add discard_oversize parameter, move optimizer to training subsection

examples/experiments/onto-joint/defaults.cfg

@@ -25,6 +25,7 @@ score_weights = {"las": 0.4, "ents_f": 0.4, "tags_acc": 0.2}
 # These settings are invalid for the transformer models.
 init_tok2vec = null
 vectors = null
+discard_oversize = false
 
 [training.batch_size]
 @schedules = "compounding.v1"
@@ -32,7 +33,7 @@ start = 1000
 stop = 1000
 compound = 1.001
 
-[optimizer]
+[training.optimizer]
 @optimizers = "Adam.v1"
 beta1 = 0.9
 beta2 = 0.999

examples/experiments/ptb-joint-pos-dep/bilstm_tok2vec.cfg

@@ -14,6 +14,7 @@ score_weights = {"las": 0.8, "tags_acc": 0.2}
 limit = 0
 seed = 0
 accumulate_gradient = 2
+discard_oversize = false
 
 [training.batch_size]
 @schedules = "compounding.v1"
@@ -21,7 +22,7 @@ start = 100
 stop = 1000
 compound = 1.001
 
-[optimizer]
+[training.optimizer]
 @optimizers = "Adam.v1"
 learn_rate = 0.001
 beta1 = 0.9

examples/experiments/ptb-joint-pos-dep/defaults.cfg

@@ -14,6 +14,7 @@ score_weights = {"las": 0.8, "tags_acc": 0.2}
 limit = 0
 seed = 0
 accumulate_gradient = 2
+discard_oversize = false
 
 [training.batch_size]
 @schedules = "compounding.v1"
@@ -21,7 +22,7 @@ start = 100
 stop = 1000
 compound = 1.001
 
-[optimizer]
+[training.optimizer]
 @optimizers = "Adam.v1"
 learn_rate = 0.001
 beta1 = 0.9

examples/experiments/tok2vec-ner/charembed_tok2vec.cfg

@@ -12,8 +12,9 @@ max_length = 0
 batch_size = 25
 seed = 0
 accumulate_gradient = 2
+discard_oversize = false
 
-[optimizer]
+[training.optimizer]
 @optimizers = "Adam.v1"
 learn_rate = 0.001
 beta1 = 0.9

examples/experiments/tok2vec-ner/multihashembed_tok2vec.cfg

@@ -11,6 +11,7 @@ gold_preproc = true
 max_length = 0
 seed = 0
 accumulate_gradient = 2
+discard_oversize = false
 
 [training.batch_size]
 @schedules = "compounding.v1"
@@ -19,7 +20,7 @@ stop = 3000
 compound = 1.001
 
 
-[optimizer]
+[training.optimizer]
 @optimizers = "Adam.v1"
 learn_rate = 0.001
 beta1 = 0.9

spacy/__main__.py

@@ -2,16 +2,15 @@ if __name__ == "__main__":
     import plac
     import sys
     from wasabi import msg
-    from spacy.cli import download, link, info, package, train, pretrain, convert
+    from spacy.cli import download, link, info, package, pretrain, convert
     from spacy.cli import init_model, profile, evaluate, validate, debug_data
-    from spacy.cli import train_from_config_cli
+    from spacy.cli import train_cli
 
     commands = {
         "download": download,
         "link": link,
         "info": info,
-        "train": train,
-        "train-from-config": train_from_config_cli,
+        "train": train_cli,
         "pretrain": pretrain,
         "debug-data": debug_data,
         "evaluate": evaluate,

spacy/cli/__init__.py

@@ -4,8 +4,7 @@ from .download import download  # noqa: F401
 from .info import info  # noqa: F401
 from .package import package  # noqa: F401
 from .profile import profile  # noqa: F401
-from .train import train  # noqa: F401
-from .train_from_config import train_from_config_cli  # noqa: F401
+from .train_from_config import train_cli  # noqa: F401
 from .pretrain import pretrain  # noqa: F401
 from .debug_data import debug_data  # noqa: F401
 from .evaluate import evaluate  # noqa: F401

spacy/cli/train_from_config.py

@@ -374,7 +374,6 @@ def train_while_improving(
         # Stop if we've exhausted our max steps (if specified)
         if max_steps and (step * accumulate_gradient) >= max_steps:
             break
-        step += 1
 
 
 def subdivide_batch(batch, accumulate_gradient):

spacy/ml/__init__.py

@@ -0,0 +1 @@
+from .models import *