mirror of
https://github.com/deepset-ai/haystack.git
synced 2025-10-31 09:49:48 +00:00
13510aa753
* Files moved, imports all broken * Fix most imports and docstrings into * Fix the paths to the modules in the API docs * Add latest docstring and tutorial changes * Add a few pipelines that were lost in the inports * Fix a bunch of mypy warnings * Add latest docstring and tutorial changes * Create a file_classifier module * Add docs for file_classifier * Fixed most circular imports, now the REST API can start * Add latest docstring and tutorial changes * Tackling more mypy issues * Reintroduce from FARM and fix last mypy issues hopefully * Re-enable old-style imports * Fix some more import from the top-level package in an attempt to sort out circular imports * Fix some imports in tests to new-style to prevent failed class equalities from breaking tests * Change document_store into document_stores * Update imports in tutorials * Add latest docstring and tutorial changes * Probably fixes summarizer tests * Improve the old-style import allowing module imports (should work) * Try to fix the docs * Remove dedicated KnowledgeGraph page from autodocs * Remove dedicated GraphRetriever page from autodocs * Fix generate_docstrings.sh with an updated list of yaml files to look for * Fix some more modules in the docs * Fix the document stores docs too * Fix a small issue on Tutorial14 * Add latest docstring and tutorial changes * Add deprecation warning to old-style imports * Remove stray folder and import Dict into dense.py * Change import path for MLFlowLogger * Add old loggers path to the import path aliases * Fix debug output of convert_ipynb.py * Fix circular import on BaseRetriever * Missed one merge block * re-run tutorial 5 * Fix imports in tutorial 5 * Re-enable squad_to_dpr CLI from the root package and move get_batches_from_generator into document_stores.base * Add latest docstring and tutorial changes * Fix typo in utils __init__ * Fix a few more imports * Fix benchmarks too * New-style imports in test_knowledge_graph * Rollback setup.py * Rollback squad_to_dpr too Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
86 lines
2.7 KiB
Python
86 lines
2.7 KiB
Python
def tutorial9_dpr_training():
|
|
# Training Your Own "Dense Passage Retrieval" Model
|
|
|
|
# Here are some imports that we'll need
|
|
|
|
from haystack.nodes import DensePassageRetriever
|
|
from haystack.utils import fetch_archive_from_http
|
|
from haystack.document_stores import InMemoryDocumentStore
|
|
|
|
# Download original DPR data
|
|
# WARNING: the train set is 7.4GB and the dev set is 800MB
|
|
|
|
doc_dir = "data/dpr_training/"
|
|
|
|
s3_url_train = "https://dl.fbaipublicfiles.com/dpr/data/retriever/biencoder-nq-train.json.gz"
|
|
s3_url_dev = "https://dl.fbaipublicfiles.com/dpr/data/retriever/biencoder-nq-dev.json.gz"
|
|
|
|
fetch_archive_from_http(s3_url_train, output_dir=doc_dir + "train/")
|
|
fetch_archive_from_http(s3_url_dev, output_dir=doc_dir + "dev/")
|
|
|
|
## Option 1: Training DPR from Scratch
|
|
|
|
# Here are the variables to specify our training data, the models that we use to initialize DPR
|
|
# and the directory where we'll be saving the model
|
|
|
|
doc_dir = "data/dpr_training/"
|
|
|
|
train_filename = "train/biencoder-nq-train.json"
|
|
dev_filename = "dev/biencoder-nq-dev.json"
|
|
|
|
query_model = "bert-base-uncased"
|
|
passage_model = "bert-base-uncased"
|
|
|
|
save_dir = "../saved_models/dpr"
|
|
|
|
# ## Option 2: Finetuning DPR
|
|
#
|
|
# # Here are the variables you might want to use instead of the set above
|
|
# # in order to perform pretraining
|
|
#
|
|
# doc_dir = "PATH_TO_YOUR_DATA_DIR"
|
|
# train_filename = "TRAIN_FILENAME"
|
|
# dev_filename = "DEV_FILENAME"
|
|
#
|
|
# query_model = "facebook/dpr-question_encoder-single-nq-base"
|
|
# passage_model = "facebook/dpr-ctx_encoder-single-nq-base"
|
|
#
|
|
# save_dir = "..saved_models/dpr"
|
|
|
|
## Initialize DPR model
|
|
|
|
retriever = DensePassageRetriever(
|
|
document_store=InMemoryDocumentStore(),
|
|
query_embedding_model=query_model,
|
|
passage_embedding_model=passage_model,
|
|
max_seq_len_query=64,
|
|
max_seq_len_passage=256
|
|
)
|
|
|
|
# Start training our model and save it when it is finished
|
|
|
|
retriever.train(
|
|
data_dir=doc_dir,
|
|
train_filename=train_filename,
|
|
dev_filename=dev_filename,
|
|
test_filename=dev_filename,
|
|
n_epochs=1,
|
|
batch_size=16,
|
|
grad_acc_steps=8,
|
|
save_dir=save_dir,
|
|
evaluate_every=3000,
|
|
embed_title=True,
|
|
num_positives=1,
|
|
num_hard_negatives=1
|
|
)
|
|
|
|
## Loading
|
|
|
|
reloaded_retriever = DensePassageRetriever.load(load_dir=save_dir, document_store=None)
|
|
|
|
if __name__ == "__main__":
|
|
tutorial9_dpr_training()
|
|
|
|
# This Haystack script was made with love by deepset in Berlin, Germany
|
|
# Haystack: https://github.com/deepset-ai/haystack
|
|
# deepset: https://deepset.ai/ |