Commit
·
66b8bd6
1
Parent(s):
3a3a0a4
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -1,47 +1,19 @@
|
|
1 |
-
|
2 |
-
dataset_info:
|
3 |
-
features:
|
4 |
-
- name: text
|
5 |
-
dtype: string
|
6 |
-
- name: token_short
|
7 |
-
dtype: string
|
8 |
-
- name: token_long
|
9 |
-
dtype: string
|
10 |
-
- name: p_short
|
11 |
-
dtype: float32
|
12 |
-
- name: p_long
|
13 |
-
dtype: float32
|
14 |
-
- name: JS
|
15 |
-
dtype: float32
|
16 |
-
- name: long_ids
|
17 |
-
sequence: int32
|
18 |
-
- name: short_max_id
|
19 |
-
dtype: int64
|
20 |
-
- name: long_max_id
|
21 |
-
dtype: int64
|
22 |
-
- name: context
|
23 |
-
dtype: string
|
24 |
-
- name: context_ids
|
25 |
-
sequence: int32
|
26 |
-
- name: p_delta_max
|
27 |
-
dtype: float32
|
28 |
-
- name: logit_excite_max
|
29 |
-
dtype: float32
|
30 |
-
- name: logit_inhibit_max
|
31 |
-
dtype: float32
|
32 |
-
- name: batch
|
33 |
-
dtype: int64
|
34 |
-
- name: sample
|
35 |
-
dtype: int64
|
36 |
-
- name: start
|
37 |
-
dtype: int64
|
38 |
-
splits:
|
39 |
-
- name: scan
|
40 |
-
num_bytes: 466393218
|
41 |
-
num_examples: 1874497
|
42 |
-
download_size: 337479388
|
43 |
-
dataset_size: 466393218
|
44 |
-
---
|
45 |
-
# Dataset Card for "pile_scan_4"
|
46 |
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# scan_4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
+
- `text`: The long prompt text with the the first token surrounded by square brackets.
|
4 |
+
- `token_short`: The model's prediction for the most likely token given the short prompt.
|
5 |
+
- `token_long`: The model's prediction for the most likely token given the long prompt.
|
6 |
+
- `p_short`: The model's prediction for the probability of `token_short`.
|
7 |
+
- `p_long`: The model's prediction for the probability of `token_long`.
|
8 |
+
- `JS`: The Jensen-Shannon divergence between the model's distribution over tokens given the short prompt and the model's distribution over tokens given the long prompt.
|
9 |
+
- `long_ids`: The ids of the tokens in the long prompt.
|
10 |
+
- `short_max_id`: The id of `token_short`.
|
11 |
+
- `long_max_id`: The id of `token_long`.
|
12 |
+
- `context`: The surrounding text of the prompt.
|
13 |
+
- `context_ids`: The ids of `context`.
|
14 |
+
- `p_delta_max`: The largest difference in probability for any token between the short and long prompt.
|
15 |
+
- `logit_excite_max`: The largest increase in logit for any token between the short and long prompt.
|
16 |
+
- `logit_inhibit_max`: The largest decrease in logit for any token between the short and long prompt.
|
17 |
+
- `batch`: The batch number of the prompt.
|
18 |
+
- `sample`: The sample number of the prompt.
|
19 |
+
- `start`: The start index of the prompt in the sample.
|