explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-11-15 06:09:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

Support cnn_maxout_pieces arg in pretrain

Browse Source
This commit is contained in:
Matthw Honnibal 2019-10-22 03:25:30 +02:00
parent 5a272d9029
commit 7ef3bcdc1c
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/cli/pretrain.py
View File

@ -38,6 +38,7 @@ from .train import _load_pretrained_tok2vec
width=("Width of CNN layers", "option", "cw", int),
depth=("Depth of CNN layers", "option", "cd", int),
cnn_window=("Window size for CNN layers", "option", "cW", int),
cnn_pieces=("Maxout size for CNN layers. 1 for Mish", "option", "cP", int),
use_chars=("Whether to use character-based embedding", "flag", "chr", bool),
sa_depth=("Depth of self-attention layers", "option", "sa", int),
bilstm_depth=("Depth of BiLSTM layers (requires PyTorch)", "option", "lstm", int),
@ -87,6 +88,7 @@ def pretrain(
width=96,
depth=4,
bilstm_depth=0,
cnn_pieces=3,
sa_depth=0,
use_chars=False,
cnn_window=1,
@ -165,7 +167,7 @@ def pretrain(
char_embed=use_chars,
self_attn_depth=sa_depth, # Experimental.
bilstm_depth=bilstm_depth, # Requires PyTorch. Experimental.
cnn_maxout_pieces=3, # You can try setting this higher
cnn_maxout_pieces=cnn_pieces, # You can try setting this higher
subword_features=not use_chars, # Set to False for Chinese etc
),
objective=loss_func