document_store.update_embeddings(retriever)
document_store.save("wiki_dump_embeddings.faiss")
# example:
# retrieved_doc = retriever.retrieve(query="Why did the revenue increase?")
# tune these parameters too
print("Running QA Reader")
reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2",
use_gpu=True,
no_ans_boost=-10,
context_window_size=500,
top_k_per_candidate=3,
top_k_per_sample=1,
num_processes=1,
max_seq_len=256,
doc_stride=128)
# example:
# reader.predict(question="Who is the father of Arya Starck?", documents=retrieved_docs, top_k=3)
print("Started pipeline")
p = Pipeline()
p.add_node(component=retriever, name="ESRetriever1", inputs=["Query"])
p.add_node(component=reader, name="QAReader", inputs=["ESRetriever1"])
res = p.run(query="What did Einstein work on?",
params={"retriever": {
"top_k": 1
}})
json.dump(res, open('answer.json', 'w'), default=dumper)
exit()
# Creating Pipeline Diagrams #
##############################
p_extractive_premade.draw("pipeline_extractive_premade.png")
p_retrieval.draw("pipeline_retrieval.png")
p_generator.draw("pipeline_generator.png")
####################
# Custom Pipelines #
####################
# Extractive QA Pipeline
########################
# Custom built extractive QA pipeline
p_extractive = Pipeline()
p_extractive.add_node(component=es_retriever,
name="Retriever",
inputs=["Query"])
p_extractive.add_node(component=reader, name="Reader", inputs=["Retriever"])
# Now we can run it
res = p_extractive.run(query="Who is the father of Arya Stark?",
top_k_retriever=10,
top_k_reader=5)
print_answers(res, details="minimal")
p_extractive.draw("pipeline_extractive.png")
# Ensembled Retriever Pipeline
##############################
def test_eval_pipeline(document_store: BaseDocumentStore, reader, retriever):
# add eval data (SQUAD format)
document_store.add_eval_data(
filename="samples/squad/tiny.json",
doc_index="haystack_test_eval_document",
label_index="haystack_test_feedback",
)
labels = document_store.get_all_labels_aggregated(index="haystack_test_feedback")
eval_retriever = EvalDocuments()
eval_reader = EvalAnswers(sas_model="sentence-transformers/paraphrase-MiniLM-L3-v2",debug=True)
eval_reader_cross = EvalAnswers(sas_model="cross-encoder/stsb-TinyBERT-L-4",debug=True)
eval_reader_vanila = EvalAnswers()
assert document_store.get_document_count(index="haystack_test_eval_document") == 2
p = Pipeline()
p.add_node(component=retriever, name="ESRetriever", inputs=["Query"])
p.add_node(component=eval_retriever, name="EvalDocuments", inputs=["ESRetriever"])
p.add_node(component=reader, name="QAReader", inputs=["EvalDocuments"])
p.add_node(component=eval_reader, name="EvalAnswers", inputs=["QAReader"])
p.add_node(component=eval_reader_cross, name="EvalAnswers_cross", inputs=["QAReader"])
p.add_node(component=eval_reader_vanila, name="EvalAnswers_vanilla", inputs=["QAReader"])
for l in labels:
res = p.run(
query=l.question,
top_k_retriever=10,
labels=l,
top_k_reader=10,
index="haystack_test_eval_document",
)
assert eval_retriever.recall == 1.0
assert round(eval_reader.top_k_f1, 4) == 0.8333
assert eval_reader.top_k_em == 0.5
assert round(eval_reader.top_k_sas, 3) == 0.800
assert round(eval_reader_cross.top_k_sas, 3) == 0.671
assert eval_reader.top_k_em == eval_reader_vanila.top_k_em
probability: Optional[float] = None
context: Optional[str]
offset_start: int
offset_end: int
offset_start_in_doc: Optional[int]
offset_end_in_doc: Optional[int]
document_id: Optional[str] = None
meta: Optional[Dict[str, Any]]
class Response(BaseModel):
query: str
answers: List[Answer]
PIPELINE = Pipeline.load_from_yaml(Path(PIPELINE_YAML_PATH),
pipeline_name=QUERY_PIPELINE_NAME)
logger.info(f"Loaded pipeline nodes: {PIPELINE.graph.nodes.keys()}")
concurrency_limiter = RequestLimiter(CONCURRENT_REQUEST_PER_WORKER)
@router.post("/query", response_model=Response)
def query(request: Request):
with concurrency_limiter.run():
result = _process_request(PIPELINE, request)
return result
def _process_request(pipeline, request) -> Response:
start_time = time.time()
filters = {}
def tutorial5_evaluation():
