Hugging Face's logo

Datasets:
tiendung
/
NAM003

Modalities:
Text
Formats:
json
Size:
10M - 100M
Libraries:
Datasets
Dask
Polars
Dataset card Data Studio Files
xet
 Community
tiendung commited on
Commit
6d2e2d1
·
1 Parent(s): 2d85795

up

Browse files
Files changed (16) hide show
  1. .gitignore +2 -0
  2. .utils/code.md +48 -0
  3. .utils/code_list.sh +26 -0
  4. .utils/epub2text.py +66 -0
  5. .utils/join.py +49 -0
  6. .utils/json2jsonl.py +60 -0
  7. .utils/lzma.py +8 -0
  8. .utils/mixsplit.py +20 -0
  9. .utils/parquet2jsonl.py +39 -0
  10. .utils/slimpajama.py +10 -0
  11. .utils/slimpajama.sh +21 -0
  12. .utils/verify.py +8 -0
  13. .utils/verify.sh +100 -0
  14. .utils/wudao.md +54 -0
  15. README.md +183 -0
  16. select/.keep +0 -0
.gitignore ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ __pycache__
2
+ .utils/bloom_tknz
.utils/code.md ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ```sh
2
+ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/bigcode/the-stack-dedup
3
+ cd the-stack-dedup
4
+
5
+ git lfs pull --include "/data/java/data-00005*"
6
+ git lfs pull --include "/data/python/data-00005*"
7
+ git lfs pull --include "/data/hcl/data-00005*"
8
+ git lfs pull --include "/data/ruby/data-00005*"
9
+ git lfs pull --include "/data/c-sharp/data-00005*"
10
+ git lfs pull --include "/data/clojure/data-00005*"
11
+ git lfs pull --include "/data/coffeescript/data-00005*"
12
+ git lfs pull --include "/data/crystal/data-00005*"
13
+ git lfs pull --include "/data/dart/data-00005*"
14
+ git lfs pull --include "/data/elixir/data-00005*"
15
+ git lfs pull --include "/data/elm/data-00005*"
16
+ git lfs pull --include "/data/f-sharp/data-00005*"
17
+ git lfs pull --include "/data/erlang/data-00005*"
18
+ git lfs pull --include "/data/go/data-00005*"
19
+ git lfs pull --include "/data/haskell/data-00005*"
20
+ git lfs pull --include "/data/julia/data-00005*"
21
+ git lfs pull --include "/data/kotlin/data-00005*"
22
+ git lfs pull --include "/data/sql/data-00005*"
23
+ git lfs pull --include "/data/toml/data-00005*"
24
+ git lfs pull --include "/data/typescript/data-00005*"
25
+
26
+
27
+ lzma -z9 c-sharp.jsonl &
28
+ lzma -z9 clojure.jsonl &
29
+ lzma -z9 coffeescript.jsonl &
30
+ lzma -z9 crystal.jsonl &
31
+ lzma -z9 dart.jsonl &
32
+ lzma -z9 elixir.jsonl &
33
+ lzma -z9 elm.jsonl &
34
+ lzma -z9 erlang.jsonl &
35
+ lzma -z9 f-sharp.jsonl &
36
+ lzma -z9 go.jsonl &
37
+ lzma -z9 haskell.jsonl &
38
+ lzma -z9 hcl.jsonl &
39
+ lzma -z9 java.jsonl &
40
+ lzma -z9 julia.jsonl &
41
+ lzma -z9 kotlin.jsonl &
42
+ lzma -z9 python.jsonl &
43
+ lzma -z9 ruby.jsonl &
44
+ lzma -z9 sql.jsonl &
45
+ lzma -z9 toml.jsonl &
46
+ lzma -z9 typescript.jsonl &
47
+ lzma -z9 zig.jsonl &
48
+ ```
.utils/code_list.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ls data/markdown/data-00000*
2
+ ls data/java/data-00000*
3
+ ls data/python/data-00000*
4
+ ls data/yaml/data-00000*
5
+ ls data/json/data-00000*
6
+ ls data/hcl/data-00000*
7
+ ls data/ruby/data-00000*
8
+ ls data/c-sharp/data-00000*
9
+ ls data/clojure/data-00000*
10
+ ls data/coffeescript/data-00000*
11
+ ls data/crystal/data-00000*
12
+ ls data/dart/data-00000*
13
+ ls data/elixir/data-00000*
14
+ ls data/elm/data-00000*
15
+ ls data/f-sharp/data-00000*
16
+ ls data/erlang/data-00000*
17
+ ls data/go/data-00000*
18
+ ls data/haskell/data-00000*
19
+ ls data/julia/data-00000*
20
+ ls data/json5/data-00000*
21
+ ls data/kotlin/data-00000*
22
+ ls data/sql/data-00000*
23
+ ls data/toml/data-00000*
24
+ ls data/typescript/data-00000*
25
+ ls data/zig/data-00000*
26
+
