Feature/remove std (#515)

* Remove std from metric defaults, clean up docs

* Update CHANGELOG.md

CHANGELOG.md

@@ -28,6 +28,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 ### Removed
 - Removed the old Model API (deprecated since version 0.2.0)
 - Removed the 'pass_state' argument from Trial, this will now be inferred
+- Removed the 'std' decorator from the default metrics
 ### Fixed
 - Fixed a bug in the weight decay callback which would result in potentially negative decay (now just uses torch.norm)
 - Fixed a bug in the cite decorator causing the citation to not show up correctly

README.md

@@ -1,6 +1,6 @@
 <img src="https://raw.githubusercontent.com/ecs-vlc/torchbearer/master/docs/_static/img/logo_dark_text.svg?sanitize=true" width="100%"/>
 
-[![PyPI version](https://badge.fury.io/py/torchbearer.svg)](https://badge.fury.io/py/torchbearer) [![Python 2.7 | 3.5 | 3.6 | 3.7](https://img.shields.io/badge/python-2.7%20%7C%203.5%20%7C%203.6%20%7C%203.7-brightgreen.svg)](https://www.python.org/) [![PyTorch 0.4.0 | 0.4.1 | 1.0.0](https://img.shields.io/badge/pytorch-0.4.0%20%7C%200.4.1%20%7C%201.0.0-brightgreen.svg)](https://pytorch.org/) [![Build Status](https://travis-ci.com/ecs-vlc/torchbearer.svg?branch=master)](https://travis-ci.com/ecs-vlc/torchbearer) [![codecov](https://codecov.io/gh/ecs-vlc/torchbearer/branch/master/graph/badge.svg)](https://codecov.io/gh/ecs-vlc/torchbearer) [![Documentation Status](https://readthedocs.org/projects/torchbearer/badge/?version=latest)](https://torchbearer.readthedocs.io/en/latest/?badge=latest) [![Codacy Badge](https://api.codacy.com/project/badge/Grade/8c9b136fbcd443fa9135d92321be480d)](https://www.codacy.com/app/ewah1g13/torchbearer?utm_source=github.com&amp;utm_medium=referral&amp;utm_content=ecs-vlc/torchbearer&amp;utm_campaign=Badge_Grade)
+[![PyPI version](https://badge.fury.io/py/torchbearer.svg)](https://badge.fury.io/py/torchbearer) [![Python 2.7 | 3.5 | 3.6 | 3.7](https://img.shields.io/badge/python-2.7%20%7C%203.5%20%7C%203.6%20%7C%203.7-brightgreen.svg)](https://www.python.org/) [![PyTorch 0.4.0 | 0.4.1 | 1.0.0](https://img.shields.io/badge/pytorch-0.4.0%20%7C%200.4.1%20%7C%201.0.0-brightgreen.svg)](https://pytorch.org/) [![Build Status](https://travis-ci.com/ecs-vlc/torchbearer.svg?branch=master)](https://travis-ci.com/ecs-vlc/torchbearer) [![codecov](https://codecov.io/gh/ecs-vlc/torchbearer/branch/master/graph/badge.svg)](https://codecov.io/gh/ecs-vlc/torchbearer) [![Documentation Status](https://readthedocs.org/projects/torchbearer/badge/?version=latest)](https://torchbearer.readthedocs.io/en/latest/?badge=latest)
 
