KevinHuSh
commited on
Commit
·
9fe9fc4
1
Parent(s):
f1ced48
add dockerfile for cuda envirement. Refine table search strategy, (#123)
Browse files- Dockerfile +1 -0
- Dockerfile.cuda +26 -0
- api/apps/conversation_app.py +69 -32
- api/db/db_models.py +2 -2
- api/db/init_data.py +57 -0
- api/settings.py +11 -4
- deepdoc/parser/excel_parser.py +1 -1
- deepdoc/parser/pdf_parser.py +2 -2
- deepdoc/parser/ppt_parser.py +2 -2
- deepdoc/vision/ocr.py +21 -4
- deepdoc/vision/recognizer.py +3 -1
- rag/app/table.py +6 -3
- rag/llm/__init__.py +8 -5
- rag/llm/chat_model.py +6 -3
- rag/llm/rpc_server.py +25 -19
- rag/nlp/search.py +5 -3
- rag/svr/task_executor.py +13 -4
- rag/utils/es_conn.py +2 -0
Dockerfile
CHANGED
@@ -14,6 +14,7 @@ ADD ./rag ./rag
|
|
14 |
ENV PYTHONPATH=/ragflow/
|
15 |
ENV HF_ENDPOINT=https://hf-mirror.com
|
16 |
|
|
|
17 |
ADD docker/entrypoint.sh ./entrypoint.sh
|
18 |
RUN chmod +x ./entrypoint.sh
|
19 |
|
|
|
14 |
ENV PYTHONPATH=/ragflow/
|
15 |
ENV HF_ENDPOINT=https://hf-mirror.com
|
16 |
|
17 |
+
/root/miniconda3/envs/py11/bin/pip install peewee==3.17.1
|
18 |
ADD docker/entrypoint.sh ./entrypoint.sh
|
19 |
RUN chmod +x ./entrypoint.sh
|
20 |
|
Dockerfile.cuda
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
FROM swr.cn-north-4.myhuaweicloud.com/infiniflow/ragflow-base:v1.0
|
2 |
+
USER root
|
3 |
+
|
4 |
+
WORKDIR /ragflow
|
5 |
+
|
6 |
+
## for cuda > 12.0
|
7 |
+
RUN /root/miniconda3/envs/py11/bin/pip uninstall -y onnxruntime-gpu
|
8 |
+
RUN /root/miniconda3/envs/py11/bin/pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
|
9 |
+
|
10 |
+
|
11 |
+
ADD ./web ./web
|
12 |
+
RUN cd ./web && npm i && npm run build
|
13 |
+
|
14 |
+
ADD ./api ./api
|
15 |
+
ADD ./conf ./conf
|
16 |
+
ADD ./deepdoc ./deepdoc
|
17 |
+
ADD ./rag ./rag
|
18 |
+
|
19 |
+
ENV PYTHONPATH=/ragflow/
|
20 |
+
ENV HF_ENDPOINT=https://hf-mirror.com
|
21 |
+
|
22 |
+
/root/miniconda3/envs/py11/bin/pip install peewee==3.17.1
|
23 |
+
ADD docker/entrypoint.sh ./entrypoint.sh
|
24 |
+
RUN chmod +x ./entrypoint.sh
|
25 |
+
|
26 |
+
ENTRYPOINT ["./entrypoint.sh"]
|
api/apps/conversation_app.py
CHANGED
@@ -21,7 +21,7 @@ from api.db.services.dialog_service import DialogService, ConversationService
|
|
21 |
from api.db import LLMType
|
22 |
from api.db.services.knowledgebase_service import KnowledgebaseService
|
23 |
from api.db.services.llm_service import LLMService, LLMBundle
|
24 |
-
from api.settings import access_logger, stat_logger, retrievaler
|
25 |
from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
|
26 |
from api.utils import get_uuid
|
27 |
from api.utils.api_utils import get_json_result
|
@@ -183,10 +183,10 @@ def chat(dialog, messages, **kwargs):
|
|
183 |
field_map = KnowledgebaseService.get_field_map(dialog.kb_ids)
|
184 |
## try to use sql if field mapping is good to go
|
185 |
if field_map:
|
186 |
-
|
187 |
-
markdown_tbl, chunks = use_sql(
|
188 |
if markdown_tbl:
|
189 |
-
return {"answer": markdown_tbl, "
|
190 |
|
191 |
prompt_config = dialog.prompt_config
|
192 |
for p in prompt_config["parameters"]:
|
@@ -201,6 +201,7 @@ def chat(dialog, messages, **kwargs):
|
|
201 |
dialog.similarity_threshold,
|
202 |
dialog.vector_similarity_weight, top=1024, aggs=False)
|
203 |
knowledges = [ck["content_with_weight"] for ck in kbinfos["chunks"]]
|
|
|
204 |
|
205 |
if not knowledges and prompt_config.get("empty_response"):
|
206 |
return {"answer": prompt_config["empty_response"], "reference": kbinfos}
|
@@ -212,7 +213,7 @@ def chat(dialog, messages, **kwargs):
|
|
212 |
if "max_tokens" in gen_conf:
|
213 |
gen_conf["max_tokens"] = min(gen_conf["max_tokens"], llm.max_tokens - used_token_count)
|
214 |
answer = chat_mdl.chat(prompt_config["system"].format(**kwargs), msg, gen_conf)
|
215 |
-
|
216 |
|
217 |
if knowledges:
|
218 |
answer, idx = retrievaler.insert_citations(answer,
|
@@ -237,47 +238,83 @@ def use_sql(question, field_map, tenant_id, chat_mdl):
|
|
237 |
|
238 |
问题如下:
|
239 |
{}
|
240 |
-
请写出SQL
|
241 |
""".format(
|
242 |
index_name(tenant_id),
|
243 |
"\n".join([f"{k}: {v}" for k, v in field_map.items()]),
|
244 |
question
|
245 |
)
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
sql = "
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
if
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
