ostris commited on
Commit
f01cba5
·
verified ·
1 Parent(s): 3f51181

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. README.md +174 -67
  2. pipeline.py +440 -0
README.md CHANGED
@@ -53,7 +53,114 @@ For the model, follow the directions for flux for the vae and text encoder model
53
 
54
  ### Diffusers
55
 
56
- _Coming Soon_
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
57
 
58
  ## Limitations
59
 
@@ -109,83 +216,83 @@ If you enjoy my projects or use them commercially, please consider sponsoring me
109
 
110
  All of these people / organizations are the ones who selflessly make this project possible. Thank you!!
111
 
112
- _Last updated: 2025-04-22 21:16 UTC_
113
 
114
  <p align="center">
115
- <a href="https://github.com/replicate" target="_blank" rel="noopener noreferrer"><img src="https://avatars.githubusercontent.com/u/60410876?v=4" alt="Replicate" width="200" height="200" style="border-radius:8px;margin:5px;display: inline-block;"></a>
116
- <a href="https://github.com/josephrocca" target="_blank" rel="noopener noreferrer"><img src="https://avatars.githubusercontent.com/u/1167575?u=92d92921b4cb5c8c7e225663fed53c4b41897736&v=4" alt="josephrocca" width="200" height="200" style="border-radius:8px;margin:5px;display: inline-block;"></a>
117
  </p>
118
  <hr style="width:100%;border:none;height:2px;background:#ddd;margin:30px 0;">
119
  <p align="center">
120
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/162524101/81a72689c3754ac5b9e38612ce5ce914/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=3XLSlLFCWAQ-0wd2_vZMikyotdQNSzKOjoyeoJiZEw0%3D" alt="Prasanth Veerina" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;">
121
- <a href="https://github.com/weights-ai" target="_blank" rel="noopener noreferrer"><img src="https://avatars.githubusercontent.com/u/185568492?v=4" alt="Weights" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;"></a>
122
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/161471720/dd330b4036d44a5985ed5985c12a5def/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=qkRvrEc5gLPxaXxLvcvbYv1W1lcmOoTwhj4A9Cq5BxQ%3D" alt="Vladimir Sotnikov" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;">
123
- <img src="https://c8.patreon.com/3/200/33158543" alt="clement Delangue" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;">
124
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/54890369/45cea21d82974c78bf43956de7fb0e12/eyJ3IjoyMDB9/2.jpeg?token-time=2145916800&token-hash=IK6OT6UpusHgdaC4y8IhK5XxXiP5TuLy3vjvgL77Fho%3D" alt="Eli Slugworth" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;">
125
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/8654302/b0f5ebedc62a47c4b56222693e1254e9/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=lpeicIh1_S-3Ji3W27gyiRB7iXurp8Bx8HAzDHftOuo%3D" alt="Misch Strotz" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;">
126
- <img src="https://c8.patreon.com/3/200/93304" alt="Joseph Rocca" width="150" height="150" style="border-radius:8px;margin:5px;display: inline-block;">
127
  </p>
128
  <hr style="width:100%;border:none;height:2px;background:#ddd;margin:30px 0;">
129
  <p align="center">
130
- <a href="https://x.com/NuxZoe" target="_blank" rel="noopener noreferrer"><img src="https://pbs.twimg.com/profile_images/1714760743273574400/tdvQjNTl_400x400.jpg" alt="tungsten" width="100" height="100" style="border-radius:8px;margin:5px;display: inline-block;"></a>
131
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/2298192/1228b69bd7d7481baf3103315183250d/eyJ3IjoyMDB9/1.jpg?token-time=2145916800&token-hash=1B7dbXy_gAcPT9WXBesLhs7z_9APiz2k1Wx4Vml_-8Q%3D" alt="Mohamed Oumoumad" width="100" height="100" style="border-radius:8px;margin:5px;display: inline-block;">
132
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/120239481/49b1ce70d3d24704b8ec34de24ec8f55/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=Dv1NPKwdv9QT8fhYYwbGnQIvfiyqTUlh52bjDW1vYxY%3D" alt="nitish PNR" width="100" height="100" style="border-radius:8px;margin:5px;display: inline-block;">
133
- <img src="https://c8.patreon.com/3/200/548524" alt="Steve Hanff" width="100" height="100" style="border-radius:8px;margin:5px;display: inline-block;">
134
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/152118848/3b15a43d71714552b5ed1c9f84e66adf/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=IEKE18CBHVZ3k-08UD7Dkb7HbiFHb84W0FATdLMI0Dg%3D" alt="Kristjan Retter" width="100" height="100" style="border-radius:8px;margin:5px;display: inline-block;">
135
- <img src="https://c8.patreon.com/3/200/83319230" alt="Miguel Lara" width="100" height="100" style="border-radius:8px;margin:5px;display: inline-block;">
136
  </p>
137
  <hr style="width:100%;border:none;height:2px;background:#ddd;margin:30px 0;">
138
  <p align="center">
139
- <img src="https://c8.patreon.com/3/200/8449560" alt="Patron" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
140
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/27288932/6c35d2d961ee4e14a7a368c990791315/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=dpFFssZXZM_KZMKQhl3uDwwusdFw1c_v9x_ChJU7_zc%3D" alt="David Garrido" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
141
- <img src="https://c8.patreon.com/3/200/2410522" alt="George Gostyshev" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
142
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/16287560/78130de30950410ca528d8a888997081/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=Ok-HSL2MthKXF09SmCOlPFCPfbMctFBZKCuTnPwxZ3A%3D" alt="Vitaly Golubenko" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
