Subject | Hash | Author | Date (UTC) |
---|---|---|---|
fix minor typo | 01d907dda5c1e4c08b5ca5607aa77365cfd70a2e | Thai Thien | 2020-03-18 17:25:32 |
forgot to change the model | cb8bde150d7e764f67f33c980759675a2f5cce8e | Thai Thien | 2020-03-18 17:23:51 |
log summary for adcrowdnet simple lrscheduler | cefb188f185dbba3dd95f5034f17bd43dd194d44 | Thai Thien | 2020-03-18 17:18:36 |
prepare exp | 4658d0a852d952f1e1f600d9c494b8d511b9d52c | Thai Thien | 2020-03-18 17:17:05 |
CompactCNNV6 with a bunch of batchnorm | a9f21eac8ab6a328fac7a23bb5eddbbaa4496d04 | Thai Thien | 2020-03-18 17:09:26 |
implement attn_can_adcrowdnet_v5 | c905bd76d3a5d3b933d1e6eab6e50da187cee3a2 | Thai Thien | 2020-03-18 16:56:02 |
forget add --momentum | 58131da12819203ec13ea62e7ad6862cf6db301d | thient | 2020-03-18 03:41:30 |
shell script | ebb46f6853a666ecdfbaf42bb927cad0ec9e548e | thient | 2020-03-18 03:38:56 |
add sgd | a37f034b6231728629586223b9c1a0ca9f8c30ba | thient | 2020-03-18 03:38:21 |
increase epoch ... because we are loading old model | 44c6dea2b2e5017f8c6f3bf87136220d6e9def42 | thient | 2020-03-18 03:16:49 |
continue train with higher lr (forgot a slash) | 982a68aa7e2cd9879c24962459db41215415833f | thient | 2020-03-18 03:13:08 |
continue train with higher lr | f2bd926e2db7f73065433ad6702c6edee31076b3 | thient | 2020-03-18 03:12:03 |
look like batch size 2 is fck up | 85e85c211dc75dd5e4bd5033e33802699b000227 | Thai Thien | 2020-03-17 18:04:19 |
100 epoch batch size 2 | 7d3cd859496dd2c268e0d9c7cb02b67c88e71de4 | Thai Thien | 2020-03-17 18:01:44 |
batch_size 1 | 57017c52dd5788f481173cefae2aab56dd15bad4 | Thai Thien | 2020-03-17 17:59:13 |
remove inplace relu, let batchsize = 4 | 5de3e36f31d3e2b000a1bd8972bda580ecf92c3e | Thai Thien | 2020-03-17 17:52:29 |
batchsize 5 | a4332d10d2a776f2b3b72c74cc706bfd9b9ff665 | Thai Thien | 2020-03-17 17:48:11 |
batchsize to 6 | 7437e22c7bd11a29421bfc645efc7ade5f0fb42a | Thai Thien | 2020-03-17 17:46:11 |
deformable ccnn | 10a75cf43918482b285c1d130a44c7f21d65b74b | Thai Thien | 2020-03-17 17:43:21 |
change cuda vis | 8f3bad803408eda16953a040ca5ec5895ad1e6cc | Thai Thien | 2020-03-16 17:23:35 |
File | Lines added | Lines deleted |
---|---|---|
models/attn_can_adcrowdnet_simple.py | 1 | 1 |
File models/attn_can_adcrowdnet_simple.py changed (mode: 100644) (index 223f463..83d7c42) | |||
... | ... | class AttnCanAdcrowdNetSimpleV5(nn.Module): | |
759 | 759 | x1 = self.dilated_conv_1_1(x) | x1 = self.dilated_conv_1_1(x) |
760 | 760 | x2 = self.dilated_conv_1_2(x) | x2 = self.dilated_conv_1_2(x) |
761 | 761 | x3 = self.dilated_conv_1_3(x) | x3 = self.dilated_conv_1_3(x) |
762 | x = F.relu(self.bn_c1(torch.cat((x1, x2, x3), 1), inplace=True)) | ||
762 | x = F.relu(self.bn_c_1(torch.cat((x1, x2, x3), 1), inplace=True)) | ||
763 | 763 | x = F.relu(self.bn1(self.concat_filter_layer_1(x), inplace=True)) | x = F.relu(self.bn1(self.concat_filter_layer_1(x), inplace=True)) |
764 | 764 | ||
765 | 765 | x1 = self.dilated_conv_2_1(x) | x1 = self.dilated_conv_2_1(x) |