You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by GitBox <gi...@apache.org> on 2017/12/27 23:14:15 UTC
[GitHub] eric-haibin-lin commented on a change in pull request #9073: Updating the README files and examples in "ctc" and "recommenders" folder.
eric-haibin-lin commented on a change in pull request #9073: Updating the README files and examples in "ctc" and "recommenders" folder.
URL: https://github.com/apache/incubator-mxnet/pull/9073#discussion_r158878842
##########
File path: example/ctc/README.md
##########
@@ -49,68 +55,26 @@ def lstm_unroll(num_lstm_layer, seq_len,
return mx.sym.Group([softmax_loss, ctc_loss])
```
-# Some Result
-If there were more training, the result would be better
+## Prerequisites
+
+Please ensure that following prerequisites are satisfied before running this examples.
+
+- ```captcha``` python package is installed.
+- ```cv2``` (or ```openCV```) python package is installed.
+- The test requires font file (```ttf``` format). The user either would need to create ```.\data\``` directory and place the font file in that directory. The user can also edit following line to specify path to the font file.
+```cython
+ # you can get this font from http://font.ubuntu.com/
+ self.captcha = ImageCaptcha(fonts=['./data/Xerox.ttf'])
```
-2017-07-08 13:22:01,155 Epoch[94] Batch [50] Speed: 4273.43 samples/sec Accuracy=0.808747
-2017-07-08 13:22:13,141 Epoch[94] Batch [100] Speed: 4271.84 samples/sec Accuracy=0.786855
-2017-07-08 13:22:25,179 Epoch[94] Batch [150] Speed: 4253.81 samples/sec Accuracy=0.810625
-2017-07-08 13:22:37,198 Epoch[94] Batch [200] Speed: 4259.96 samples/sec Accuracy=0.808809
-2017-07-08 13:22:49,233 Epoch[94] Batch [250] Speed: 4254.13 samples/sec Accuracy=0.806426
-2017-07-08 13:23:01,308 Epoch[94] Batch [300] Speed: 4239.98 samples/sec Accuracy=0.817305
-2017-07-08 13:23:02,030 Epoch[94] Train-Accuracy=0.819336
-2017-07-08 13:23:02,030 Epoch[94] Time cost=73.092
-2017-07-08 13:23:02,101 Saved checkpoint to "ocr-0095.params"
-2017-07-08 13:23:07,192 Epoch[94] Validation-Accuracy=0.819417
-2017-07-08 13:23:20,579 Epoch[95] Batch [50] Speed: 4288.76 samples/sec Accuracy=0.817459
-2017-07-08 13:23:32,573 Epoch[95] Batch [100] Speed: 4268.75 samples/sec Accuracy=0.815215
-2017-07-08 13:23:44,635 Epoch[95] Batch [150] Speed: 4244.85 samples/sec Accuracy=0.820215
-2017-07-08 13:23:56,670 Epoch[95] Batch [200] Speed: 4254.38 samples/sec Accuracy=0.823613
-2017-07-08 13:24:08,650 Epoch[95] Batch [250] Speed: 4273.83 samples/sec Accuracy=0.827109
-2017-07-08 13:24:20,680 Epoch[95] Batch [300] Speed: 4256.49 samples/sec Accuracy=0.824961
-2017-07-08 13:24:21,401 Epoch[95] Train-Accuracy=0.840495
-2017-07-08 13:24:21,401 Epoch[95] Time cost=73.008
-2017-07-08 13:24:21,441 Saved checkpoint to "ocr-0096.params"
-2017-07-08 13:24:26,508 Epoch[95] Validation-Accuracy=0.834798
-2017-07-08 13:24:39,938 Epoch[96] Batch [50] Speed: 4259.32 samples/sec Accuracy=0.825578
-2017-07-08 13:24:51,987 Epoch[96] Batch [100] Speed: 4249.67 samples/sec Accuracy=0.826562
-2017-07-08 13:25:04,041 Epoch[96] Batch [150] Speed: 4247.44 samples/sec Accuracy=0.831855