143
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/570742/4ceb33453a5a4745b430a216aba9280f/eyJ3IjoyMDB9/1.jpg?token-time=2145916800&token-hash=wUzsI5cO5Evp2ukIGdSgBbvKeYgv5LSOQMa6Br33Rrs%3D" alt="Al H" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
144
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/131773947/eda3405aa582437db4582fce908c8739/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=S4Bh0sMqTNmJlo3uRr7co5d_kxvBjITemDTfi_1KrCA%3D" alt="Jodh Singh" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
145
- <img src="https://c8.patreon.com/3/200/22809690" alt="Michael Levine" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
146
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/99036356/7ae9c4d80e604e739b68cca12ee2ed01/eyJ3IjoyMDB9/3.png?token-time=2145916800&token-hash=zK0dHe6A937WtNlrGdefoXFTPPzHUCfn__23HP8-Ui0%3D" alt="Noctre" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
147
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/141098579/1a9f0a1249d447a7a0df718a57343912/eyJ3IjoyMDB9/2.png?token-time=2145916800&token-hash=Rd_AjZGhMATVkZDf8E95ILc0n93gvvFWe1Ig0_dxwf4%3D" alt="The Local Lab" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
148
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/98811435/3a3632d1795b4c2b9f8f0270f2f6a650/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=93w8RMxwXlcM4X74t03u6P5_SrKvlm1IpjnD2SzVpJk%3D" alt="EmmanuelMr18" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
149
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/338551/e8f257d8d3dd46c38272b391a5785948/eyJ3IjoyMDB9/1.jpg?token-time=2145916800&token-hash=GLom1rGgOZjBeO7I1OnjiIgWmjl6PO9ZjBB8YTvc7AM%3D" alt="Plaidam" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
150
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/82763/f99cc484361d4b9d94fe4f0814ada303/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=BpwC020pR3TRZ4r0RSCiSIOh-jmatkrpy1h2XU4sGa4%3D" alt="Doron Adler" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
151
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/103077711/bb215761cc004e80bd9cec7d4bcd636d/eyJ3IjoyMDB9/2.jpeg?token-time=2145916800&token-hash=zvtBie29rRTKTXvAA2KhOI-l3mSMk9xxr-mg_CksLtc%3D" alt="John Dopamine" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
152
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/93348210/5c650f32a0bc481d80900d2674528777/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=PpXK9B_iy288annlNdLOexhiQHbTftPEDeCh-sTQ2KA%3D" alt="Armin Behjati" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
153
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/155963250/6f8fd7075c3b4247bfeb054ba49172d6/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=twmKs4mADF_h7bKh5jBuigYVScMeaeHv2pEPin9K0Dg%3D" alt="Un Defined" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
154
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/45562978/0de33cf52ec642ae8a2f612cddec4ca6/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=hSAvaD4phiLcF0pvX7FP0juI5NQWCon-_TZSNpJzQJg%3D" alt="Jack English" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
155
- <img src="https://c8.patreon.com/3/200/27791680" alt="Jean-Tristan Marin" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
156
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/60995694/92e0e8f336eb4a5bb8d99b940247d1d1/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=pj6Tm8XRdpGJcAEdnCakqYSNiSjoAYjvZescX7d0ic0%3D" alt="Abraham Irawan" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
157
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/164958178/4eb7a37baa0541bab7a091f2b14615b7/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=_aaum7fBJAGaJhMBhlR8vqYavDhExdVxmO9mwd3_XMw%3D" alt="Austin Robinson" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
158
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/134129880/680c7e14cd1a4d1a9face921fb010f88/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=vNKojv67krNqx7gdpKBX1R_stX2TkMRYvRc0xZrbY6s%3D" alt="Bharat Prabhakar" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
159
- <img src="https://c8.patreon.com/3/200/70218846" alt="Cosmosis" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
160
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/83054970/13de6cb103ad41a5841edf549e66cd51/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=wU_Eke9VYcfI40FAQvdEV84Xspqlo5VSiafLqhg_FOE%3D" alt="Gili Ben Shahar" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
161
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/30931983/54ab4e4ceab946e79a6418d205f9ed51/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=LBmsSsMQZhO6yRZ_YyRwTgE6a7BVWrGNsAVveLXHXR0%3D" alt="HestoySeghuro ." width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
162
- <img src="https://c8.patreon.com/3/200/4105384" alt="Jack Blakely" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
163
- <img src="https://c8.patreon.com/3/200/494309" alt="Julian Tsependa" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
164
- <img src="https://c8.patreon.com/3/200/24653779" alt="RayHell" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
165
- <img src="https://c8.patreon.com/3/200/4541423" alt="Sören " width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
166
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/31950857/c567dc648f6144be9f6234946df05da2/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=3Vx4R1eOfD4X_ZPPd40MsZ-3lyknLM35XmaHRELnWjM%3D" alt="Trent Hunter" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