-
|
266 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
267 |
|
268 |
docid_idx = set([ii for ii, c in enumerate(tbl["columns"]) if c["name"] == "doc_id"])
|
269 |
docnm_idx = set([ii for ii, c in enumerate(tbl["columns"]) if c["name"] == "docnm_kwd"])
|
270 |
clmn_idx = [ii for ii in range(len(tbl["columns"])) if ii not in (docid_idx | docnm_idx)]
|
271 |
|
272 |
# compose markdown table
|
273 |
-
clmns = "|".join([re.sub(r"(/.*|([^()]+))", "", field_map.get(tbl["columns"][i]["name"],
|
274 |
-
line = "|".join(["------" for _ in range(len(clmn_idx))]) + "
|
275 |
-
rows = ["|".join([rmSpace(str(r[i])) for i in clmn_idx]).replace("None", " ") + "|" for r in tbl["rows"]]
|
276 |
if not docid_idx or not docnm_idx:
|
277 |
-
|
278 |
return "\n".join([clmns, line, "\n".join(rows)]), []
|
279 |
|
280 |
-
rows = "\n".join([r + f"##{ii}$$" for ii, r in enumerate(rows)])
|
281 |
docid_idx = list(docid_idx)[0]
|
282 |
docnm_idx = list(docnm_idx)[0]
|
283 |
return "\n".join([clmns, line, rows]), [{"doc_id": r[docid_idx], "docnm_kwd": r[docnm_idx]} for r in tbl["rows"]]
|
|
|
21 |
from api.db import LLMType
|
22 |
from api.db.services.knowledgebase_service import KnowledgebaseService
|
23 |
from api.db.services.llm_service import LLMService, LLMBundle
|
24 |
+
from api.settings import access_logger, stat_logger, retrievaler, chat_logger
|
25 |
from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
|
26 |
from api.utils import get_uuid
|
27 |
from api.utils.api_utils import get_json_result
|
|
|
183 |
field_map = KnowledgebaseService.get_field_map(dialog.kb_ids)
|
184 |
## try to use sql if field mapping is good to go
|
185 |
if field_map:
|
186 |
+
chat_logger.info("Use SQL to retrieval:{}".format(questions[-1]))
|
187 |
+
markdown_tbl, chunks = use_sql(questions[-1], field_map, dialog.tenant_id, chat_mdl)
|
188 |
if markdown_tbl:
|
189 |
+
return {"answer": markdown_tbl, "reference": {"chunks": chunks, "doc_aggs": []}}
|
190 |
|
191 |
prompt_config = dialog.prompt_config
|
192 |
for p in prompt_config["parameters"]:
|
|
|
201 |
dialog.similarity_threshold,
|
202 |
dialog.vector_similarity_weight, top=1024, aggs=False)
|
203 |
knowledges = [ck["content_with_weight"] for ck in kbinfos["chunks"]]
|
204 |
+
chat_logger.info("{}->{}".format(" ".join(questions), "\n->".join(knowledges)))
|
205 |
|
206 |
if not knowledges and prompt_config.get("empty_response"):
|
207 |
return {"answer": prompt_config["empty_response"], "reference": kbinfos}
|
|
|
213 |
if "max_tokens" in gen_conf:
|
214 |
gen_conf["max_tokens"] = min(gen_conf["max_tokens"], llm.max_tokens - used_token_count)
|
215 |
answer = chat_mdl.chat(prompt_config["system"].format(**kwargs), msg, gen_conf)
|
216 |
+
chat_logger.info("User: {}|Assistant: {}".format(msg[-1]["content"], answer))
|
217 |
|
218 |
if knowledges:
|
219 |
answer, idx = retrievaler.insert_citations(answer,
|
|
|
238 |
|
239 |
问题如下:
|
240 |
{}
|
241 |
+
请写出SQL, 且只要SQL,不要有其他说明及文字。
|
242 |
""".format(
|
243 |
index_name(tenant_id),
|
244 |
"\n".join([f"{k}: {v}" for k, v in field_map.items()]),
|
245 |
question
|
246 |
)
|
247 |
+
tried_times = 0
|
248 |
+
def get_table():
|
249 |
+
nonlocal sys_prompt, user_promt, question, tried_times
|
250 |
+
sql = chat_mdl.chat(sys_prompt, [{"role": "user", "content": user_promt}], {"temperature": 0.06})
|
251 |
+
print(user_promt, sql)
|
252 |
+
chat_logger.info(f"“{question}”==>{user_promt} get SQL: {sql}")
|
253 |
+
sql = re.sub(r"[\r\n]+", " ", sql.lower())
|
254 |
+
sql = re.sub(r".*select ", "select ", sql.lower())
|
255 |
+
sql = re.sub(r" +", " ", sql)
|
256 |
+
sql = re.sub(r"([;;]|```).*", "", sql)
|
257 |
+
if sql[:len("select ")] != "select ":
|
258 |
+
return None, None
|
259 |
+
if not re.search(r"((sum|avg|max|min)\(|group by )", sql.lower()):
|
260 |
+
if sql[:len("select *")] != "select *":
|
261 |
+
sql = "select doc_id,docnm_kwd," + sql[6:]
|
262 |
+
else:
|
263 |
+
flds = []
|
264 |
+
for k in field_map.keys():
|
265 |
+
if k in forbidden_select_fields4resume:continue
|
266 |
+
if len(flds) > 11:break
|
267 |
+
flds.append(k)
|
268 |
+
sql = "select doc_id,docnm_kwd," + ",".join(flds) + sql[8:]
|
269 |
+
|
270 |
+
print(f"“{question}” get SQL(refined): {sql}")
|
271 |
+
|
272 |
+
chat_logger.info(f"“{question}” get SQL(refined): {sql}")
|
273 |
+
tried_times += 1
|
274 |
+
return retrievaler.sql_retrieval(sql, format="json"), sql
|
275 |
+
|
276 |
+
tbl, sql = get_table()
|
277 |
+
if tbl.get("error") and tried_times <= 2:
|
278 |
+
user_promt = """
|
279 |
+
表名:{};
|
280 |
+
数据库表字段说明如下:
|
281 |
+
{}
|
282 |
+
|
283 |
+
问题如下:
|
284 |
+
{}
|
285 |
+
|
286 |
+
你上一次给出的错误SQL如下:
|
287 |
+
{}
|
288 |
+
|
289 |
+
后台报错如下:
|
290 |
+
{}
|
291 |
+
|
292 |
+
请纠正SQL中的错误再写一遍,且只要SQL,不要有其他说明及文字。
|
293 |
+
""".format(
|
294 |
+
index_name(tenant_id),
|
295 |
+
"\n".join([f"{k}: {v}" for k, v in field_map.items()]),
|
296 |
+
question, sql, tbl["error"]
|
297 |
+
)
|
298 |
+
tbl, sql = get_table()
|
299 |
+
chat_logger.info("TRY it again: {}".format(sql))
|
300 |
+
|
301 |
+
chat_logger.info("GET table: {}".format(tbl))
|
302 |
+
print(tbl)
|
303 |
+
if tbl.get("error") or len(tbl["rows"]) == 0: return None, None
|
304 |
|
305 |
docid_idx = set([ii for ii, c in enumerate(tbl["columns"]) if c["name"] == "doc_id"])
|
306 |
docnm_idx = set([ii for ii, c in enumerate(tbl["columns"]) if c["name"] == "docnm_kwd"])
|
307 |
clmn_idx = [ii for ii in range(len(tbl["columns"])) if ii not in (docid_idx | docnm_idx)]
|
308 |
|
309 |
# compose markdown table
|
310 |
+
clmns = "|"+"|".join([re.sub(r"(/.*|([^()]+))", "", field_map.get(tbl["columns"][i]["name"], tbl["columns"][i]["name"])) for i in clmn_idx]) + ("|原文|" if docid_idx and docid_idx else "|")
|
311 |
+
line = "|"+"|".join(["------" for _ in range(len(clmn_idx))]) + ("|------|" if docid_idx and docid_idx else "")
|
312 |
+
rows = ["|"+"|".join([rmSpace(str(r[i])) for i in clmn_idx]).replace("None", " ") + "|" for r in tbl["rows"]]
|
313 |
if not docid_idx or not docnm_idx:
|
314 |
+
chat_logger.warning("SQL missing field: " + sql)
|
315 |
return "\n".join([clmns, line, "\n".join(rows)]), []
|
316 |
|
317 |
+
rows = "\n".join([r + f" ##{ii}$$ |" for ii, r in enumerate(rows)])
|
318 |
docid_idx = list(docid_idx)[0]
|
319 |
docnm_idx = list(docnm_idx)[0]
|
320 |
return "\n".join([clmns, line, rows]), [{"doc_id": r[docid_idx], "docnm_kwd": r[docnm_idx]} for r in tbl["rows"]]
|
api/db/db_models.py
CHANGED
@@ -502,7 +502,7 @@ class Document(DataBaseModel):
|
|
502 |
token_num = IntegerField(default=0)
|
503 |
chunk_num = IntegerField(default=0)
|
504 |
progress = FloatField(default=0)
|
505 |
-
progress_msg =
|
506 |
process_begin_at = DateTimeField(null=True)
|
507 |
process_duation = FloatField(default=0)
|
508 |
run = CharField(max_length=1, null=True, help_text="start to run processing or cancel.(1: run it; 2: cancel)", default="0")
|
@@ -520,7 +520,7 @@ class Task(DataBaseModel):
|
|
520 |
begin_at = DateTimeField(null=True)
|
521 |
process_duation = FloatField(default=0)
|
522 |
progress = FloatField(default=0)
|
523 |
-
progress_msg = TextField(
|
524 |
|
525 |
|
526 |
class Dialog(DataBaseModel):
|
|
|
502 |
token_num = IntegerField(default=0)
|
503 |
chunk_num = IntegerField(default=0)
|
504 |
progress = FloatField(default=0)
|
505 |
+
progress_msg = TextField(null=True, help_text="process message", default="")
|
506 |
process_begin_at = DateTimeField(null=True)
|
507 |
process_duation = FloatField(default=0)
|
508 |
run = CharField(max_length=1, null=True, help_text="start to run processing or cancel.(1: run it; 2: cancel)", default="0")
|
|
|
520 |
begin_at = DateTimeField(null=True)
|
521 |
process_duation = FloatField(default=0)
|
522 |
progress = FloatField(default=0)
|
523 |
+
progress_msg = TextField(null=True, help_text="process message", default="")
|
524 |
|
525 |
|
526 |
class Dialog(DataBaseModel):
|
api/db/init_data.py
CHANGED
@@ -90,6 +90,17 @@ def init_llm_factory():
|
|
90 |
"tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
|
91 |
"status": "1",
|
92 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
# {
|
94 |
# "name": "文心一言",
|
95 |
# "logo": "",
|
@@ -155,6 +166,12 @@ def init_llm_factory():
|
|
155 |
"tags": "LLM,CHAT,32K",
|
156 |
"max_tokens": 32768,
|
157 |
"model_type": LLMType.CHAT.value
|
|
|
|
|
|
|
|
|
|
|
|
|
158 |
},{
|
159 |
"fid": factory_infos[1]["name"],
|
160 |
"llm_name": "text-embedding-v2",
|
@@ -201,6 +218,46 @@ def init_llm_factory():
|
|
201 |
"max_tokens": 512,
|
202 |
"model_type": LLMType.EMBEDDING.value
|
203 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
204 |
]
|
205 |
for info in factory_infos:
|
206 |
LLMFactoriesService.save(**info)
|
|
|
90 |
"tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
|
91 |
"status": "1",
|
92 |
},
|
93 |
+
{
|
94 |
+
"name": "Local",
|
95 |
+
"logo": "",
|
96 |
+
"tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
|
97 |
+
"status": "0",
|
98 |
+
},{
|
99 |
+
"name": "Moonshot",
|
100 |
+
"logo": "",
|
101 |
+
"tags": "LLM,TEXT EMBEDDING",
|
102 |
+
"status": "1",
|
103 |
+
}
|
104 |
# {
|
105 |
# "name": "文心一言",
|
106 |
# "logo": "",
|
|
|
166 |
"tags": "LLM,CHAT,32K",
|
167 |
"max_tokens": 32768,
|
168 |
"model_type": LLMType.CHAT.value
|
169 |
+
},{
|
170 |
+
"fid": factory_infos[1]["name"],
|
171 |
+
"llm_name": "qwen-max-1201",
|
172 |
+
"tags": "LLM,CHAT,6K",
|
173 |
+
"max_tokens": 5899,
|
174 |
+
"model_type": LLMType.CHAT.value
|
175 |
},{
|
176 |
"fid": factory_infos[1]["name"],
|
177 |
"llm_name": "text-embedding-v2",
|
|
|
218 |
"max_tokens": 512,
|
219 |
"model_type": LLMType.EMBEDDING.value
|
220 |
},
|
221 |
+
# ---------------------- 本地 ----------------------
|
222 |
+
{
|
223 |
+
"fid": factory_infos[3]["name"],
|
224 |
+
"llm_name": "qwen-14B-chat",
|
225 |
+
"tags": "LLM,CHAT,",
|
226 |
+
"max_tokens": 8191,
|
227 |
+
"model_type": LLMType.CHAT.value
|
228 |
+
}, {
|
229 |
+
"fid": factory_infos[3]["name"],
|
230 |
+
"llm_name": "flag-enbedding",
|
231 |
+
"tags": "TEXT EMBEDDING,",
|
232 |
+
"max_tokens": 128 * 1000,
|
233 |
+
"model_type": LLMType.EMBEDDING.value
|
234 |
+
},
|
235 |
+
# ------------------------ Moonshot -----------------------
|
236 |
+
{
|
237 |
+
"fid": factory_infos[4]["name"],
|
238 |
+
"llm_name": "moonshot-v1-8k",
|
239 |
+
"tags": "LLM,CHAT,",
|
240 |
+
"max_tokens": 7900,
|
241 |
+
"model_type": LLMType.CHAT.value
|
242 |
+
}, {
|
243 |
+
"fid": factory_infos[4]["name"],
|
244 |
+
"llm_name": "flag-enbedding",
|
245 |
+
"tags": "TEXT EMBEDDING,",
|
246 |
+
"max_tokens": 128 * 1000,
|
247 |
+
"model_type": LLMType.EMBEDDING.value
|
248 |
+
},{
|
249 |
+
"fid": factory_infos[4]["name"],
|
250 |
+
"llm_name": "moonshot-v1-32k",
|
251 |
+
"tags": "LLM,CHAT,",
|
252 |
+
"max_tokens": 32768,
|
253 |
+
"model_type": LLMType.CHAT.value
|
254 |
+
},{
|
255 |
+
"fid": factory_infos[4]["name"],
|
256 |
+
"llm_name": "moonshot-v1-128k",
|
257 |
+
"tags": "LLM,CHAT",
|
258 |
+
"max_tokens": 128 * 1000,
|
259 |
+
"model_type": LLMType.CHAT.value
|
260 |
+
},
|
261 |
]
|
262 |
for info in factory_infos:
|
263 |
LLMFactoriesService.save(**info)
|
api/settings.py
CHANGED
@@ -29,6 +29,7 @@ LoggerFactory.LEVEL = 10
|
|
29 |
stat_logger = getLogger("stat")
|
30 |
access_logger = getLogger("access")
|
31 |
database_logger = getLogger("database")
|
|
|
32 |
|
33 |
API_VERSION = "v1"
|
34 |
RAG_FLOW_SERVICE_NAME = "ragflow"
|
@@ -69,9 +70,15 @@ default_llm = {
|
|
69 |
"image2text_model": "glm-4v",
|
70 |
"asr_model": "",
|
71 |
},
|
72 |
-
"
|
73 |
-
"chat_model": "",
|
74 |
-
"embedding_model": "",
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
"image2text_model": "",
|
76 |
"asr_model": "",
|
77 |
}
|
@@ -86,7 +93,7 @@ EMBEDDING_MDL = default_llm[LLM_FACTORY]["embedding_model"]
|
|
86 |
ASR_MDL = default_llm[LLM_FACTORY]["asr_model"]
|
87 |
IMAGE2TEXT_MDL = default_llm[LLM_FACTORY]["image2text_model"]
|
88 |
|
89 |
-
API_KEY = LLM.get("api_key", "
|
90 |
PARSERS = LLM.get("parsers", "naive:General,qa:Q&A,resume:Resume,table:Table,laws:Laws,manual:Manual,book:Book,paper:Paper,presentation:Presentation,picture:Picture")
|
91 |
|
92 |
# distribution
|
|
|
29 |
stat_logger = getLogger("stat")
|
30 |
access_logger = getLogger("access")
|
31 |
database_logger = getLogger("database")
|
32 |
+
chat_logger = getLogger("chat")
|
33 |
|
34 |
API_VERSION = "v1"
|
35 |
RAG_FLOW_SERVICE_NAME = "ragflow"
|
|
|
70 |
"image2text_model": "glm-4v",
|
71 |
"asr_model": "",
|
72 |
},
|
73 |
+
"Local": {
|
74 |
+
"chat_model": "qwen-14B-chat",
|
75 |
+
"embedding_model": "flag-enbedding",
|
76 |
+
"image2text_model": "",
|
77 |
+
"asr_model": "",
|
78 |
+
},
|
79 |
+
"Moonshot": {
|
80 |
+
"chat_model": "moonshot-v1-8k",
|
81 |
+
"embedding_model": "flag-enbedding",
|
82 |
"image2text_model": "",
|
83 |
"asr_model": "",
|
84 |
}
|
|
|
93 |
ASR_MDL = default_llm[LLM_FACTORY]["asr_model"]
|
94 |
IMAGE2TEXT_MDL = default_llm[LLM_FACTORY]["image2text_model"]
|
95 |
|
96 |
+
API_KEY = LLM.get("api_key", "")
|
97 |
PARSERS = LLM.get("parsers", "naive:General,qa:Q&A,resume:Resume,table:Table,laws:Laws,manual:Manual,book:Book,paper:Paper,presentation:Presentation,picture:Picture")
|
98 |
|
99 |
# distribution
|
deepdoc/parser/excel_parser.py
CHANGED
@@ -34,7 +34,7 @@ class HuExcelParser:
|
|
34 |
total = 0
|
35 |
for sheetname in wb.sheetnames:
|
36 |
ws = wb[sheetname]
|
37 |
-
total += len(ws.rows)
|
38 |
return total
|
39 |
|
40 |
if fnm.split(".")[-1].lower() in ["csv", "txt"]:
|
|
|
34 |
total = 0
|
35 |
for sheetname in wb.sheetnames:
|
36 |
ws = wb[sheetname]
|
37 |
+
total += len(list(ws.rows))
|
38 |
return total
|
39 |
|
40 |
if fnm.split(".")[-1].lower() in ["csv", "txt"]:
|
deepdoc/parser/pdf_parser.py
CHANGED
@@ -655,14 +655,14 @@ class HuParser:
|
|
655 |
#if min(tv, fv) > 2000:
|
656 |
# i += 1
|
657 |
# continue
|
658 |
-
if tv < fv:
|
659 |
tables[tk].insert(0, c)
|
660 |
logging.debug(
|
661 |
"TABLE:" +
|
662 |
self.boxes[i]["text"] +
|
663 |
"; Cap: " +
|
664 |
tk)
|
665 |
-
|
666 |
figures[fk].insert(0, c)
|
667 |
logging.debug(
|
668 |
"FIGURE:" +
|
|
|
655 |
#if min(tv, fv) > 2000:
|
656 |
# i += 1
|
657 |
# continue
|
658 |
+
if tv < fv and tk:
|
659 |
tables[tk].insert(0, c)
|
660 |
logging.debug(
|
661 |
"TABLE:" +
|
662 |
self.boxes[i]["text"] +
|
663 |
"; Cap: " +
|
664 |
tk)
|
665 |
+
elif fk:
|
666 |
figures[fk].insert(0, c)
|
667 |
logging.debug(
|
668 |
"FIGURE:" +
|
deepdoc/parser/ppt_parser.py
CHANGED
@@ -31,7 +31,7 @@ class HuPptParser(object):
|
|
31 |
|
32 |
if shape.shape_type == 6:
|
33 |
texts = []
|
34 |
-
for p in shape.shapes:
|
35 |
t = self.__extract(p)
|
36 |
if t: texts.append(t)
|
37 |
return "\n".join(texts)
|
@@ -46,7 +46,7 @@ class HuPptParser(object):
|
|
46 |
if i < from_page: continue
|
47 |
if i >= to_page:break
|
48 |
texts = []
|
49 |
-
for shape in slide.shapes:
|
50 |
txt = self.__extract(shape)
|
51 |
if txt: texts.append(txt)
|
52 |
txts.append("\n".join(texts))
|
|
|
31 |
|
32 |
if shape.shape_type == 6:
|
33 |
texts = []
|
34 |
+
for p in sorted(shape.shapes, key=lambda x: (x.top//10, x.left)):
|
35 |
t = self.__extract(p)
|
36 |
if t: texts.append(t)
|
37 |
return "\n".join(texts)
|
|
|
46 |
if i < from_page: continue
|
47 |
if i >= to_page:break
|
48 |
texts = []
|
49 |
+
for shape in sorted(slide.shapes, key=lambda x: (x.top//10, x.left)):
|
50 |
txt = self.__extract(shape)
|
51 |
if txt: texts.append(txt)
|
52 |
txts.append("\n".join(texts))
|
deepdoc/vision/ocr.py
CHANGED
@@ -64,10 +64,15 @@ def load_model(model_dir, nm):
|
|
64 |
raise ValueError("not find model file path {}".format(
|
65 |
model_file_path))
|
66 |
|
|
|
|
|
|
|
|
|
|
|
67 |
if ort.get_device() == "GPU":
|
68 |
-
sess = ort.InferenceSession(model_file_path, providers=['CUDAExecutionProvider'])
|
69 |
else:
|
70 |
-
sess = ort.InferenceSession(model_file_path, providers=['CPUExecutionProvider'])
|
71 |
return sess, sess.get_inputs()[0]
|
72 |
|
73 |
|
@@ -325,7 +330,13 @@ class TextRecognizer(object):
|
|
325 |
|
326 |
input_dict = {}
|
327 |
input_dict[self.input_tensor.name] = norm_img_batch
|
328 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
329 |
preds = outputs[0]
|
330 |
rec_result = self.postprocess_op(preds)
|
331 |
for rno in range(len(rec_result)):
|
@@ -430,7 +441,13 @@ class TextDetector(object):
|
|
430 |
img = img.copy()
|
431 |
input_dict = {}
|
432 |
input_dict[self.input_tensor.name] = img
|
433 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
434 |
|
435 |
post_result = self.postprocess_op({"maps": outputs[0]}, shape_list)
|
436 |
dt_boxes = post_result[0]['points']
|
|
|
64 |
raise ValueError("not find model file path {}".format(
|
65 |
model_file_path))
|
66 |
|
67 |
+
options = ort.SessionOptions()
|
68 |
+
options.enable_cpu_mem_arena = False
|
69 |
+
options.execution_mode = ort.ExecutionMode.ORT_SEQUENTIAL
|
70 |
+
options.intra_op_num_threads = 2
|
71 |
+
options.inter_op_num_threads = 2
|
72 |
if ort.get_device() == "GPU":
|
73 |
+
sess = ort.InferenceSession(model_file_path, options=options, providers=['CUDAExecutionProvider'])
|
74 |
else:
|
75 |
+
sess = ort.InferenceSession(model_file_path, options=options, providers=['CPUExecutionProvider'])
|
76 |
return sess, sess.get_inputs()[0]
|
77 |
|
78 |
|
|
|
330 |
|
331 |
input_dict = {}
|
332 |
input_dict[self.input_tensor.name] = norm_img_batch
|
333 |
+
for i in range(100000):
|
334 |
+
try:
|
335 |
+
outputs = self.predictor.run(None, input_dict)
|
336 |
+
break
|
337 |
+
except Exception as e:
|
338 |
+
if i >= 3: raise e
|
339 |
+
time.sleep(5)
|
340 |
preds = outputs[0]
|
341 |
rec_result = self.postprocess_op(preds)
|
342 |
for rno in range(len(rec_result)):
|
|
|
441 |
img = img.copy()
|
442 |
input_dict = {}
|
443 |
input_dict[self.input_tensor.name] = img
|
444 |
+
for i in range(100000):
|
445 |
+
try:
|
446 |
+
outputs = self.predictor.run(None, input_dict)
|
447 |
+
break
|
448 |
+
except Exception as e:
|
449 |
+
if i >= 3: raise e
|
450 |
+
time.sleep(5)
|
451 |
|
452 |
post_result = self.postprocess_op({"maps": outputs[0]}, shape_list)
|
453 |
dt_boxes = post_result[0]['points']
|
deepdoc/vision/recognizer.py
CHANGED
@@ -42,7 +42,9 @@ class Recognizer(object):
|
|
42 |
raise ValueError("not find model file path {}".format(
|
43 |
model_file_path))
|
44 |
if ort.get_device() == "GPU":
|
45 |
-
|
|
|
|
|
46 |
else:
|
47 |
self.ort_sess = ort.InferenceSession(model_file_path, providers=['CPUExecutionProvider'])
|
48 |
self.input_names = [node.name for node in self.ort_sess.get_inputs()]
|
|
|
42 |
raise ValueError("not find model file path {}".format(
|
43 |
model_file_path))
|
44 |
if ort.get_device() == "GPU":
|
45 |
+
options = ort.SessionOptions()
|
46 |
+
options.enable_cpu_mem_arena = False
|
47 |
+
self.ort_sess = ort.InferenceSession(model_file_path, options=options, providers=[('CUDAExecutionProvider')])
|
48 |
else:
|
49 |
self.ort_sess = ort.InferenceSession(model_file_path, providers=['CPUExecutionProvider'])
|
50 |
self.input_names = [node.name for node in self.ort_sess.get_inputs()]
|
rag/app/table.py
CHANGED
@@ -67,7 +67,7 @@ class Excel(ExcelParser):
|
|
67 |
|
68 |
def trans_datatime(s):
|
69 |
try:
|
70 |
-
return datetime_parse(s.strip()).strftime("%Y-%m-%
|
71 |
except Exception as e:
|
72 |
pass
|
73 |
|
@@ -80,6 +80,7 @@ def trans_bool(s):
|
|
80 |
|
81 |
|
82 |
def column_data_type(arr):
|
|
|
83 |
uni = len(set([a for a in arr if a is not None]))
|
84 |
counts = {"int": 0, "float": 0, "text": 0, "datetime": 0, "bool": 0}
|
85 |
trans = {t: f for f, t in
|
@@ -130,7 +131,7 @@ def chunk(filename, binary=None, from_page=0, to_page=10000000000, lang="Chinese
|
|
130 |
if re.search(r"\.xlsx?$", filename, re.IGNORECASE):
|
131 |
callback(0.1, "Start to parse.")
|
132 |
excel_parser = Excel()
|
133 |
-
dfs = excel_parser(filename, binary, callback)
|
134 |
elif re.search(r"\.(txt|csv)$", filename, re.IGNORECASE):
|
135 |
callback(0.1, "Start to parse.")
|
136 |
txt = ""
|
@@ -188,7 +189,7 @@ def chunk(filename, binary=None, from_page=0, to_page=10000000000, lang="Chinese
|
|
188 |
df[clmns[j]] = cln
|
189 |
if ty == "text":
|
190 |
txts.extend([str(c) for c in cln if c])
|
191 |
-
clmns_map = [(py_clmns[i] + fieds_map[clmn_tys[i]], clmns[i])
|
192 |
for i in range(len(clmns))]
|
193 |
|
194 |
eng = lang.lower() == "english"#is_english(txts)
|
@@ -201,6 +202,8 @@ def chunk(filename, binary=None, from_page=0, to_page=10000000000, lang="Chinese
|
|
201 |
for j in range(len(clmns)):
|
202 |
if row[clmns[j]] is None:
|
203 |
continue
|
|
|
|
|
204 |
fld = clmns_map[j][0]
|
205 |
d[fld] = row[clmns[j]] if clmn_tys[j] != "text" else huqie.qie(
|
206 |
row[clmns[j]])
|
|
|
67 |
|
68 |
def trans_datatime(s):
|
69 |
try:
|
70 |
+
return datetime_parse(s.strip()).strftime("%Y-%m-%d %H:%M:%S")
|
71 |
except Exception as e:
|
72 |
pass
|
73 |
|
|
|
80 |
|
81 |
|
82 |
def column_data_type(arr):
|
83 |
+
arr = list(arr)
|
84 |
uni = len(set([a for a in arr if a is not None]))
|
85 |
counts = {"int": 0, "float": 0, "text": 0, "datetime": 0, "bool": 0}
|
86 |
trans = {t: f for f, t in
|
|
|
131 |
if re.search(r"\.xlsx?$", filename, re.IGNORECASE):
|
132 |
callback(0.1, "Start to parse.")
|
133 |
excel_parser = Excel()
|
134 |
+
dfs = excel_parser(filename, binary, from_page=from_page, to_page=to_page, callback=callback)
|
135 |
elif re.search(r"\.(txt|csv)$", filename, re.IGNORECASE):
|
136 |
callback(0.1, "Start to parse.")
