update docs

.github/ISSUE_TEMPLATE.md

@@ -22,7 +22,7 @@ Feature Requests:
 + You can implement a lot of features by extending tensorpack
 	(See http://tensorpack.readthedocs.io/en/latest/tutorial/index.html#extend-tensorpack).
 	It does not have to be added to tensorpack unless you have a good reason.
-+ We don't take example requests.
++ We don't take feature requests for examples.
 
 Usage Questions:
 

docs/tutorial/inference.md

@@ -6,7 +6,8 @@
 There are two ways to do inference during training.
 
 1. The easiest way is to write a callback, and use
-  `self.trainer.get_predictor()` to get a callable under inference mode.
+  [self.trainer.get_predictor()](../modules/modules/train.html#tensorpack.train.TowerTrainer.get_predictor)
+	to get a callable under inference mode.
 	See [Write a Callback](extend/callback.html).
 
 2. If your inference follows the paradigm of:
@@ -29,15 +30,16 @@ Please note that, the metagraph saved during training is the training graph.
 But sometimes you need a different one for inference.
 For example, you may need a different data layout for CPU inference,
 or you may need placeholders in the inference graph, or the training graph contains multi-GPU replication
-which you want to remove.
+which you want to remove. In fact, directly import a huge training metagraph is usually not a good idea for deployment.
+
 In this case, you can always construct a new graph by simply:
 ```python
 a, b = tf.placeholder(...), tf.placeholder(...)
 # call symbolic functions on a, b
 ```
 
-The only tool tensorpack has for after-training inference is `OfflinePredictor`,
+The only tool tensorpack has for after-training inference is [OfflinePredictor](../modules/predict.html#tensorpack.predict.OfflinePredictor),
 a simple function to build the graph and return a callable for you.
 It is mainly for quick demo purposes.
-It only runs inference on Python data, therefore may not be the most efficient way.
-Check out some examples for its usage.
+It only runs inference on numpy arrays, therefore may not be the most efficient way.
+Check out examples and docs for its usage.

examples/GAN/Improved-WGAN.py

@@ -4,6 +4,7 @@
 # Author: Yuxin Wu <ppwwyyxxc@gmail.com>
 
 from tensorpack import *
+from tensorpack.tfutils import get_tf_version_number
 from tensorpack.tfutils.summary import add_moving_summary
 from tensorpack.tfutils.scope_utils import auto_reuse_variable_scope
 import tensorflow as tf
@@ -83,6 +84,7 @@ class Model(DCGAN.Model):
 
 
 if __name__ == '__main__':
+    assert get_tf_version_number() >= 1.4
     args = DCGAN.get_args(default_batch=64, default_z_dim=128)
     M = Model(shape=args.final_size, batch=args.batch, z_dim=args.z_dim)
     if args.sample:

tensorpack/predict/config.py

@@ -40,7 +40,7 @@ class PredictConfig(object):
                 tensors can be any computable tensor in the graph.
             return_input (bool): same as in :attr:`PredictorBase.return_input`.
             create_graph (bool): create a new graph, or use the default graph
-                when then predictor is first initialized.
+                when predictor is first initialized.
 
         You need to set either `model`, or `inputs_desc` plus `tower_func`.
         """