167
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/110407414/30f9e9d88ef945ddb0f47fd23a8cbac2/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=QQRWOkMyOfDBERHn4O8N2wMB32zeiIEsydVTbSNUw-I%3D" alt="Wesley Reitzfeld" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
168
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/162398691/89d78d89eecb4d6b981ce8c3c6a3d4b8/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=SWhI-0jGpY6Nc_bUQeXz4pa9DRURi9VnnnJ3Mxjg1po%3D" alt="Zoltán-Csaba Nyiró" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
169
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/97985240/3d1d0e6905d045aba713e8132cab4a30/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=pG3X2m-py2lRYI2aoJiXI47_4ArD78ZHdSm6jCAHA_w%3D" alt="עומר מכלוף" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
170
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/140599287/cff037fb93804af28bc3a4f1e91154f8/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=vkscmpmFoM5wq7GnsLmOEgNhvyXe-774kNGNqD0wurE%3D" alt="Lukas" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
171
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/96561218/b0694642d13a49faa75aec9762ff2aeb/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=sLQXomYm1iMYpknvGwKQ49f30TKQ0B1R2W3EZfCJqr8%3D" alt="Ultimate Golf Archives" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
172
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/81275465/1e4148fe9c47452b838949d02dd9a70f/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=uzJzkUq9rte3wx8wDLjGAgvSoxdtZcAnH7HctDhdYEo%3D" alt="Aaron Amortegui" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
173
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/44568304/a9d83a0e786b41b4bdada150f7c9271c/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=SBphTD654nwr-OTrvIBIJBEQho7GE2PtRre8nyaG1Fk%3D" alt="Albert Bukoski" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
174
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/49304261/d0a730de1c3349e585c49288b9f419c6/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=C2BMZ3ci-Ty2nhnSwKZqsR-5hOGsUNDYcvXps0Geq9w%3D" alt="Arvin Flores" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
175
- <img src="https://c8.patreon.com/3/200/5048649" alt="Ben Ward" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
176
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/130338124/f904a3bb76cd4588ac8d8f595c6cb486/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=k-inISRUtYDu9q7fNAKc3S2S7qcaw26fr1pj7PqU28Q%3D" alt="Bnp" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
177
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/111904990/08b1cf65be6a4de091c9b73b693b3468/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=OAJc9W5Ak0uJfQ2COlo1Upo38K3aj1fMQFCMC7ft5tM%3D" alt="Brian Smith" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
178
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/113207022/d4a67cc113e84fb69032bef71d068720/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=mu-tIg88VwoQdgLEOmxuVkhVm9JT59DdnHXJstmkkLU%3D" alt="Fagem X" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
179
- <img src="https://c8.patreon.com/3/200/5602036" alt="Kelevra" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
180
- <img src="https://c8.patreon.com/3/200/358350" alt="L D" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
181
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/159203973/36c817f941ac4fa18103a4b8c0cb9cae/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=9toslDfsO14QyaOiu6vIf--d4marBsWCZWN3gdPqbIU%3D" alt="Marko jak" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
182
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/11198131/e696d9647feb4318bcf16243c2425805/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=o6Hrpzw9rf2Ucd4cZ-hdUkGejLNv44-pqF8smeOF3ts%3D" alt="Nicholas Agranoff" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
183
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/785333/bdb9ede5765d42e5a2021a86eebf0d8f/eyJ3IjoyMDB9/2.jpg?token-time=2145916800&token-hash=dr5eaMg3Ua0wyCy40Qv3F-ZFajWZmuz2fWG55FskREc%3D" alt="Sapjes " width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
184
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/44738426/b01ff676da864d4ab9c21f226275b63e/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=54nIkcxFaGszJ3q0jNhtrVSBbV3WNK9e5WX9VzXltYk%3D" alt="Shakeel Saleemi" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
185
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/76566911/6485eaf5ec6249a7b524ee0b979372f0/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=S1QK78ief5byQU7tB_reqnw4V2zhW_cpwTqHThk-tGc%3D" alt="the biitz" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
186
- <img src="https://c8.patreon.com/3/200/83034" alt="william tatum" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
187
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/32633822/1ab5612efe80417cbebfe91e871fc052/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=RHYMcjr0UGIYw5FBrUfJdKMGuoYWhBQlLIykccEFJvo%3D" alt="Zack Abrams" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
188
- <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/138787189/2b5662dcb638466282ac758e3ac651b4/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=IlUAs9JAlVRphfx81V-Jt-nMiSBS8mPewRr9u6pQjaQ%3D" alt="Антон Антонио" width="60" height="60" style="border-radius:8px;margin:5px;display: inline-block;">
189
  </p>
190
 
