34309, acc1 0.00000, acc5 0.01562, lr 0.10000, elapse 0.1619 sec
2020-04-03 11:39:55,174-INFO: [Pass 0, train batch 230] loss 7.40224, acc1 0.00781, acc5 0.01562, lr 0.10000, elapse 0.1711 sec
2020-04-03 11:39:56,786-INFO: [Pass 0, train batch 240] loss 7.90558, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1612 sec
2020-04-03 11:39:58,426-INFO: [Pass 0, train batch 250] loss 7.40587, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1619 sec
2020-04-03 11:40:00,059-INFO: [Pass 0, train batch 260] loss 7.36620, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1616 sec
2020-04-03 11:40:01,684-INFO: [Pass 0, train batch 270] loss 7.76474, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1630 sec
2020-04-03 11:40:03,296-INFO: [Pass 0, train batch 280] loss 7.36287, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1602 sec
2020-04-03 11:40:04,904-INFO: [Pass 0, train batch 290] loss 7.39509, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1609 sec
2020-04-03 11:40:06,510-INFO: [Pass 0, train batch 300] loss 8.02802, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1605 sec
2020-04-03 11:40:08,196-INFO: [Pass 0, train batch 310] loss 7.82140, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1630 sec
2020-04-03 11:40:09,819-INFO: [Pass 0, train batch 320] loss 7.41313, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1612 sec
2020-04-03 11:40:11,461-INFO: [Pass 0, train batch 330] loss 7.74391, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1667 sec
2020-04-03 11:40:13,089-INFO: [Pass 0, train batch 340] loss 7.72739, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1634 sec
2020-04-03 11:40:14,711-INFO: [Pass 0, train batch 350] loss 7.33674, acc1 0.01562, acc5 0.01562, lr 0.10000, elapse 0.1613 secdevelop with fusion group:
speed
I0420 07:50:40.972506 108885 fusion_group_pass.cc:36] Detect 32 elementwise fusion groups.
2020-04-20 07:50:41,567-INFO: [Pass 0, train batch 0] loss 58.35946, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 8.1403 sec
2020-04-20 07:50:43,376-INFO: [Pass 0, train batch 10] loss 41.27401, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1948 sec
2020-04-20 07:50:45,259-INFO: [Pass 0, train batch 20] loss 45.37650, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1811 sec
2020-04-20 07:50:47,071-INFO: [Pass 0, train batch 30] loss 42.26008, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1727 sec
2020-04-20 07:50:48,845-INFO: [Pass 0, train batch 40] loss 27.94326, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1724 sec
2020-04-20 07:50:50,590-INFO: [Pass 0, train batch 50] loss 16.90263, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1726 sec
2020-04-20 07:50:52,349-INFO: [Pass 0, train batch 60] loss 9.08802, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1772 sec
2020-04-20 07:50:54,191-INFO: [Pass 0, train batch 70] loss 7.46220, acc1 0.00781, acc5 0.01562, lr 0.10000, elapse 0.1848 sec
2020-04-20 07:50:56,104-INFO: [Pass 0, train batch 80] loss 7.42559, acc1 0.00000, acc5 0.01562, lr 0.10000, elapse 0.1744 sec
2020-04-20 07:50:57,995-INFO: [Pass 0, train batch 90] loss 7.34507, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1734 sec
2020-04-20 07:50:59,815-INFO: [Pass 0, train batch 100] loss 7.41156, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1815 sec
2020-04-20 07:51:01,643-INFO: [Pass 0, train batch 110] loss 7.40986, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1735 sec
2020-04-20 07:51:03,443-INFO: [Pass 0, train batch 120] loss 7.85518, acc1 0.00000, acc5 0.02344, lr 0.10000, elapse 0.1727 sec
2020-04-20 07:51:05,287-INFO: [Pass 0, train batch 130] loss 7.41322, acc1 0.01562, acc5 0.01562, lr 0.10000, elapse 0.1751 sec
2020-04-20 07:51:07,112-INFO: [Pass 0, train batch 140] loss 7.39413, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.2064 sec
2020-04-20 07:51:08,908-INFO: [Pass 0, train batch 150] loss 7.34854, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1745 sec
2020-04-20 07:51:10,649-INFO: [Pass 0, train batch 160] loss 7.37619, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1737 sec
2020-04-20 07:51:12,389-INFO: [Pass 0, train batch 170] loss 7.36869, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1736 sec
2020-04-20 07:51:14,230-INFO: [Pass 0, train batch 180] loss 7.66026, acc1 0.01562, acc5 0.02344, lr 0.10000, elapse 0.1809 sec
2020-04-20 07:51:16,205-INFO: [Pass 0, train batch 190] loss 7.76989, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.2118 sec
2020-04-20 07:51:18,048-INFO: [Pass 0, train batch 200] loss 7.63018, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1853 sec
2020-04-20 07:51:19,829-INFO: [Pass 0, train batch 210] loss 7.36925, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1708 sec
2020-04-20 07:51:21,713-INFO: [Pass 0, train batch 220] loss 7.35503, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1850 sec
2020-04-20 07:51:23,582-INFO: [Pass 0, train batch 230] loss 7.47642, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.2238 sec
2020-04-20 07:51:25,521-INFO: [Pass 0, train batch 240] loss 7.36822, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1909 sec
2020-04-20 07:51:27,313-INFO: [Pass 0, train batch 250] loss 7.36916, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1748 sec
2020-04-20 07:51:29,134-INFO: [Pass 0, train batch 260] loss 7.66580, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1734 sec
2020-04-20 07:51:30,877-INFO: [Pass 0, train batch 270] loss 7.35389, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1744 sec
this pr with fusion group
speed:
I0420 07:37:51.626042 102071 graph_pattern_detector.cc:101] --- detected 33 subgraphs
I0420 07:37:51.824518 102071 graph_pattern_detector.cc:101] --- detected 33 subgraphs
I0420 07:37:59.413408 102071 fusion_group_pass.cc:36] Detect 32 elementwise fusion groups.
