Spaces:
Running on Zero
Running on Zero
Return to the previous compatible pair: transformers 5 collides with the vendored xcodec config
Browse files- README.md +1 -1
- app.py +46 -47
- requirements.txt +2 -2
README.md
CHANGED
|
@@ -4,7 +4,7 @@ emoji: 🎤
|
|
| 4 |
colorFrom: pink
|
| 5 |
colorTo: purple
|
| 6 |
sdk: gradio
|
| 7 |
-
sdk_version:
|
| 8 |
python_version: '3.11'
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
|
|
|
| 4 |
colorFrom: pink
|
| 5 |
colorTo: purple
|
| 6 |
sdk: gradio
|
| 7 |
+
sdk_version: 5.28.0
|
| 8 |
python_version: '3.11'
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
app.py
CHANGED
|
@@ -28,7 +28,6 @@ import soundfile as sf
|
|
| 28 |
import torch
|
| 29 |
from huggingface_hub import hf_hub_download
|
| 30 |
from pyharp import ModelCard, build_endpoint
|
| 31 |
-
from pyharp.tags import Subcategory
|
| 32 |
|
| 33 |
from songgen import (
|
| 34 |
SongGenDualTrackForConditionalGeneration,
|
|
@@ -90,7 +89,7 @@ model_card = ModelCard(
|
|
| 90 |
"voice, and it generates a full song."
|
| 91 |
),
|
| 92 |
author="Zihan Liu, Shuangrui Ding, Zhixiong Zhang, Xiaoyi Dong, Pan Zhang, Yuhang Zang, Yuhang Cao, Dahua Lin, Jiaqi Wang",
|
| 93 |
-
tags=[
|
| 94 |
)
|
| 95 |
|
| 96 |
|
|
@@ -156,49 +155,49 @@ def process_fn(mode_name, description, lyrics, ref_voice_path, extract_vocals, t
|
|
| 156 |
return song_path, vocals_path
|
| 157 |
|
| 158 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 159 |
if __name__ == "__main__":
|
| 160 |
-
|
| 161 |
-
input_components = [
|
| 162 |
-
gr.Dropdown(
|
| 163 |
-
choices=list(MODES.keys()),
|
| 164 |
-
value="Mixed",
|
| 165 |
-
label="Mode",
|
| 166 |
-
info="Mixed outputs one finished track; Dual-track also gives you an isolated vocal stem",
|
| 167 |
-
),
|
| 168 |
-
gr.Textbox(
|
| 169 |
-
label="Music Description",
|
| 170 |
-
info="Genre, mood, instruments (e.g. \"upbeat pop song with piano and drums\")",
|
| 171 |
-
),
|
| 172 |
-
gr.Textbox(
|
| 173 |
-
label="Lyrics",
|
| 174 |
-
info="English lyrics for the song (required)",
|
| 175 |
-
),
|
| 176 |
-
gr.Audio(type="filepath", label="Reference Voice (optional)").harp_required(False),
|
| 177 |
-
gr.Checkbox(
|
| 178 |
-
value=True,
|
| 179 |
-
label="Extract vocals from reference",
|
| 180 |
-
info="(default: True, per repo example) turn off only if the reference audio is already an isolated vocal, not a full mix",
|
| 181 |
-
),
|
| 182 |
-
gr.Slider(
|
| 183 |
-
minimum=0.1, maximum=2.0, step=0.1, value=1.0,
|
| 184 |
-
label="Creativity",
|
| 185 |
-
info="(default: 1.0, per repo config) sampling temperature — higher is more varied/unpredictable, lower is safer/more repetitive",
|
| 186 |
-
),
|
| 187 |
-
]
|
| 188 |
-
output_components = [
|
| 189 |
-
gr.Audio(type="filepath", label="Song").set_info(
|
| 190 |
-
"The generated song. In Dual-track mode this is vocals + accompaniment combined."
|
| 191 |
-
),
|
| 192 |
-
gr.Audio(type="filepath", label="Isolated Vocals").set_info(
|
| 193 |
-
"Vocal-only stem. Only populated in Dual-track mode; silent in Mixed mode."