191
  ---
 
53
 
54
  ### Diffusers
55
 
56
+ For diffusers, you can use `AutoPipelineForText2Image`. This will load the model with the pipeline located [here](https://huggingface.co/ostris/Flex.2-preview/blob/main/pipeline.py). We will run this example using the `diffusers` library.
57
+
58
+
59
+ <img src="https://ostris.com/wp-content/uploads/2025/04/diffusers_example.jpg" alt="Diffusers Flex.2" style="max-width: 100%; height: auto;"/>
60
+
61
+ First install the requirements.
62
+
63
+ ```bash
64
+ pip install --upgrade torch accelerate transformers diffusers
65
+ ```
66
+
67
+ Then you can load the model and use it like this:
68
+
69
+ ```python
70
+ import torch
71
+ from diffusers import AutoPipelineForText2Image
72
+ from diffusers.utils import load_image
73
+
74
+ name_or_path = "ostris/Flex.2-preview"
75
+
76
+ inpaint_image = load_image("https://ostris.com/wp-content/uploads/2025/04/dog.jpg")
77
+ inpaint_mask = load_image("https://ostris.com/wp-content/uploads/2025/04/dog_mask.jpg")
78
+ control_image = load_image("https://ostris.com/wp-content/uploads/2025/04/dog_depth.jpg")
79
+
80
+ dtype = torch.bfloat16
81
+
82
+ pipe = AutoPipelineForText2Image.from_pretrained(
83
+ name_or_path,
84
+ custom_pipeline=name_or_path,
85
+ torch_dtype=dtype
86
+ ).to("cuda")
87
+
88
+ image = pipe(
89
+ prompt="A white friendly robotic dog sitting on a bench",
90
+ inpaint_image=inpaint_image,
91
+ inpaint_mask=inpaint_mask,
92
+ control_image=control_image,
93
+ control_strength=0.5,
94
+ control_stop=0.33,
95
+ height=1024,
96
+ width=1024,
97
+ guidance_scale=3.5,
98
+ num_inference_steps=50,
99
+ generator=torch.Generator("cpu").manual_seed(42)
100
+ ).images[0]
101
+ image.save(f"robot_dog.png")
102
+
103
+ ```
104
+
105
+ For consumer cards < 24GB, you can use torchao to
106
+
107
+ ```bash
108
+ pip install --upgrade torchao
109
+ ```
110
+
111
+ Then you can load the model and use it like this:
112
+
113
+ ```python
114
+ import torch
115
+ from diffusers import AutoPipelineForText2Image
116
+ from diffusers.utils import load_image
117
+ from transformers import T5EncoderModel, TorchAoConfig
118
+ from diffusers import FluxTransformer2DModel
119
+
120
+ name_or_path = "ostris/Flex.2-preview"
121
+
122
+ inpaint_image = load_image("https://ostris.com/wp-content/uploads/2025/04/dog.jpg")
123
+ inpaint_mask = load_image("https://ostris.com/wp-content/uploads/2025/04/dog_mask.jpg")
124
+ control_image = load_image("https://ostris.com/wp-content/uploads/2025/04/dog_depth.jpg")
125
+
126
+ dtype = torch.bfloat16
127
+
128
+ quant_config = TorchAoConfig("int8_weight_only")
129
+
130
+ text_encoder_2 = T5EncoderModel.from_pretrained(
131
+ name_or_path, subfolder="text_encoder_2", torch_dtype=dtype, quantization_config=quant_config
132
+ ).to("cuda")
133
+
134
+ transformer = FluxTransformer2DModel.from_pretrained(
135
+ name_or_path, subfolder="transformer", torch_dtype=dtype, quantization_config=quant_config
136
+ ).to("cuda")
137
+
138
+
139
+ pipe = AutoPipelineForText2Image.from_pretrained(
140
+ name_or_path,
141
+ transformer=transformer,
142
+ text_encoder_2=text_encoder_2,
143
+ custom_pipeline=name_or_path,
144
+ torch_dtype=dtype
145
+ ).to("cuda")
146
+
147
+ image = pipe(
148
+ prompt="A white friendly robotic dog sitting on a bench",
149
+ inpaint_image=inpaint_image,
150
+ inpaint_mask=inpaint_mask,
151
+ control_image=control_image,
152
+ control_strength=0.5,
153
+ control_stop=0.33,
154
+ height=1024,
155
+ width=1024,
156
+ guidance_scale=3.5,
157
+ num_inference_steps=50,
158
+ generator=torch.Generator("cpu").manual_seed(42)
159
+ ).images[0]
160
+ image.save(f"robot_dog.png")
161
+ ```
162
+
163
+
164
 
