DeepPavlov is an open-source conversational AI library built on TensorFlow, Keras and PyTorch.
DeepPavlov is designed for
- development of production ready chat-bots and complex conversational systems,
- research in the area of NLP and, particularly, of dialog systems.
Quick Links
- Demo demo.deeppavlov.ai
- Documentation docs.deeppavlov.ai
- Model List docs:features/
- Contribution Guide docs:contribution_guide/
- Issues github/issues/
- Forum forum.deeppavlov.ai
- Blogs medium.com/deeppavlov
- Tutorials examples/ and extended colab tutorials
- Docker Hub hub.docker.com/u/deeppavlov/
- Docker Images Documentation docs:docker-images/
Please leave us your feedback on how we can improve the DeepPavlov framework.
Models
Named Entity Recognition | Slot filling
Intent/Sentence Classification | Question Answering over Text (SQuAD)
Knowledge Base Question Answering
Sentence Similarity/Ranking | TF-IDF Ranking
Morphological tagging | Syntactic parsing
Automatic Spelling Correction | ELMo training and fine-tuning
Speech recognition and synthesis (ASR and TTS) based on NVIDIA NeMo
Entity Linking | Multitask BERT
Skills
Goal(Task)-oriented Bot | Seq2seq Goal-Oriented bot
Open Domain Questions Answering | eCommerce Bot
Frequently Asked Questions Answering | Pattern Matching
Embeddings
BERT embeddings for the Russian, Polish, Bulgarian, Czech, and informal English
ELMo embeddings for the Russian language
FastText embeddings for the Russian language
Auto ML
Tuning Models with Evolutionary Algorithm
Integrations
REST API | Socket API | Yandex Alice
Telegram | Microsoft Bot Framework
Installation
-
We support
LinuxandWindowsplatforms,Python 3.6andPython 3.7Python 3.5is not supported!- installation for
WindowsrequiresGit(for example, git) andVisual Studio 2015/2017withC++build tools installed!
-
Create and activate a virtual environment:
Linux
python -m venv env source ./env/bin/activateWindows
python -m venv env .\env\Scripts\activate.bat -
Install the package inside the environment:
pip install deeppavlov
QuickStart
There is a bunch of great pre-trained NLP models in DeepPavlov. Each model is determined by its config file.
List of models is available on the doc page in the deeppavlov.configs (Python):
from deeppavlov import configs
When you're decided on the model (+ config file), there are two ways to train, evaluate and infer it:
- via Command line interface (CLI) and
- via Python.
GPU requirements
To run supported DeepPavlov models on GPU you should have CUDA 10.0 installed on your host machine and TensorFlow with GPU support (tensorflow-gpu) installed in your python environment. Current supported TensorFlow version is 1.15.2. Run
pip install tensorflow-gpu==1.15.2
before installing model's package requirements to install supported tensorflow-gpu version.
Before making choice of an interface, install model's package requirements (CLI):
python -m deeppavlov install <config_path>
- where
<config_path>is path to the chosen model's config file (e.g.deeppavlov/configs/ner/slotfill_dstc2.json) or just name without .json extension (e.g.slotfill_dstc2)
Command line interface (CLI)
To get predictions from a model interactively through CLI, run
python -m deeppavlov interact <config_path> [-d]
-ddownloads required data -- pretrained model files and embeddings (optional).
You can train it in the same simple way:
python -m deeppavlov train <config_path> [-d]
Dataset will be downloaded regardless of whether there was -d flag or not.
To train on your own data you need to modify dataset reader path in the train config doc. The data format is specified in the corresponding model doc page.
There are even more actions you can perform with configs:
python -m deeppavlov <action> <config_path> [-d]
<action>can bedownloadto download model's data (same as-d),trainto train the model on the data specified in the config file,evaluateto calculate metrics on the same dataset,interactto interact via CLI,riseapito run a REST API server (see doc),telegramto run as a Telegram bot (see doc),msbotto run a Miscrosoft Bot Framework server (see doc),predictto get prediction for samples from stdin or from <file_path> if-f <file_path>is specified.