-2017-07-08 13:25:16,058 Epoch[96] Batch [200] Speed: 4260.77 samples/sec Accuracy=0.830840
-2017-07-08 13:25:28,109 Epoch[96] Batch [250] Speed: 4248.44 samples/sec Accuracy=0.827168
-2017-07-08 13:25:40,057 Epoch[96] Batch [300] Speed: 4285.23 samples/sec Accuracy=0.832715
-2017-07-08 13:25:40,782 Epoch[96] Train-Accuracy=0.830729
-2017-07-08 13:25:40,782 Epoch[96] Time cost=73.098
-2017-07-08 13:25:40,821 Saved checkpoint to "ocr-0097.params"
-2017-07-08 13:25:45,886 Epoch[96] Validation-Accuracy=0.840820
-2017-07-08 13:25:59,283 Epoch[97] Batch [50] Speed: 4271.85 samples/sec Accuracy=0.831648
-2017-07-08 13:26:11,243 Epoch[97] Batch [100] Speed: 4280.89 samples/sec Accuracy=0.835371
-2017-07-08 13:26:23,263 Epoch[97] Batch [150] Speed: 4259.89 samples/sec Accuracy=0.831094
-2017-07-08 13:26:35,230 Epoch[97] Batch [200] Speed: 4278.40 samples/sec Accuracy=0.827129
-2017-07-08 13:26:47,199 Epoch[97] Batch [250] Speed: 4277.77 samples/sec Accuracy=0.834258
-2017-07-08 13:26:59,257 Epoch[97] Batch [300] Speed: 4245.93 samples/sec Accuracy=0.833770
-2017-07-08 13:26:59,971 Epoch[97] Train-Accuracy=0.844727
-2017-07-08 13:26:59,971 Epoch[97] Time cost=72.908
-2017-07-08 13:27:00,020 Saved checkpoint to "ocr-0098.params"
-2017-07-08 13:27:05,130 Epoch[97] Validation-Accuracy=0.827962
-2017-07-08 13:27:18,521 Epoch[98] Batch [50] Speed: 4281.06 samples/sec Accuracy=0.834118
-2017-07-08 13:27:30,537 Epoch[98] Batch [100] Speed: 4261.20 samples/sec Accuracy=0.835352
-2017-07-08 13:27:42,542 Epoch[98] Batch [150] Speed: 4264.88 samples/sec Accuracy=0.839395
-2017-07-08 13:27:54,544 Epoch[98] Batch [200] Speed: 4266.31 samples/sec Accuracy=0.836328
-2017-07-08 13:28:06,550 Epoch[98] Batch [250] Speed: 4264.50 samples/sec Accuracy=0.841465
-2017-07-08 13:28:18,622 Epoch[98] Batch [300] Speed: 4241.11 samples/sec Accuracy=0.831680
-2017-07-08 13:28:19,349 Epoch[98] Train-Accuracy=0.833984
-2017-07-08 13:28:19,349 Epoch[98] Time cost=73.018
-2017-07-08 13:28:19,393 Saved checkpoint to "ocr-0099.params"
-2017-07-08 13:28:24,472 Epoch[98] Validation-Accuracy=0.818034
-2017-07-08 13:28:37,961 Epoch[99] Batch [50] Speed: 4242.14 samples/sec Accuracy=0.835861
-2017-07-08 13:28:50,031 Epoch[99] Batch [100] Speed: 4241.94 samples/sec Accuracy=0.846543
-2017-07-08 13:29:02,108 Epoch[99] Batch [150] Speed: 4239.22 samples/sec Accuracy=0.850645
-2017-07-08 13:29:14,160 Epoch[99] Batch [200] Speed: 4248.34 samples/sec Accuracy=0.844141
-2017-07-08 13:29:26,225 Epoch[99] Batch [250] Speed: 4243.71 samples/sec Accuracy=0.842129
-2017-07-08 13:29:38,277 Epoch[99] Batch [300] Speed: 4248.07 samples/sec Accuracy=0.851250
-2017-07-08 13:29:38,975 Epoch[99] Train-Accuracy=0.854492
-2017-07-08 13:29:38,976 Epoch[99] Time cost=73.315
-2017-07-08 13:29:39,023 Saved checkpoint to "ocr-0100.params"
-2017-07-08 13:29:44,110 Epoch[99] Validation-Accuracy=0.851969
Review comment:
The expected accuracy is important information to validate if the example works. Can we either keep it in readme, or add assertion in the code to ensure the example actually learns?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services