2020-04-20 07:38:00,071-INFO: [Pass 0, train batch 0] loss 57.01205, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 8.5414 sec
2020-04-20 07:38:01,711-INFO: [Pass 0, train batch 10] loss 22.20854, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1618 sec
2020-04-20 07:38:03,335-INFO: [Pass 0, train batch 20] loss 19.15162, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1574 sec
2020-04-20 07:38:05,019-INFO: [Pass 0, train batch 30] loss 13.94458, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1713 sec
2020-04-20 07:38:06,685-INFO: [Pass 0, train batch 40] loss 10.60628, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1652 sec
2020-04-20 07:38:08,333-INFO: [Pass 0, train batch 50] loss 9.11628, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1645 sec
2020-04-20 07:38:09,945-INFO: [Pass 0, train batch 60] loss 7.43096, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1627 sec
2020-04-20 07:38:11,536-INFO: [Pass 0, train batch 70] loss 7.36789, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1575 sec
2020-04-20 07:38:13,124-INFO: [Pass 0, train batch 80] loss 7.35075, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1582 sec
2020-04-20 07:38:14,728-INFO: [Pass 0, train batch 90] loss 7.36890, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1604 sec
2020-04-20 07:38:16,361-INFO: [Pass 0, train batch 100] loss 7.79940, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1691 sec
2020-04-20 07:38:17,966-INFO: [Pass 0, train batch 110] loss 7.52226, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1568 sec
2020-04-20 07:38:19,578-INFO: [Pass 0, train batch 120] loss 7.58882, acc1 0.00000, acc5 0.01562, lr 0.10000, elapse 0.1596 sec
2020-04-20 07:38:21,168-INFO: [Pass 0, train batch 130] loss 7.56549, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1570 sec
2020-04-20 07:38:22,782-INFO: [Pass 0, train batch 140] loss 7.39170, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1597 sec
2020-04-20 07:38:24,377-INFO: [Pass 0, train batch 150] loss 7.35504, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1580 sec
2020-04-20 07:38:25,978-INFO: [Pass 0, train batch 160] loss 7.35626, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1593 sec
2020-04-20 07:38:27,590-INFO: [Pass 0, train batch 170] loss 7.35559, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1613 sec
2020-04-20 07:38:29,176-INFO: [Pass 0, train batch 180] loss 7.35849, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1579 sec
2020-04-20 07:38:30,805-INFO: [Pass 0, train batch 190] loss 7.45062, acc1 0.00781, acc5 0.00781, lr 0.10000, elapse 0.1638 sec
2020-04-20 07:38:32,392-INFO: [Pass 0, train batch 200] loss 7.45199, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1574 sec
2020-04-20 07:38:33,985-INFO: [Pass 0, train batch 210] loss 7.35009, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1587 sec
2020-04-20 07:38:35,579-INFO: [Pass 0, train batch 220] loss 7.35454, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1589 sec
2020-04-20 07:38:37,160-INFO: [Pass 0, train batch 230] loss 7.35036, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1579 sec
2020-04-20 07:38:38,813-INFO: [Pass 0, train batch 240] loss 7.66524, acc1 0.00000, acc5 0.00781, lr 0.10000, elapse 0.1589 sec
2020-04-20 07:38:40,396-INFO: [Pass 0, train batch 250] loss 7.47834, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1583 sec
2020-04-20 07:38:41,992-INFO: [Pass 0, train batch 260] loss 7.36098, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1587 sec
2020-04-20 07:38:43,582-INFO: [Pass 0, train batch 270] loss 7.46548, acc1 0.00000, acc5 0.00000, lr 0.10000, elapse 0.1582 sec
2020-04-20 07:38:45,166-INFO: [Pass 0, train batch 280]