Skip to content

Add ModelCheckpoints to TrainKerasWidget. Closes #1987 #1990

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 3 commits into
base: master
Choose a base branch
from
Draft

Add ModelCheckpoints to TrainKerasWidget. Closes #1987 #1990

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
afb8271
WIP- #1987 generate schemas for callbacks
umesh-timalsina Jan 27, 2021
ecdec13
WIP- Fix list style arguments in callbacks
umesh-timalsina Jan 27, 2021
c398281
WIP- Add History to callback skip names
umesh-timalsina Jan 27, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 24 additions & 8 deletions src/visualizers/widgets/TrainKeras/schemas/generate-schemas.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,43 +1,58 @@
import tensorflow as tf
from tensorflow import keras
import inspect
import json

def is_class_name(n, skip_names=[]):
from tensorflow import keras


def is_class_name(n, skip_names=None):
if skip_names is None:
skip_names = []
return n[0].isupper() and n not in skip_names


def parse_schema(mod_name, module, name):
class_ = getattr(module, name)
spec = inspect.getfullargspec(class_)
spec = inspect.getfullargspec(class_.__init__)
ctor_args = spec.args[1:]
kw_arg_start_index = len(ctor_args)-len(spec.defaults)
kw_args = list(zip(ctor_args[kw_arg_start_index:], spec.defaults))
kw_arg_start_index = -1
if spec.defaults is not None:
kw_arg_start_index = len(ctor_args)-len(spec.defaults)
kw_args = list(zip(ctor_args[kw_arg_start_index:], spec.defaults if spec.defaults is not None else []))
pos_args = list(zip(ctor_args[0:kw_arg_start_index]))
args = [ (name, None) for name in pos_args ]
args.extend(kw_args)

return {
'name': name,
#'docstring': inspect.getdoc(class_),
'arguments': [ {'name': n, 'default': d} for (n, d) in args ],
'arguments': [ {'name': n if isinstance(n, str) else n[0], 'default': d} for (n, d) in args ],
#'url': f'https://keras.io/api/{mod_name}/{name.lower()}/'
}

def parse_module_schemas(module, skip_names=[]):

def parse_module_schemas(module, skip_names=None):
if skip_names is None:
skip_names = []

mod_name = module.__name__.split('.')[-1]
mod_names = ( n for n in dir(module) if is_class_name(n, skip_names) )
class_names = ( n for n in mod_names if True ) # type(getattr(module, n)) is type)
schemas = ( parse_schema(mod_name, module, n) for n in class_names )
return [ schema for schema in schemas if schema is not None ]


all_schemas = {}
all_schemas['optimizers'] = parse_module_schemas(keras.optimizers, ['Optimizer'])
all_schemas['losses'] = parse_module_schemas(keras.losses, ['Loss', 'Reduction', 'KLD', 'MAE', 'MAPE', 'MSE', 'MSLE'])
all_schemas['reductions'] = [ getattr(keras.losses.Reduction, name) for name in dir(keras.losses.Reduction) if name[0].isupper() ]
all_schemas['callbacks'] = parse_module_schemas(keras.callbacks, ['BaseLogger', 'Callback', 'History', 'CallbackList', 'ProgbarLogger'])


def is_regression(loss_name):
other_losses = ['CosineSimilarity', 'LogCosh', 'Huber']
return 'Error' in loss_name or loss_name in other_losses


def add_loss_category(loss):
if 'Hinge' in loss['name']:
category = 'Hinge'
Expand All @@ -48,6 +63,7 @@ def add_loss_category(loss):

loss['category'] = category + ' losses'


for loss in all_schemas['losses']:
add_loss_category(loss)

Expand Down
2 changes: 1 addition & 1 deletion src/visualizers/widgets/TrainKeras/schemas/index.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1 +1 @@
{"optimizers": [{"name": "Adadelta", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "rho", "default": 0.95}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Adadelta"}]}, {"name": "Adagrad", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "initial_accumulator_value", "default": 0.1}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Adagrad"}]}, {"name": "Adam", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "beta_1", "default": 0.9}, {"name": "beta_2", "default": 0.999}, {"name": "epsilon", "default": 1e-07}, {"name": "amsgrad", "default": false}, {"name": "name", "default": "Adam"}]}, {"name": "Adamax", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "beta_1", "default": 0.9}, {"name": "beta_2", "default": 0.999}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Adamax"}]}, {"name": "Ftrl", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "learning_rate_power", "default": -0.5}, {"name": "initial_accumulator_value", "default": 0.1}, {"name": "l1_regularization_strength", "default": 0.0}, {"name": "l2_regularization_strength", "default": 0.0}, {"name": "name", "default": "Ftrl"}, {"name": "l2_shrinkage_regularization_strength", "default": 0.0}]}, {"name": "Nadam", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "beta_1", "default": 0.9}, {"name": "beta_2", "default": 0.999}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Nadam"}]}, {"name": "RMSprop", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "rho", "default": 0.9}, {"name": "momentum", "default": 0.0}, {"name": "epsilon", "default": 1e-07}, {"name": "centered", "default": false}, {"name": "name", "default": "RMSprop"}]}, {"name": "SGD", "arguments": [{"name": "learning_rate", "default": 0.01}, {"name": "momentum", "default": 0.0}, {"name": "nesterov", "default": false}, {"name": "name", "default": "SGD"}]}], "losses": [{"name": "BinaryCrossentropy", "arguments": [{"name": "from_logits", "default": false}, {"name": "label_smoothing", "default": 0}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "binary_crossentropy"}], "category": "Probabilistic losses"}, {"name": "CategoricalCrossentropy", "arguments": [{"name": "from_logits", "default": false}, {"name": "label_smoothing", "default": 0}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "categorical_crossentropy"}], "category": "Probabilistic losses"}, {"name": "CategoricalHinge", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "categorical_hinge"}], "category": "Hinge losses"}, {"name": "CosineSimilarity", "arguments": [{"name": "axis", "default": -1}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "cosine_similarity"}], "category": "Regression losses"}, {"name": "Hinge", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "hinge"}], "category": "Hinge losses"}, {"name": "Huber", "arguments": [{"name": "delta", "default": 1.0}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "huber_loss"}], "category": "Regression losses"}, {"name": "KLDivergence", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "kullback_leibler_divergence"}], "category": "Probabilistic losses"}, {"name": "LogCosh", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "logcosh"}], "category": "Regression losses"}, {"name": "MeanAbsoluteError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_absolute_error"}], "category": "Regression losses"}, {"name": "MeanAbsolutePercentageError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_absolute_percentage_error"}], "category": "Regression losses"}, {"name": "MeanSquaredError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_squared_error"}], "category": "Regression losses"}, {"name": "MeanSquaredLogarithmicError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_squared_logarithmic_error"}], "category": "Regression losses"}, {"name": "Poisson", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "poisson"}], "category": "Probabilistic losses"}, {"name": "SparseCategoricalCrossentropy", "arguments": [{"name": "from_logits", "default": false}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "sparse_categorical_crossentropy"}], "category": "Probabilistic losses"}, {"name": "SquaredHinge", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "squared_hinge"}], "category": "Hinge losses"}], "reductions": ["auto", "none", "sum", "sum_over_batch_size"]}
{"optimizers": [{"name": "Adadelta", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "rho", "default": 0.95}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Adadelta"}]}, {"name": "Adagrad", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "initial_accumulator_value", "default": 0.1}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Adagrad"}]}, {"name": "Adam", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "beta_1", "default": 0.9}, {"name": "beta_2", "default": 0.999}, {"name": "epsilon", "default": 1e-07}, {"name": "amsgrad", "default": false}, {"name": "name", "default": "Adam"}]}, {"name": "Adamax", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "beta_1", "default": 0.9}, {"name": "beta_2", "default": 0.999}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Adamax"}]}, {"name": "Ftrl", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "learning_rate_power", "default": -0.5}, {"name": "initial_accumulator_value", "default": 0.1}, {"name": "l1_regularization_strength", "default": 0.0}, {"name": "l2_regularization_strength", "default": 