|
137 |
txt = ""
|
|
|
189 |
df[clmns[j]] = cln
|
190 |
if ty == "text":
|
191 |
txts.extend([str(c) for c in cln if c])
|
192 |
+
clmns_map = [(py_clmns[i] + fieds_map[clmn_tys[i]], clmns[i].replace("_", " "))
|
193 |
for i in range(len(clmns))]
|
194 |
|
195 |
eng = lang.lower() == "english"#is_english(txts)
|
|
|
202 |
for j in range(len(clmns)):
|
203 |
if row[clmns[j]] is None:
|
204 |
continue
|
205 |
+
if not str(row[clmns[j]]):
|
206 |
+
continue
|
207 |
fld = clmns_map[j][0]
|
208 |
d[fld] = row[clmns[j]] if clmn_tys[j] != "text" else huqie.qie(
|
209 |
row[clmns[j]])
|
rag/llm/__init__.py
CHANGED
@@ -19,18 +19,20 @@ from .cv_model import *
|
|
19 |
|
20 |
|
21 |
EmbeddingModel = {
|
22 |
-
"
|
23 |
"OpenAI": OpenAIEmbed,
|
24 |
"通义千问": HuEmbedding, #QWenEmbed,
|
25 |
-
"智谱AI": ZhipuEmbed
|
|
|
26 |
}
|
27 |
|
28 |
|
29 |
CvModel = {
|
30 |
"OpenAI": GptV4,
|
31 |
-
"
|
32 |
"通义千问": QWenCV,
|
33 |
-
"智谱AI": Zhipu4V
|
|
|
34 |
}
|
35 |
|
36 |
|
@@ -38,6 +40,7 @@ ChatModel = {
|
|
38 |
"OpenAI": GptTurbo,
|
39 |
"智谱AI": ZhipuChat,
|
40 |
"通义千问": QWenChat,
|
41 |
-
"
|
|
|
42 |
}
|
43 |
|
|
|
19 |
|
20 |
|
21 |
EmbeddingModel = {
|
22 |
+
"Local": HuEmbedding,
|
23 |
"OpenAI": OpenAIEmbed,
|
24 |
"通义千问": HuEmbedding, #QWenEmbed,
|
25 |
+
"智谱AI": ZhipuEmbed,
|
26 |
+
"Moonshot": HuEmbedding
|
27 |
}
|
28 |
|
29 |
|
30 |
CvModel = {
|
31 |
"OpenAI": GptV4,
|
32 |
+
"Local": LocalCV,
|
33 |
"通义千问": QWenCV,
|
34 |
+
"智谱AI": Zhipu4V,
|
35 |
+
"Moonshot": LocalCV
|
36 |
}
|
37 |
|
38 |
|
|
|
40 |
"OpenAI": GptTurbo,
|
41 |
"智谱AI": ZhipuChat,
|
42 |
"通义千问": QWenChat,
|
43 |
+
"Local": LocalLLM,
|
44 |
+
"Moonshot": MoonshotChat
|
45 |
}
|
46 |
|
rag/llm/chat_model.py
CHANGED
@@ -14,11 +14,8 @@
|
|
14 |
# limitations under the License.
|
15 |
#
|
16 |
from abc import ABC
|
17 |
-
from copy import deepcopy
|
18 |
-
|
19 |
from openai import OpenAI
|
20 |
import openai
|
21 |
-
|
22 |
from rag.nlp import is_english
|
23 |
from rag.utils import num_tokens_from_string
|
24 |
|
@@ -52,6 +49,12 @@ class GptTurbo(Base):
|
|
52 |
return "**ERROR**: "+str(e), 0
|
53 |
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
from dashscope import Generation
|
56 |
class QWenChat(Base):
|
57 |
def __init__(self, key, model_name=Generation.Models.qwen_turbo):
|
|
|
14 |
# limitations under the License.
|
15 |
#
|
16 |
from abc import ABC
|
|
|
|
|
17 |
from openai import OpenAI
|
18 |
import openai
|
|
|
19 |
from rag.nlp import is_english
|
20 |
from rag.utils import num_tokens_from_string
|
21 |
|
|
|
49 |
return "**ERROR**: "+str(e), 0
|
50 |
|
51 |
|
52 |
+
class MoonshotChat(GptTurbo):
|
53 |
+
def __init__(self, key, model_name="moonshot-v1-8k"):
|
54 |
+
self.client = OpenAI(api_key=key, base_url="https://api.moonshot.cn/v1",)
|
55 |
+
self.model_name = model_name
|
56 |
+
|
57 |
+
|
58 |
from dashscope import Generation
|
59 |
class QWenChat(Base):
|
60 |
def __init__(self, key, model_name=Generation.Models.qwen_turbo):
|
rag/llm/rpc_server.py
CHANGED
@@ -4,7 +4,7 @@ import random
|
|
4 |
import time
|
5 |
from multiprocessing.connection import Listener
|
6 |
from threading import Thread
|
7 |
-
import
|
8 |
|
9 |
|
10 |
class RPCHandler:
|
@@ -47,14 +47,27 @@ tokenizer = None
|
|
47 |
def chat(messages, gen_conf):
|
48 |
global tokenizer
|
49 |
model = Model()
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
|
60 |
def Model():
|
@@ -71,20 +84,13 @@ if __name__ == "__main__":
|
|
71 |
handler = RPCHandler()
|
72 |
handler.register_function(chat)
|
73 |
|
74 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
75 |
-
from transformers.generation.utils import GenerationConfig
|
76 |
-
|
77 |
models = []
|
78 |
-
for _ in range(
|
79 |
m = AutoModelForCausalLM.from_pretrained(args.model_name,
|
80 |
device_map="auto",
|
81 |
-
torch_dtype='auto'
|
82 |
-
trust_remote_code=True)
|
83 |
-
m.generation_config = GenerationConfig.from_pretrained(args.model_name)
|
84 |
-
m.generation_config.pad_token_id = m.generation_config.eos_token_id
|
85 |
models.append(m)
|
86 |
-
tokenizer = AutoTokenizer.from_pretrained(args.model_name
|
87 |
-
trust_remote_code=True)
|
88 |
|
89 |
# Run the server
|
90 |
rpc_server(handler, ('0.0.0.0', args.port), authkey=b'infiniflow-token4kevinhu')
|
|
|
4 |
import time
|
5 |
from multiprocessing.connection import Listener
|
6 |
from threading import Thread
|
7 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
8 |
|
9 |
|
10 |
class RPCHandler:
|
|
|
47 |
def chat(messages, gen_conf):
|
48 |
global tokenizer
|
49 |
model = Model()
|
50 |
+
try:
|
51 |
+
conf = {"max_new_tokens": int(gen_conf.get("max_tokens", 256)), "temperature": float(gen_conf.get("temperature", 0.1))}
|
52 |
+
print(messages, conf)
|
53 |
+
text = tokenizer.apply_chat_template(
|
54 |
+
messages,
|
55 |
+
tokenize=False,
|
56 |
+
add_generation_prompt=True
|
57 |
