Overview
Latent Consistency Models (LCMs) inzira inoita kuti majenareta emifananidzo awedzere kuburitsa mifananidzo yemhando yepamusoro munhanho imwe chete kusvika ina pachinzvimbo chezvakawanda. Ivo vanoita pedyo-chaiyo-nguva, inopindirana mufananidzo chizvarwa chinoshanda kunyangwe pane zvine mwero Hardware.
Latent Consistency Models ndeyekombuta-yekuona workflows inodudzira kana kuburitsa inooneka midhiya yekuongorora, mashandiro, uye kugadzira.
Deep Dive
Standard latent diffusion modhi seStable Diffusion inotanga kubva muruzha uye kuita ruzha nguva nenguva, kazhinji ichida makumi maviri kusvika makumi mashanu kuongororwa netiweki kugadzira mufananidzo mumwe, unononoka. MaLCM, akaunzwa naLuo uye vaanoshanda navo muna 2023, vanoisa kusawirirana kwe distillation munzvimbo yakadzikama yemhando yakadzidziswa yekuparadzira modhi. Pfungwa inokosha: dzidzisa mudzidzi network kuti asvetuke akananga kumhedzisiro yakachena kubva kune chero nzvimbo padivi pe denoising trajectory, saka mhinduro imwechete inosvikwa mune imwe nhanho hombe yakambotora akawanda madiki. Mhedzisiro yacho mifananidzo yakapinza mune ingangoita 1 kusvika 4 nhanho. Imwe nzira yekufambidzana, iyo LCM-LoRA, inorongedza kukwidziridzwa uku sediki plug-in adapta inogona kudonhedzwa pane iripo yakanyatso-tuned Stable Diffusion modhi pasina kudzidzisazve network yese.
Technical Insight
Consistency modhi inomanikidza 'self-consistency' pfuma: chero mapoinzi maviri ari munzira imwechete yekurevera (iyo mukana-kuyerera ODE trajectory) inofanira mepu kune imwechete yekupedzisira yakachena mufananidzo. Mudzidzi anonyungudutswa kubva kumudzidzisi wekuparadzira modhi kuti agutse izvi, achidzidza kufanotaura magumo enzira zvakananga. Kushanda munzvimbo yakamanikidzwa yakadzikama pane mapixels kunoita kuti distillation idhure. Nekuti ongororo imwe inogona kusvetuka munzira, inorema iterative sampling inodonha kuita mashoma ematanho.
Mastering Latent Consistency Models
Latent Consistency Models (LCMs) inzira inoita kuti majenareta emifananidzo awedzere kuburitsa mifananidzo yemhando yepamusoro munhanho imwe chete kusvika ina pachinzvimbo chezvakawanda. Ivo vanoita pedyo-chaiyo-nguva, inopindirana mufananidzo chizvarwa chinoshanda kunyangwe pane zvine mwero Hardware. Latent Consistency Models ndeyekombuta-yekuona workflows inodudzira kana kuburitsa inooneka midhiya yekuongorora, mashandiro, uye kugadzira. Kuti uvake kunzwisisa kwakadzama, bata Latent Consistency Models semuenzaniso wekushandisa, kwete chinhu chimwe chete: tsanangura zvinodikanwa, jekesa fungidziro, uye patsanura zvinogona kuitwa nehurongwa hwakavimbika kubva kune zvichiri kuda kutonga kwenyanzvi.
Mukuita, zvikwata zvakasimba zvinoshandisa Latent Consistency Models kuenzanirana nezviri kuitika semhando yedata, kusiyana kwemwenje, uye kuenderana kwemazita. Ivo vanonyora zvakajeka maitiro ebudiriro, bvunzo vachipokana ne data rechokwadi uye mafambiro ebasa, uye iterate zvichibva pane zvakacherechedzwa maitiro ekutadza kwete kuhwina-nguva imwe chete yebhenji. Apa ndipo apo kunzwisisa kwe theoretical kunoshanduka kuve kugona kwakasimba pane chigadzirwa, mutemo, uye mashandiro.
Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero. Panguva imwecheteyo, kodzero dzeMufananidzo uye kubvumirwa kunogona kuve njodzi dzepamutemo kana hunhu husina kujeka. Nzira yakatsiga ndeyekubatanidza kukurumidza kuyedza nekutonga: mhanyisa vatyairi vendege, tora humbowo, buritsa matanda esarudzo, uye urambe uchivandudza chengetedzo semaitiro emuenzaniso, zvinotarisirwa nemushandisi, uye zvinodikanwa zvekutonga.
Strategic Impact
Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero.
Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero. Mukutumirwa kwemhando yepamusoro, izvi zvinoshandurirwa kuita mitemo inoyerwa yekushanda, miganhu yevaridzi, uye tsika dzekudzokorora dzinodzokororwa kuitira kuti zvikwata zvikwire kuvimba pane kukwidza kusajeka.
Zvikwata zvekugadzira zvinogona prototype pfungwa nekukurumidza nekudzokororwa kwemaoko mashoma.
Zvikwata zvekugadzira zvinogona prototype pfungwa nekukurumidza nekudzokororwa kwemaoko mashoma. Mukutumirwa kwemhando yepamusoro, izvi zvinoshandurirwa kuita mitemo inoyerwa yekushanda, miganhu yevaridzi, uye tsika dzekudzokorora dzinodzokororwa kuitira kuti zvikwata zvikwire kuvimba pane kukwidza kusajeka.
Mashandisirwo anogona kushandisa masaini emifananidzo nemavhidhiyo ayo aimbove akaoma kugadzirisa.
Mashandisirwo anogona kushandisa masaini emifananidzo nemavhidhiyo ayo aimbove akaoma kugadzirisa. Mukutumirwa kwemhando yepamusoro, izvi zvinoshandurirwa kuita mitemo inoyerwa yekushanda, miganhu yevaridzi, uye tsika dzekudzokorora dzinodzokororwa kuitira kuti zvikwata zvikwire kuvimba pane kukwidza kusajeka.
Real-World Implementation
Real-time canvas maturusi anovandudza mufananidzo wakagadzirwa paunenge uchinyora kana sketch, ine padyo-zero lag.
Kumhanya Yakagadzika Diffusion mufananidzo chizvarwa pane laptop kana foni GPU muchidimbu chesekondi
Kudonhedza iyo LCM-LoRA adapta pane iripo yakakwenenzverwa modhi kuti ikurumidze nekukurumidza pasina kudzidziswazve.
Kugadzira mabheji mahombe emifananidzo zvakachipa yekuongorora dhizaini nekucheka matanho kubva ~ 30 pasi kusvika ~ 4
Maitiro Ekuita
Latent Consistency Models mukuita
Real-time canvas maturusi anovandudza mufananidzo wakagadzirwa paunenge uchinyora kana kudhirowa, nepedyo-zero lag.
Real-time canvas maturusi anovandudza mufananidzo wakagadzirwa paunenge uchinyora kana kudhirowa, nepedyo-zero lag Matimu anowanzo kuwana zvirinani kana achinge atsanangura emhando yepamusoro kumberi, chengetedza nzira yekukwira kwevanhu yemakesi emupendero, uye kuteedzera zvese zvakawanikwa zvechigadzirwa uye mutengo wekukanganisa nekufamba kwenguva.
Latent Consistency Models mukuita
Kumhanya Yakagadzika Diffusion mufananidzo chizvarwa pane laptop kana foni GPU muchidimbu chesekondi.
Kumhanya Yakagadzika Diffusion mufananidzo chizvarwa palaptop kana foni GPU muchikamu chechipiri Matimu anowanzo kuwana mhedzisiro iri nani kana achinge atsanangura emhando yepamusoro kumberi, chengetedza nzira yekukwira kwevanhu yemakesi emupendero, uye kuteedzera zvese zvakawanikwa zvechigadzirwa uye mutengo wekukanganisa nekufamba kwenguva.
Latent Consistency Models mukuita
Kudonhedza iyo LCM-LoRA adapta pane iripo yakanyatso-tuned modhi kuti ikurumidze ipapo pasina kudzidziswazve.
Kudonhedza adapta yeLCM-LoRA pane iripo modhi yakanyatso gadziridzwa kuti ikurumidze kukurumidza kudzidzisa Matimu anowanzo kuwana mhedzisiro iri nani kana achinge atsanangura emhando yepamusoro kumberi, chengetedza nzira yekukwira kwevanhu yemakesi emupendero, uye kuteedzera zvese zvakawanikwa zvechigadzirwa uye mutengo wekukanganisa nekufamba kwenguva.
Latent Consistency Models mukuita
Kugadzira mabheji mahombe emifananidzo zvakachipa yekuongorora dhizaini nekucheka matanho kubva ~ 30 pasi kusvika ~ 4.
Kugadzira mabheji mahombe emifananidzo zvakachipa yekuongorora dhizaini nekucheka nhanho kubva ~ 30 zvichidzika kusvika ~ 4 Matimu anowanzo kuwana mhedzisiro iri nani kana achinge atsanangura emhando yepamusoro kumberi, chengetedza nzira yekukwira kwevanhu yemakesi ekumucheto, uye kuteedzera zvese zvakawanikwa zvechigadzirwa nemitengo yekukanganisa nekufamba kwenguva.
Njodzi & Guardrails
Kodzero dzemifananidzo uye kubvumirwa kunogona kuve njodzi dzepamutemo kana provenance isina kujeka.
Kuita kwemuenzaniso kunogona kusiyanisa kupenya, huwandu hwevanhu, uye nharaunda.
Manyepo enhema anogona kusacherechedzwa kunze kwekunge zvikumbaridzo zvekuvimba zvikatariswa.
Implementation Roadmap
Tsanangura maitiro ekugamuchirwa echokwadi, kurangarira, uye mutengo wekukanganisa.
Tsanangura maitiro ekugamuchirwa echokwadi, kurangarira, uye mutengo wekukanganisa. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.
Edzai nedata rinoenderana nemamiriro chaiwo ekugadzira.
Edzai nedata rinoenderana nemamiriro chaiwo ekugadzira. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.
Wedzera ongororo yemunhu kune yakaderera-kusavimbika kana yakakwirira-inokanganisa kufanotaura.
Wedzera ongororo yemunhu kune yakaderera-kusavimbika kana yakakwirira-inokanganisa kufanotaura. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.
Tevera modhi kudonha uye simbisa mushure mekuchinja kwekamera kana dataset.
Tevera modhi kudonha uye simbisa mushure mekuchinja kwekamera kana dataset. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.