 A model fitting library for PyTorch
 ## Contents
@@ -132,14 +132,14 @@ trial.evaluate(data_key=torchbearer.TEST_DATA)
 - Running that code gives output using Tqdm and providing running accuracies and losses during the training phase:
 
 ```
-0/10(t): 100%|██████████| 352/352 [00:01<00:00, 176.55it/s, running_acc=0.526, running_loss=1.31, acc=0.453, acc_std=0.498, loss=1.53, loss_std=0.25]
-0/10(v): 100%|██████████| 40/40 [00:00<00:00, 201.14it/s, val_acc=0.528, val_acc_std=0.499, val_loss=1.32, val_loss_std=0.0874]
+0/10(t): 100%|██████████| 352/352 [00:02<00:00, 163.98it/s, acc=0.4339, loss=1.5776, running_acc=0.5202, running_loss=1.3494]
+0/10(v): 100%|██████████| 40/40 [00:00<00:00, 365.42it/s, val_acc=0.5266, val_loss=1.3208]
 .
 .
 .
-9/10(t): 100%|██████████| 352/352 [00:02<00:00, 171.22it/s, running_acc=0.738, running_loss=0.737, acc=0.749, acc_std=0.434, loss=0.723, loss_std=0.0885]
-9/10(v): 100%|██████████| 40/40 [00:00<00:00, 188.51it/s, val_acc=0.669, val_acc_std=0.471, val_loss=0.97, val_loss_std=0.173]
-0/1(e): 100%|██████████| 79/79 [00:00<00:00, 241.00it/s, test_acc=0.675, test_acc_std=0.468, test_loss=0.952, test_loss_std=0.109]
+9/10(t): 100%|██████████| 352/352 [00:02<00:00, 165.28it/s, acc=0.7515, loss=0.715, running_acc=0.7352, running_loss=0.7492]
+9/10(v): 100%|██████████| 40/40 [00:00<00:00, 310.76it/s, val_acc=0.6792, val_loss=0.9743]
+0/1(e): 100%|██████████| 79/79 [00:00<00:00, 233.06it/s, test_acc=0.6673, test_loss=0.9741]
 ```
 
 <a name="docs"/>

docs/_static/examples/basic_opt.py

@@ -39,7 +39,7 @@ def loss(y_pred, y_true):
 model = Net(p)
 optim = torch.optim.SGD(model.parameters(), lr=0.0001)
 
-tbtrial = tb.Trial(model, optim, loss, [tb.metrics.running_mean(ESTIMATE, dim=1), 'loss'], pass_state=True)
+tbtrial = tb.Trial(model, optim, loss, [tb.metrics.running_mean(ESTIMATE, dim=1), 'loss'])
 tbtrial.for_train_steps(training_steps).to('cuda')
 tbtrial.run()
 print(list(model.parameters())[0].data)

docs/_static/examples/gan.py

@@ -158,7 +158,7 @@ def process(self, state):
 
 
 torchbearertrial = tb.Trial(model, optim, criterion=None, metrics=['loss', g_loss(), d_loss()],
-                            callbacks=[loss_callback, saver_callback], pass_state=True)
+                            callbacks=[loss_callback, saver_callback])
 torchbearertrial.with_train_generator(dataloader)
 torchbearertrial.to(device)
 torchbearertrial.run(epochs=200)

docs/_static/examples/vae.py

@@ -122,6 +122,6 @@ def saver(state):
 from torchbearer import Trial
 
 torchbearer_trial = Trial(model, optimizer, loss, metrics=['acc', 'loss'],
-                          callbacks=[add_kld_loss_callback, save_reconstruction_callback()], pass_state=True).to('cuda')
+                          callbacks=[add_kld_loss_callback, save_reconstruction_callback()]).to('cuda')
 torchbearer_trial.with_generators(train_generator=traingen, val_generator=valgen)
 torchbearer_trial.run(epochs=10)

docs/_static/examples/vae_standard.py

@@ -104,6 +104,6 @@ def loss_function(y_pred, y_true):
 
 from torchbearer import Trial
 
-torchbearer_trial = Trial(model, optimizer, loss, metrics=['loss'], pass_state=False).to('cuda')
+torchbearer_trial = Trial(model, optimizer, loss, metrics=['loss']).to('cuda')
 torchbearer_trial.with_generators(train_generator=traingen, val_generator=valgen)
 torchbearer_trial.run(epochs=10)

docs/examples/basic_opt.rst

@@ -70,7 +70,7 @@ Putting all of it together and running provides the following output:
 
 .. code::
 
-    0/1(t): 100%|██████████| 50000/50000 [00:54<00:00, 912.37it/s, est=[4.9988 0.     1.0004], running_loss=1.6e-06, loss=4.55, loss_std=13.7]
+    0/1(t): 100%|██████████| 50000/50000 [00:53<00:00, 931.36it/s, loss=4.5502, running_est=[4.9988, 0.0, 1.0004], running_loss=0.0]
 
 The final estimate is very close to the true minimum at [5, 0, 1]:
 

docs/examples/quickstart.rst

@@ -30,27 +30,27 @@ Running the above produces the following output:
 
    Files already downloaded and verified
    Files already downloaded and verified
-   0/10(t): 100%|██████████| 352/352 [00:01<00:00, 233.36it/s, running_acc=0.536, running_loss=1.32, acc=0.459, acc_std=0.498, loss=1.52, loss_std=0.239]
-   0/10(v): 100%|██████████| 40/40 [00:00<00:00, 239.40it/s, val_acc=0.536, val_acc_std=0.499, val_loss=1.29, val_loss_std=0.0731]
-   1/10(t): 100%|██████████| 352/352 [00:01<00:00, 211.19it/s, running_acc=0.599, running_loss=1.13, acc=0.578, acc_std=0.494, loss=1.18, loss_std=0.096]
-   1/10(v): 100%|██████████| 40/40 [00:00<00:00, 232.97it/s, val_acc=0.594, val_acc_std=0.491, val_loss=1.14, val_loss_std=0.101]
-   2/10(t): 100%|██████████| 352/352 [00:01<00:00, 216.68it/s, running_acc=0.636, running_loss=1.04, acc=0.631, acc_std=0.482, loss=1.04, loss_std=0.0944]
-   2/10(v): 100%|██████████| 40/40 [00:00<00:00, 210.73it/s, val_acc=0.626, val_acc_std=0.484, val_loss=1.07, val_loss_std=0.0974]
-   3/10(t): 100%|██████████| 352/352 [00:01<00:00, 190.88it/s, running_acc=0.671, running_loss=0.929, acc=0.664, acc_std=0.472, loss=0.957, loss_std=0.0929]
-   3/10(v): 100%|██████████| 40/40 [00:00<00:00, 221.79it/s, val_acc=0.639, val_acc_std=0.48, val_loss=1.02, val_loss_std=0.103]
-   4/10(t): 100%|██████████| 352/352 [00:01<00:00, 212.43it/s, running_acc=0.685, running_loss=0.897, acc=0.689, acc_std=0.463, loss=0.891, loss_std=0.0888]
-   4/10(v): 100%|██████████| 40/40 [00:00<00:00, 249.99it/s, val_acc=0.655, val_acc_std=0.475, val_loss=0.983, val_loss_std=0.113]
-   5/10(t): 100%|██████████| 352/352 [00:01<00:00, 209.45it/s, running_acc=0.711, running_loss=0.835, acc=0.706, acc_std=0.456, loss=0.844, loss_std=0.088]
-   5/10(v): 100%|██████████| 40/40 [00:00<00:00, 240.80it/s, val_acc=0.648, val_acc_std=0.477, val_loss=0.965, val_loss_std=0.107]
-   6/10(t): 100%|██████████| 352/352 [00:01<00:00, 216.89it/s, running_acc=0.713, running_loss=0.826, acc=0.72, acc_std=0.449, loss=0.802, loss_std=0.0903]
-   6/10(v): 100%|██████████| 40/40 [00:00<00:00, 238.17it/s, val_acc=0.655, val_acc_std=0.475, val_loss=0.97, val_loss_std=0.0997]
-   7/10(t): 100%|██████████| 352/352 [00:01<00:00, 213.82it/s, running_acc=0.737, running_loss=0.773, acc=0.734, acc_std=0.442, loss=0.765, loss_std=0.0878]
-   7/10(v): 100%|██████████| 40/40 [00:00<00:00, 202.45it/s, val_acc=0.677, val_acc_std=0.468, val_loss=0.936, val_loss_std=0.0985]
-   8/10(t): 100%|██████████| 352/352 [00:01<00:00, 211.36it/s, running_acc=0.732, running_loss=0.744, acc=0.746, acc_std=0.435, loss=0.728, loss_std=0.0902]
-   8/10(v): 100%|██████████| 40/40 [00:00<00:00, 204.52it/s, val_acc=0.674, val_acc_std=0.469, val_loss=0.949, val_loss_std=0.124]
-   9/10(t): 100%|██████████| 352/352 [00:02<00:00, 171.22it/s, running_acc=0.738, running_loss=0.737, acc=0.749, acc_std=0.434, loss=0.723, loss_std=0.0885]
-   9/10(v): 100%|██████████| 40/40 [00:00<00:00, 188.51it/s, val_acc=0.669, val_acc_std=0.471, val_loss=0.97, val_loss_std=0.173]
-   0/1(e): 100%|██████████| 79/79 [00:00<00:00, 241.00it/s, test_acc=0.675, test_acc_std=0.468, test_loss=0.952, test_loss_std=0.109]
+   0/10(t): 100%|██████████| 352/352 [00:02<00:00, 163.98it/s, acc=0.4339, loss=1.5776, running_acc=0.5202, running_loss=1.3494]
+   0/10(v): 100%|██████████| 40/40 [00:00<00:00, 365.42it/s, val_acc=0.5266, val_loss=1.3208]
+   1/10(t): 100%|██████████| 352/352 [00:02<00:00, 171.36it/s, acc=0.5636, loss=1.2176, running_acc=0.5922, running_loss=1.1418]
+   1/10(v): 100%|██████████| 40/40 [00:00<00:00, 292.15it/s, val_acc=0.5888, val_loss=1.1657]
+   2/10(t): 100%|██████████| 352/352 [00:02<00:00, 124.04it/s, acc=0.6226, loss=1.0671, running_acc=0.6222, running_loss=1.0566]
+   2/10(v): 100%|██████████| 40/40 [00:00<00:00, 359.21it/s, val_acc=0.626, val_loss=1.0555]
+   3/10(t): 100%|██████████| 352/352 [00:02<00:00, 151.69it/s, acc=0.6587, loss=0.972, running_acc=0.6634, running_loss=0.9589]
+   3/10(v): 100%|██████████| 40/40 [00:00<00:00, 222.62it/s, val_acc=0.6414, val_loss=1.0064]
+   4/10(t): 100%|██████████| 352/352 [00:02<00:00, 131.49it/s, acc=0.6829, loss=0.9061, running_acc=0.6764, running_loss=0.918]
+   4/10(v): 100%|██████████| 40/40 [00:00<00:00, 346.88it/s, val_acc=0.6636, val_loss=0.9449]
+   5/10(t): 100%|██████████| 352/352 [00:02<00:00, 164.28it/s, acc=0.6988, loss=0.8563, running_acc=0.6919, running_loss=0.858]
+   5/10(v): 100%|██████████| 40/40 [00:00<00:00, 244.97it/s, val_acc=0.663, val_loss=0.9404]
+   6/10(t): 100%|██████████| 352/352 [00:02<00:00, 149.52it/s, acc=0.7169, loss=0.8131, running_acc=0.7095, running_loss=0.8421]
+   6/10(v): 100%|██████████| 40/40 [00:00<00:00, 329.26it/s, val_acc=0.6704, val_loss=0.9209]
+   7/10(t): 100%|██████████| 352/352 [00:02<00:00, 160.60it/s, acc=0.7302, loss=0.7756, running_acc=0.738, running_loss=0.767]
+   7/10(v): 100%|██████████| 40/40 [00:00<00:00, 349.86it/s, val_acc=0.6716, val_loss=0.9313]
+   8/10(t): 100%|██████████| 352/352 [00:02<00:00, 155.08it/s, acc=0.7412, loss=0.7444, running_acc=0.7347, running_loss=0.7547]
+   8/10(v): 100%|██████████| 40/40 [00:00<00:00, 350.05it/s, val_acc=0.673, val_loss=0.9324]
+   9/10(t): 100%|██████████| 352/352 [00:02<00:00, 165.28it/s, acc=0.7515, loss=0.715, running_acc=0.7352, running_loss=0.7492]
+   9/10(v): 100%|██████████| 40/40 [00:00<00:00, 310.76it/s, val_acc=0.6792, val_loss=0.9743]
+   0/1(e): 100%|██████████| 79/79 [00:00<00:00, 233.06it/s, test_acc=0.6673, test_loss=0.9741]
 