##############################################
# Settings
##############################################
# Choose from Evaluation style from ['retriever_closed', 'reader_closed', 'retriever_reader_open']
# 'retriever_closed' - evaluates only the retriever, based on whether the gold_label document is retrieved.
# 'reader_closed' - evaluates only the reader in a closed domain fashion i.e. the reader is given one query
# and one document and metrics are calculated on whether the right position in this text is selected by
# the model as the answer span (i.e. SQuAD style)
# 'retriever_reader_open' - evaluates retriever and reader in open domain fashion i.e. a document is considered
# correctly retrieved if it contains the answer string within it. The reader is evaluated based purely on the
# predicted string, regardless of which document this came from and the position of the extracted span.
style = "retriever_reader_open"
# make sure these indices do not collide with existing ones, the indices will be wiped clean before data is inserted
doc_index = "tutorial5_docs"
label_index = "tutorial5_labels"
##############################################
# Code
##############################################
launch_es()
device, n_gpu = initialize_device_settings(use_cuda=True)
# Download evaluation data, which is a subset of Natural Questions development set containing 50 documents
doc_dir = "../data/nq"
s3_url = "https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/datasets/nq_dev_subset_v2.json.zip"
fetch_archive_from_http(url=s3_url, output_dir=doc_dir)
# Connect to Elasticsearch
document_store = ElasticsearchDocumentStore(host="localhost",
username="",
password="",
index="document",
create_index=False,
embedding_field="emb",
embedding_dim=768,
excluded_meta_data=["emb"])
# Add evaluation data to Elasticsearch document store
# We first delete the custom tutorial indices to not have duplicate elements
# and also split our documents into shorter passages using the PreProcessor
preprocessor = PreProcessor(split_by="word",
split_length=500,
split_overlap=0,
split_respect_sentence_boundary=False,
clean_empty_lines=False,
clean_whitespace=False)
document_store.delete_all_documents(index=doc_index)
document_store.delete_all_documents(index=label_index)
document_store.add_eval_data(filename="../data/nq/nq_dev_subset_v2.json",
doc_index=doc_index,
label_index=label_index,
preprocessor=preprocessor)
# Let's prepare the labels that we need for the retriever and the reader
labels = document_store.get_all_labels_aggregated(index=label_index)
# Initialize Retriever
retriever = ElasticsearchRetriever(document_store=document_store)
# Alternative: Evaluate DensePassageRetriever
# Note, that DPR works best when you index short passages < 512 tokens as only those tokens will be used for the embedding.
# Here, for nq_dev_subset_v2.json we have avg. num of tokens = 5220(!).
# DPR still outperforms Elastic's BM25 by a small margin here.
# retriever = DensePassageRetriever(document_store=document_store,
# query_embedding_model="facebook/dpr-question_encoder-single-nq-base",
# passage_embedding_model="facebook/dpr-ctx_encoder-single-nq-base",
# use_gpu=True,
# embed_title=True,
# remove_sep_tok_from_untitled_passages=True)
# document_store.update_embeddings(retriever, index=doc_index)
# Initialize Reader
reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2",
top_k=4,
return_no_answer=True)
# Here we initialize the nodes that perform evaluation
eval_retriever = EvalDocuments()
eval_reader = EvalAnswers(
sas_model="sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
)
## Evaluate Retriever on its own in closed domain fashion
if style == "retriever_closed":
retriever_eval_results = retriever.eval(top_k=10,
label_index=label_index,
doc_index=doc_index)
## Retriever Recall is the proportion of questions for which the correct document containing the answer is
## among the correct documents
print("Retriever Recall:", retriever_eval_results["recall"])
## Retriever Mean Avg Precision rewards retrievers that give relevant documents a higher rank
print("Retriever Mean Avg Precision:", retriever_eval_results["map"])
# Evaluate Reader on its own in closed domain fashion (i.e. SQuAD style)
elif style == "reader_closed":
reader_eval_results = reader.eval(document_store=document_store,
device=device,
label_index=label_index,
doc_index=doc_index)
# Evaluation of Reader can also be done directly on a SQuAD-formatted file without passing the data to Elasticsearch
#reader_eval_results = reader.eval_on_file("../data/nq", "nq_dev_subset_v2.json", device=device)
## Reader Top-N-Accuracy is the proportion of predicted answers that match with their corresponding correct answer
print("Reader Top-N-Accuracy:", reader_eval_results["top_n_accuracy"])
## Reader Exact Match is the proportion of questions where the predicted answer is exactly the same as the correct answer
print("Reader Exact Match:", reader_eval_results["EM"])
## Reader F1-Score is the average overlap between the predicted answers and the correct answers
print("Reader F1-Score:", reader_eval_results["f1"])
# Evaluate combination of Reader and Retriever in open domain fashion
elif style == "retriever_reader_open":
# Here is the pipeline definition
p = Pipeline()
p.add_node(component=retriever, name="ESRetriever", inputs=["Query"])
p.add_node(component=eval_retriever,
name="EvalDocuments",
inputs=["ESRetriever"])
p.add_node(component=reader, name="QAReader", inputs=["EvalDocuments"])
p.add_node(component=eval_reader,
name="EvalAnswers",
inputs=["QAReader"])
results = []
for l in labels:
res = p.run(
query=l.question,
top_k_retriever=10,
labels=l,
top_k_reader=10,
index=doc_index,
)
results.append(res)
eval_retriever.print()
print()
retriever.print_time()
print()
eval_reader.print(mode="reader")
print()
reader.print_time()
print()
eval_reader.print(mode="pipeline")
else:
raise ValueError(
f'style={style} is not a valid option. Choose from retriever_closed, reader_closed, retriever_reader_open'
)
def test_eval_pipeline(document_store: BaseDocumentStore, reader, retriever):
# add eval data (SQUAD format)
document_store.add_eval_data(
filename="samples/squad/tiny.json",
doc_index="haystack_test_eval_document",
label_index="haystack_test_feedback",
)
labels = document_store.get_all_labels_aggregated(
index="haystack_test_feedback")
q_to_l_dict = {l.question: {"retriever": l, "reader": l} for l in labels}
eval_retriever = EvalRetriever()
eval_reader = EvalReader()
assert document_store.get_document_count(
index="haystack_test_eval_document") == 2
p = Pipeline()
p.add_node(component=retriever, name="ESRetriever", inputs=["Query"])
p.add_node(component=eval_retriever,
name="EvalRetriever",
inputs=["ESRetriever"])
p.add_node(component=reader, name="QAReader", inputs=["EvalRetriever"])
p.add_node(component=eval_reader, name="EvalReader", inputs=["QAReader"])
for q, l in q_to_l_dict.items():
res = p.run(
query=q,
top_k_retriever=10,
labels=l,
top_k_reader=10,
index="haystack_test_eval_document",
)
assert eval_retriever.recall == 1.0
assert round(eval_reader.top_k_f1, 4) == 0.8333
assert eval_reader.top_k_em == 0.5
def main():
launch_es()
document_store = ElasticsearchDocumentStore()
es_retriever = ElasticsearchRetriever(document_store=document_store)
eval_retriever = EvalRetriever(open_domain=open_domain)
reader = FARMReader("deepset/roberta-base-squad2",
top_k_per_candidate=4,
num_processes=1,
return_no_answer=True)
eval_reader = EvalReader(debug=True, open_domain=open_domain)
# Download evaluation data, which is a subset of Natural Questions development set containing 50 documents
doc_dir = "../data/nq"
s3_url = "https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/datasets/nq_dev_subset_v2.json.zip"
fetch_archive_from_http(url=s3_url, output_dir=doc_dir)
# Add evaluation data to Elasticsearch document store
# We first delete the custom tutorial indices to not have duplicate elements
preprocessor = PreProcessor(split_length=500,
split_overlap=0,
split_respect_sentence_boundary=False,
clean_empty_lines=False,
clean_whitespace=False)
document_store.delete_all_documents(index=doc_index)
document_store.delete_all_documents(index=label_index)
document_store.add_eval_data(filename="../data/nq/nq_dev_subset_v2.json",
doc_index=doc_index,
label_index=label_index,
preprocessor=preprocessor)
labels = document_store.get_all_labels_aggregated(index=label_index)
q_to_l_dict = {l.question: {"retriever": l, "reader": l} for l in labels}
# Here is the pipeline definition
p = Pipeline()
p.add_node(component=es_retriever, name="ESRetriever", inputs=["Query"])
p.add_node(component=eval_retriever,
name="EvalRetriever",
inputs=["ESRetriever"])
p.add_node(component=reader, name="QAReader", inputs=["EvalRetriever"])
p.add_node(component=eval_reader, name="EvalReader", inputs=["QAReader"])
results = []
for i, (q, l) in enumerate(q_to_l_dict.items()):
res = p.run(
query=q,
top_k_retriever=top_k_retriever,
labels=l,
top_k_reader=10,
index=doc_index,
# skip_incorrect_retrieval=True
)
results.append(res)
eval_retriever.print()
print()
es_retriever.print_time()
print()
eval_reader.print(mode="reader")
print()
reader.print_time()
print()
eval_reader.print(mode="pipeline")
def setup(self):
print("SETTING UP PIPELINE")
self.document_store = ElasticsearchDocumentStore(
similarity="dot_product", host="elasticsearch", username="", password="", index="document")
self.document_store_faiss = FAISSDocumentStore(
index="document",
faiss_index_factory_str="Flat",
return_embedding=True,
sql_url=f"postgresql://{config('POSTGRES_USER')}:{config('POSTGRES_PASSWORD')}@{config('POSTGRES_HOST')}:{config('POSTGRES_PORT')}/faiss"
)
processor, converter = self.write_as4_docs()
table_data = self.write_table_docs(converter, processor)
es_retriever = ElasticsearchRetriever(
document_store=self.document_store)
print("SETTING UP DPR")
dpr_retriever = DPRTrainingManager.get_current_retriever(
self.document_store_faiss)
print("SETTING UP EMBEDDINGS")
embedding_retriever = EmbeddingRetriever(
document_store=self.document_store_faiss,
embedding_model="deepset/sentence_bert"
)
query_classifier = QueryClassifier()
print("SETTING UP TABLE")
table_retriever = TableRetriever(table_data)
print("SETUP RETRIEVERS")
self.question_generator = FurtherQuestionGenerator()
print("UPDATING EMBEDDINGS")
self.document_store_faiss.update_embeddings(dpr_retriever)
print("UPDATED EMBEDDINGS")
self.dpr_node = ContinualDPRNode(
dpr_retriever, self.document_store_faiss)
result = Result()
self.trainer = DPRTrainingManager(
self.document_store_faiss, self.dpr_node)
print("SETUP COMPONENTS")
pipeline = Pipeline()
pipeline.add_node(component=es_retriever,
name="ESRetriever", inputs=["Query"])
pipeline.add_node(component=self.dpr_node,
name="DPRRetriever", inputs=["Query"])
pipeline.add_node(component=embedding_retriever,
name="EmbeddingRetriever", inputs=["Query"])
pipeline.add_node(component=JoinDocuments(join_mode="merge"), name="JoinResults", inputs=[
"DPRRetriever", "EmbeddingRetriever", "ESRetriever"])
pipeline.add_node(component=query_classifier,
name="QueryClassifier", inputs=["JoinResults"])
pipeline.add_node(component=self.question_generator,
name="QnGenerator", inputs=["QueryClassifier.output_1"])
pipeline.add_node(component=table_retriever, name="TableRetriever", inputs=[
"QueryClassifier.output_2"])
pipeline.add_node(component=result, name="Result", inputs=[
"QnGenerator", "TableRetriever"])
self.pipeline = pipeline
print("SETUP PIPELINE")
def tutorial14_query_classifier():
#Download and prepare data - 517 Wikipedia articles for Game of Thrones
doc_dir = "data/article_txt_got"
s3_url = "https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/datasets/documents/wiki_gameofthrones_txt.zip"
fetch_archive_from_http(url=s3_url, output_dir=doc_dir)
# convert files to dicts containing documents that can be indexed to our datastore
got_dicts = convert_files_to_dicts(
dir_path=doc_dir,
clean_func=clean_wiki_text,
split_paragraphs=True
)
# Initialize DocumentStore and index documents
launch_es()
document_store = ElasticsearchDocumentStore()
document_store.delete_all_documents()
document_store.write_documents(got_dicts)
# Initialize Sparse retriever
es_retriever = ElasticsearchRetriever(document_store=document_store)
# Initialize dense retriever
dpr_retriever = DensePassageRetriever(document_store)
document_store.update_embeddings(dpr_retriever, update_existing_embeddings=False)
reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2")
# Here we build the pipeline
sklearn_keyword_classifier = Pipeline()
sklearn_keyword_classifier.add_node(component=SklearnQueryClassifier(), name="QueryClassifier", inputs=["Query"])
sklearn_keyword_classifier.add_node(component=dpr_retriever, name="DPRRetriever", inputs=["QueryClassifier.output_1"])
sklearn_keyword_classifier.add_node(component=es_retriever, name="ESRetriever", inputs=["QueryClassifier.output_2"])
