explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-07-12 17:22:25 +03:00
Code Issues Packages Projects Releases Wiki Activity

Auto-format

Browse Source
This commit is contained in:
Ines Montani 2019-07-10 10:27:29 +02:00
parent 205c73a589
commit 881f5bc401
1 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
website/docs/usage/training.md
View File

@ -630,13 +630,13 @@ should be somewhat larger, especially if your documents are long.
### Learning rate, regularization and gradient clipping {#tips-hyperparams} ### Learning rate, regularization and gradient clipping {#tips-hyperparams}
By default spaCy uses the Adam solver, with default settings (`learn_rate=0.001`, By default spaCy uses the Adam solver, with default settings
`beta1=0.9`, `beta2=0.999`). Some researchers have said they found (`learn_rate=0.001`, `beta1=0.9`, `beta2=0.999`). Some researchers have said
these settings terrible on their problems – but they've always performed very they found these settings terrible on their problems – but they've always
well in training spaCy's models, in combination with the rest of our recipe. You performed very well in training spaCy's models, in combination with the rest of
can change these settings directly, by modifying the corresponding attributes on our recipe. You can change these settings directly, by modifying the
the `optimizer` object. You can also set environment variables, to adjust the corresponding attributes on the `optimizer` object. You can also set environment
defaults. variables, to adjust the defaults.
There are two other key hyper-parameters of the solver: `L2` **regularization**, There are two other key hyper-parameters of the solver: `L2` **regularization**,
and **gradient clipping** (`max_grad_norm`). Gradient clipping is a hack that's and **gradient clipping** (`max_grad_norm`). Gradient clipping is a hack that's