 Source Code
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

docs/examples/vae.rst

@@ -77,7 +77,7 @@ We can (and should) generate unique state keys for interacting with state:
    :lines: 48-49
 
 
-By default the model forward pass does not have access to the state dictionary, but setting the ``pass_state`` flag to true when initialising Trial_ gives the model access to state on forward.
+By default the model forward pass does not have access to the state dictionary, but torchbearer will infer the ``state`` argument from the model forward.
 
 .. _Trial: https://torchbearer.readthedocs.io/en/latest/code/main.html#torchbearer.trial.Trial
 
@@ -130,26 +130,26 @@ This gives the following output:
 
 .. code::
 
-    0/10(t): 100%|██████████| 422/422 [00:01<00:00, 219.71it/s, binary_acc=0.9139, loss=2.139e+4, loss_std=6582, running_binary_acc=0.9416, running_loss=1.685e+4]
-    0/10(v): 100%|██████████| 47/47 [00:00<00:00, 269.77it/s, val_binary_acc=0.9505, val_loss=1.558e+4, val_loss_std=470.8]
-    1/10(t): 100%|██████████| 422/422 [00:01<00:00, 219.80it/s, binary_acc=0.9492, loss=1.573e+4, loss_std=573.6, running_binary_acc=0.9531, running_loss=1.52e+4]
-    1/10(v): 100%|██████████| 47/47 [00:00<00:00, 300.54it/s, val_binary_acc=0.9614, val_loss=1.399e+4, val_loss_std=427.7]
-    2/10(t): 100%|██████████| 422/422 [00:01<00:00, 232.41it/s, binary_acc=0.9558, loss=1.476e+4, loss_std=407.3, running_binary_acc=0.9571, running_loss=1.457e+4]
-    2/10(v): 100%|██████████| 47/47 [00:00<00:00, 296.49it/s, val_binary_acc=0.9652, val_loss=1.336e+4, val_loss_std=338.2]
-    3/10(t): 100%|██████████| 422/422 [00:01<00:00, 213.10it/s, binary_acc=0.9585, loss=1.437e+4, loss_std=339.6, running_binary_acc=0.9595, running_loss=1.423e+4]
-    3/10(v): 100%|██████████| 47/47 [00:00<00:00, 313.42it/s, val_binary_acc=0.9672, val_loss=1.304e+4, val_loss_std=372.3]
-    4/10(t): 100%|██████████| 422/422 [00:01<00:00, 213.43it/s, binary_acc=0.9601, loss=1.413e+4, loss_std=332.5, running_binary_acc=0.9605, running_loss=1.409e+4]
-    4/10(v): 100%|██████████| 47/47 [00:00<00:00, 242.23it/s, val_binary_acc=0.9683, val_loss=1.282e+4, val_loss_std=369.3]
-    5/10(t): 100%|██████████| 422/422 [00:01<00:00, 220.94it/s, binary_acc=0.9611, loss=1.398e+4, loss_std=300.9, running_binary_acc=0.9614, running_loss=1.397e+4]
-    5/10(v): 100%|██████████| 47/47 [00:00<00:00, 316.69it/s, val_binary_acc=0.9689, val_loss=1.281e+4, val_loss_std=423.6]
-    6/10(t): 100%|██████████| 422/422 [00:01<00:00, 230.53it/s, binary_acc=0.9619, loss=1.385e+4, loss_std=292.1, running_binary_acc=0.9621, running_loss=1.38e+4]
-    6/10(v): 100%|██████████| 47/47 [00:00<00:00, 241.06it/s, val_binary_acc=0.9695, val_loss=1.275e+4, val_loss_std=459.9]
-    7/10(t): 100%|██████████| 422/422 [00:01<00:00, 227.49it/s, binary_acc=0.9624, loss=1.377e+4, loss_std=306.9, running_binary_acc=0.9624, running_loss=1.381e+4]
-    7/10(v): 100%|██████████| 47/47 [00:00<00:00, 237.75it/s, val_binary_acc=0.97, val_loss=1.258e+4, val_loss_std=353.8]
