Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

Solved toy_ctc example error by debugging input data shape #5864

Merged
merged 1 commit into from
Apr 17, 2017
Merged

Solved toy_ctc example error by debugging input data shape #5864

merged 1 commit into from
Apr 17, 2017

Conversation

Soonhwan-Kwon
Copy link
Contributor

Solved related issue #5435 which happened
because data shape of symbol(seq_len=80) and real data shape(seq_len=800) doesn't met.
I modified data_iter's provide_data and gen_rand to support lstm_unroll's symbol(seq_len=80),
and confirmed it works well.
Below is the training log.

2017-04-15 16:27:06,025 Start training with [gpu(0)]
2017-04-15 16:28:11,823 Epoch[0] Batch [50] Speed: 1257.73 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:13,122 Epoch[0] Batch [100] Speed: 1231.47 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:14,348 Epoch[0] Batch [150] Speed: 1305.72 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:15,595 Epoch[0] Batch [200] Speed: 1283.31 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:16,912 Epoch[0] Batch [250] Speed: 1215.02 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:18,191 Epoch[0] Batch [300] Speed: 1251.47 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:19,430 Epoch[0] Batch [350] Speed: 1291.18 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:20,736 Epoch[0] Batch [400] Speed: 1225.21 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:22,045 Epoch[0] Batch [450] Speed: 1222.62 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:23,342 Epoch[0] Batch [500] Speed: 1233.66 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:24,615 Epoch[0] Batch [550] Speed: 1257.35 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:25,893 Epoch[0] Batch [600] Speed: 1252.21 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:27,153 Epoch[0] Batch [650] Speed: 1269.36 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:28,392 Epoch[0] Batch [700] Speed: 1291.39 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:29,692 Epoch[0] Batch [750] Speed: 1231.25 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:30,935 Epoch[0] Batch [800] Speed: 1287.14 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:32,168 Epoch[0] Batch [850] Speed: 1297.67 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:33,460 Epoch[0] Batch [900] Speed: 1239.32 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:34,744 Epoch[0] Batch [950] Speed: 1246.30 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:36,027 Epoch[0] Batch [1000] Speed: 1246.97 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:37,325 Epoch[0] Batch [1050] Speed: 1233.15 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:38,567 Epoch[0] Batch [1100] Speed: 1287.80 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:39,782 Epoch[0] Batch [1150] Speed: 1317.17 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:40,994 Epoch[0] Batch [1200] Speed: 1320.13 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:42,213 Epoch[0] Batch [1250] Speed: 1313.24 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:43,451 Epoch[0] Batch [1300] Speed: 1292.53 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:44,686 Epoch[0] Batch [1350] Speed: 1295.47 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:46,028 Epoch[0] Batch [1400] Speed: 1192.42 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:47,366 Epoch[0] Batch [1450] Speed: 1196.41 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:48,595 Epoch[0] Batch [1500] Speed: 1301.52 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:49,823 Epoch[0] Batch [1550] Speed: 1303.15 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:51,119 Epoch[0] Batch [1600] Speed: 1234.43 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:52,353 Epoch[0] Batch [1650] Speed: 1297.16 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:53,619 Epoch[0] Batch [1700] Speed: 1264.41 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:54,920 Epoch[0] Batch [1750] Speed: 1229.41 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:56,172 Epoch[0] Batch [1800] Speed: 1278.26 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:57,424 Epoch[0] Batch [1850] Speed: 1278.35 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:58,635 Epoch[0] Batch [1900] Speed: 1321.37 samples/sec Train-Accuracy=0.000000
2017-04-15 16:28:59,862 Epoch[0] Batch [1950] Speed: 1304.36 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:01,146 Epoch[0] Batch [2000] Speed: 1245.59 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:02,449 Epoch[0] Batch [2050] Speed: 1228.02 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:03,660 Epoch[0] Batch [2100] Speed: 1322.37 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:04,930 Epoch[0] Batch [2150] Speed: 1259.39 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:06,205 Epoch[0] Batch [2200] Speed: 1255.53 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:07,429 Epoch[0] Batch [2250] Speed: 1306.73 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:08,753 Epoch[0] Batch [2300] Speed: 1208.87 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:09,988 Epoch[0] Batch [2350] Speed: 1295.36 samples/sec Train-Accuracy=0.000000
2017-04-15 16:29:11,220 Epoch[0] Batch [2400] Speed: 1299.24 samples/sec Train-Accuracy=0.017500
2017-04-15 16:29:12,540 Epoch[0] Batch [2450] Speed: 1212.11 samples/sec Train-Accuracy=0.103750
2017-04-15 16:29:13,785 Epoch[0] Batch [2500] Speed: 1286.16 samples/sec Train-Accuracy=0.468750
2017-04-15 16:29:15,019 Epoch[0] Batch [2550] Speed: 1296.61 samples/sec Train-Accuracy=0.635625
2017-04-15 16:29:16,245 Epoch[0] Batch [2600] Speed: 1304.37 samples/sec Train-Accuracy=0.824375
2017-04-15 16:29:17,482 Epoch[0] Batch [2650] Speed: 1294.61 samples/sec Train-Accuracy=0.968125
2017-04-15 16:29:18,708 Epoch[0] Batch [2700] Speed: 1305.09 samples/sec Train-Accuracy=0.998750
2017-04-15 16:29:19,963 Epoch[0] Batch [2750] Speed: 1274.84 samples/sec Train-Accuracy=0.999375
2017-04-15 16:29:21,220 Epoch[0] Batch [2800] Speed: 1272.97 samples/sec Train-Accuracy=0.998125
2017-04-15 16:29:22,493 Epoch[0] Batch [2850] Speed: 1257.39 samples/sec Train-Accuracy=1.000000

@piiswrong piiswrong merged commit d4e611b into apache:master Apr 17, 2017
Guneet-Dhillon pushed a commit to Guneet-Dhillon/mxnet that referenced this pull request Sep 13, 2017
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants