File size: 2,695 Bytes
bd2d17d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 |
# Copyright (2023) Tsinghua University, Bytedance Ltd. and/or its affiliates
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import torch
import argparse
from model import SALMONN
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--device", type=str, default="cuda")
parser.add_argument("--ckpt_path", type=str, default='./salomnn_7b.bin')
parser.add_argument("--whisper_path", type=str, default='whisper-large-v2')
parser.add_argument("--beats_path", type=str, default='BEATs_iter3_plus_AS2M_finetuned_on_AS2M_cpt2.pt')
parser.add_argument("--vicuna_path", type=str, default='vicuna-7b-v1.5')
parser.add_argument("--low_resource", action='store_true', default=False)
parser.add_argument("--debug", action="store_true", default=False)
args = parser.parse_args()
model = SALMONN(
ckpt=args.ckpt_path,
whisper_path=args.whisper_path,
beats_path=args.beats_path,
vicuna_path=args.vicuna_path
).to(torch.float16).cuda()
prompt = 'First describe the music in general in terms of mood, theme, tempo, melody, instruments and chord progression. Then provide a detailed music analysis by describing each functional segment and its time boundaries.'
prompt_tmp = 'This is a Pop music of 69 beat-per-minute (BPM). First describe the music in general in terms of mood, theme, tempo, melody, instruments and chord progression. Then provide a detailed music analysis by describing each functional segment and its time boundaries. Please note that the music boundaries are [0, 41, 58, 83, 100].'
model.eval()
while True:
print("=====================================")
wav_path = input("Your Wav Path:\n")
prompt = input("Your Prompt:\n")
try:
print("Output:")
# for environment with cuda>=117
with torch.cuda.amp.autocast(dtype=torch.float16):
print(model.generate(wav_path, prompt=prompt, repetition_penalty=1.5, num_beams=10, top_p=.7, temperature=.2)[0])
except Exception as e:
print(e)
if args.debug:
import pdb
pdb.set_trace()
|