PyTorch demo (wip)

#1120

Author: albertz

demos/demo-torch.config

@@ -0,0 +1,38 @@
+#!returnn.py
+# kate: syntax python;
+# -*- mode: python -*-
+# sublime: syntax 'Packages/Python Improved/PythonImproved.tmLanguage'
+# vim:set expandtab tabstop=4 fenc=utf-8 ff=unix ft=python:
+
+import os
+from returnn.util.basic import get_login_username
+
+demo_name, _ = os.path.splitext(__file__)
+print("Hello, experiment: %s" % demo_name)
+
+backend = "torch"
+
+task = "train"
+train = {"class": "Task12AXDataset", "num_seqs": 1000}
+dev = {"class": "Task12AXDataset", "num_seqs": 100, "fixed_random_seed": 1}
+
+num_inputs = 9
+num_outputs = 2
+batching = "random"
+batch_size = 5000
+max_seqs = 10
+chunking = "200:200"
+
+# TODO define network and loss...
+
+# training
+# TODO maybe remove this...
+optimizer = {"class": "adam"}
+learning_rate = 0.01
+
+model = "/tmp/%s/returnn/%s/model" % (get_login_username(), demo_name)
+num_epochs = 5
+
+# log
+#log_verbosity = 3
+log_verbosity = 5