.utils/epub2text.py ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ '''
2
+ pip install beautifulsoup4
3
+ pip install re
4
+ pip install html2text
5
+ pip install epub-conversion
6
+ pip install xml_cleaner
7
+ '''
8
+
9
+ from bs4 import BeautifulSoup
10
+ import html2text
11
+ from epub_conversion.utils import open_book, convert_epub_to_lines, convert_lines_to_text
12
+ from xml_cleaner import to_raw_text
13
+ import re, html
14
+ from tqdm import tqdm
15
+ import glob, os
16
+
17
+ def clean_string(st):
18
+ return '\n'.join(re.sub(r'([\n\t]+)', '\n', re.sub(r'\[[^\]]*\]', '', st)).replace('\r', ' ').split('\n \n'))
19
+
20
+ TCVN3TAB = "µ¸¶·¹¨»¾¼½Æ©ÇÊÈÉË®ÌÐÎÏѪÒÕÓÔÖ×ÝØÜÞßãáâä«åèæçé¬êíëìîïóñòôõøö÷ùúýûüþ¡¢§£¤¥¦−" # NOQA
21
+ TCVN3TAB = [ch for ch in TCVN3TAB]
22
+
23
+ UNICODETAB = "àáảãạăằắẳẵặâầấẩẫậđèéẻẽẹêềếểễệìíỉĩịòóỏõọôồốổỗộơờớởỡợùúủũụừứửữựỳýỷỹỵĂÂĐÊÔƠƯư" # NOQA
24
+ UNICODETAB = [ch for ch in UNICODETAB]
25
+
26
+ r = re.compile("|".join(TCVN3TAB))
27
+ replaces_dict = dict(zip(TCVN3TAB, UNICODETAB))
28
+
29
+ def TCVN3_to_unicode(tcvn3str):
30
+ return r.sub(lambda m: replaces_dict[m.group(0)], tcvn3str)
31
+
32
+
33
+ def unicode_to_TCVN3(unicodestr):
34
+ return r.sub(lambda m: replaces_dict[m.group(0)], unicodestr)
35
+
36
+ os.chdir(r'D:\Python\ePub\Sách\output')
37
+ path = 'D:\Python\ePub\Sách\epub'
38
+ text_files = [f for f in os.listdir(path) if f.endswith('.epub')]
39
+ for i in tqdm(range(len(text_files))):
40
+ file_path = os.path.join(path, text_files[i])
41
+ new_file = str(text_files[i])[:-4] + 'txt'
42
+ for r in (('@', ''), ('#', ''), ('$', ''), ('&', ''), ('^', '')):
43
+ new_file = new_file.replace(*r)
44
+
45
+ try:
46
+ book = open_book(file_path)
47
+ lines = convert_epub_to_lines(book)
48
+ except:
49
+ lines = []
50
+ data = ''.join(lines)
51
+ new_string = re.sub(r'>\s+<', '><', data)
52
+ new_string = new_string.split('><')
53
+ n = len(new_string)
54
+ new_string[0] = new_string[0] + ">"
55
+ new_string[n - 1] = "<" + new_string[n - 1]
56
+ for j in range(1, n):
57
+ new_string[j] = "<" + new_string[j] + ">"
58
+ data = '\n'.join(new_string)
59
+ tag_re = re.compile(r'(<!--.*?-->|<[^>]*>)')
60
+ no_tags = tag_re.sub('', data)
61
+ ready_for_web = html.escape(no_tags)
62
+ clean_text = re.sub(r'\r\s+', '\n', ready_for_web)
63
+ clean_text = re.sub(r'([a-zA-Z])\n([a-zA-Z])', r'\1\2', clean_text)
64
+
65
+ with open(new_file, "w+", encoding = 'utf-8') as file:
66
+ file.write(clean_text)
.utils/join.py ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import glob, json, sys, gzip, random
2
+
3
+ b = int(sys.argv[1])
4
+ e = int(sys.argv[2])
5
+ parts = "aa ab ac ad ae af ag ah ai aj ak al am an ao ap aq ar as at au av aw ax ay az ba bb bc bd be bf bg bh bi bj bk bl bm bn bo bp bq br bs bt bu bv bw bx by bz ca cb cc cd ce cf cg ch ci cj ck cl cm cn co cp cq cr cs ct cu cv cw cx cy cz da db dc dd de df dg dh di dj dk dl dm dn do dp dq dr ds dt du dv".split()
6
+
7
+ for p in parts[b:b+e]:
8
+ lines = []
9
+ files = glob.glob(f"*/*.{p}")
10
+ for file in files:
11
+ with open(file, "rt") as fin:
12
+ for i, line in enumerate(fin):
13
+ try: text = json.loads(line)["text"]
14
+ except: print(i, line, end="", flush=True); continue
15
+ if text is None: print(i, line, end="", flush=True); continue
16
+ if len(text) < 5: print(i, line, end="", flush=True); continue
17
+ if len(text) < 500: continue
18
+ ss = json.dumps({"text": text}, ensure_ascii=False)
19
+ lines.append(ss)
20
+
21
+ random.shuffle(lines)
22
+ with gzip.open(f"nam_{p}.jsonl.gz", "wt") as fout:
23
+ for ss in lines:
24
+ fout.write(ss + "\n")
25
+
26
+ '''
27
+
28
+ nohup python3 join.py 5 5 &