sklearn_keyword_classifier.add_node(component=reader, name="QAReader", inputs=["ESRetriever", "DPRRetriever"])
sklearn_keyword_classifier.draw("pipeline_classifier.png")
# Run only the dense retriever on the full sentence query
res_1 = sklearn_keyword_classifier.run(
query="Who is the father of Arya Stark?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_1)
# Run only the sparse retriever on a keyword based query
res_2 = sklearn_keyword_classifier.run(
query="arya stark father",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
print_answers(res_2)
# Run only the dense retriever on the full sentence query
res_3 = sklearn_keyword_classifier.run(
query="which country was jon snow filmed ?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_3)
# Run only the sparse retriever on a keyword based query
res_4 = sklearn_keyword_classifier.run(
query="jon snow country",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
print_answers(res_4)
# Run only the dense retriever on the full sentence query
res_5 = sklearn_keyword_classifier.run(
query="who are the younger brothers of arya stark ?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_5)
# Run only the sparse retriever on a keyword based query
res_6 = sklearn_keyword_classifier.run(
query="arya stark younger brothers",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
print_answers(res_6)
# Here we build the pipeline
transformer_keyword_classifier = Pipeline()
transformer_keyword_classifier.add_node(component=TransformersQueryClassifier(), name="QueryClassifier", inputs=["Query"])
transformer_keyword_classifier.add_node(component=dpr_retriever, name="DPRRetriever", inputs=["QueryClassifier.output_1"])
transformer_keyword_classifier.add_node(component=es_retriever, name="ESRetriever", inputs=["QueryClassifier.output_2"])
transformer_keyword_classifier.add_node(component=reader, name="QAReader", inputs=["ESRetriever", "DPRRetriever"])
transformer_keyword_classifier.draw("pipeline_classifier.png")
# Run only the dense retriever on the full sentence query
res_1 = transformer_keyword_classifier.run(
query="Who is the father of Arya Stark?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_1)
# Run only the sparse retriever on a keyword based query
res_2 = transformer_keyword_classifier.run(
query="arya stark father",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
print_answers(res_2)
# Run only the dense retriever on the full sentence query
res_3 = transformer_keyword_classifier.run(
query="which country was jon snow filmed ?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_3)
# Run only the sparse retriever on a keyword based query
res_4 = transformer_keyword_classifier.run(
query="jon snow country",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
print_answers(res_4)
# Run only the dense retriever on the full sentence query
res_5 = transformer_keyword_classifier.run(
query="who are the younger brothers of arya stark ?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_5)
# Run only the sparse retriever on a keyword based query
res_6 = transformer_keyword_classifier.run(
query="arya stark younger brothers",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
print_answers(res_6)
# Here we build the pipeline
transformer_question_classifier = Pipeline()
transformer_question_classifier.add_node(component=dpr_retriever, name="DPRRetriever", inputs=["Query"])
transformer_question_classifier.add_node(component=TransformersQueryClassifier(model_name_or_path="shahrukhx01/question-vs-statement-classifier"), name="QueryClassifier", inputs=["DPRRetriever"])
transformer_question_classifier.add_node(component=reader, name="QAReader", inputs=["QueryClassifier.output_1"])
transformer_question_classifier.draw("question_classifier.png")
# Run only the QA reader on the question query
res_1 = transformer_question_classifier.run(
query="Who is the father of Arya Stark?",
top_k_retriever=10
)
print("DPR Results" + "\n" + "="*15)
print_answers(res_1)
# Show only DPR results
res_2 = transformer_question_classifier.run(
query="Arya Stark was the daughter of a Lord.",
top_k_retriever=10
)
print("ES Results" + "\n" + "="*15)
res_2
# Here we create the keyword vs question/statement query classifier
queries = ["arya stark father","jon snow country",
"who is the father of arya stark","which country was jon snow filmed?"]