0.0}, {"name": "name", "default": "Ftrl"}, {"name": "l2_shrinkage_regularization_strength", "default": 0.0}]}, {"name": "Nadam", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "beta_1", "default": 0.9}, {"name": "beta_2", "default": 0.999}, {"name": "epsilon", "default": 1e-07}, {"name": "name", "default": "Nadam"}]}, {"name": "RMSprop", "arguments": [{"name": "learning_rate", "default": 0.001}, {"name": "rho", "default": 0.9}, {"name": "momentum", "default": 0.0}, {"name": "epsilon", "default": 1e-07}, {"name": "centered", "default": false}, {"name": "name", "default": "RMSprop"}]}, {"name": "SGD", "arguments": [{"name": "learning_rate", "default": 0.01}, {"name": "momentum", "default": 0.0}, {"name": "nesterov", "default": false}, {"name": "name", "default": "SGD"}]}], "losses": [{"name": "BinaryCrossentropy", "arguments": [{"name": "from_logits", "default": false}, {"name": "label_smoothing", "default": 0}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "binary_crossentropy"}], "category": "Probabilistic losses"}, {"name": "CategoricalCrossentropy", "arguments": [{"name": "from_logits", "default": false}, {"name": "label_smoothing", "default": 0}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "categorical_crossentropy"}], "category": "Probabilistic losses"}, {"name": "CategoricalHinge", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "categorical_hinge"}], "category": "Hinge losses"}, {"name": "CosineSimilarity", "arguments": [{"name": "axis", "default": -1}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "cosine_similarity"}], "category": "Regression losses"}, {"name": "Hinge", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "hinge"}], "category": "Hinge losses"}, {"name": "Huber", "arguments": [{"name": "delta", "default": 1.0}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "huber_loss"}], "category": "Regression losses"}, {"name": "KLDivergence", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "kl_divergence"}], "category": "Probabilistic losses"}, {"name": "LogCosh", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "log_cosh"}], "category": "Regression losses"}, {"name": "MeanAbsoluteError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_absolute_error"}], "category": "Regression losses"}, {"name": "MeanAbsolutePercentageError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_absolute_percentage_error"}], "category": "Regression losses"}, {"name": "MeanSquaredError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_squared_error"}], "category": "Regression losses"}, {"name": "MeanSquaredLogarithmicError", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "mean_squared_logarithmic_error"}], "category": "Regression losses"}, {"name": "Poisson", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "poisson"}], "category": "Probabilistic losses"}, {"name": "SparseCategoricalCrossentropy", "arguments": [{"name": "from_logits", "default": false}, {"name": "reduction", "default": "auto"}, {"name": "name", "default": "sparse_categorical_crossentropy"}], "category": "Probabilistic losses"}, {"name": "SquaredHinge", "arguments": [{"name": "reduction", "default": "auto"}, {"name": "name", "default": "squared_hinge"}], "category": "Hinge losses"}], "reductions": ["auto", "none", "sum", "sum_over_batch_size"], "callbacks": [{"name": "CSVLogger", "arguments": [{"name": "filename", "default": null}, {"name": "separator", "default": ","}, {"name": "append", "default": false}]}, {"name": "EarlyStopping", "arguments": [{"name": "monitor", "default": "val_loss"}, {"name": "min_delta", "default": 0}, {"name": "patience", "default": 0}, {"name": "verbose", "default": 0}, {"name": "mode", "default": "auto"}, {"name": "baseline", "default": null}, {"name": "restore_best_weights", "default": false}]}, {"name": "LambdaCallback", "arguments": [{"name": "on_epoch_begin", "default": null}, {"name": "on_epoch_end", "default": null}, {"name": "on_batch_begin", "default": null}, {"name": "on_batch_end", "default": null}, {"name": "on_train_begin", "default": null}, {"name": "on_train_end", "default": null}]}, {"name": "LearningRateScheduler", "arguments": [{"name": "schedule", "default": null}, {"name": "verbose", "default": 0}]}, {"name": "ModelCheckpoint", "arguments": [{"name": "filepath", "default": null}, {"name": "monitor", "default": "val_loss"}, {"name": "verbose", "default": 0}, {"name": "save_best_only", "default": false}, {"name": "save_weights_only", "default": false}, {"name": "mode", "default": "auto"}, {"name": "save_freq", "default": "epoch"}, {"name": "options", "default": null}]}, {"name": "ReduceLROnPlateau", "arguments": [{"name": "monitor", "default": "val_loss"}, {"name": "factor", "default": 0.1}, {"name": "patience", "default": 10}, {"name": "verbose", "default": 0}, {"name": "mode", "default": "auto"}, {"name": "min_delta", "default": 0.0001}, {"name": "cooldown", "default": 0}, {"name": "min_lr", "default": 0}]}, {"name": "RemoteMonitor", "arguments": [{"name": "root", "default": "http://localhost:9000"}, {"name": "path", "default": "/publish/epoch/end/"}, {"name": "field", "default": "data"}, {"name": "headers", "default": null}, {"name": "send_as_json", "default": false}]}, {"name": "TensorBoard", "arguments": [{"name": "log_dir", "default": "logs"}, {"name": "histogram_freq", "default": 0}, {"name": "write_graph", "default": true}, {"name": "write_images", "default": false}, {"name": "update_freq", "default": "epoch"}, {"name": "profile_batch", "default": 2}, {"name": "embeddings_freq", "default": 0}, {"name": "embeddings_metadata", "default": null}]}, {"name": "TerminateOnNaN", "arguments": []}]}