Exploring compressibility of transformer based text-to-music (TTM) models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917703753662464 |
|---|---|
| author | Moschopoulos, Vasileios Kotsiopoulos, Thanasis Parada, Pablo Peso Nikiforidis, Konstantinos Stergiadis, Alexandros Papakostas, Gerasimos Jalal, Md Asif Zhang, Jisi Drosou, Anastasios Saravanan, Karthikeyan |
| author_facet | Moschopoulos, Vasileios Kotsiopoulos, Thanasis Parada, Pablo Peso Nikiforidis, Konstantinos Stergiadis, Alexandros Papakostas, Gerasimos Jalal, Md Asif Zhang, Jisi Drosou, Anastasios Saravanan, Karthikeyan |
| contents | State-of-the art Text-To-Music (TTM) generative AI models are large and require desktop or server class compute, making them infeasible for deployment on mobile phones. This paper presents an analysis of trade-offs between model compression and generation performance of TTM models. We study compression through knowledge distillation and specific modifications that enable applicability over the various components of the TTM model (encoder, generative model and the decoder). Leveraging these methods we create TinyTTM (89.2M params) that achieves a FAD of 3.66 and KL of 1.32 on MusicBench dataset, better than MusicGen-Small (557.6M params) but not lower than MusicGen-small fine-tuned on MusicBench. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_17159 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Exploring compressibility of transformer based text-to-music (TTM) models Moschopoulos, Vasileios Kotsiopoulos, Thanasis Parada, Pablo Peso Nikiforidis, Konstantinos Stergiadis, Alexandros Papakostas, Gerasimos Jalal, Md Asif Zhang, Jisi Drosou, Anastasios Saravanan, Karthikeyan Audio and Speech Processing Multimedia Sound State-of-the art Text-To-Music (TTM) generative AI models are large and require desktop or server class compute, making them infeasible for deployment on mobile phones. This paper presents an analysis of trade-offs between model compression and generation performance of TTM models. We study compression through knowledge distillation and specific modifications that enable applicability over the various components of the TTM model (encoder, generative model and the decoder). Leveraging these methods we create TinyTTM (89.2M params) that achieves a FAD of 3.66 and KL of 1.32 on MusicBench dataset, better than MusicGen-Small (557.6M params) but not lower than MusicGen-small fine-tuned on MusicBench. |
| title | Exploring compressibility of transformer based text-to-music (TTM) models |
| topic | Audio and Speech Processing Multimedia Sound |
| url | https://arxiv.org/abs/2406.17159 |