returnn 1.20230418.134656__tar.gz → 1.20230418.140214__tar.gz

{returnn-1.20230418.134656/returnn.egg-info → returnn-1.20230418.140214}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: returnn
-Version: 1.20230418.134656
+Version: 1.20230418.140214
 Summary: The RWTH extensible training framework for universal recurrent neural networks
 Home-page: https://github.com/rwth-i6/returnn/
 Author: Albert Zeyer

returnn-1.20230418.140214/_setup_info_generated.py

@@ -0,0 +1,2 @@
+version = '1.20230418.140214'
+long_version = '1.20230418.140214+git.7f606bc'

{returnn-1.20230418.134656 → returnn-1.20230418.140214/returnn.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: returnn
-Version: 1.20230418.134656
+Version: 1.20230418.140214
 Summary: The RWTH extensible training framework for universal recurrent neural networks
 Home-page: https://github.com/rwth-i6/returnn/
 Author: Albert Zeyer

{returnn-1.20230418.134656 → returnn-1.20230418.140214}/returnn.egg-info/SOURCES.txt

@@ -313,6 +313,7 @@ tests/test_rf_attention.py
 tests/test_rf_base.py
 tests/test_rf_container.py
 tests/test_rf_conv.py
+tests/test_rf_encoder_conformer.py
 tests/test_rf_math.py
 tests/test_rf_normalization.py
 tests/test_tensor.py

returnn-1.20230418.140214/tests/test_rf_encoder_conformer.py

@@ -0,0 +1,57 @@
+"""
+RETURNN frontend (returnn.frontend) tests
+"""
+
+from __future__ import annotations
+import _setup_test_env  # noqa
+from returnn.tensor import Tensor, Dim, TensorDict, batch_dim
+from rf_utils import run_model
+
+
+def test_conformer():
+    # This test needs a huge stack size currently, due to the way RETURNN layer construction works currently.
+    # On RETURNN side, there is the option flat_net_construction to solve this,
+    # however, it's experimental and also does not work for this case.
+    # https://github.com/rwth-i6/returnn/issues/957
+    # https://stackoverflow.com/a/16248113/133374
+    import resource
+    import sys
+
+    try:
+        resource.setrlimit(resource.RLIMIT_STACK, (2**29, -1))
+    except Exception as exc:
+        print(f"resource.setrlimit {type(exc).__name__}: {exc}")
+    sys.setrecursionlimit(10**6)
+
+    time_dim = Dim(Tensor("time", [batch_dim], dtype="int32"))
+    in_dim = Dim(7, name="in")
+    extern_data = TensorDict(
+        {
+            "data": Tensor("data", [batch_dim, time_dim, in_dim], dtype="float32"),
+        }
+    )
+
+    from returnn.frontend.encoder.conformer import ConformerEncoder, ConformerConvSubsample
+
+    # noinspection PyShadowingNames
+    def _forward_step(*, model: ConformerEncoder, extern_data: TensorDict):
+        out, out_spatial_dim = model(extern_data["data"], in_spatial_dim=time_dim)
+        out.mark_as_default_output(shape=(batch_dim, out_spatial_dim, model.out_dim))
+
+    run_model(
+        extern_data,
+        lambda *, epoch, step: ConformerEncoder(
+            in_dim,
+            Dim(14, name="out"),
+            ff_dim=Dim(17, name="ff"),
+            input_layer=ConformerConvSubsample(
+                in_dim,
+                out_dims=[Dim(32, name="conv1"), Dim(64, name="conv2")],
+                filter_sizes=[(3, 3), (3, 3)],
+                pool_sizes=[(2, 1), (2, 1)],
+            ),
+            num_heads=2,
+            num_layers=2,
+        ),
+        _forward_step,
+    )

returnn-1.20230418.134656/_setup_info_generated.py

@@ -1,2 +0,0 @@
-version = '1.20230418.134656'
-long_version = '1.20230418.134656+git.79116cf'