-    8/10(t): 100%|██████████| 422/422 [00:01<00:00, 220.68it/s, binary_acc=0.9629, loss=1.37e+4, loss_std=300.8, running_binary_acc=0.9629, running_loss=1.369e+4]
-    8/10(v): 100%|██████████| 47/47 [00:00<00:00, 301.59it/s, val_binary_acc=0.9704, val_loss=1.255e+4, val_loss_std=347.7]
-    9/10(t): 100%|██████████| 422/422 [00:01<00:00, 215.23it/s, binary_acc=0.9633, loss=1.364e+4, loss_std=310, running_binary_acc=0.9633, running_loss=1.366e+4]
-    9/10(v): 100%|██████████| 47/47 [00:00<00:00, 309.51it/s, val_binary_acc=0.9707, val_loss=1.25e+4, val_loss_std=358.9]
+    0/10(t): 100%|██████████| 422/422 [00:01<00:00, 219.71it/s, binary_acc=0.9139, loss=2.139e+4, running_binary_acc=0.9416, running_loss=1.685e+4]
+    0/10(v): 100%|██████████| 47/47 [00:00<00:00, 269.77it/s, val_binary_acc=0.9505, val_loss=1.558e+4]
+    1/10(t): 100%|██████████| 422/422 [00:01<00:00, 219.80it/s, binary_acc=0.9492, loss=1.573e+4, running_binary_acc=0.9531, running_loss=1.52e+4]
+    1/10(v): 100%|██████████| 47/47 [00:00<00:00, 300.54it/s, val_binary_acc=0.9614, val_loss=1.399e+4]
+    2/10(t): 100%|██████████| 422/422 [00:01<00:00, 232.41it/s, binary_acc=0.9558, loss=1.476e+4, running_binary_acc=0.9571, running_loss=1.457e+4]
+    2/10(v): 100%|██████████| 47/47 [00:00<00:00, 296.49it/s, val_binary_acc=0.9652, val_loss=1.336e+4]
+    3/10(t): 100%|██████████| 422/422 [00:01<00:00, 213.10it/s, binary_acc=0.9585, loss=1.437e+4, running_binary_acc=0.9595, running_loss=1.423e+4]
+    3/10(v): 100%|██████████| 47/47 [00:00<00:00, 313.42it/s, val_binary_acc=0.9672, val_loss=1.304e+4]
+    4/10(t): 100%|██████████| 422/422 [00:01<00:00, 213.43it/s, binary_acc=0.9601, loss=1.413e+4, running_binary_acc=0.9605, running_loss=1.409e+4]
+    4/10(v): 100%|██████████| 47/47 [00:00<00:00, 242.23it/s, val_binary_acc=0.9683, val_loss=1.282e+4]
+    5/10(t): 100%|██████████| 422/422 [00:01<00:00, 220.94it/s, binary_acc=0.9611, loss=1.398e+4, running_binary_acc=0.9614, running_loss=1.397e+4]
+    5/10(v): 100%|██████████| 47/47 [00:00<00:00, 316.69it/s, val_binary_acc=0.9689, val_loss=1.281e+4]
+    6/10(t): 100%|██████████| 422/422 [00:01<00:00, 230.53it/s, binary_acc=0.9619, loss=1.385e+4, running_binary_acc=0.9621, running_loss=1.38e+4]
+    6/10(v): 100%|██████████| 47/47 [00:00<00:00, 241.06it/s, val_binary_acc=0.9695, val_loss=1.275e+4]
+    7/10(t): 100%|██████████| 422/422 [00:01<00:00, 227.49it/s, binary_acc=0.9624, loss=1.377e+4, running_binary_acc=0.9624, running_loss=1.381e+4]
+    7/10(v): 100%|██████████| 47/47 [00:00<00:00, 237.75it/s, val_binary_acc=0.97, val_loss=1.258e+4]
+    8/10(t): 100%|██████████| 422/422 [00:01<00:00, 220.68it/s, binary_acc=0.9629, loss=1.37e+4, running_binary_acc=0.9629, running_loss=1.369e+4]
+    8/10(v): 100%|██████████| 47/47 [00:00<00:00, 301.59it/s, val_binary_acc=0.9704, val_loss=1.255e+4]
+    9/10(t): 100%|██████████| 422/422 [00:01<00:00, 215.23it/s, binary_acc=0.9633, loss=1.364e+4, running_binary_acc=0.9633, running_loss=1.366e+4]
+    9/10(v): 100%|██████████| 47/47 [00:00<00:00, 309.51it/s, val_binary_acc=0.9707, val_loss=1.25e+4]
 
 The visualised results after ten epochs then look like this: