update configs and tests to reflect changes

* add tests for empty embedding
* use better tests for overriding embedding `kwargs`

biogtr/training/configs/base.yaml

@@ -14,11 +14,11 @@ model:
  dropout_attn_head: 0.1
  embedding_meta: 
  pos:
- type: "learned"
+ mode: "learned"
  emb_num: 16
  over_boxes: false
  temp:
- type: "learned"
+ mode: "learned"
  emb_num: 16
  return_embedding: False
  decoder_self_attn: False

biogtr/training/configs/params.yaml

@@ -3,10 +3,11 @@ model:
  num_decoder_layers: 2
  embedding_meta:
  pos:
- type: learned
+ mode: learned
  emb_num: 16
  over_boxes: True
- temp: null
+ temp: 
+ mode: "off"
 dataset:
  train_dataset:
  slp_files: ['190612_110405_wt_18159111_rig2.2@11730.slp']

tests/configs/params.yaml

@@ -7,7 +7,8 @@ model:
  mode: 'learned'
  emb_num: 16
  over_boxes: True
- temp: null
+ temp: 
+ mode: "off"
 
 dataset:
  train_dataset:
@@ -29,3 +30,4 @@ trainer:
  limit_test_batches: 1
  limit_val_batches: 1
  max_epochs: 1
+ enable_checkpointing: true 

tests/test_models.py

@@ -66,41 +66,41 @@ def test_embedding_validity():
 
  # this would throw assertion since embedding should be "pos"
  with pytest.raises(Exception):
- _ = Embedding(type="position", mode="learned")
+ _ = Embedding(emb_type="position", mode="learned", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="position", mode="fixed")
+ _ = Embedding(emb_type="position", mode="fixed", features=128)
 
  with pytest.raises(Exception):
- _ = Embedding(type="temporal", mode="learned")
+ _ = Embedding(emb_type="temporal", mode="learned", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="position", mode="fixed")
+ _ = Embedding(emb_type="position", mode="fixed", features=128)
 
  with pytest.raises(Exception):
- _ = Embedding(type="pos", mode="learn")
+ _ = Embedding(emb_type="pos", mode="learn", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="temp", mode="learn")
+ _ = Embedding(emb_type="temp", mode="learn", features=128)
 
  with pytest.raises(Exception):
- _ = Embedding(type="pos", mode="fix")
+ _ = Embedding(emb_type="pos", mode="fix", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="temp", mode="fix")
+ _ = Embedding(emb_type="temp", mode="fix", features=128)
 
  with pytest.raises(Exception):
- _ = Embedding(type="position", mode="learn")
+ _ = Embedding(emb_type="position", mode="learn", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="temporal", mode="learn")
+ _ = Embedding(emb_type="temporal", mode="learn", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="position", mode="fix")
+ _ = Embedding(emb_type="position", mode="fix", features=128)
  with pytest.raises(Exception):
- _ = Embedding(type="temporal", mode="learn")
+ _ = Embedding(emb_type="temporal", mode="learn", features=128)
 
  with pytest.raises(Exception):
- _ = Embedding(type="temp", mode="fixed")
+ _ = Embedding(emb_type="temp", mode="fixed", features=128)
 
- _ = Embedding(type="temp", mode="learned")
- _ = Embedding(type="pos", mode="learned")
+ _ = Embedding(emb_type="temp", mode="learned", features=128)
+ _ = Embedding(emb_type="pos", mode="learned", features=128)
 
- _ = Embedding(type="pos", mode="learned")
+ _ = Embedding(emb_type="pos", mode="learned", features=128)
 
 
 def test_embedding():
@@ -116,7 +116,7 @@ def test_embedding():
  times = torch.rand(size=(N,))
 
  pos_emb = Embedding(
- type="pos",
+ emb_type="pos",
  mode="fixed",
  features=d_model,
  temperature=objects,
@@ -126,12 +126,30 @@ def test_embedding():
 
  sine_pos_emb = pos_emb(boxes)
 
- pos_emb = Embedding(type="pos", mode="learned", features=d_model, emb_num=100)
+ pos_emb = Embedding(emb_type="pos", mode="learned", features=d_model, emb_num=100)
  learned_pos_emb = pos_emb(boxes)
 
- temp_emb = Embedding(type="temp", mode="learned", features=d_model, emb_num=16)
+ temp_emb = Embedding(emb_type="temp", mode="learned", features=d_model, emb_num=16)
  learned_temp_emb = temp_emb(times)
 
+ pos_emb_off = Embedding(emb_type="pos", mode="off", features=d_model)
+ off_pos_emb = pos_emb_off(boxes)
+
+ temp_emb_off = Embedding(emb_type="temp", mode="off", features=d_model)
+ off_temp_emb = temp_emb_off(times)
+
+ learned_emb_off = Embedding(emb_type="off", mode="learned", features=d_model)
+ off_learned_emb_boxes = learned_emb_off(boxes)
+ off_learned_emb_times = learned_emb_off(times)
+
+ fixed_emb_off = Embedding(emb_type="off", mode="fixed", features=d_model)
+ off_fixed_emb_boxes = fixed_emb_off(boxes)
+ off_fixed_emb_times = fixed_emb_off(times)
+
+ off_emb = Embedding(emb_type="off", mode="off", features=d_model)
+ off_emb_boxes = off_emb(boxes)
+ off_emb_times = off_emb(times)
+
  assert sine_pos_emb.size() == (N, d_model)
  assert learned_pos_emb.size() == (N, d_model)
  assert learned_temp_emb.size() == (N, d_model)
@@ -140,6 +158,24 @@ def test_embedding():
  assert not torch.equal(sine_pos_emb, learned_temp_emb)
  assert not torch.equal(learned_pos_emb, learned_temp_emb)
 
+ assert off_pos_emb.size() == (N, d_model)
+ assert off_temp_emb.size() == (N, d_model)
+ assert off_learned_emb_boxes.size() == (N, d_model)
+ assert off_learned_emb_times.size() == (N, d_model)
+ assert off_fixed_emb_boxes.size() == (N, d_model)
+ assert off_fixed_emb_times.size() == (N, d_model)
+ assert off_emb_boxes.size() == (N, d_model)
+ assert off_emb_times.size() == (N, d_model)
+
+ assert not off_pos_emb.any()
+ assert not off_temp_emb.any()
+ assert not off_learned_emb_boxes.any()
+ assert not off_learned_emb_times.any()
+ assert not off_fixed_emb_boxes.any()
+ assert not off_fixed_emb_times.any()
+ assert not off_emb_boxes.any()
+ assert not off_emb_times.any()
+
 
 def test_embedding_kwargs():
  """Test embedding config logic."""
@@ -154,52 +190,35 @@ def test_embedding_kwargs():
 
  # sine embedding
 
- sine_no_args = Embedding("pos", "fixed")(boxes)
+ sine_no_args = Embedding("pos", "fixed", 128)(boxes)
 
  sine_args = {
  "temperature": objects,
  "scale": frames,
  "normalize": True,
  }
 
- sine_with_args = Embedding("pos", "fixed", **sine_args)(boxes)
+ sine_with_args = Embedding("pos", "fixed", 128, **sine_args)(boxes)
 
  assert not torch.equal(sine_no_args, sine_with_args)
 
  # learned pos embedding
 
- lp_no_args = Embedding("pos", "learned")(boxes)
+ lp_no_args = Embedding("pos", "learned", 128)
 
  lp_args = {"emb_num": 100, "over_boxes": False}
 
- lp_with_args = Embedding("pos", "learned", **lp_args)(boxes)
+ lp_with_args = Embedding("pos", "learned", 128, **lp_args)
+ assert lp_no_args.lookup.weight.shape != lp_with_args.lookup.weight.shape
 
- assert not torch.equal(lp_no_args, lp_with_args)
- assert not torch.equal(lp_no_args, sine_no_args)
- assert not torch.equal(lp_no_args, sine_with_args)
- assert not torch.equal(lp_with_args, sine_no_args)
- assert not torch.equal(lp_with_args, sine_with_args)
  # learned temp embedding
 
- lt_no_args = Embedding("temp", "learned")(times)
+ lt_no_args = Embedding("temp", "learned", 128)
 
  lt_args = {"emb_num": 100}
 
- lt_with_args = Embedding("temp", "learned", **lt_args)(times)
-
- assert not torch.equal(lt_no_args, lt_with_args)
-
- assert not torch.equal(lt_no_args, lp_no_args)
- assert not torch.equal(lt_no_args, lp_with_args)
-
- assert not torch.equal(lt_no_args, sine_no_args)
- assert not torch.equal(lt_no_args, sine_with_args)
-
- assert not torch.equal(lt_with_args, lp_no_args)
- assert not torch.equal(lt_with_args, lp_with_args)
-
- assert not torch.equal(lt_with_args, sine_no_args)
- assert not torch.equal(lt_with_args, sine_with_args)
+ lt_with_args = Embedding("temp", "learned", 128, **lt_args)
+ assert lt_no_args.lookup.weight.shape != lt_with_args.lookup.weight.shape
 
 
 def test_transformer_encoder():

tests/test_training.py

@@ -11,7 +11,8 @@
 from biogtr.config import Config
 from biogtr.training.train import main
 
-# todo: add named tensor tests
+# TODO: add named tensor tests
+# TODO: use temp dir and cleanup after tests (https://docs.pytest.org/en/7.1.x/how-to/tmp_path.html)
 
 
 def test_asso_loss():