29
+ nohup python3 join.py 10 5 &
30
+ nohup python3 join.py 15 5 &
31
+ nohup python3 join.py 20 5 &
32
+ nohup python3 join.py 25 5 &
33
+ nohup python3 join.py 30 5 &
34
+ nohup python3 join.py 35 5 &
35
+ nohup python3 join.py 40 5 &
36
+ nohup python3 join.py 45 5 &
37
+
38
+ nohup python3 join.py 50 5 &
39
+ nohup python3 join.py 55 5 &
40
+ nohup python3 join.py 60 5 &
41
+ nohup python3 join.py 65 5 &
42
+ nohup python3 join.py 70 5 &
43
+ nohup python3 join.py 75 5 &
44
+ nohup python3 join.py 80 5 &
45
+ nohup python3 join.py 85 5 &
46
+ nohup python3 join.py 90 5 &
47
+ nohup python3 join.py 95 5 &
48
+
49
+ '''
.utils/json2jsonl.py ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import json, sys
2
+
3
+ docs = json.loads(open(sys.argv[1]).read())
4
+ for doc in docs:
5
+ ss = json.dumps({"text": doc["content"]}, ensure_ascii=False)
6
+ print(ss)
7
+
8
+ '''
9
+
10
+ python3 json2jsonl.py part-2021012502.json > ../NAM002/wudao/part-2021012502.json
11
+ python3 json2jsonl.py part-2021012506.json > ../NAM002/wudao/part-2021012506.json
12
+ python3 json2jsonl.py part-2021012518.json > ../NAM002/wudao/part-2021012518.json
13
+ python3 json2jsonl.py part-2021012618.json > ../NAM002/wudao/part-2021012618.json
14
+ python3 json2jsonl.py part-2021012619.json > ../NAM002/wudao/part-2021012619.json
15
+ python3 json2jsonl.py part-2021012620.json > ../NAM002/wudao/part-2021012620.json
16
+ python3 json2jsonl.py part-2021012621.json > ../NAM002/wudao/part-2021012621.json
17
+ python3 json2jsonl.py part-2021012707.json > ../NAM002/wudao/part-2021012707.json
18
+ python3 json2jsonl.py part-2021012713.json > ../NAM002/wudao/part-2021012713.json
19
+ python3 json2jsonl.py part-202101281a.json > ../NAM002/wudao/part-202101281a.json
20
+ python3 json2jsonl.py part-202101281b.json > ../NAM002/wudao/part-202101281b.json
21
+ python3 json2jsonl.py part-202101281c.json > ../NAM002/wudao/part-202101281c.json
22
+ python3 json2jsonl.py part-2021020127.json > ../NAM002/wudao/part-2021020127.json
23
+ python3 json2jsonl.py part-2021020401.json > ../NAM002/wudao/part-2021020401.json
24
+ python3 json2jsonl.py part-2021020428.json > ../NAM002/wudao/part-2021020428.json
25
+ python3 json2jsonl.py part-2021020726.json > ../NAM002/wudao/part-2021020726.json
26
+ python3 json2jsonl.py part-2021020765.json > ../NAM002/wudao/part-2021020765.json
27
+ python3 json2jsonl.py part-2021020789.json > ../NAM002/wudao/part-2021020789.json
28
+ python3 json2jsonl.py part-2021020809.json > ../NAM002/wudao/part-2021020809.json
29
+ python3 json2jsonl.py part-2021021333.json > ../NAM002/wudao/part-2021021333.json
30
+ python3 json2jsonl.py part-2021021428.json > ../NAM002/wudao/part-2021021428.json
31
+ python3 json2jsonl.py part-2021021745.json > ../NAM002/wudao/part-2021021745.json
32
+ python3 json2jsonl.py part-2021021792.json > ../NAM002/wudao/part-2021021792.json
33
+ python3 json2jsonl.py part-2021021841.json > ../NAM002/wudao/part-2021021841.json
34
+ python3 json2jsonl.py part-2021021861.json > ../NAM002/wudao/part-2021021861.json
35
+ python3 json2jsonl.py part-2021021924.json > ../NAM002/wudao/part-2021021924.json
36
+ python3 json2jsonl.py part-2021022340.json > ../NAM002/wudao/part-2021022340.json
37
+ python3 json2jsonl.py part-2021022386.json > ../NAM002/wudao/part-2021022386.json
38
+ python3 json2jsonl.py part-2021022428.json > ../NAM002/wudao/part-2021022428.json
39
+ python3 json2jsonl.py part-2021022522.json > ../NAM002/wudao/part-2021022522.json
40
+ python3 json2jsonl.py part-2021022586.json > ../NAM002/wudao/part-2021022586.json
41
+ python3 json2jsonl.py part-2021022591.json > ../NAM002/wudao/part-2021022591.json
42
+ python3 json2jsonl.py part-2021022605.json > ../NAM002/wudao/part-2021022605.json
43
+ python3 json2jsonl.py part-2021022607.json > ../NAM002/wudao/part-2021022607.json
44
+ python3 json2jsonl.py part-2021022637.json > ../NAM002/wudao/part-2021022637.json
45
+ python3 json2jsonl.py part-2021022642.json > ../NAM002/wudao/part-2021022642.json
46
+ python3 json2jsonl.py part-2021022698.json > ../NAM002/wudao/part-2021022698.json
47
+ python3 json2jsonl.py part-2021022838.json > ../NAM002/wudao/part-2021022838.json
48
+ python3 json2jsonl.py part-2021023162.json > ../NAM002/wudao/part-2021023162.json
49
+ python3 json2jsonl.py part-2021023294.json > ../NAM002/wudao/part-2021023294.json
50
+ python3 json2jsonl.py part-2021023329.json > ../NAM002/wudao/part-2021023329.json
51
+ python3 json2jsonl.py part-2021023358.json > ../NAM002/wudao/part-2021023358.json
52
+ python3 json2jsonl.py part-2021023461.json > ../NAM002/wudao/part-2021023461.json
53
+ python3 json2jsonl.py part-2021023478.json > ../NAM002/wudao/part-2021023478.json
54
+ python3 json2jsonl.py part-2021023885.json > ../NAM002/wudao/part-2021023885.json
55
+ python3 json2jsonl.py part-2021023912.json > ../NAM002/wudao/part-2021023912.json
56
+ python3 json2jsonl.py part-2021024151.json > ../NAM002/wudao/part-2021024151.json
57
+ python3 json2jsonl.py part-2021024167.json > ../NAM002/wudao/part-2021024167.json
58
+ python3 json2jsonl.py part-2021024638.json > ../NAM002/wudao/part-2021024638.json
59
+
60
+ '''
.utils/lzma.py ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ import os, sys, glob, subprocess
2
+
3
+ for file in glob.glob("*.lzma"):
4
+ prefix = file.split(".lzma")[0]
5
+ if not os.path.exists(f"{prefix}"):
6
+ cmd = f"nohup lzma -d {file} &"
7
+ print(cmd, flush=True)
8
+ subprocess.run(cmd, shell=True)
.utils/mixsplit.py ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os, sys, glob, subprocess, math
2
+
3
+ curr_dir = os.getcwd()
4
+ x = curr_dir.split("/")[-1]
5
+ x = x.split("_")[0]
6
+ x = int(x)
7
+
8
+ if x > 100: times = math.ceil(x / 100)
9
+ else: times = 1
10
+
11
+ for jsonl_file in glob.glob("*.jsonl"):
12
+ prefix = jsonl_file.split(".jsonl")[0]
13
+ if not os.path.exists(f"{prefix}.aa"):
14
+ if times == 1:
15
+ cmd = f"shuf {jsonl_file} | split -l $(( $(wc -l < {jsonl_file})*{x}/10000+1 )) - {prefix}. &"
16
+ else:
17
+ cmd = f"yes {jsonl_file} | head -n {times} | xargs cat | shuf | \
18
+ split -l $(( $(wc -l < {jsonl_file})*{times}/100+1 )) - {prefix}. &"
19
+ print(cmd, flush=True)
20
+ subprocess.run(cmd, shell=True)
.utils/parquet2jsonl.py ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ files = """
2
+ data/java/data-00000-of-00208.parquet
3
+ data/python/data-00000-of-00144.parquet
4
+ data/hcl/data-00000-of-00002.parquet
5
+ data/ruby/data-00000-of-00019.parquet
6
+ data/c-sharp/data-00000-of-00107.parquet
7
+ data/clojure/data-00000-of-00002.parquet
8
+ data/coffeescript/data-00000-of-00002.parquet
9
+ data/crystal/data-00000-of-00001.parquet
10
+ data/dart/data-00000-of-00009.parquet
11
+ data/elixir/data-00000-of-00002.parquet
12
+ data/elm/data-00000-of-00001.parquet
13
+ data/f-sharp/data-00000-of-00002.parquet
14
+ data/erlang/data-00000-of-00002.parquet
15
+ data/go/data-00000-of-00057.parquet
16
+ data/haskell/data-00000-of-00006.parquet
17
+ data/julia/data-00000-of-00004.parquet
18
+ data/kotlin/data-00000-of-00015.parquet
19
+ data/sql/data-00000-of-00027.parquet
20
+ data/toml/data-00000-of-00002.parquet
21
+ data/typescript/data-00000-of-00071.parquet
22
+ data/zig/data-00000-of-00001.parquet