keyword_classifier = TransformersQueryClassifier()
for query in queries:
result = keyword_classifier.run(query=query)
if result[1] == "output_1":
category = "question/statement"
else:
category = "keyword"
print(f"Query: {query}, raw_output: {result}, class: {category}")
# Here we create the question vs statement query classifier
queries = ["Lord Eddard was the father of Arya Stark.","Jon Snow was filmed in United Kingdom.",
"who is the father of arya stark?","Which country was jon snow filmed in?"]
question_classifier = TransformersQueryClassifier(model_name_or_path="shahrukhx01/question-vs-statement-classifier")
for query in queries:
result = question_classifier.run(query=query)
if result[1] == "output_1":
category = "question"
else:
category = "statement"
print(f"Query: {query}, raw_output: {result}, class: {category}")
document_store = FAISSDocumentStore.load(faiss,
sql_url=sql_url,
index='document')
print("Loaded document store")
retriever = DensePassageRetriever(
document_store=document_store,
query_embedding_model="facebook/dpr-question_encoder-single-nq-base",
passage_embedding_model="facebook/dpr-ctx_encoder-single-nq-base",
use_gpu=True,
batch_size=64,
embed_title=True)
print("Loaded retriever")
reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2",
use_gpu=True,
no_ans_boost=-10,
context_window_size=500,
top_k_per_candidate=3,
top_k_per_sample=1,
num_processes=1,
max_seq_len=256,
doc_stride=128,
progress_bar=False)
print("Loaded reader")
p = Pipeline()
p.add_node(component=retriever, name="ESRetriever1", inputs=["Query"])
p.add_node(component=reader, name="QAReader", inputs=["ESRetriever1"])
testing.test(p, dataset, jsonify=True, output_file=output_path)
def tutorial11_pipelines():
#Download and prepare data - 517 Wikipedia articles for Game of Thrones
doc_dir = "data/article_txt_got"
s3_url = "https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/datasets/documents/wiki_gameofthrones_txt.zip"
fetch_archive_from_http(url=s3_url, output_dir=doc_dir)
# convert files to dicts containing documents that can be indexed to our datastore
got_dicts = convert_files_to_dicts(dir_path=doc_dir,
clean_func=clean_wiki_text,
split_paragraphs=True)
# Initialize DocumentStore and index documents
launch_es()
document_store = ElasticsearchDocumentStore()
document_store.delete_all_documents()
document_store.write_documents(got_dicts)
# Initialize Sparse retriever
es_retriever = ElasticsearchRetriever(document_store=document_store)
# Initialize dense retriever
dpr_retriever = DensePassageRetriever(document_store)
document_store.update_embeddings(dpr_retriever,
update_existing_embeddings=False)
reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2")
######################
# Prebuilt Pipelines #
######################
# Extractive QA Pipeline
########################
p_extractive_premade = ExtractiveQAPipeline(reader=reader,
retriever=es_retriever)
res = p_extractive_premade.run(query="Who is the father of Arya Stark?",
top_k_retriever=10,
top_k_reader=5)
print_answers(res, details="minimal")
# Document Search Pipeline
##########################
p_retrieval = DocumentSearchPipeline(es_retriever)
res = p_retrieval.run(query="Who is the father of Arya Stark?",
top_k_retriever=10)
print_documents(res, max_text_len=200)
# Generator Pipeline
##########################
# We set this to True so that the document store returns document embeddings
# with each document, this is needed by the Generator
document_store.return_embedding = True
# Initialize generator
rag_generator = RAGenerator()
# Generative QA
p_generator = GenerativeQAPipeline(generator=rag_generator,
retriever=dpr_retriever)
res = p_generator.run(query="Who is the father of Arya Stark?",
top_k_retriever=10)
print_answers(res, details="minimal")
# We are setting this to False so that in later pipelines,
# we get a cleaner printout
document_store.return_embedding = False
##############################
# Creating Pipeline Diagrams #
##############################
