Datasets:

Modalities:
Text
Languages:
English
Libraries:
Datasets
License:
Zhangir Azerbayev commited on
Commit
72455b2
1 Parent(s): af62b4d

added stats

Browse files
Files changed (2) hide show
  1. count_tokens.py +49 -0
  2. stats.json +26 -0
count_tokens.py ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from datasets import load_dataset
2
+ from itertools import islice
3
+ import sys
4
+ import time
5
+ from tqdm import tqdm
6
+ from transformers import AutoTokenizer
7
+ from itertools import islice
8
+ import json
9
+
10
+ NUM_PROC = 12
11
+
12
+ dataset = load_dataset("hoskinson-center/proof-pile")
13
+ tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neox-20B")
14
+
15
+ def length(example):
16
+ return {"length": [len(x) for x in tokenizer(example["text"])["input_ids"]]}
17
+
18
+ dataset = dataset.map(length, batched=True, num_proc=NUM_PROC)
19
+
20
+ stats = dict()
21
+
22
+ for x in tqdm(dataset["train"]):
23
+ meta = json.loads(x["meta"])
24
+
25
+ if "config" in meta.keys():
26
+ config = meta["config"]
27
+ elif "set_name" in meta.keys():
28
+ config = meta["set_name"]
29
+ elif "subset_name" in meta.keys():
30
+ path = meta["file"]
31
+ config = path[:path.index("/")]
32
+ else:
33
+ print(x)
34
+ raise KeyError()
35
+
36
+ if config not in stats.keys():
37
+ stats[config] = dict()
38
+ stats[config]["bytes"] = 0
39
+ stats[config]["tokens"] = 0
40
+
41
+ stats[config]["bytes"] += len(x["text"].encode("utf-8"))
42
+ stats[config]["tokens"] += x["length"]
43
+
44
+
45
+ print(json.dumps(stats, indent=2))
46
+ print("saving stats...")
47
+
48
+ with open("stats.json", "w") as f:
49
+ f.write(json.dumps(stats, indent=2))
stats.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "arxiv": {
3
+ "bytes": 13609690878,
4
+ "tokens": 4905230148
5
+ },
6
+ "stack_exchange": {
7
+ "bytes": 962813093,
8
+ "tokens": 306853916
9
+ },
10
+ "formal": {
11
+ "bytes": 141328715,
12
+ "tokens": 59043874
13
+ },
14
+ "wiki": {
15
+ "bytes": 20573531,
16
+ "tokens": 6561329
17
+ },
18
+ "MATH": {
19
+ "bytes": 2429955,
20
+ "tokens": 898410
21
+ },
22
+ "books": {
23
+ "bytes": 15776008,
24
+ "tokens": 5553483
25
+ }
26
+ }