23
+ """.strip().split()
24
+
25
+ import json, sys, os
26
+ import pandas as pd
27
+
28
+ for input_file in files:
29
+ output_file = input_file.split("/data-000")[0].replace("data/", "../NAM002/code/") + ".jsonl"
30
+ with open(output_file, "wt") as f:
31
+ for i in range(0,8):
32
+ file = input_file.replace("00000", f"0000{i}")
33
+ if os.path.exists(file):
34
+ try: df = pd.read_parquet(file)
35
+ except: continue
36
+ print(file)
37
+ for index, row in df.iterrows():
38
+ ss = json.dumps({"text":row[25]}, ensure_ascii=False)
39
+ f.write(ss + "\n")
.utils/slimpajama.py ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ import subprocess, sys, os
2
+ chunk = int(sys.argv[1])
3
+ c = chunk + 1
4
+ for i in range(1, 59000, 40):
5
+ file = f"example_train_{i}.jsonl.zst"
6
+ if os.path.exists(file): continue
7
+ cmd = f"wget https://huggingface.co/datasets/cerebras/SlimPajama-627B\
8
+ /resolve/main/train/chunk{c}/{file}"
9
+ print(cmd, flush=True)
10
+ subprocess.run(cmd, shell=True)
.utils/slimpajama.sh ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ mkdir -p 0; cd 0; nohup python3 ../load.py 0 &
2
+ cd ..; mkdir -p 1; cd 1; nohup python3 ../load.py 1 &
3
+ cd ..; mkdir -p 2; cd 2; nohup python3 ../load.py 2 &
4
+ cd ..; mkdir -p 3; cd 3; nohup python3 ../load.py 3 &
5
+ cd ..; mkdir -p 4; cd 4; nohup python3 ../load.py 4 &
6
+ cd ..; mkdir -p 5; cd 5; nohup python3 ../load.py 5 &
7
+ cd ..; mkdir -p 6; cd 6; nohup python3 ../load.py 6 &
8
+ cd ..; mkdir -p 7; cd 7; nohup python3 ../load.py 7 &
9
+ cd ..; mkdir -p 8; cd 8; nohup python3 ../load.py 8 &
10
+ cd ..; mkdir -p 9; cd 9; nohup python3 ../load.py 9 &
11
+
12
+ cat 0/* > 0.jsonl.zst &
13
+ cat 1/* > 1.jsonl.zst &
14
+ cat 2/* > 2.jsonl.zst &
15
+ cat 3/* > 3.jsonl.zst &
16
+ cat 4/* > 4.jsonl.zst &
17
+ cat 5/* > 5.jsonl.zst &
18
+ cat 6/* > 6.jsonl.zst &
19
+ cat 7/* > 7.jsonl.zst &
20
+ cat 8/* > 8.jsonl.zst &
21
+ cat 9/* > 9.jsonl.zst &
.utils/verify.py ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ import gzip, json, sys
2
+
3
+ with gzip.open(sys.argv[1], "rt") as f:
4
+ for count, line in enumerate(f):
5
+ t = json.loads(line)["text"]
6
+ if t is None: print(line)
7
+ if len(t) == 0: print(line)
8
+ print(sys.argv[1], count)
.utils/verify.sh ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ nohup python3 verify.py nam_aa.jsonl.gz &
2
+ nohup python3 verify.py nam_ab.jsonl.gz &
3
+ nohup python3 verify.py nam_ac.jsonl.gz &
4
+ nohup python3 verify.py nam_ad.jsonl.gz &
5
+ nohup python3 verify.py nam_ae.jsonl.gz &
6
+ nohup python3 verify.py nam_af.jsonl.gz &
7
+ nohup python3 verify.py nam_ag.jsonl.gz &
8
+ nohup python3 verify.py nam_ah.jsonl.gz &
9
+ nohup python3 verify.py nam_ai.jsonl.gz &
10
+ nohup python3 verify.py nam_aj.jsonl.gz &
11
+ nohup python3 verify.py nam_ak.jsonl.gz &
12
+ nohup python3 verify.py nam_al.jsonl.gz &
13
+ nohup python3 verify.py nam_am.jsonl.gz &
14
+ nohup python3 verify.py nam_an.jsonl.gz &
15
+ nohup python3 verify.py nam_ao.jsonl.gz &
16
+ nohup python3 verify.py nam_ap.jsonl.gz &
17
+ nohup python3 verify.py nam_aq.jsonl.gz &
18
+ nohup python3 verify.py nam_ar.jsonl.gz &
19
+ nohup python3 verify.py nam_as.jsonl.gz &
20
+ nohup python3 verify.py nam_at.jsonl.gz &
21
+ nohup python3 verify.py nam_au.jsonl.gz &
22
+ nohup python3 verify.py nam_av.jsonl.gz &
23
+ nohup python3 verify.py nam_aw.jsonl.gz &
24
+ nohup python3 verify.py nam_ax.jsonl.gz &
25
+ nohup python3 verify.py nam_ay.jsonl.gz &
26
+ nohup python3 verify.py nam_az.jsonl.gz &
27
+ nohup python3 verify.py nam_ba.jsonl.gz &
28
+ nohup python3 verify.py nam_bb.jsonl.gz &
