46 lines
5.9 KiB
Plaintext
46 lines
5.9 KiB
Plaintext
|
Ë
|
|||
|
|
=û g ã ó– — d dl Z d dlmZ d dlmZ d dlZd dlmZ ddlm Z ddl
|
|||
|
|
m Z dd l m
|
|||
|
|
Z
|
|||
|
|
mZmZmZmZmZmZ ej( d
|
|||
|
|
d d d œ¬
|
|||
|
|
« e
|
|||
|
|
ddd dd ¬« e
|
|||
|
|
dd¬« edddd¬« edddd¬« edddd ¬« ed d d d ¬« edd!d"d#¬« fd$ej* d%ed&ed'ee d(ee d)ee d*ed+efd,„« Zd-„ Zy).é N)ÚPath)ÚOptional)Úmsgé )Úpretrain)Ú load_configé )ÚArgÚOptÚappÚ import_codeÚparse_config_overridesÚ setup_gpuÚshow_validation_errorr T)Úallow_extra_argsÚignore_unknown_options)Úcontext_settings.zPath to config fileF)ÚhelpÚexistsÚdir_okayÚ
|
|||
|
|
allow_dashz+Directory to write weights to on each epoch)r z--codez-czNPath to Python file with additional code (registered functions) to be importedz
|
|||
|
|
--resume-pathz-rz;Path to pretrained weights from which to resume pretrainingz--epoch-resumez-erzuThe epoch to resume counting from when using --resume-path. Prevents unintended overwriting of existing weight files.éÿÿÿÿz--gpu-idz-gzGPU ID or -1 for CPUz --skip-lastz-LzSkip saving model-last.binÚctxÚ config_pathÚ
|
|||
|
|
output_dirÚ code_pathÚ resume_pathÚ epoch_resumeÚuse_gpuÚ skip_lastc ó† — t | j « }t |« t ||||« t |« t j d|› �« t |« 5 t ||d¬« } ddd« j « }
|
|||
|
|
|
|
|||
|
|
j d« st j dd¬« |j « s*|j d ¬
|
|||
|
|
« t j d |› �« | j |d z « t j d
|
|||
|
|
« t! |
|
|||
|
|
||||d|¬« t j d« y# 1 sw Y ŒÌxY w)a
|
|||
|
|
Pre-train the 'token-to-vector' (tok2vec) layer of pipeline components,
|
|||
|
|
using an approximate language-modelling objective. Two objective types
|
|||
|
|
are available, vector-based and character-based.
|
|||
|
|
|
|||
|
|
In the vector-based objective, we load word vectors that have been trained
|
|||
|
|
using a word2vec-style distributional similarity algorithm, and train a
|
|||
|
|
component like a CNN, BiLSTM, etc to predict vectors which match the
|
|||
|
|
pretrained ones. The weights are saved to a directory after each epoch. You
|
|||
|
|
can then pass a path to one of these pretrained weights files to the
|
|||
|
|
'spacy train' command.
|
|||
|
|
|
|||
|
|
This technique may be especially helpful if you have little labelled data.
|
|||
|
|
However, it's still quite experimental, so your mileage may vary.
|
|||
|
|
|
|||
|
|
To load the weights back in during 'spacy train', you need to ensure
|
|||
|
|
all settings are the same between pretraining and training. Ideally,
|
|||
|
|
this is done by using the same config file for both commands.
|
|||
|
|
|
|||
|
|
DOCS: https://spacy.io/api/cli#pretrain
|
|||
|
|
zLoading config from: F)Ú overridesÚ interpolateNÚ pretrainingz/The [pretraining] block in your config is emptyr ©ÚexitsT)ÚparentszCreated output directory: z
|
|||
|
|
config.cfgz)Saved config file in the output directory)r r r Úsilentr z Successfully finished pretrain)r Úargsr
|
|||
|
|
Úverify_cli_argsr r Úinfor r r# ÚgetÚfailr ÚmkdirÚgoodÚto_diskr ) r r r r r r r r Úconfig_overridesÚ
|
|||
|
|
raw_configÚconfigs úSC:\Users\garci\AppData\Roaming\Python\Python312\site-packages\spacy/cli/pretrain.pyÚ pretrain_clir5 s € ôJ .¨c¯h©hÓ7ÐÜ� ÔÜ�K ¨[¸,ÔGÜ
|
|||
|
|
ˆgÔ܇H�HÐ$ [ MÐ
|
|||
|
|
2Ô3ä ˜{Õ +Ü Ø Ð#3Àô
|
|||
|
|
ˆ
|
|||
|
|
÷
|
|||
|
|
,ð ×
|
|||
|
|
#Ñ
|