Upload 2 files
Browse files- .gitattributes +1 -0
- derive_guancao.py +55 -0
- guanaco.jsonl +3 -0
.gitattributes
CHANGED
@@ -52,3 +52,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
52 |
*.jpg filter=lfs diff=lfs merge=lfs -text
|
53 |
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
54 |
*.webp filter=lfs diff=lfs merge=lfs -text
|
|
|
|
52 |
*.jpg filter=lfs diff=lfs merge=lfs -text
|
53 |
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
54 |
*.webp filter=lfs diff=lfs merge=lfs -text
|
55 |
+
guanaco.jsonl filter=lfs diff=lfs merge=lfs -text
|
derive_guancao.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import logging
|
3 |
+
|
4 |
+
import pandas as pd
|
5 |
+
from datasets import load_dataset, Dataset
|
6 |
+
|
7 |
+
train_dataset = load_dataset("OpenAssistant/oasst1")["train"]
|
8 |
+
|
9 |
+
|
10 |
+
def get_children(df, parent_ids):
|
11 |
+
children = df[df['parent_id'].isin(parent_ids)]
|
12 |
+
return children.sort_values('rank', ascending=True).drop_duplicates('parent_id')
|
13 |
+
|
14 |
+
|
15 |
+
def trace_conversations(df, parent_ids):
|
16 |
+
conversations = []
|
17 |
+
children = get_children(df, parent_ids)
|
18 |
+
|
19 |
+
while not children.empty:
|
20 |
+
conversations.extend(children.to_dict('records'))
|
21 |
+
parent_ids = children['message_id']
|
22 |
+
children = get_children(df, parent_ids)
|
23 |
+
|
24 |
+
return conversations
|
25 |
+
|
26 |
+
# Convert the HuggingFace's dataset to pandas dataframe
|
27 |
+
df = pd.DataFrame.from_records(train_dataset)
|
28 |
+
|
29 |
+
# Get the root nodes
|
30 |
+
root_nodes = df[df['parent_id'].isnull()]
|
31 |
+
|
32 |
+
conversations = []
|
33 |
+
for idx, root in root_nodes.iterrows():
|
34 |
+
conversation_chain = [root.to_dict()]
|
35 |
+
conversation_chain.extend(trace_conversations(df, [root['message_id']]))
|
36 |
+
conversations.append(conversation_chain)
|
37 |
+
|
38 |
+
# Select only necessary columns for each conversation
|
39 |
+
for conversation in conversations:
|
40 |
+
for message in conversation:
|
41 |
+
keys_to_delete = set(message.keys()) - {'message_id', 'parent_id', 'role', 'text'}
|
42 |
+
for key in keys_to_delete:
|
43 |
+
del message[key]
|
44 |
+
|
45 |
+
# Create a new dataframe with only the 'conversations' field
|
46 |
+
result_df = pd.DataFrame({'conversations': conversations})
|
47 |
+
|
48 |
+
# Convert dataframe back to HuggingFace's dataset
|
49 |
+
result_dataset = Dataset.from_pandas(result_df)
|
50 |
+
|
51 |
+
logging.info(result_dataset)
|
52 |
+
|
53 |
+
with open("guanaco.jsonl", "w") as f:
|
54 |
+
for row in result_dataset:
|
55 |
+
f.write(json.dumps(row))
|
guanaco.jsonl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:109f3b5c63c30a340bec3a30bda94b22f8c05c1df1073623be8a6467270f8672
|
3 |
+
size 24376150
|