better log & h5py

examples/mnist_convnet.py

@@ -18,6 +18,7 @@
 from tensorpack.tfutils import *
 from tensorpack.callbacks import *
 from tensorpack.dataflow import *
+from IPython import embed; embed()
 
 """
 MNIST ConvNet example.

tensorpack/dataflow/dataset/ilsvrc.py

@@ -16,6 +16,11 @@
 CAFFE_ILSVRC12_URL = "http://dl.caffe.berkeleyvision.org/caffe_ilsvrc12.tar.gz"
 CAFFE_PROTO_URL = "https://github.com/BVLC/caffe/raw/master/src/caffe/proto/caffe.proto"
 
+"""
+cd train
+for i in *.tar; do dir=${i%.tar}; echo $dir; mkdir -p $dir; tar xf $i -C $dir; done
+"""
+
 class ILSVRCMeta(object):
     """
     Provide metadata for ILSVRC dataset.
@@ -119,7 +124,7 @@ def get_data(self):
 
 if __name__ == '__main__':
     meta = ILSVRCMeta()
-    print meta.get_per_pixel_mean()
+    print(meta.get_per_pixel_mean())
     #print(meta.get_synset_words_1000())
 
     #ds = ILSVRC12('/home/wyx/data/imagenet', 'val')

tensorpack/dataflow/format.py

@@ -11,15 +11,16 @@
 try:
     import h5py
 except ImportError:
-    logger.error("Error in `import h5py`. HDF5Data cannot function.")
+    logger.error("Error in 'import h5py'. HDF5Data won't be imported.")
+    __all__ = []
+else:
+    __all__ = ['HDF5Data']
 
 
 """
-Adapter for different data format.
+Adapters for different data format.
 """
 
-__all__ = ['HDF5Data']
-
 class HDF5Data(DataFlow):
     """
     Zip data from different paths in an HDF5 file. Will load all data into memory.

tensorpack/models/_common.py

@@ -27,31 +27,37 @@ def wrapper(func):
         def wrapped_func(*args, **kwargs):
             name = args[0]
             assert isinstance(name, six.string_types), \
-                    'name must be the first argument. Args: {}'.format(str(args))
+                    'name must be the first argument. Args: {}'.format(args)
             args = args[1:]
             do_summary = kwargs.pop(
                 'summary_activation', summary_activation)
             inputs = args[0]
 
+            # update from current argument scope
             actual_args = get_arg_scope()[func.__name__]
             actual_args.update(kwargs)
 
             with tf.variable_scope(name) as scope:
+                do_log_shape = log_shape and scope.name not in _layer_logged
+                do_summary = do_summary and scope.name not in _layer_logged
+                if do_log_shape:
+                    logger.info("{} input: {}".format(scope.name, get_shape_str(inputs)))
+
+                # run the actual function
                 outputs = func(*args, **actual_args)
-                if log_shape and scope.name not in _layer_logged:
+
+                if do_log_shape:
                     # log shape info and add activation
-                    logger.info("{} input: {}".format(
-                        scope.name, get_shape_str(inputs)))
                     logger.info("{} output: {}".format(
                         scope.name, get_shape_str(outputs)))
-
-                    if do_summary:
-                        if isinstance(outputs, list):
-                            for x in outputs:
-                                add_activation_summary(x, scope.name)
-                        else:
-                            add_activation_summary(outputs, scope.name)
                     _layer_logged.add(scope.name)
+
+                if do_summary:
+                    if isinstance(outputs, list):
+                        for x in outputs:
+                            add_activation_summary(x, scope.name)
+                    else:
+                        add_activation_summary(outputs, scope.name)
                 return outputs
         wrapped_func.f = func   # attribute to access the underlining function object
         return wrapped_func