29
+ nohup python3 verify.py nam_bc.jsonl.gz &
30
+ nohup python3 verify.py nam_bd.jsonl.gz &
31
+ nohup python3 verify.py nam_be.jsonl.gz &
32
+ nohup python3 verify.py nam_bf.jsonl.gz &
33
+ nohup python3 verify.py nam_bg.jsonl.gz &
34
+ nohup python3 verify.py nam_bh.jsonl.gz &
35
+ nohup python3 verify.py nam_bi.jsonl.gz &
36
+ nohup python3 verify.py nam_bj.jsonl.gz &
37
+ nohup python3 verify.py nam_bk.jsonl.gz &
38
+ nohup python3 verify.py nam_bl.jsonl.gz &
39
+ nohup python3 verify.py nam_bm.jsonl.gz &
40
+ nohup python3 verify.py nam_bn.jsonl.gz &
41
+ nohup python3 verify.py nam_bo.jsonl.gz &
42
+ nohup python3 verify.py nam_bp.jsonl.gz &
43
+ nohup python3 verify.py nam_bq.jsonl.gz &
44
+ nohup python3 verify.py nam_br.jsonl.gz &
45
+ nohup python3 verify.py nam_bs.jsonl.gz &
46
+ nohup python3 verify.py nam_bt.jsonl.gz &
47
+ nohup python3 verify.py nam_bu.jsonl.gz &
48
+ nohup python3 verify.py nam_bv.jsonl.gz &
49
+ nohup python3 verify.py nam_bw.jsonl.gz &
50
+ nohup python3 verify.py nam_bx.jsonl.gz &
51
+ nohup python3 verify.py nam_by.jsonl.gz &
52
+ nohup python3 verify.py nam_bz.jsonl.gz &
53
+ nohup python3 verify.py nam_ca.jsonl.gz &
54
+ nohup python3 verify.py nam_cb.jsonl.gz &
55
+ nohup python3 verify.py nam_cc.jsonl.gz &
56
+ nohup python3 verify.py nam_cd.jsonl.gz &
57
+ nohup python3 verify.py nam_ce.jsonl.gz &
58
+ nohup python3 verify.py nam_cf.jsonl.gz &
59
+ nohup python3 verify.py nam_cg.jsonl.gz &
60
+ nohup python3 verify.py nam_ch.jsonl.gz &
61
+ nohup python3 verify.py nam_ci.jsonl.gz &
62
+ nohup python3 verify.py nam_cj.jsonl.gz &
63
+ nohup python3 verify.py nam_ck.jsonl.gz &
64
+ nohup python3 verify.py nam_cl.jsonl.gz &
65
+ nohup python3 verify.py nam_cm.jsonl.gz &
66
+ nohup python3 verify.py nam_cn.jsonl.gz &
67
+ nohup python3 verify.py nam_co.jsonl.gz &
68
+ nohup python3 verify.py nam_cp.jsonl.gz &
69
+ nohup python3 verify.py nam_cq.jsonl.gz &
70
+ nohup python3 verify.py nam_cr.jsonl.gz &
71
+ nohup python3 verify.py nam_cs.jsonl.gz &
72
+ nohup python3 verify.py nam_ct.jsonl.gz &
73
+ nohup python3 verify.py nam_cu.jsonl.gz &
74
+ nohup python3 verify.py nam_cv.jsonl.gz &
75
+ nohup python3 verify.py nam_cw.jsonl.gz &
76
+ nohup python3 verify.py nam_cx.jsonl.gz &
77
+ nohup python3 verify.py nam_cy.jsonl.gz &
78
+ nohup python3 verify.py nam_cz.jsonl.gz &
79
+ nohup python3 verify.py nam_da.jsonl.gz &
80
+ nohup python3 verify.py nam_db.jsonl.gz &
81
+ nohup python3 verify.py nam_dc.jsonl.gz &
82
+ nohup python3 verify.py nam_dd.jsonl.gz &
83
+ nohup python3 verify.py nam_de.jsonl.gz &
84
+ nohup python3 verify.py nam_df.jsonl.gz &
85
+ nohup python3 verify.py nam_dg.jsonl.gz &
86
+ nohup python3 verify.py nam_dh.jsonl.gz &
87
+ nohup python3 verify.py nam_di.jsonl.gz &
88
+ nohup python3 verify.py nam_dj.jsonl.gz &
89
+ nohup python3 verify.py nam_dk.jsonl.gz &
90
+ nohup python3 verify.py nam_dl.jsonl.gz &
91
+ nohup python3 verify.py nam_dm.jsonl.gz &
92
+ nohup python3 verify.py nam_dn.jsonl.gz &
93
+ nohup python3 verify.py nam_do.jsonl.gz &
94
+ nohup python3 verify.py nam_dp.jsonl.gz &
95
+ nohup python3 verify.py nam_dq.jsonl.gz &
96
+ nohup python3 verify.py nam_dr.jsonl.gz &
97
+ nohup python3 verify.py nam_ds.jsonl.gz &
98
+ nohup python3 verify.py nam_dt.jsonl.gz &
99
+ nohup python3 verify.py nam_du.jsonl.gz &
100
+ nohup python3 verify.py nam_dv.jsonl.gz &
.utils/wudao.md ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ```sh
2
+
3
+ lzma -z9 part-2021012502.jsonl &
4
+ lzma -z9 part-2021012506.jsonl &
5
+ lzma -z9 part-2021012518.jsonl &
6
+ lzma -z9 part-2021012618.jsonl &
7
+ lzma -z9 part-2021012619.jsonl &
8
+ lzma -z9 part-2021012620.jsonl &
9
+ lzma -z9 part-2021012621.jsonl &
10
+ lzma -z9 part-2021012707.jsonl &
11
+ lzma -z9 part-2021012713.jsonl &
12
+ lzma -z9 part-202101281a.jsonl &
13
+ lzma -z9 part-202101281b.jsonl &
14
+ lzma -z9 part-202101281c.jsonl &
15
+ lzma -z9 part-2021020127.jsonl &
16
+ lzma -z9 part-2021020401.jsonl &
17
+ lzma -z9 part-2021020428.jsonl &
18
+ lzma -z9 part-2021020726.jsonl &
19
+ lzma -z9 part-2021020765.jsonl &
20
+ lzma -z9 part-2021020789.jsonl &
21
+ lzma -z9 part-2021020809.jsonl &
22
+ lzma -z9 part-2021021333.jsonl &
23
+ lzma -z9 part-2021021428.jsonl &
24
+ lzma -z9 part-2021021745.jsonl &
25
+ lzma -z9 part-2021021792.jsonl &
26
+ lzma -z9 part-2021021841.jsonl &
27
+ lzma -z9 part-2021021861.jsonl &
28
+ lzma -z9 part-2021021924.jsonl &
29
+ lzma -z9 part-2021022340.jsonl &
30
+ lzma -z9 part-2021022386.jsonl &
31
+ lzma -z9 part-2021022428.jsonl &
32
+ lzma -z9 part-2021022522.jsonl &
33
+ lzma -z9 part-2021022586.jsonl &
34
+ lzma -z9 part-2021022591.jsonl &
35
+ lzma -z9 part-2021022605.jsonl &
36
+ lzma -z9 part-2021022607.jsonl &
37
+ lzma -z9 part-2021022637.jsonl &
38
+ lzma -z9 part-2021022642.jsonl &
39
+ lzma -z9 part-2021022698.jsonl &
40
+ lzma -z9 part-2021022838.jsonl &
41
+ lzma -z9 part-2021023162.jsonl &
42
+ lzma -z9 part-2021023294.jsonl &
43
+ lzma -z9 part-2021023329.jsonl &
44
+ lzma -z9 part-2021023358.jsonl &
45
+ lzma -z9 part-2021023461.jsonl &
46
+ lzma -z9 part-2021023478.jsonl &
47
+ lzma -z9 part-2021023885.jsonl &
48
+ lzma -z9 part-2021023912.jsonl &
49
+ lzma -z9 part-2021024151.jsonl &
50
+ lzma -z9 part-2021024167.jsonl &
51
+ lzma -z9 part-2021024638.jsonl &
52
+
53
+
54
+ ```
README.md ADDED
@@ -0,0 +1,183 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ```
2
+ 30G 060_slimpajama
3
+ 40G 100_book
4
+ 29G 100_code
5
+ - - - - - - - -
6
+ 99G En+code
7
+
8
+ 57G 080_c4-vi
9
+ 13G 100_news_2020-2021
10
+ 18G 100_news_2022-2023
11
+ 18G 300_laws
12
+ 55G 500_epub
13
+ 16G 999_gov
14
+ - - - - - - - -
15
+ 176G Vietnam
16
+
17
+ 12G 100_instructs
18
+ 4G 200_instructs
19
+ 65M 500_instructs
20
+ - - - - - - - - - - - -
21
+ 16G chat /instructs
22
+
23
+ - - - - - -
24
+ 291GB Total
25
+ ```
26
+
27
+ # TODOs
28
+ - [x] Sưu tập mọi top instructs data tiếng Anh để dạy model khả năng suy luận
29
+ - [x] Loại bỏ instructs có chatgpt, openai, as an ai language model ...
30
+ - [ ] Lọc `gov/datachinhphu.jsonl,datatieusu.jsonl,canbo_bo_nganh_vietnam.jsonl`
31
+ - [ ] Thêm secret data
32
+
33
+ (( Trộn đều data ))
34
+ - [x] Trộn ngẫu nhiên data trong cùng 1 thư mục
35
+ - [x] Phân chia data trong cùng 1 thư mục ra 100 phần bằng nhau
36
+ - [x] Đọc dữ liệu của 100 phần và chuẩn hóa về `{ "text": ... }` write chung về 1 file
37
+ - [x] `zcat nam_ds.jsonl.gz | shuf | gzip -c > shuf_ds.jsonl.gz`
38
+
39
+ (( thêm data muộn vào 100 file đã trộn, shuf 1 lần nữa ))
40
+ - `gzip -c late-data >> filename.gz`
41
+ - `zcat file.gz | shuf | gzip -c > new_file.gz`
42
+ - [ ] upload lên NAM003
43
+
44
+ (( optional ))
45
+ - [x] Review lại code sinh binidx để bỏ `{ "text": ... }`
46
+ - [x] Convert vài mẫu sang binidx để đếm số tokens
47
+
48
+ ```py
49
+ import subprocess
50
+ for i in range(0, 100, 4):
51
+ cmd = f"nohup python3 join.py {i} 4 &"
52
+ print(cmd)
53
+ subprocess.run(cmd, shell=True)
54
+ ```
55
+
56
+ ```sh
57
+ gzip -c file1 > foo.gz
58
+ gzip -c file2 >> foo.gz
59
+ gunzip -c foo
60
+ # is equivalent to
61
+ cat file1 file2
62
+
63
+ zcat file.txt.gz | shuf -n1
64
+
65
+ cat 1/* | gzip -c > 1.jsonl.gz &
66
+ cat 2/* | gzip -c > 2.jsonl.gz &