p_extractive_premade.draw("pipeline_extractive_premade.png")
p_retrieval.draw("pipeline_retrieval.png")
p_generator.draw("pipeline_generator.png")
####################
# Custom Pipelines #
####################
# Extractive QA Pipeline
########################
# Custom built extractive QA pipeline
p_extractive = Pipeline()
p_extractive.add_node(component=es_retriever,
name="Retriever",
inputs=["Query"])
p_extractive.add_node(component=reader,
name="Reader",
inputs=["Retriever"])
# Now we can run it
res = p_extractive.run(query="Who is the father of Arya Stark?",
top_k_retriever=10,
top_k_reader=5)
print_answers(res, details="minimal")
p_extractive.draw("pipeline_extractive.png")
# Ensembled Retriever Pipeline
##############################
# Create ensembled pipeline
p_ensemble = Pipeline()
p_ensemble.add_node(component=es_retriever,
name="ESRetriever",
inputs=["Query"])
p_ensemble.add_node(component=dpr_retriever,
name="DPRRetriever",
inputs=["Query"])
p_ensemble.add_node(component=JoinDocuments(join_mode="concatenate"),
name="JoinResults",
inputs=["ESRetriever", "DPRRetriever"])
p_ensemble.add_node(component=reader,
name="Reader",
inputs=["JoinResults"])
p_ensemble.draw("pipeline_ensemble.png")
# Run pipeline
res = p_ensemble.run(
query="Who is the father of Arya Stark?",
top_k_retriever=5 #This is top_k per retriever
)
print_answers(res, details="minimal")
# Query Classification Pipeline
###############################
# Decision Nodes help you route your data so that only certain branches of your `Pipeline` are run.
# Though this looks very similar to the ensembled pipeline shown above,
# the key difference is that only one of the retrievers is run for each request.
# By contrast both retrievers are always run in the ensembled approach.
class QueryClassifier():
outgoing_edges = 2
def run(self, **kwargs):
if "?" in kwargs["query"]:
return (kwargs, "output_2")
else:
return (kwargs, "output_1")
# Here we build the pipeline
p_classifier = Pipeline()
p_classifier.add_node(component=QueryClassifier(),
name="QueryClassifier",
inputs=["Query"])
p_classifier.add_node(component=es_retriever,
name="ESRetriever",
inputs=["QueryClassifier.output_1"])
p_classifier.add_node(component=dpr_retriever,
name="DPRRetriever",
inputs=["QueryClassifier.output_2"])
p_classifier.add_node(component=reader,
name="QAReader",
inputs=["ESRetriever", "DPRRetriever"])
p_classifier.draw("pipeline_classifier.png")
# Run only the dense retriever on the full sentence query
res_1 = p_classifier.run(query="Who is the father of Arya Stark?",
top_k_retriever=10)
print("DPR Results" + "\n" + "=" * 15)
print_answers(res_1)
# Run only the sparse retriever on a keyword based query
res_2 = p_classifier.run(query="Arya Stark father", top_k_retriever=10)
print("ES Results" + "\n" + "=" * 15)
print_answers(res_2)
docs = [doc[1] for doc in kwargs["result"]]
return (docs, "output_1")
else:
return (kwargs["result"]["answer"], "output_1")
result = Result()
"""## Assembling into Pipeline"""
# Current approach
from haystack import Pipeline
from haystack.pipeline import JoinDocuments
# Building new pipeline with multiple retrievers
pipeline = Pipeline()
pipeline.add_node(component=es_retriever, name="ESRetriever", inputs=["Query"])
pipeline.add_node(component=dpr_retriever,
name="DPRRetriever", inputs=["Query"])
pipeline.add_node(component=embedding_retriever,
name="EmbeddingRetriever", inputs=["Query"])
pipeline.add_node(component=JoinDocuments(join_mode="merge"), name="JoinResults", inputs=[
"DPRRetriever", "EmbeddingRetriever", "ESRetriever"])
pipeline.add_node(component=query_classifier,
name="QueryClassifier", inputs=["JoinResults"])
pipeline.add_node(component=question_generator,
name="QnGenerator", inputs=["QueryClassifier.output_1"])
pipeline.add_node(component=table_retriever, name="TableRetriever", inputs=[
"QueryClassifier.output_2"])
pipeline.add_node(component=result, name="Result", inputs=[
"QnGenerator", "TableRetriever"])