forked from dataelement/bisheng
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
patch, fix chatmessage bug
- Loading branch information
Showing
7 changed files
with
167 additions
and
116 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,5 +1,6 @@ | ||
| import json | ||
|
|
||
| import requests | ||
| from bisheng_langchain.embeddings import HostEmbeddings | ||
| from bisheng_langchain.vectorstores import Milvus | ||
| from pymilvus import Collection, MilvusException | ||
|
|
@@ -20,121 +21,143 @@ | |
| database_url = 'mysql+pymysql://root:[email protected]:3306/bisheng?charset=utf8mb4' | ||
| engine = create_engine(database_url, connect_args={}, pool_pre_ping=True) | ||
|
|
||
| params['collection_name'] = 'partition_textembeddingada002_knowledge_1' | ||
| openai_target = Milvus.from_documents(embedding=embedding, **params) | ||
| params['collection_name'] = 'partition_multilinguale5large_knowledge_1' | ||
| host_targe = Milvus.from_documents(embedding=embedding, **params) | ||
| with Session(engine) as session: | ||
| db_knowledge = session.exec( | ||
| 'select id, collection_name, model, index_name from knowledge').all() | ||
| for knowledge in db_knowledge: | ||
| # if not knowledge[1].startswith('col'): | ||
| # if knowledge[3].startswith('col'): | ||
| # # 迁移完 | ||
| # print(f"drop id={knowledge}") | ||
| # params['collection_name'] = knowledge[3] | ||
| # cli = Milvus.from_documents(embedding=embedding, **params) | ||
| # if cli.col: | ||
| # cli.col.drop() | ||
| # time.sleep(1) | ||
| # continue | ||
| if knowledge[1].startswith('col'): | ||
| print(f'deal id={knowledge[0]} model={knowledge[2]} col={knowledge[1]}') | ||
| params['collection_name'] = knowledge[1] | ||
| cli = Milvus.from_documents(embedding=embedding, **params) | ||
| if not cli.col: | ||
| print(f'escape id={knowledge[0]} col={knowledge[1]}') | ||
|
|
||
| def milvus_clean(): | ||
| params['collection_name'] = 'partition_textembeddingada002_knowledge_1' | ||
| openai_target = Milvus.from_documents(embedding=embedding, **params) | ||
| params['collection_name'] = 'partition_multilinguale5large_knowledge_1' | ||
| host_targe = Milvus.from_documents(embedding=embedding, **params) | ||
| with Session(engine) as session: | ||
| db_knowledge = session.exec( | ||
| 'select id, collection_name, model, index_name from knowledge').all() | ||
| for knowledge in db_knowledge: | ||
| # if not knowledge[1].startswith('col'): | ||
| # if knowledge[3].startswith('col'): | ||
| # # 迁移完 | ||
| # print(f"drop id={knowledge}") | ||
| # params['collection_name'] = knowledge[3] | ||
| # cli = Milvus.from_documents(embedding=embedding, **params) | ||
| # if cli.col: | ||
| # cli.col.drop() | ||
| # time.sleep(1) | ||
| # continue | ||
| if knowledge[1].startswith('col'): | ||
| print(f'deal id={knowledge[0]} model={knowledge[2]} col={knowledge[1]}') | ||
| params['collection_name'] = knowledge[1] | ||
| cli = Milvus.from_documents(embedding=embedding, **params) | ||
| if not cli.col: | ||
| print(f'escape id={knowledge[0]} col={knowledge[1]}') | ||
| index_name = knowledge[1] | ||
| col_name = f'partition_{knowledge[2]}_knowledge_1'.replace('-', '') | ||
| sql = 'update knowledge set collection_name="%s", index_name="%s" where id=%d' % ( | ||
| col_name, index_name, knowledge[0]) | ||
| session.exec(sql) | ||
| session.commit() | ||
| continue | ||
| fields = [s.name for s in cli.col.schema.fields if s.name != 'pk'] | ||
| print(fields) | ||
| pks = cli.col.query(expr='file_id>1') | ||
| pk_len = len(pks) | ||
| if pk_len == 0: | ||
| continue | ||
| li = [] | ||
| batch_size = 500 | ||
| for i in range(0, pk_len, batch_size): | ||
| end = min(i + batch_size, pk_len) | ||
| pk_ids = [str(pk.get('pk')) for pk in pks[i:end]] | ||
| pk_with_fields = cli.col.query(f"pk in [{','.join(pk_ids)}]", | ||
| output_fields=fields) | ||
| li.extend(pk_with_fields) | ||
| if knowledge[2] == 'text-embedding-ada-002': | ||
| target = openai_target | ||
| elif knowledge[2] == 'multilingual-e5-large': | ||
| target = host_targe | ||
| else: | ||
| continue | ||
|
|
||
| insert_fields = [s.name for s in target.col.schema.fields if s.name != 'pk'] | ||
| insert_dict = { | ||
| 'text': [], | ||
| 'vector': [], | ||
| 'file_id': [], | ||
| 'knowledge_id': [], | ||
| 'page': [], | ||
| 'source': [], | ||
| 'bbox': [], | ||
| 'extra': [] | ||
| } | ||
| for data in li: | ||
| insert_dict.get('text').append(data.get('text')) | ||
| insert_dict.get('vector').append(data.get('vector')) | ||
| insert_dict.get('file_id').append(data.get('file_id')) | ||
| insert_dict.get('knowledge_id').append(f'{knowledge[0]}') | ||
|
|
||
| if 'bbox' in fields: | ||
| if data.get('bbox'): | ||
| insert_dict.get('bbox').append( | ||
| '{"chunk_bboxes":%s}' % | ||
| (json.loads(data.get('bbox')).get('chunk_bboxes'))) | ||
| if json.loads(data.get('bbox')).get('source'): | ||
| insert_dict.get('source').append( | ||
| json.loads(data.get('bbox')).get('source')) | ||
| if json.loads(data.get('bbox')).get('chunk_bboxes')[0].get('page'): | ||
| insert_dict.get('page').append( | ||
| json.loads( | ||
| data.get('bbox')).get('chunk_bboxes')[0].get('page')) | ||
| else: | ||
| insert_dict.get('bbox').append('') | ||
| else: | ||
| insert_dict.get('bbox').append('') | ||
| if 'source' in fields: | ||
| insert_dict.get('source').append(data.get('source')) | ||
| if len(insert_dict.get('source')) != len(insert_dict.get('bbox')): | ||
| insert_dict.get('source').append('') | ||
| if 'page' in fields: | ||
| insert_dict.get('page').append(data.get('page') if data.get('page') else 1) | ||
|
|
||
| insert_dict.get('extra').append('') | ||
|
|
||
| total_count = len(li) | ||
| batch_size = 1000 | ||
| for i in range(0, total_count, batch_size): | ||
| # Grab end index | ||
| end = min(i + batch_size, total_count) | ||
| # Convert dict to list of lists batch for insertion | ||
| insert_list = [insert_dict[x][i:end] for x in insert_fields] | ||
| # Insert into the collection. | ||
| try: | ||
| res: Collection | ||
| res = target.col.insert(insert_list, timeout=100) | ||
| print(res) | ||
| except MilvusException as e: | ||
| print('Failed to insert batch starting at entity: %s/%s', i, total_count) | ||
| raise e | ||
|
|
||
| index_name = knowledge[1] | ||
| col_name = f'partition_{knowledge[2]}_knowledge_1'.replace('-', '') | ||
| sql = 'update knowledge set collection_name="%s", index_name="%s" where id=%d' % ( | ||
| col_name, index_name, knowledge[0]) | ||
| session.exec(sql) | ||
| session.commit() | ||
| continue | ||
| fields = [s.name for s in cli.col.schema.fields if s.name != 'pk'] | ||
| print(fields) | ||
| pks = cli.col.query(expr='file_id>1') | ||
| pk_len = len(pks) | ||
| if pk_len == 0: | ||
| continue | ||
| li = [] | ||
| batch_size = 500 | ||
| for i in range(0, pk_len, batch_size): | ||
| end = min(i + batch_size, pk_len) | ||
| pk_ids = [str(pk.get('pk')) for pk in pks[i:end]] | ||
| pk_with_fields = cli.col.query(f"pk in [{','.join(pk_ids)}]", output_fields=fields) | ||
| li.extend(pk_with_fields) | ||
| if knowledge[2] == 'text-embedding-ada-002': | ||
| target = openai_target | ||
| elif knowledge[2] == 'multilingual-e5-large': | ||
| target = host_targe | ||
| else: | ||
| continue | ||
|
|
||
| insert_fields = [s.name for s in target.col.schema.fields if s.name != 'pk'] | ||
| insert_dict = { | ||
| 'text': [], | ||
| 'vector': [], | ||
| 'file_id': [], | ||
| 'knowledge_id': [], | ||
| 'page': [], | ||
| 'source': [], | ||
| 'bbox': [], | ||
| 'extra': [] | ||
| } | ||
| for data in li: | ||
| insert_dict.get('text').append(data.get('text')) | ||
| insert_dict.get('vector').append(data.get('vector')) | ||
| insert_dict.get('file_id').append(data.get('file_id')) | ||
| insert_dict.get('knowledge_id').append(f'{knowledge[0]}') | ||
|
|
||
| if 'bbox' in fields: | ||
| if data.get('bbox'): | ||
| insert_dict.get('bbox').append( | ||
| '{"chunk_bboxes":%s}' % | ||
| (json.loads(data.get('bbox')).get('chunk_bboxes'))) | ||
| if json.loads(data.get('bbox')).get('source'): | ||
| insert_dict.get('source').append( | ||
| json.loads(data.get('bbox')).get('source')) | ||
| if json.loads(data.get('bbox')).get('chunk_bboxes')[0].get('page'): | ||
| insert_dict.get('page').append( | ||
| json.loads(data.get('bbox')).get('chunk_bboxes')[0].get('page')) | ||
| else: | ||
| insert_dict.get('bbox').append('') | ||
| else: | ||
| insert_dict.get('bbox').append('') | ||
| if 'source' in fields: | ||
| insert_dict.get('source').append(data.get('source')) | ||
| if len(insert_dict.get('source')) != len(insert_dict.get('bbox')): | ||
| insert_dict.get('source').append('') | ||
| if 'page' in fields: | ||
| insert_dict.get('page').append(data.get('page') if data.get('page') else 1) | ||
| print(f'deal_done id={knowledge[0]} index={index_name}') | ||
| cli.col.drop() | ||
| pass | ||
|
|
||
| insert_dict.get('extra').append('') | ||
|
|
||
| insert = [] | ||
| total_count = len(li) | ||
| batch_size = 1000 | ||
| for i in range(0, total_count, batch_size): | ||
| # Grab end index | ||
| end = min(i + batch_size, total_count) | ||
| # Convert dict to list of lists batch for insertion | ||
| insert_list = [insert_dict[x][i:end] for x in insert_fields] | ||
| # Insert into the collection. | ||
| try: | ||
| res: Collection | ||
| res = target.col.insert(insert_list, timeout=100) | ||
| def elastic_clean(): | ||
| url = 'http://192.168.106.116:9200/_stats' | ||
| headers = {'Authorization': 'Basic ZWxhc3RpYzpvU0dMLXpWdlo1UDNUbTdxa0RMQw=='} | ||
| del_url = 'http://192.168.106.116:9200/%s' | ||
| col = requests.get(url, headers=headers).json() | ||
| for c in col.get('indices').keys(): | ||
| if c.startswith('tmp'): | ||
| print(c) | ||
| x = requests.delete(del_url % c, headers=headers) | ||
| # url = f'http:// | ||
| elif col.get('indices').get(c).get('primaries').get('docs').get('count') == 0: | ||
| print(c) | ||
| x = requests.delete(del_url % c, headers=headers) | ||
| print(x) | ||
|
|
||
| except MilvusException as e: | ||
| print('Failed to insert batch starting at entity: %s/%s', i, total_count) | ||
| raise e | ||
|
|
||
| index_name = knowledge[1] | ||
| col_name = f'partition_{knowledge[2]}_knowledge_1'.replace('-', '') | ||
| sql = 'update knowledge set collection_name="%s", index_name="%s" where id=%d' % ( | ||
| col_name, index_name, knowledge[0]) | ||
| session.exec(sql) | ||
| session.commit() | ||
| print(f'deal_done id={knowledge[0]} index={index_name}') | ||
| cli.col.drop() | ||
| pass | ||
| elastic_clean() | ||
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters