froggeric commited on
Commit
959115a
1 Parent(s): e5187a8

added datasets descriptions and links

Browse files
Files changed (1) hide show
  1. README.md +21 -6
README.md CHANGED
@@ -20,13 +20,25 @@ You will find below descriptions for the various input files provided, to help y
20
 
21
  ## Community provided files
22
 
23
- **8k_random_data**\
24
-
25
- **20k_random_data**\
26
-
27
  **groups_merged**\
 
 
 
 
28
 
29
  **group_10_merged**\
 
 
 
 
 
 
 
 
 
 
 
 
30
 
31
  **ptb.train**\
32
  Penn Treebank (PTB) is a widely used preprocessed large dataset designed for language training. Casing,
@@ -48,8 +60,11 @@ https://huggingface.co/datasets/wikitext
48
  or "good articles".\
49
  https://huggingface.co/datasets/asi/wikitext_fr
50
 
51
- **c4** (exllamav2)\
52
- Constructed from news articles?
 
 
 
53
 
54
  **code** (exllamav2)\
55
  Programming
 
20
 
21
  ## Community provided files
22
 
 
 
 
 
23
  **groups_merged**\
24
+ _"Here is a decent general purpose imatrix calibration dataset. It should be more diverse than wikitext at ~30k tokens, as it is excerpts of a larger dataset which includes coding examples (which seems quite important!)
25
+ This means it's generally higher entropy data compared to wikitext, and it's real data rather than pseudo-randomly generated data.
26
+ I get lower KL div than wikitext for the same length and the outputs seem qualitatively better."_ (kalomaze)\
27
+ https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8395384
28
 
29
  **group_10_merged**\
30
+ (superseeded by groups_merged)\
31
+ _"This is about ~50k pseudo-random tokens.
32
+ I am getting the best balance between the maximum divergence and the other divergence statistics using this file when quantizing 7b"_ (kalomaze)\
33
+ https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8349233
34
+
35
+ **20k_random_data**\
36
+ (superseeded by groups_10_merged)\
37
+ https://github.com/ggerganov/llama.cpp/discussions/5006#discussioncomment-8163190
38
+
39
+ **8k_random_data**\
40
+ (superseeded by 20k_random_data)\
41
+ https://github.com/ggerganov/llama.cpp/discussions/5006#discussion-6087829
42
 
43
  **ptb.train**\
44
  Penn Treebank (PTB) is a widely used preprocessed large dataset designed for language training. Casing,
 
60
  or "good articles".\
61
  https://huggingface.co/datasets/asi/wikitext_fr
62
 
63
+ **c4**\
64
+ The C4 dataset is a collection text sourced from the public Common Crawl web scrape.
65
+ It includes heuristics to extract only natural language (as opposed to boilerplate and other gibberish)
66
+ in addition to extensive deduplication. C4 dataset was explicitly designed to be English only:
67
+ any page that was not given a probability of at least 99% of being English by langdetect was discarded.
68
 
69
  **code** (exllamav2)\
70
  Programming