enlarge docker memory usage (#501)

### What problem does this PR solve?

### Type of change

- [x] Refactoring

deepdoc/parser/pdf_parser.py

@@ -11,7 +11,7 @@ import pdfplumber
 import logging
 from PIL import Image, ImageDraw
 import numpy as np
-
+from timeit import default_timer as timer
 from PyPDF2 import PdfReader as pdf2_read
 
 from api.utils.file_utils import get_project_base_directory
@@ -936,6 +936,7 @@ class HuParser:
         self.page_cum_height = [0]
         self.page_layout = []
         self.page_from = page_from
+        st = timer()
         try:
             self.pdf = pdfplumber.open(fnm) if isinstance(
                 fnm, str) else pdfplumber.open(BytesIO(fnm))
@@ -989,7 +990,9 @@ class HuParser:
             self.is_english = True
         else:
             self.is_english = False
+        self.is_english = False
 
+        st = timer()
         for i, img in enumerate(self.page_images):
             chars = self.page_chars[i] if not self.is_english else []
             self.mean_height.append(
@@ -1007,15 +1010,11 @@ class HuParser:
                                                                        chars[j]["width"]) / 2:
                     chars[j]["text"] += " "
                 j += 1
-            # if i > 0:
-            #     if not chars:
-            #         self.page_cum_height.append(img.size[1] / zoomin)
-            #     else:
-            #         self.page_cum_height.append(
-            #             np.max([c["bottom"] for c in chars]))
+
             self.__ocr(i + 1, img, chars, zoomin)
-            if callback:
-                callback(prog=(i + 1) * 0.6 / len(self.page_images), msg="")
+            #if callback:
+            #    callback(prog=(i + 1) * 0.6 / len(self.page_images), msg="")
+        #print("OCR:", timer()-st)
 
         if not self.is_english and not any(
                 [c for c in self.page_chars]) and self.boxes:

docker/.env

@@ -11,7 +11,9 @@ ES_PORT=1200
 KIBANA_PORT=6601
 
 # Increase or decrease based on the available host memory (in bytes)
-MEM_LIMIT=12073741824
+
+MEM_LIMIT=8073741824
+
 
 MYSQL_PASSWORD=infini_rag_flow
 MYSQL_PORT=5455

docker/docker-compose-base.yml

@@ -29,23 +29,23 @@ services:
       - ragflow
     restart: always
 
-  kibana:
-    depends_on:
-        es01:
-          condition: service_healthy
-    image: docker.elastic.co/kibana/kibana:${STACK_VERSION}
-    container_name: ragflow-kibana
-    volumes:
-      - kibanadata:/usr/share/kibana/data
-    ports:
-      - ${KIBANA_PORT}:5601
-    environment:
-      - SERVERNAME=kibana
-      - ELASTICSEARCH_HOSTS=http://es01:9200
-      - TZ=${TIMEZONE}
-    mem_limit: ${MEM_LIMIT}
-    networks:
-      - ragflow
+  #kibana:
+  #  depends_on:
+  #      es01:
+  #        condition: service_healthy
+  #  image: docker.elastic.co/kibana/kibana:${STACK_VERSION}
+  #  container_name: ragflow-kibana
+  #  volumes:
+  #    - kibanadata:/usr/share/kibana/data
+  #  ports:
+  #    - ${KIBANA_PORT}:5601
+  #  environment:
+  #    - SERVERNAME=kibana
+  #    - ELASTICSEARCH_HOSTS=http://es01:9200
+  #    - TZ=${TIMEZONE}
+  #  mem_limit: ${MEM_LIMIT}
+  #  networks:
+  #    - ragflow
 
   mysql:
     image: mysql:5.7.18

docker/entrypoint.sh

@@ -29,7 +29,7 @@ function task_bro(){
 
 task_bro &
 
-WS=2
+WS=1
 for ((i=0;i<WS;i++))
 do
   task_exe $i $WS &

rag/app/book.py

@@ -37,7 +37,7 @@ class Pdf(PdfParser):
         start = timer()
         self._layouts_rec(zoomin)
         callback(0.67, "Layout analysis finished")
-        print("paddle layouts:", timer() - start)
+        print("layouts:", timer() - start)
         self._table_transformer_job(zoomin)
         callback(0.68, "Table analysis finished")
         self._text_merge()

rag/app/laws.py

@@ -71,7 +71,7 @@ class Pdf(PdfParser):
         start = timer()
         self._layouts_rec(zoomin)
         callback(0.67, "Layout analysis finished")
-        cron_logger.info("paddle layouts:".format(
+        cron_logger.info("layouts:".format(
             (timer() - start) / (self.total_page + 0.1)))
         self._naive_vertical_merge()
 

rag/app/manual.py

@@ -32,7 +32,7 @@ class Pdf(PdfParser):
 
         self._layouts_rec(zoomin)
         callback(0.65, "Layout analysis finished.")
-        print("paddle layouts:", timer() - start)
+        print("layouts:", timer() - start)
         self._table_transformer_job(zoomin)
         callback(0.67, "Table analysis finished.")
         self._text_merge()

rag/app/naive.py

@@ -77,12 +77,12 @@ class Pdf(PdfParser):
             callback
         )
         callback(msg="OCR finished")
-        cron_logger.info("OCR: {}".format(timer() - start))
+        cron_logger.info("OCR({}~{}): {}".format(from_page, to_page, timer() - start))
 
         start = timer()
         self._layouts_rec(zoomin)
         callback(0.63, "Layout analysis finished.")
-        print("paddle layouts:", timer() - start)
+        print("layouts:", timer() - start)
         self._table_transformer_job(zoomin)
         callback(0.65, "Table analysis finished.")
         self._text_merge()
@@ -92,7 +92,7 @@ class Pdf(PdfParser):
         self._concat_downward()
         #self._filter_forpages()
 
-        cron_logger.info("paddle layouts: {}".format(
+        cron_logger.info("layouts: {}".format(
             (timer() - start) / (self.total_page + 0.1)))
         return [(b["text"], self._line_tag(b, zoomin))
                 for b in self.boxes], tbls

rag/app/one.py

@@ -33,7 +33,7 @@ class Pdf(PdfParser):
         start = timer()
         self._layouts_rec(zoomin, drop=False)
         callback(0.63, "Layout analysis finished.")
-        print("paddle layouts:", timer() - start)
+        print("layouts:", timer() - start)
         self._table_transformer_job(zoomin)
         callback(0.65, "Table analysis finished.")
         self._text_merge()

rag/app/paper.py

@@ -42,7 +42,7 @@ class Pdf(PdfParser):
         start = timer()
         self._layouts_rec(zoomin)
         callback(0.63, "Layout analysis finished")
-        print("paddle layouts:", timer() - start)
+        print("layouts:", timer() - start)
         self._table_transformer_job(zoomin)
         callback(0.68, "Table analysis finished")
         self._text_merge()

rag/svr/task_broker.py

@@ -33,6 +33,8 @@ from api.settings import database_logger
 from api.utils import get_format_time, get_uuid
 from api.utils.file_utils import get_project_base_directory
 from rag.utils.redis_conn import REDIS_CONN
+from api.db.db_models import init_database_tables as init_web_db
+from api.db.init_data import init_web_data
 
 
 def collect(tm):
@@ -181,6 +183,9 @@ if __name__ == "__main__":
     peewee_logger.propagate = False
     peewee_logger.addHandler(database_logger.handlers[0])
     peewee_logger.setLevel(database_logger.level)
+    # init db
+    init_web_db()
+    init_web_data()
 
     while True:
         dispatch()

rag/svr/task_executor.py

@@ -163,6 +163,7 @@ def build(row):
         "doc_id": row["doc_id"],
         "kb_id": [str(row["kb_id"])]
     }
+    el = 0
     for ck in cks:
         d = copy.deepcopy(doc)
         d.update(ck)
@@ -182,10 +183,13 @@ def build(row):
         else:
             d["image"].save(output_buffer, format='JPEG')
 
+        st = timer()
         MINIO.put(row["kb_id"], d["_id"], output_buffer.getvalue())
+        el += timer() - st
         d["img_id"] = "{}-{}".format(row["kb_id"], d["_id"])
         del d["image"]
         docs.append(d)
+    cron_logger.info("MINIO PUT({}):{}".format(row["name"], el))
 
     return docs
 
@@ -258,7 +262,9 @@ def main(comm, mod):
             callback(prog=-1, msg=str(e))
             continue
 
+        st = timer()
         cks = build(r)
+        cron_logger.info("Build chunks({}): {}".format(r["name"], timer()-st))
         if cks is None:
             continue
         if not cks:
@@ -277,12 +283,14 @@ def main(comm, mod):
             callback(-1, "Embedding error:{}".format(str(e)))
             cron_logger.error(str(e))
             tk_count = 0
+        cron_logger.info("Embedding elapsed({}): {}".format(r["name"], timer()-st))
 
         callback(msg="Finished embedding({})! Start to build index!".format(timer()-st))
         init_kb(r)
         chunk_count = len(set([c["_id"] for c in cks]))
         st = timer()
         es_r = ELASTICSEARCH.bulk(cks, search.index_name(r["tenant_id"]))
+        cron_logger.info("Indexing elapsed({}): {}".format(r["name"], timer()-st))
         if es_r:
             callback(-1, "Index failure!")
             ELASTICSEARCH.deleteByQuery(