Spaces:
Sleeping
Sleeping
Update app.py
#5
by
ChengsongHuang
- opened
app.py
CHANGED
@@ -12,6 +12,12 @@ import torch
|
|
12 |
import shutil
|
13 |
import os
|
14 |
import uuid
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
|
17 |
css = """
|
@@ -21,7 +27,6 @@ css = """
|
|
21 |
"""
|
22 |
st.markdown(css, unsafe_allow_html=True)
|
23 |
|
24 |
-
|
25 |
def main():
|
26 |
st.title("π‘ LoraHub")
|
27 |
st.markdown("Low-rank adaptations (LoRA) are techniques for fine-tuning large language models on new tasks. We propose LoraHub, a framework that allows composing multiple LoRA modules trained on different tasks. The goal is to achieve good performance on unseen tasks using just a few examples, without needing extra parameters or training. And we want to build a marketplace where users can share their trained LoRA modules, thereby facilitating the application of these modules to new tasks.")
|
@@ -105,12 +110,28 @@ Infer the date from context. Q: Today is the second day of the third month of 1
|
|
105 |
txt_input, txt_output, max_inference_step=max_step)
|
106 |
|
107 |
st.success("Lorahub learning finished! You got the following recommendation:")
|
|
|
108 |
df = {
|
109 |
"modules": [LORA_HUB_NAMES[i] for i in st.session_state["select_names"]],
|
110 |
"weights": recommendation.value,
|
111 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
st.table(df)
|
113 |
-
|
114 |
random_id = uuid.uuid4().hex
|
115 |
os.makedirs(f"lora/{random_id}")
|
116 |
# copy config file
|
@@ -126,7 +147,16 @@ Infer the date from context. Q: Today is the second day of the third month of 1
|
|
126 |
file_name=f"lora_{random_id}.zip",
|
127 |
mime="application/zip"
|
128 |
)
|
129 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
130 |
|
131 |
|
132 |
|
|
|
12 |
import shutil
|
13 |
import os
|
14 |
import uuid
|
15 |
+
import json
|
16 |
+
|
17 |
+
|
18 |
+
from google.oauth2 import service_account
|
19 |
+
import gspread
|
20 |
+
from google.oauth2.service_account import Credentials
|
21 |
|
22 |
|
23 |
css = """
|
|
|
27 |
"""
|
28 |
st.markdown(css, unsafe_allow_html=True)
|
29 |
|
|
|
30 |
def main():
|
31 |
st.title("π‘ LoraHub")
|
32 |
st.markdown("Low-rank adaptations (LoRA) are techniques for fine-tuning large language models on new tasks. We propose LoraHub, a framework that allows composing multiple LoRA modules trained on different tasks. The goal is to achieve good performance on unseen tasks using just a few examples, without needing extra parameters or training. And we want to build a marketplace where users can share their trained LoRA modules, thereby facilitating the application of these modules to new tasks.")
|
|
|
110 |
txt_input, txt_output, max_inference_step=max_step)
|
111 |
|
112 |
st.success("Lorahub learning finished! You got the following recommendation:")
|
113 |
+
|
114 |
df = {
|
115 |
"modules": [LORA_HUB_NAMES[i] for i in st.session_state["select_names"]],
|
116 |
"weights": recommendation.value,
|
117 |
}
|
118 |
+
|
119 |
+
|
120 |
+
|
121 |
+
def share():
|
122 |
+
credentials = service_account.Credentials.from_service_account_info(
|
123 |
+
json.loads(st.secrets["gcp_service_account"]),
|
124 |
+
scopes=[
|
125 |
+
"https://www.googleapis.com/auth/spreadsheets",
|
126 |
+
]
|
127 |
+
)
|
128 |
+
gsheet_url = st.secrets["private_gsheets_url"]
|
129 |
+
gc = gspread.authorize(credentials)
|
130 |
+
sh = gc.open_by_url(gsheet_url)
|
131 |
+
|
132 |
+
ws = sh.sheet1
|
133 |
+
ws.insert_rows([[LORA_HUB_NAMES[i] for i in st.session_state["select_names"]],recommendation.value.tolist(),[]])
|
134 |
st.table(df)
|
|
|
135 |
random_id = uuid.uuid4().hex
|
136 |
os.makedirs(f"lora/{random_id}")
|
137 |
# copy config file
|
|
|
147 |
file_name=f"lora_{random_id}.zip",
|
148 |
mime="application/zip"
|
149 |
)
|
150 |
+
with open(f"lora_{random_id}.zip", "rb") as fp:
|
151 |
+
btn = st.download_button(
|
152 |
+
label="π₯ Download the final LoRA Module and share your results",
|
153 |
+
data=fp,
|
154 |
+
file_name=f"lora_{random_id}.zip",
|
155 |
+
mime="application/zip",
|
156 |
+
on_click=share
|
157 |
+
)
|
158 |
+
st.button("π₯ Share your results",on_click=share)
|
159 |
+
st.warning("The page will be refreshed once you click the download button. Share results may cost 1-2 mins.")
|
160 |
|
161 |
|
162 |
|