InoteveraGaidhi rinotevera
X-Vector Mutauri Embeddings
AI Yenyaya
Chidzidzo cheAI yeRuzha
Residual Vector Quantization
Residual vector quantization (RVQ) ndiyo dhizaini inoshandura kuenderera mberi kwekumisikidzwa kwemaodhiyo kuita compact stack ye discrete macode nekudzokorora kuverengera kukanganisa kwasara.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
It matters because it is the engine behind modern neural codecs like SoundStream and EnCodec and the tokenizer for generative audio.
Kudzika Kwakadzika
Plain vector quantization (VQ) inotsiva inoenderera vhekita neinopinda iri padyo mubhuku rekodhi rakadzidzwa, asi bhuku rekodhi rimwe chete rakakwana zvemhando yepamusoro rinoda nhamba huru yezvenyeredzi. RVQ inogadzirisa izvi nekukanda akati wandei madiki macodebook. Bhuku rekutanga rekodhi rinoburitsa fungidziro yakaoma; unoibvisa kuti utore chikanganiso chasara, wedzera izvo zvakasara nerechipiri codebook, bvisa zvakare, uye enderera kune N nhanho. Kodhi yekupedzisira ndiyo rondedzero yemandikisi akasarudzwa pamatanho ese, uye kuvaka patsva ndiyo huwandu hweese akasarudzwa makodhi mavheti. Izvi zvinogadzira hombe rinoshanda rekodhi bhuku mune akawanda madiki, zvinoshamisa kucheka ndangariro uye komputa uku uchirega bitrate kuyera nekushandisa akawanda kana mashoma nhanho. Quantizer kudonhedza panguva yekudzidziswa kunoita kuti ekutanga macodebook atakure ruzivo rwakawanda, achigonesa kunaka kuderera.
Kuitwa muNyika yeChokwadi
Discretizing encoder embeddings mukati meSoundStream, EnCodec, uye DAC neural codecs
Kugadzira akaturikidzana odhiyo tokeni ayo AudioLM neMusicLM anoburitsa pamusoro
Kuyera bitrate yecodec kumusoro kana pasi nekuita mamwe kana mashoma quantizer matanho
Kudzvanya embeddings yakakwira-dimensional mukutora uye kuchengetedza masisitimu uchishandisa akaturikidzana macodebook
Ramangwana reResidual Vector Quantization
RVQ yave yakajairwa discretization layer inobatanidza inoenderera neural inomiririra kune token-based generative modhi, uye kunatsiridzwa kunoenderera mberi: nani codebook mashandisirwo kudzivirira 'akafa' manyorero, efactorized uye akaderera-dimensional macodebook, uye semantically ane chinangwa chechiratidzo hierarchies. Kupfuura odhiyo, iyo imwechete yakasara-stacking zano iri kupararira kune mufananidzo uye vhidhiyo tokenizer, inomisikidza RVQ sebhiriji rakajairika pakati peanoenderera encoder uye mutauro-modhi-maitiro ekuteedzera majenareta.
Kuzvishandisa
Nei zvichikosha
- Svika uye svika. Inonatsiridza kusvikika kuburikidza nekunyora, kurondedzera, uye mazwi ekubatanidza.
- Mutengo uye bhajeti. Zvikwata zveMedia zvinogona kutumira odhiyo yakakwenenzverwa nekukurumidza nemabhajeti madiki.
- Kumhanya uye chiyero. Masisitimu anotarisana nevatengi anogona kugadzirisa kutaurirana kwekutaura pamwero mukuru.
Chenjera
- Kushandisa izwi zvisizvo uye njodzi dzekuedzesera dzinowedzera kana chibvumirano chisipo.
- Kururama kunogona kudonha mumitauro, mataurirwo, kana nharaunda dzine ruzha.
- Synthetic audio inogona kukanganisa kutaura kwechokwadi isina mavara akajeka.
Matanho anotevera
- Wana mvumo yakajeka yekutora inzwi, kugadzira, uye kushandisa zvakare.
- Yedza mhando pavatauri vakasiyana uye mamiriro ekumashure.
- Tsanangura apo munhu anofanira kuongorora kana kubvumidza zvabuda.
- Label synthetic odhiyo uye chengetedza marekodhi ekuzvidavirira.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Take the Residual Vector Quantization quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
What is Residual Vector Quantization?
Residual vector quantization (RVQ) ndiyo dhizaini inoshandura kuenderera mberi kwekumisikidzwa kwemaodhiyo kuita compact stack ye discrete macode nekudzokorora kuverengera kukanganisa kwasara. Izvo zvine basa nekuti ndiyo injini kuseri kwemazuva ano neural codecs seSoundStream uye EnCodec uye tokenizer yekugadzira odhiyo.
Ko rimwe nerimwe rinotevedzana codebook muRVQ rinoverengera chii?
Mushure mekunge bhuku rekutanga rekodhi rapa fungidziro yakakasharara, RVQ inoibvisa uye kuverengera zvakasara nekodekodhi inotevera, ichidzokorora mumatanho.
Sei uchishandisa RVQ pachinzvimbo chebhuku rekodhi rimwe chete rakakura?
Bhuku rekodhi rimwechete rakakwana zvakakwana zvemhando yepamusoro raizove rakakura zvisingaite; kurongedza M macodebook eK ekupinda kunopa K ^ M misanganiswa ine kudiki kwazvo kuchengetedza.
Kuvakazve kwekupedzisira kunoumbwa sei kubva kuRVQ makodhi?
Kuvaka patsva ndiko kuwanda kweakasarudzwa codebook vectors pamatanho ese, imwe neimwe ichigadzirisa zvakasara zvepakutanga.
NeM nhanho dzeK mapindiro ega ega, mangani anoshanda macode musanganiswa anomiririrwa neRVQ?
Kusarudza imwe yezvinyorwa zveK pane imwe neimwe yeM yakazvimiririra inoburitsa K ^ M inogoneka musanganiswa, uchichengeta chete M*K mavheji.
Ndechipi chinangwa chakajairika che'kuzvipira kurasikirwa' paunenge uchidzidzisa RVQ macodebook?
Kurasikirwa kwekuzvipira kunoranga iyo encoder kana zvabuda zvichidonha kubva kune yakasarudzwa codebook vectors, kuchengetedza quantization yakagadzikana; macodebook pachawo anowanzo gadziridza kuburikidza neexponential inofamba pakati.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi