alihan/llm-examples
1
0
Fork 0
mirror of https://github.com/pinecone-io/examples.git synced 2023-10-11 20:04:54 +03:00
Code Issues Packages Projects Releases Wiki Activity

added updated notebooks + code

Browse Source
This commit is contained in:
jamescalam
2022-07-13 23:51:04 +02:00
parent a5cafb6e47
commit 51e36ccd37
4 changed files with 952 additions and 635 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
integrations/haystack/nhs-search/README.md
View File

@@ -1,3 +1,11 @@
# NHS Search App \[Pinecone x Haystack\]
# Doc Search App [Pinecone x Streamlit]
Here is the code covering the NHS search app presented at the Pinecone x Haystack webinar on 14th July 2022.
## Ideas
* Doc search app
* Using scraped technical docs from Streamlit
* Could be cool to add technical docs from other places \[HuggingFace, Pinecone\]?
* * With additional docs, we could show off metadata filtering
* UI nice simple search bar, metadata filtering could be dropdown/checkboxes, or a recommendation after performing a search that returns mostly docs from one of the sections
* Q&A style search would be best, if not possible a vanilla semantic search, which we could fine-tune easily w/ TSDAE if performance is not optimal
# Search Tool

1376
integrations/haystack/nhs-search/notebooks/00_indexing.ipynb
View File

File diff suppressed because it is too large Load Diff

193
integrations/haystack/nhs-search/notebooks/01_test_pipeline.ipynb
View File

@@ -36,7 +36,7 @@
{
"data": {
"text/plain": [
"(100, 100)"
"(8521, 8521)"
]
},
"execution_count": 2,
@@ -59,16 +59,16 @@
"text": [
"INFO - haystack.modeling.utils - Using devices: CPU\n",
"INFO - haystack.modeling.utils - Number of GPUs: 0\n",
"INFO - haystack.retriever.dense - Init retriever using embeddings of model flax-sentence-embeddings/all_datasets_v3_mpnet-base\n"
"INFO - haystack.nodes.retriever.dense - Init retriever using embeddings of model sentence-transformers/multi-qa-mpnet-base-dot-v1\n"
]
}
],
"source": [
"from haystack.retriever.dense import EmbeddingRetriever\n",
"from haystack.nodes.retriever import EmbeddingRetriever\n",
"\n",
"retriever = EmbeddingRetriever(\n",
" document_store=document_store,\n",
" embedding_model='flax-sentence-embeddings/all_datasets_v3_mpnet-base',\n",
" embedding_model='sentence-transformers/multi-qa-mpnet-base-dot-v1',\n",
" model_format=\"sentence_transformers\"\n",
")"
]
@@ -86,9 +86,119 @@
"INFO - haystack.modeling.utils - Number of GPUs: 0\n",
"INFO - haystack.modeling.model.language_model - LOADING MODEL\n",
"INFO - haystack.modeling.model.language_model - =============\n",
"INFO - haystack.modeling.model.language_model - Could not find deepset/electra-base-squad2 locally.\n",
"INFO - haystack.modeling.model.language_model - Looking on Transformers Model Hub (in local cache and online)...\n",
"INFO - haystack.modeling.model.language_model - Loaded deepset/electra-base-squad2\n",
"INFO - haystack.modeling.model.language_model - Could not find deepset/roberta-base-squad2-distilled locally.\n",
"INFO - haystack.modeling.model.language_model - Looking on Transformers Model Hub (in local cache and online)...\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "fae8519fb44d405bacae249f96bb9d69",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/729 [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "098a78dd37da48b5bccdf95c85e27bbc",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/473M [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"INFO - haystack.modeling.model.language_model - Loaded deepset/roberta-base-squad2-distilled\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "da97f59cc2854872b11751048ab910bc",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/295 [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "4670b856e53645eeb59a22922d3d082e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/780k [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "938ab35200e54240870e2eee72d6bda2",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/446k [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "33780f9f6e0f495893004257e95d7c5e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/1.29M [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "bc09919fe2d64cae89b573fc93042b6b",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/239 [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"INFO - haystack.modeling.utils - Using devices: CPU\n",
"INFO - haystack.modeling.utils - Number of GPUs: 0\n",
"INFO - haystack.modeling.infer - Got ya 9 parallel workers to do inference ...\n",
@@ -103,8 +213,7 @@
"from haystack.pipelines import ExtractiveQAPipeline\n",
"\n",
"reader = FARMReader(\n",
" model_name_or_path='deepset/electra-base-squad2',\n",
" use_gpu=True\n",
" model_name_or_path='deepset/roberta-base-squad2-distilled'\n",
")\n",
"pipe = ExtractiveQAPipeline(reader, retriever)"
]
@@ -124,7 +233,7 @@
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "f03771f42ee040668e73c919b9ef6aeb",
"model_id": "baca227f02f647c8ad8f76bd121a220c",
"version_major": 2,
"version_minor": 0
},
@@ -166,8 +275,8 @@
"huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",
"To disable this warning, you can either:\n",
"\t- Avoid using `tokenizers` before the fork if possible\n",
"huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",
"\t- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)\n",
"huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",
"To disable this warning, you can either:\n",
"\t- Avoid using `tokenizers` before the fork if possible\n",