165
  ## Limitations
166
 
 
216
 
217
  All of these people / organizations are the ones who selflessly make this project possible. Thank you!!
218
 
219
+ _Last updated: 2025-04-22 16:45 UTC_
220
 
221
  <p align="center">
222
+ <a href="https://github.com/replicate" target="_blank" rel="noopener noreferrer"><img src="https://avatars.githubusercontent.com/u/60410876?v=4" alt="Replicate" width="200" height="200" style="border-radius:8px;margin:5px;"></a>
223
+ <a href="https://github.com/josephrocca" target="_blank" rel="noopener noreferrer"><img src="https://avatars.githubusercontent.com/u/1167575?u=92d92921b4cb5c8c7e225663fed53c4b41897736&v=4" alt="josephrocca" width="200" height="200" style="border-radius:8px;margin:5px;"></a>
224
  </p>
225
  <hr style="width:100%;border:none;height:2px;background:#ddd;margin:30px 0;">
226
  <p align="center">
227
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/162524101/81a72689c3754ac5b9e38612ce5ce914/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=3XLSlLFCWAQ-0wd2_vZMikyotdQNSzKOjoyeoJiZEw0%3D" alt="Prasanth Veerina" width="150" height="150" style="border-radius:8px;margin:5px;">
228
+ <a href="https://github.com/weights-ai" target="_blank" rel="noopener noreferrer"><img src="https://avatars.githubusercontent.com/u/185568492?v=4" alt="Weights" width="150" height="150" style="border-radius:8px;margin:5px;"></a>
229
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/161471720/dd330b4036d44a5985ed5985c12a5def/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=qkRvrEc5gLPxaXxLvcvbYv1W1lcmOoTwhj4A9Cq5BxQ%3D" alt="Vladimir Sotnikov" width="150" height="150" style="border-radius:8px;margin:5px;">
230
+ <img src="https://c8.patreon.com/3/200/33158543" alt="clement Delangue" width="150" height="150" style="border-radius:8px;margin:5px;">
231
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/54890369/45cea21d82974c78bf43956de7fb0e12/eyJ3IjoyMDB9/2.jpeg?token-time=2145916800&token-hash=IK6OT6UpusHgdaC4y8IhK5XxXiP5TuLy3vjvgL77Fho%3D" alt="Eli Slugworth" width="150" height="150" style="border-radius:8px;margin:5px;">
232
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/8654302/b0f5ebedc62a47c4b56222693e1254e9/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=lpeicIh1_S-3Ji3W27gyiRB7iXurp8Bx8HAzDHftOuo%3D" alt="Misch Strotz" width="150" height="150" style="border-radius:8px;margin:5px;">
233
+ <img src="https://c8.patreon.com/3/200/93304" alt="Joseph Rocca" width="150" height="150" style="border-radius:8px;margin:5px;">
234
  </p>
235
  <hr style="width:100%;border:none;height:2px;background:#ddd;margin:30px 0;">
236
  <p align="center">
237
+ <a href="https://x.com/NuxZoe" target="_blank" rel="noopener noreferrer"><img src="https://pbs.twimg.com/profile_images/1714760743273574400/tdvQjNTl_400x400.jpg" alt="tungsten" width="100" height="100" style="border-radius:8px;margin:5px;"></a>
238
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/2298192/1228b69bd7d7481baf3103315183250d/eyJ3IjoyMDB9/1.jpg?token-time=2145916800&token-hash=1B7dbXy_gAcPT9WXBesLhs7z_9APiz2k1Wx4Vml_-8Q%3D" alt="Mohamed Oumoumad" width="100" height="100" style="border-radius:8px;margin:5px;">
239
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/120239481/49b1ce70d3d24704b8ec34de24ec8f55/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=Dv1NPKwdv9QT8fhYYwbGnQIvfiyqTUlh52bjDW1vYxY%3D" alt="nitish PNR" width="100" height="100" style="border-radius:8px;margin:5px;">
240
+ <img src="https://c8.patreon.com/3/200/548524" alt="Steve Hanff" width="100" height="100" style="border-radius:8px;margin:5px;">
241
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/152118848/3b15a43d71714552b5ed1c9f84e66adf/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=IEKE18CBHVZ3k-08UD7Dkb7HbiFHb84W0FATdLMI0Dg%3D" alt="Kristjan Retter" width="100" height="100" style="border-radius:8px;margin:5px;">
242
+ <img src="https://c8.patreon.com/3/200/83319230" alt="Miguel Lara" width="100" height="100" style="border-radius:8px;margin:5px;">
243
  </p>
244
  <hr style="width:100%;border:none;height:2px;background:#ddd;margin:30px 0;">
245
  <p align="center">
246
+ <img src="https://c8.patreon.com/3/200/8449560" alt="Patron" width="60" height="60" style="border-radius:8px;margin:5px;">
247
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/27288932/6c35d2d961ee4e14a7a368c990791315/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=dpFFssZXZM_KZMKQhl3uDwwusdFw1c_v9x_ChJU7_zc%3D" alt="David Garrido" width="60" height="60" style="border-radius:8px;margin:5px;">
248
+ <img src="https://c8.patreon.com/3/200/2410522" alt="George Gostyshev" width="60" height="60" style="border-radius:8px;margin:5px;">
249
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/16287560/78130de30950410ca528d8a888997081/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=Ok-HSL2MthKXF09SmCOlPFCPfbMctFBZKCuTnPwxZ3A%3D" alt="Vitaly Golubenko" width="60" height="60" style="border-radius:8px;margin:5px;">
250
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/570742/4ceb33453a5a4745b430a216aba9280f/eyJ3IjoyMDB9/1.jpg?token-time=2145916800&token-hash=wUzsI5cO5Evp2ukIGdSgBbvKeYgv5LSOQMa6Br33Rrs%3D" alt="Al H" width="60" height="60" style="border-radius:8px;margin:5px;">
251
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/131773947/eda3405aa582437db4582fce908c8739/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=S4Bh0sMqTNmJlo3uRr7co5d_kxvBjITemDTfi_1KrCA%3D" alt="Jodh Singh" width="60" height="60" style="border-radius:8px;margin:5px;">
252
+ <img src="https://c8.patreon.com/3/200/22809690" alt="Michael Levine" width="60" height="60" style="border-radius:8px;margin:5px;">
253
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/99036356/7ae9c4d80e604e739b68cca12ee2ed01/eyJ3IjoyMDB9/3.png?token-time=2145916800&token-hash=zK0dHe6A937WtNlrGdefoXFTPPzHUCfn__23HP8-Ui0%3D" alt="Noctre" width="60" height="60" style="border-radius:8px;margin:5px;">
254
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/141098579/1a9f0a1249d447a7a0df718a57343912/eyJ3IjoyMDB9/2.png?token-time=2145916800&token-hash=Rd_AjZGhMATVkZDf8E95ILc0n93gvvFWe1Ig0_dxwf4%3D" alt="The Local Lab" width="60" height="60" style="border-radius:8px;margin:5px;">
255
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/98811435/3a3632d1795b4c2b9f8f0270f2f6a650/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=93w8RMxwXlcM4X74t03u6P5_SrKvlm1IpjnD2SzVpJk%3D" alt="EmmanuelMr18" width="60" height="60" style="border-radius:8px;margin:5px;">
256
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/338551/e8f257d8d3dd46c38272b391a5785948/eyJ3IjoyMDB9/1.jpg?token-time=2145916800&token-hash=GLom1rGgOZjBeO7I1OnjiIgWmjl6PO9ZjBB8YTvc7AM%3D" alt="Plaidam" width="60" height="60" style="border-radius:8px;margin:5px;">
257
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/82763/f99cc484361d4b9d94fe4f0814ada303/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=BpwC020pR3TRZ4r0RSCiSIOh-jmatkrpy1h2XU4sGa4%3D" alt="Doron Adler" width="60" height="60" style="border-radius:8px;margin:5px;">
258
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/103077711/bb215761cc004e80bd9cec7d4bcd636d/eyJ3IjoyMDB9/2.jpeg?token-time=2145916800&token-hash=zvtBie29rRTKTXvAA2KhOI-l3mSMk9xxr-mg_CksLtc%3D" alt="John Dopamine" width="60" height="60" style="border-radius:8px;margin:5px;">
259
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/93348210/5c650f32a0bc481d80900d2674528777/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=PpXK9B_iy288annlNdLOexhiQHbTftPEDeCh-sTQ2KA%3D" alt="Armin Behjati" width="60" height="60" style="border-radius:8px;margin:5px;">
260
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/155963250/6f8fd7075c3b4247bfeb054ba49172d6/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=twmKs4mADF_h7bKh5jBuigYVScMeaeHv2pEPin9K0Dg%3D" alt="Un Defined" width="60" height="60" style="border-radius:8px;margin:5px;">
261
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/45562978/0de33cf52ec642ae8a2f612cddec4ca6/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=hSAvaD4phiLcF0pvX7FP0juI5NQWCon-_TZSNpJzQJg%3D" alt="Jack English" width="60" height="60" style="border-radius:8px;margin:5px;">