67
+ cat 3/* | gzip -c > 3.jsonl.gz &
68
+ cat 4/* | gzip -c > 4.jsonl.gz &
69
+ cat 5/* | gzip -c > 5.jsonl.gz &
70
+ cat 6/* | gzip -c > 6.jsonl.gz &
71
+ cat 7/* | gzip -c > 7.jsonl.gz &
72
+ cat 8/* | gzip -c > 8.jsonl.gz &
73
+ ```
74
+
75
+ - - - (( DONE )) - - -
76
+
77
+ # Chọn lọc 121G từ SlimPajama
78
+ ```
79
+ ls .utils/slimpajama.*
80
+ unzstd *.zst
81
+ ```
82
+
83
+ # Lọc lại dữ liệu forum (BỎ)
84
+
85
+ ```
86
+ mongorestore --db databasename --verbose \path\dump\<dumpfolder>
87
+ ```
88
+ Xem file `tinhte.py`
89
+
90
+
91
+ # Cân bằng c4_vi filtered
92
+
93
+ https://huggingface.co/tiendung/c4_vi_filtered/tree/main/classified
94
+
95
+ Lọc bớt để fit 100b token limitation
96
+
97
+ => Sau khi cân bằng từ 200GB còn 122GB (giảm 39%)
98
+
99
+
100
+ ```sh
101
+ 7z -y x "*.7z"
102
+
103
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/cong_nghe.7z &
104
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/doi_song.7z &
105
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/giai_tri.7z &
106
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/giao_duc.7z &
107
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/khoa_hoc.7z &
108
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/kinh_te.7z &
109
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/nha_dat.7z &
110
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/phap_luat.7z &
111
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/the_gioi.7z &
112
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/the_thao.7z &
113
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/unknown.7z &
114
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/van_hoa.7z &
115
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/xa_hoi.7z &
116
+ wget https://huggingface.co/tiendung/c4_vi_filtered/resolve/main/classified/xe_co.7z &
117
+
118
+ ```
119
+
120
+
121
+ # Cân bằng News
122
+
123
+ 31GB (balanced)
124
+
125
+ x2 giai_tri
126
+
127
+ ```
128
+ 18G news-2020-2021/ -5G (27%)
129
+ 26G news-2022-2023/ -8G (30%)
130
+ - - - - - - - - - - - -
131
+ 44G total => 31G
132
+ ```
133
+
134
+ ```sh
135
+ 2022-2023
136
+
137
+ BY-LINE-COUNT
138
+ 6364652 total
139
+ 1296014 xa_hoi.jsonl
140
+ 952884 kinh_te.jsonl
141
+ 740499 phap_luat.jsonl
142
+ 602891 doi_song.jsonl
143
+ 514224 the_thao.jsonl
144
+ 479060 giai_tri.jsonl
145
+ 380726 the_gioi.jsonl
146
+ 360872 van_hoa.jsonl
147
+ 343918 nha_dat.jsonl
148
+ 268879 cong_nghe.jsonl
149
+ 246832 giao_duc.jsonl
150
+ 91774 xe_co.jsonl
151
+ 85400 khoa_hoc.jsonl
152
+ 679 unknown.jsonl
153
+
154
+
155
+ shuf xa_hoi.jsonl | split -l $(( $(wc -l <xa_hoi.jsonl) * 66 / 100 ))
156
+ mv xaa .save/xa_hoi.jsonl; mv xab xa_hoi.jsonl
157
+
158
+ shuf kinh_te.jsonl | split -l $(( $(wc -l <kinh_te.jsonl) * 60 / 100 ))
159
+ mv xaa .save/kinh_te.jsonl; mv xab kinh_te.jsonl
160
+
161
+ shuf doi_song.jsonl | split -l $(( $(wc -l <doi_song.jsonl) * 66 / 100 ))
162
+ mv xab .save/doi_song.jsonl; mv xaa doi_song.jsonl
163
+
164
+ shuf nha_dat.jsonl | split -l $(( $(wc -l <nha_dat.jsonl) * 50 / 100 ))
165
+ mv xaa .save/nha_dat.jsonl; mv xab nha_dat.jsonl
166
+
167
+ BY-FILE-SIZE
168
+ 6023720846 xa_hoi.jsonl
169
+ 5012502465 kinh_te.jsonl
170
+ 2739936418 doi_song.jsonl
171
+ 2460018829 phap_luat.jsonl
172
+ 1752318410 nha_dat.jsonl
173
+ 1734247739 van_hoa.jsonl
174
+ 1627705588 the_thao.jsonl
175
+ 1680150745 giai_tri.jsonl
176
+ 1199792764 cong_nghe.jsonl
177
+ 1187043681 giao_duc.jsonl
178
+ 1344303086 the_gioi.jsonl
179
+ 341947997 khoa_hoc.jsonl
180
+ 340266911 xe_co.jsonl
181
+ 3287153 unknown.jsonl
182
+
183
+ ```
select/.keep ADDED
File without changes