"\t- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)\n",
@@ -183,11 +292,11 @@
"text": [
"Inferencing Samples: 0%| | 0/1 [00:00<?, ? Batches/s]/Users/jamesbriggs/opt/anaconda3/envs/ml/lib/python3.9/site-packages/haystack/modeling/model/prediction_head.py:483: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').\n",
" start_indices = flat_sorted_indices // max_seq_len\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 4.17 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 4.61 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 4.45 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 4.51 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 4.63 Batches/s]\n"
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 8.38 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 9.36 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 9.45 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 9.58 Batches/s]\n",
"Inferencing Samples: 100%|██████████| 1/1 [00:00<00:00, 9.61 Batches/s]\n"
]
}
],
@@ -213,16 +322,14 @@
"\n",
"Query: Who is affected by pre-eclampsia?\n",
"Answers:\n",
"[ { 'answer': 'the mother and her baby',\n",
" 'context': 'cause no problems and improve soon after the baby is '\n",
" \"delivered, there's a risk of serious complications that \"\n",
" 'can affect both the mother and her baby. '},\n",
" { 'answer': 'before you were pregnant having an autoimmune condition, '\n",
" 'such as lupus or antiphospholipid syndrome having high '\n",
" 'blood pressure or pre-eclampsia in a previous pregnancy',\n",
" 'context': ' before you were pregnant having an autoimmune condition, '\n",
" 'such as lupus or antiphospholipid syndrome having high '\n",
" 'blood pressure or pre-eclampsia in a previous pregnancy'}]\n"
"[ { 'answer': 'pregnant women',\n",
" 'context': 'atment Complications Pre-eclampsia is a condition that '\n",
" 'affects some pregnant women, usually during the second '\n",
" 'half of pregnancy (from 20 weeks) or soo'},\n",
" { 'answer': 'mother and baby',\n",
" 'context': ' are mild, the condition can lead to serious complications '\n",
" \"for both mother and baby if it's not monitored and \"\n",
" 'treated. The earlier pre-eclampsia is d'}]\n"
]
}
],
@@ -240,16 +347,16 @@
},
{
"cell_type": "code",
"execution_count": 14,
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\"cause no problems and improve soon after the baby is delivered, there's a risk of serious complications that can affect both the mother and her baby. \""
"'atment Complications Pre-eclampsia is a condition that affects some pregnant women, usually during the second half of pregnancy (from 20 weeks) or soo'"
]
},
"execution_count": 14,
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
@@ -260,16 +367,16 @@
},
{
"cell_type": "code",
"execution_count": 26,
"execution_count": 8,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(125, 148)"
"(68, 82)"
]
},
"execution_count": 26,
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
@@ -283,16 +390,16 @@
},
{
"cell_type": "code",
"execution_count": 27,
"execution_count": 9,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'the mother and her baby'"
"'pregnant women'"
]
},
"execution_count": 27,
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
@@ -303,16 +410,16 @@
},
{
"cell_type": "code",
"execution_count": 28,
"execution_count": 10,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<Answer {'answer': 'the mother and her baby', 'type': 'extractive', 'score': 0.9425204694271088, 'context': \"cause no problems and improve soon after the baby is delivered, there's a risk of serious complications that can affect both the mother and her baby. \", 'offsets_in_document': [{'start': 631, 'end': 654}], 'offsets_in_context': [{'start': 125, 'end': 148}], 'document_id': 'f65f2a23620b9ea4761c43683729fc6c', 'meta': {'url': 'www.nhs.uk_conditions_pre-eclampsia_.txt'}}>"
"<Answer {'answer': 'pregnant women', 'type': 'extractive', 'score': 0.8104832470417023, 'context': 'atment Complications Pre-eclampsia is a condition that affects some pregnant women, usually during the second half of pregnancy (from 20 weeks) or soo', 'offsets_in_document': [{'start': 140, 'end': 154}], 'offsets_in_context': [{'start': 68, 'end': 82}], 'document_id': '3bc401b213c2720c83ee9bddb0e769b8', 'meta': {'url': 'www.nhs.uk_conditions_pre-eclampsia_.txt'}}>"
]
},
"execution_count": 28,
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
@@ -323,16 +430,16 @@
},
{
"cell_type": "code",
"execution_count": 29,
"execution_count": 11,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"0.9425204694271088"
"0.8104832470417023"
]
},
"execution_count": 29,
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
@@ -343,7 +450,7 @@
},
{
"cell_type": "code",
"execution_count": 31,
"execution_count": 12,
"metadata": {},
"outputs": [
{
@@ -352,7 +459,7 @@
"'www.nhs.uk_conditions_pre-eclampsia_.txt'"
]
},
"execution_count": 31,
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}

6
integrations/haystack/nhs-search/src/doc_search/app.py
View File

@@ -7,9 +7,9 @@ from haystack.pipelines import ExtractiveQAPipeline
PINECONE_API_KEY = st.secrets["PINECONE_KEY"]
RETRIEVER = 'mpnet'
RETRIEVER_URL = 'flax-sentence-embeddings/all_datasets_v3_mpnet-base'
READER = 'electra'
READER_URL = 'deepset/electra-base-squad2'
RETRIEVER_URL = 'sentence-transformers/multi-qa-mpnet-base-dot-v1'
READER = 'roberta-distilled'
READER_URL = 'deepset/roberta-base-squad2-distilled'
INDEX = 'haystack-nhs-jul'
DIMS = 768