|
| 194 |
-
),
|
| 195 |
-
]
|
| 196 |
-
|
| 197 |
-
build_endpoint(
|
| 198 |
-
model_card=model_card,
|
| 199 |
-
input_components=input_components,
|
| 200 |
-
output_components=output_components,
|
| 201 |
-
process_fn=process_fn,
|
| 202 |
-
)
|
| 203 |
-
|
| 204 |
-
demo.queue().launch(show_error=True, share=True, pwa=True)
|
|
|
|
| 28 |
import torch
|
| 29 |
from huggingface_hub import hf_hub_download
|
| 30 |
from pyharp import ModelCard, build_endpoint
|
|
|
|
| 31 |
|
| 32 |
from songgen import (
|
| 33 |
SongGenDualTrackForConditionalGeneration,
|
|
|
|
| 89 |
"voice, and it generates a full song."
|
| 90 |
),
|
| 91 |
author="Zihan Liu, Shuangrui Ding, Zhixiong Zhang, Xiaoyi Dong, Pan Zhang, Yuhang Zang, Yuhang Cao, Dahua Lin, Jiaqi Wang",
|
| 92 |
+
tags=["text-to-music", "song-generation"],
|
| 93 |
)
|
| 94 |
|
| 95 |
|
|
|
|
| 155 |
return song_path, vocals_path
|
| 156 |
|
| 157 |
|
| 158 |
+
with gr.Blocks() as demo:
|
| 159 |
+
input_components = [
|
| 160 |
+
gr.Dropdown(
|
| 161 |
+
choices=list(MODES.keys()),
|
| 162 |
+
value="Mixed",
|
| 163 |
+
label="Mode",
|
| 164 |
+
info="Mixed outputs one finished track; Dual-track also gives you an isolated vocal stem",
|
| 165 |
+
),
|
| 166 |
+
gr.Textbox(
|
| 167 |
+
label="Music Description",
|
| 168 |
+
info="Genre, mood, instruments (e.g. \"upbeat pop song with piano and drums\")",
|
| 169 |
+
),
|
| 170 |
+
gr.Textbox(
|
| 171 |
+
label="Lyrics",
|
| 172 |
+
info="English lyrics for the song (required)",
|
| 173 |
+
),
|
| 174 |
+
gr.Audio(type="filepath", label="Reference Voice (optional)").harp_required(False),
|
| 175 |
+
gr.Checkbox(
|
| 176 |
+
value=True,
|
| 177 |
+
label="Extract vocals from reference",
|
| 178 |
+
info="(default: True, per repo example) turn off only if the reference audio is already an isolated vocal, not a full mix",
|
| 179 |
+
),
|
| 180 |
+
gr.Slider(
|
| 181 |
+
minimum=0.1, maximum=2.0, step=0.1, value=1.0,
|
| 182 |
+
label="Creativity",
|
| 183 |
+
info="(default: 1.0, per repo config) sampling temperature — higher is more varied/unpredictable, lower is safer/more repetitive",
|
| 184 |
+
),
|
| 185 |
+
]
|
| 186 |
+
output_components = [
|
| 187 |
+
gr.Audio(type="filepath", label="Song").set_info(
|
| 188 |
+
"The generated song. In Dual-track mode this is vocals + accompaniment combined."
|
| 189 |
+
),
|
| 190 |
+
gr.Audio(type="filepath", label="Isolated Vocals").set_info(
|
| 191 |
+
"Vocal-only stem. Only populated in Dual-track mode; silent in Mixed mode."
|
| 192 |
+
),
|
| 193 |
+
]
|
| 194 |
+
|
| 195 |
+
build_endpoint(
|
| 196 |
+
model_card=model_card,
|
| 197 |
+
input_components=input_components,
|
| 198 |
+
output_components=output_components,
|
| 199 |
+
process_fn=process_fn,
|
| 200 |
+
)
|
| 201 |
+
|
| 202 |
if __name__ == "__main__":
|
| 203 |
+
demo.queue().launch(pwa=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
requirements.txt
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
-
git+https://github.com/TEAMuP-dev/pyharp.git@v0.
|
| 2 |
# model-specific deps below:
|
| 3 |
torch==2.8.0
|
| 4 |
torchaudio==2.8.0
|
| 5 |
-
transformers==
|
| 6 |
sentencepiece==0.2.2
|
| 7 |
descript-audio-codec==1.0.0
|
| 8 |
descript-audiotools==0.7.2
|
|
|
|
| 1 |
+
git+https://github.com/TEAMuP-dev/pyharp.git@v0.3.0
|
| 2 |
# model-specific deps below:
|
| 3 |
torch==2.8.0
|
| 4 |
torchaudio==2.8.0
|
| 5 |
+
transformers==4.43.3
|
| 6 |
sentencepiece==0.2.2
|
| 7 |
descript-audio-codec==1.0.0
|
| 8 |
descript-audiotools==0.7.2
|