<config_path>specifies path (or name) of model's config file-ddownloads required data
Python
To get predictions from a model interactively through Python, run
from deeppavlov import build_model
model = build_model(<config_path>, download=True)
# get predictions for 'input_text1', 'input_text2'
model(['input_text1', 'input_text2'])
- where
download=Truedownloads required data from web -- pretrained model files and embeddings (optional), <config_path>is path to the chosen model's config file (e.g."deeppavlov/configs/ner/ner_ontonotes_bert_mult.json") ordeeppavlov.configsattribute (e.g.deeppavlov.configs.ner.ner_ontonotes_bert_multwithout quotation marks).
You can train it in the same simple way:
from deeppavlov import train_model
model = train_model(<config_path>, download=True)
download=Truedownloads pretrained model, therefore the pretrained model will be, first, loaded and then train (optional).
Dataset will be downloaded regardless of whether there was -d flag or not.
To train on your own data you need to modify dataset reader path in the train config doc. The data format is specified in the corresponding model doc page.
You can also calculate metrics on the dataset specified in your config file:
from deeppavlov import evaluate_model
model = evaluate_model(<config_path>, download=True)
There are also available integrations with various messengers, see Telegram Bot doc page and others in the Integrations section for more info.
Breaking Changes
Breaking changes in version 0.7.0
- in dialog logger config file dialog_logger_config.json
agent_nameparameter was renamed tologger_name, the default value was changed - Agent, Skill, eCommerce Bot and Pattern Matching classes were moved to deeppavlov.deprecated
- AIML Skill, RASA Skill, Yandex Alice, Amazon Alexa, Microsoft Bot Framework and Telegram integration interfaces were changed
/startand/helpTelegram messages were moved frommodels_info.jsonto server_config.json- risesocket request and response format was changed
- riseapi and risesocket model-specific properties parametrization was changed
Breaking changes in version 0.6.0
- REST API:
- all models default endpoints were renamed to
/model - by default model arguments names are taken from
chainer.inconfiguration parameter instead of pre-set names from a settings file - swagger api endpoint moved from
/apidocsto/docs
- all models default endpoints were renamed to
- when using
"max_proba": truein aproba2labelscomponent for classification, it will return single label for every batch element instead of a list. One can set"top_n": 1to get batches of single item lists as before
Breaking changes in version 0.5.0
- dependencies have to be reinstalled for most pipeline configurations
- models depending on
tensorflowrequireCUDA 10.0to run on GPU instead ofCUDA 9.0 - scikit-learn models have to be redownloaded or retrained
Breaking changes in version 0.4.0!
- default target variable name for neural evolution was changed from
MODELS_PATHtoMODEL_PATH.
Breaking changes in version 0.3.0!
- component option
fit_on_batchin configuration files was removed and replaced with adaptive usage of thefit_onparameter.
Breaking changes in version 0.2.0!
utilsmodule was moved from repository root in todeeppavlovmodulems_bot_framework_utils,server_utils,telegram utilsmodules was renamed toms_bot_framework,serverandtelegramcorrespondingly- rename metric functions
exact_matchtosquad_v2_emandsquad_f1tosquad_v2_f1 - replace dashes in configs name with underscores
Breaking changes in version 0.1.0!
-
As of
version 0.1.0all models, embeddings and other downloaded data for provided configurations are by default downloaded to the.deeppavlovdirectory in current user's home directory. This can be changed on per-model basis by modifying aROOT_PATHvariable or related fields one by one in model's configuration file. -
In configuration files, for all features/models, dataset readers and iterators
"name"and"class"fields are combined into the"class_name"field. -
deeppavlov.core.commands.infer.build_model_from_config()was renamed tobuild_modeland can be imported from thedeeppavlovmodule directly. -
The way arguments are passed to metrics functions during training and evaluation was changed and documented.
License
DeepPavlov is Apache 2.0 - licensed.
The Team
DeepPavlov is built and maintained by Neural Networks and Deep Learning Lab at MIPT.
42 Dec 20, 2022
14 Aug 24, 2022
3 Nov 17, 2022
106 Jan 01, 2023
102 Dec 19, 2022
48 Dec 14, 2022
409 Oct 28, 2022
475 Jan 04, 2023
110 Jan 01, 2023
99 Jan 06, 2023
95 Jan 05, 2023
2 Feb 08, 2022
37 Dec 25, 2022
504 Dec 20, 2022
177 Jan 04, 2023
0 Jun 28, 2022
3 Dec 20, 2022
16 Nov 12, 2022
1.1k Jan 03, 2023
9 Jan 01, 2023