Datasets:

Modalities:
Audio
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
Dask
hungchiayu commited on
Commit
e9c166e
·
verified ·
1 Parent(s): 2b5cf26

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +52 -2
README.md CHANGED
@@ -22,8 +22,58 @@ dataset_info:
22
  dtype: int32
23
  splits:
24
  - name: train
25
- num_bytes: 180239660645.0
26
  num_examples: 100000
27
  download_size: 172620977911
28
- dataset_size: 180239660645.0
 
 
 
 
 
29
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
  dtype: int32
23
  splits:
24
  - name: train
25
+ num_bytes: 180239660645
26
  num_examples: 100000
27
  download_size: 172620977911
28
+ dataset_size: 180239660645
29
+ task_categories:
30
+ - text-to-audio
31
+ tags:
32
+ - DPO
33
+ - text-to-audio
34
  ---
35
+
36
+
37
+ ### Dataset Description
38
+
39
+ <!-- Provide a longer summary of what this dataset is. -->
40
+
41
+ This dataset consists of 100k audio preference pairs generated by TangoFlux during the CRPO stage. Specifically, TangoFlux performed five iterations of CRPO. In each iteration, 20k prompts were sampled from a prompt bank. For each prompt, audio samples with the highest and lowest CLAP scores were selected to form the "chosen" and "rejected" pairs, respectively. This process resulted in a total of 100k preference pairs.
42
+
43
+
44
+ Since every iteration contains 20k prompts sampled from audiocaps prompts, some prompts are the same across iterations.
45
+
46
+ ### Dataset Sources
47
+
48
+ <!-- Provide the basic links for the dataset. -->
49
+
50
+ - **Repository:** https://github.com/declare-lab/TangoFlux
51
+ - **Paper :** https://arxiv.org/abs/2412.21037
52
+ - **Demo :** https://huggingface.co/spaces/declare-lab/TangoFlux
53
+
54
+ ## Uses
55
+
56
+ <!-- Address questions around how the dataset is intended to be used. -->
57
+
58
+ You can directly download the dataset and use them for preference optimization in text-to-audio.
59
+
60
+
61
+
62
+
63
+ ## Citation
64
+
65
+ If you find our dataset useful, please cite us! Thanks!
66
+
67
+ **BibTeX:**
68
+
69
+ ```
70
+ @misc{hung2024tangofluxsuperfastfaithful,
71
+ title={TangoFlux: Super Fast and Faithful Text to Audio Generation with Flow Matching and Clap-Ranked Preference Optimization},
72
+ author={Chia-Yu Hung and Navonil Majumder and Zhifeng Kong and Ambuj Mehrish and Rafael Valle and Bryan Catanzaro and Soujanya Poria},
73
+ year={2024},
74
+ eprint={2412.21037},
75
+ archivePrefix={arXiv},
76
+ primaryClass={cs.SD},
77
+ url={https://arxiv.org/abs/2412.21037},
78
+ }
79
+ ```