tbenthompson
commited on
Commit
•
db06d31
1
Parent(s):
aa5d78d
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -1,49 +1,16 @@
|
|
1 |
-
---
|
2 |
-
dataset_info:
|
3 |
-
features:
|
4 |
-
- name: id0
|
5 |
-
dtype: uint16
|
6 |
-
- name: id1
|
7 |
-
dtype: uint16
|
8 |
-
- name: id2
|
9 |
-
dtype: uint16
|
10 |
-
- name: sum_count
|
11 |
-
dtype: uint64
|
12 |
-
- name: max_count
|
13 |
-
dtype: uint32
|
14 |
-
- name: frac_max
|
15 |
-
dtype: float64
|
16 |
-
- name: token0
|
17 |
-
dtype: string
|
18 |
-
- name: token1
|
19 |
-
dtype: string
|
20 |
-
- name: token2
|
21 |
-
dtype: string
|
22 |
-
- name: seq
|
23 |
-
dtype: string
|
24 |
-
- name: p_70m
|
25 |
-
dtype: float64
|
26 |
-
- name: p_160m
|
27 |
-
dtype: float64
|
28 |
-
- name: p_410m
|
29 |
-
dtype: float64
|
30 |
-
- name: p_1b
|
31 |
-
dtype: float64
|
32 |
-
- name: p_1.4b
|
33 |
-
dtype: float64
|
34 |
-
- name: p_2.8b
|
35 |
-
dtype: float64
|
36 |
-
- name: p_6.9b
|
37 |
-
dtype: float64
|
38 |
-
- name: p_12b
|
39 |
-
dtype: float64
|
40 |
-
splits:
|
41 |
-
- name: top_trigrams
|
42 |
-
num_bytes: 203881376
|
43 |
-
num_examples: 1542074
|
44 |
-
download_size: 159272559
|
45 |
-
dataset_size: 203881376
|
46 |
-
---
|
47 |
-
# Dataset Card for "pile_ngrams_top_trigrams"
|
48 |
|
49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
|
2 |
+
# top_trigrams
|
3 |
+
|
4 |
+
See https://confirmlabs.org/posts/catalog.html for details.
|
5 |
+
|
6 |
+
- `id0`: the first token in the trigram
|
7 |
+
- `id1`: the second token in the trigram
|
8 |
+
- `id2`: the most common token following `(id0, id1)` in The Pile
|
9 |
+
- `sum_count`: the number of times that `(id0, id1)` appears in The Pile.
|
10 |
+
- `max_count`: the number of times that `id2` appears after `(id0, id1)` in The Pile.
|
11 |
+
- `frac_max`: `max_count / sum_count`
|
12 |
+
- `token0`: the string representation of `id0`
|
13 |
+
- `token1`: the string representation of `id1`
|
14 |
+
- `token2`: the string representation of `id2`
|
15 |
+
- `seq`: the string representation of the trigram, `token0 token1 token2`
|
16 |
+
- `p_{model_size}`: the probability of the trigram under Pythia-{model_size} when prompted with `(id0, id1)`.
|