Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Add a few docs to the default_config.cfg (#9981) 

* Clarify patience hyperparameter

The current value for patience doesn't seem to indicate that it's
pointing to the number of steps. It may be useful to specify that
explicitly.

Ref: https://github.com/explosion/spaCy/discussions/7450
Ref: https://github.com/explosion/spaCy/discussions/7465

* Update docs for max_steps
This commit is contained in:
Lj Miranda 2022-01-05 16:16:40 +08:00 committed by GitHub
parent 55cf492218
commit 00e7bf5ffd
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
spacy/default_config.cfg
View File

@ -68,12 +68,14 @@ seed = ${system.seed}
gpu_allocator = ${system.gpu_allocator} gpu_allocator = ${system.gpu_allocator}
dropout = 0.1 dropout = 0.1
accumulate_gradient = 1 accumulate_gradient = 1
# Controls early-stopping. 0 disables early stopping. # Controls early-stopping, i.e., the number of steps to continue without
# improvement before stopping. 0 disables early stopping.
patience = 1600 patience = 1600
# Number of epochs. 0 means unlimited. If >= 0, train corpus is loaded once in # Number of epochs. 0 means unlimited. If >= 0, train corpus is loaded once in
# memory and shuffled within the training loop. -1 means stream train corpus # memory and shuffled within the training loop. -1 means stream train corpus
# rather than loading in memory with no shuffling within the training loop. # rather than loading in memory with no shuffling within the training loop.
max_epochs = 0 max_epochs = 0
# Maximum number of update steps to train for. 0 means an unlimited number of steps.
max_steps = 20000 max_steps = 20000
eval_frequency = 200 eval_frequency = 200
# Control how scores are printed and checkpoints are evaluated. # Control how scores are printed and checkpoints are evaluated.