+
)
|
58 |
+
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
|
59 |
+
|
60 |
+
generated_ids = model.generate(
|
61 |
+
model_inputs.input_ids,
|
62 |
+
**conf
|
63 |
+
)
|
64 |
+
generated_ids = [
|
65 |
+
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
|
66 |
+
]
|
67 |
+
|
68 |
+
return tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
69 |
+
except Exception as e:
|
70 |
+
return str(e)
|
71 |
|
72 |
|
73 |
def Model():
|
|
|
84 |
handler = RPCHandler()
|
85 |
handler.register_function(chat)
|
86 |
|
|
|
|
|
|
|
87 |
models = []
|
88 |
+
for _ in range(1):
|
89 |
m = AutoModelForCausalLM.from_pretrained(args.model_name,
|
90 |
device_map="auto",
|
91 |
+
torch_dtype='auto')
|
|
|
|
|
|
|
92 |
models.append(m)
|
93 |
+
tokenizer = AutoTokenizer.from_pretrained(args.model_name)
|
|
|
94 |
|
95 |
# Run the server
|
96 |
rpc_server(handler, ('0.0.0.0', args.port), authkey=b'infiniflow-token4kevinhu')
|
rag/nlp/search.py
CHANGED
@@ -7,6 +7,7 @@ from elasticsearch_dsl import Q, Search
|
|
7 |
from typing import List, Optional, Dict, Union
|
8 |
from dataclasses import dataclass
|
9 |
|
|
|
10 |
from rag.settings import es_logger
|
11 |
from rag.utils import rmSpace
|
12 |
from rag.nlp import huqie, query
|
@@ -333,15 +334,16 @@ class Dealer:
|
|
333 |
replaces = []
|
334 |
for r in re.finditer(r" ([a-z_]+_l?tks)( like | ?= ?)'([^']+)'", sql):
|
335 |
fld, v = r.group(1), r.group(3)
|
336 |
-
match = " MATCH({}, '{}', 'operator=OR;
|
337 |
replaces.append(("{}{}'{}'".format(r.group(1), r.group(2), r.group(3)), match))
|
338 |
|
339 |
for p, r in replaces: sql = sql.replace(p, r, 1)
|
340 |
-
|
341 |
|
342 |
try:
|
343 |
tbl = self.es.sql(sql, fetch_size, format)
|
344 |
return tbl
|
345 |
except Exception as e:
|
346 |
-
|
|
|
347 |
|
|
|
7 |
from typing import List, Optional, Dict, Union
|
8 |
from dataclasses import dataclass
|
9 |
|
10 |
+
from api.settings import chat_logger
|
11 |
from rag.settings import es_logger
|
12 |
from rag.utils import rmSpace
|
13 |
from rag.nlp import huqie, query
|
|
|
334 |
replaces = []
|
335 |
for r in re.finditer(r" ([a-z_]+_l?tks)( like | ?= ?)'([^']+)'", sql):
|
336 |
fld, v = r.group(1), r.group(3)
|
337 |
+
match = " MATCH({}, '{}', 'operator=OR;minimum_should_match=30%') ".format(fld, huqie.qieqie(huqie.qie(v)))
|
338 |
replaces.append(("{}{}'{}'".format(r.group(1), r.group(2), r.group(3)), match))
|
339 |
|
340 |
for p, r in replaces: sql = sql.replace(p, r, 1)
|
341 |
+
chat_logger.info(f"To es: {sql}")
|
342 |
|
343 |
try:
|
344 |
tbl = self.es.sql(sql, fetch_size, format)
|
345 |
return tbl
|
346 |
except Exception as e:
|
347 |
+
chat_logger.error(f"SQL failure: {sql} =>" + str(e))
|
348 |
+
return {"error": str(e)}
|
349 |
|
rag/svr/task_executor.py
CHANGED
@@ -169,16 +169,25 @@ def init_kb(row):
|
|
169 |
|
170 |
|
171 |
def embedding(docs, mdl, parser_config={}, callback=None):
|
|
|
172 |
tts, cnts = [rmSpace(d["title_tks"]) for d in docs if d.get("title_tks")], [
|
173 |
d["content_with_weight"] for d in docs]
|
174 |
tk_count = 0
|
175 |
if len(tts) == len(cnts):
|
176 |
-
|
177 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
|
179 |
cnts_ = np.array([])
|
180 |
-
for i in range(0, len(cnts),
|
181 |
-
vts, c = mdl.encode(cnts[i: i+
|
182 |
if len(cnts_) == 0: cnts_ = vts
|
183 |
else: cnts_ = np.concatenate((cnts_, vts), axis=0)
|
184 |
tk_count += c
|
|
|
169 |
|
170 |
|
171 |
def embedding(docs, mdl, parser_config={}, callback=None):
|
172 |
+
batch_size = 32
|
173 |
tts, cnts = [rmSpace(d["title_tks"]) for d in docs if d.get("title_tks")], [
|
174 |
d["content_with_weight"] for d in docs]
|
175 |
tk_count = 0
|
176 |
if len(tts) == len(cnts):
|
177 |
+
tts_ = np.array([])
|
178 |
+
for i in range(0, len(tts), batch_size):
|
179 |
+
vts, c = mdl.encode(tts[i: i + batch_size])
|
180 |
+
if len(tts_) == 0:
|
181 |
+
tts_ = vts
|
182 |
+
else:
|
183 |
+
tts_ = np.concatenate((tts_, vts), axis=0)
|
184 |
+
tk_count += c
|
185 |
+
callback(prog=0.6 + 0.1 * (i + 1) / len(tts), msg="")
|
186 |
+
tts = tts_
|
187 |
|
188 |
cnts_ = np.array([])
|
189 |
+
for i in range(0, len(cnts), batch_size):
|
190 |
+
vts, c = mdl.encode(cnts[i: i+batch_size])
|
191 |
if len(cnts_) == 0: cnts_ = vts
|
192 |
else: cnts_ = np.concatenate((cnts_, vts), axis=0)
|
193 |
tk_count += c
|
rag/utils/es_conn.py
CHANGED
@@ -249,6 +249,8 @@ class HuEs:
|
|
249 |
except ConnectionTimeout as e:
|
250 |
es_logger.error("Timeout【Q】:" + sql)
|
251 |
continue
|
|
|
|
|
252 |
es_logger.error("ES search timeout for 3 times!")
|
253 |
raise ConnectionTimeout()
|
254 |
|
|
|
249 |
except ConnectionTimeout as e:
|
250 |
es_logger.error("Timeout【Q】:" + sql)
|
251 |
continue
|
252 |
+
except Exception as e:
|
253 |
+
raise e
|
254 |
es_logger.error("ES search timeout for 3 times!")
|
255 |
raise ConnectionTimeout()
|
256 |
|