262
+ <img src="https://c8.patreon.com/3/200/27791680" alt="Jean-Tristan Marin" width="60" height="60" style="border-radius:8px;margin:5px;">
263
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/60995694/92e0e8f336eb4a5bb8d99b940247d1d1/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=pj6Tm8XRdpGJcAEdnCakqYSNiSjoAYjvZescX7d0ic0%3D" alt="Abraham Irawan" width="60" height="60" style="border-radius:8px;margin:5px;">
264
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/164958178/4eb7a37baa0541bab7a091f2b14615b7/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=_aaum7fBJAGaJhMBhlR8vqYavDhExdVxmO9mwd3_XMw%3D" alt="Austin Robinson" width="60" height="60" style="border-radius:8px;margin:5px;">
265
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/134129880/680c7e14cd1a4d1a9face921fb010f88/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=vNKojv67krNqx7gdpKBX1R_stX2TkMRYvRc0xZrbY6s%3D" alt="Bharat Prabhakar" width="60" height="60" style="border-radius:8px;margin:5px;">
266
+ <img src="https://c8.patreon.com/3/200/70218846" alt="Cosmosis" width="60" height="60" style="border-radius:8px;margin:5px;">
267
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/83054970/13de6cb103ad41a5841edf549e66cd51/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=wU_Eke9VYcfI40FAQvdEV84Xspqlo5VSiafLqhg_FOE%3D" alt="Gili Ben Shahar" width="60" height="60" style="border-radius:8px;margin:5px;">
268
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/30931983/54ab4e4ceab946e79a6418d205f9ed51/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=LBmsSsMQZhO6yRZ_YyRwTgE6a7BVWrGNsAVveLXHXR0%3D" alt="HestoySeghuro ." width="60" height="60" style="border-radius:8px;margin:5px;">
269
+ <img src="https://c8.patreon.com/3/200/4105384" alt="Jack Blakely" width="60" height="60" style="border-radius:8px;margin:5px;">
270
+ <img src="https://c8.patreon.com/3/200/494309" alt="Julian Tsependa" width="60" height="60" style="border-radius:8px;margin:5px;">
271
+ <img src="https://c8.patreon.com/3/200/24653779" alt="RayHell" width="60" height="60" style="border-radius:8px;margin:5px;">
272
+ <img src="https://c8.patreon.com/3/200/4541423" alt="Sören " width="60" height="60" style="border-radius:8px;margin:5px;">
273
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/31950857/c567dc648f6144be9f6234946df05da2/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=3Vx4R1eOfD4X_ZPPd40MsZ-3lyknLM35XmaHRELnWjM%3D" alt="Trent Hunter" width="60" height="60" style="border-radius:8px;margin:5px;">
274
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/110407414/30f9e9d88ef945ddb0f47fd23a8cbac2/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=QQRWOkMyOfDBERHn4O8N2wMB32zeiIEsydVTbSNUw-I%3D" alt="Wesley Reitzfeld" width="60" height="60" style="border-radius:8px;margin:5px;">
275
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/162398691/89d78d89eecb4d6b981ce8c3c6a3d4b8/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=SWhI-0jGpY6Nc_bUQeXz4pa9DRURi9VnnnJ3Mxjg1po%3D" alt="Zoltán-Csaba Nyiró" width="60" height="60" style="border-radius:8px;margin:5px;">
276
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/97985240/3d1d0e6905d045aba713e8132cab4a30/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=pG3X2m-py2lRYI2aoJiXI47_4ArD78ZHdSm6jCAHA_w%3D" alt="עומר מכלוף" width="60" height="60" style="border-radius:8px;margin:5px;">
277
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/140599287/cff037fb93804af28bc3a4f1e91154f8/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=vkscmpmFoM5wq7GnsLmOEgNhvyXe-774kNGNqD0wurE%3D" alt="Lukas" width="60" height="60" style="border-radius:8px;margin:5px;">
278
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/96561218/b0694642d13a49faa75aec9762ff2aeb/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=sLQXomYm1iMYpknvGwKQ49f30TKQ0B1R2W3EZfCJqr8%3D" alt="Ultimate Golf Archives" width="60" height="60" style="border-radius:8px;margin:5px;">
279
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/81275465/1e4148fe9c47452b838949d02dd9a70f/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=uzJzkUq9rte3wx8wDLjGAgvSoxdtZcAnH7HctDhdYEo%3D" alt="Aaron Amortegui" width="60" height="60" style="border-radius:8px;margin:5px;">
280
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/44568304/a9d83a0e786b41b4bdada150f7c9271c/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=SBphTD654nwr-OTrvIBIJBEQho7GE2PtRre8nyaG1Fk%3D" alt="Albert Bukoski" width="60" height="60" style="border-radius:8px;margin:5px;">
281
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/49304261/d0a730de1c3349e585c49288b9f419c6/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=C2BMZ3ci-Ty2nhnSwKZqsR-5hOGsUNDYcvXps0Geq9w%3D" alt="Arvin Flores" width="60" height="60" style="border-radius:8px;margin:5px;">
282
+ <img src="https://c8.patreon.com/3/200/5048649" alt="Ben Ward" width="60" height="60" style="border-radius:8px;margin:5px;">
283
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/130338124/f904a3bb76cd4588ac8d8f595c6cb486/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=k-inISRUtYDu9q7fNAKc3S2S7qcaw26fr1pj7PqU28Q%3D" alt="Bnp" width="60" height="60" style="border-radius:8px;margin:5px;">
284
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/111904990/08b1cf65be6a4de091c9b73b693b3468/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=OAJc9W5Ak0uJfQ2COlo1Upo38K3aj1fMQFCMC7ft5tM%3D" alt="Brian Smith" width="60" height="60" style="border-radius:8px;margin:5px;">
285
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/113207022/d4a67cc113e84fb69032bef71d068720/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=mu-tIg88VwoQdgLEOmxuVkhVm9JT59DdnHXJstmkkLU%3D" alt="Fagem X" width="60" height="60" style="border-radius:8px;margin:5px;">
286
+ <img src="https://c8.patreon.com/3/200/5602036" alt="Kelevra" width="60" height="60" style="border-radius:8px;margin:5px;">
287
+ <img src="https://c8.patreon.com/3/200/358350" alt="L D" width="60" height="60" style="border-radius:8px;margin:5px;">
288
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/159203973/36c817f941ac4fa18103a4b8c0cb9cae/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=9toslDfsO14QyaOiu6vIf--d4marBsWCZWN3gdPqbIU%3D" alt="Marko jak" width="60" height="60" style="border-radius:8px;margin:5px;">
289
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/11198131/e696d9647feb4318bcf16243c2425805/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=o6Hrpzw9rf2Ucd4cZ-hdUkGejLNv44-pqF8smeOF3ts%3D" alt="Nicholas Agranoff" width="60" height="60" style="border-radius:8px;margin:5px;">
290
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/785333/bdb9ede5765d42e5a2021a86eebf0d8f/eyJ3IjoyMDB9/2.jpg?token-time=2145916800&token-hash=dr5eaMg3Ua0wyCy40Qv3F-ZFajWZmuz2fWG55FskREc%3D" alt="Sapjes " width="60" height="60" style="border-radius:8px;margin:5px;">
291
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/44738426/b01ff676da864d4ab9c21f226275b63e/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=54nIkcxFaGszJ3q0jNhtrVSBbV3WNK9e5WX9VzXltYk%3D" alt="Shakeel Saleemi" width="60" height="60" style="border-radius:8px;margin:5px;">
292
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/76566911/6485eaf5ec6249a7b524ee0b979372f0/eyJ3IjoyMDB9/1.jpeg?token-time=2145916800&token-hash=S1QK78ief5byQU7tB_reqnw4V2zhW_cpwTqHThk-tGc%3D" alt="the biitz" width="60" height="60" style="border-radius:8px;margin:5px;">
293
+ <img src="https://c8.patreon.com/3/200/83034" alt="william tatum" width="60" height="60" style="border-radius:8px;margin:5px;">
294
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/32633822/1ab5612efe80417cbebfe91e871fc052/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=RHYMcjr0UGIYw5FBrUfJdKMGuoYWhBQlLIykccEFJvo%3D" alt="Zack Abrams" width="60" height="60" style="border-radius:8px;margin:5px;">
295
+ <img src="https://c10.patreonusercontent.com/4/patreon-media/p/user/138787189/2b5662dcb638466282ac758e3ac651b4/eyJ3IjoyMDB9/1.png?token-time=2145916800&token-hash=IlUAs9JAlVRphfx81V-Jt-nMiSBS8mPewRr9u6pQjaQ%3D" alt="Антон Антонио" width="60" height="60" style="border-radius:8px;margin:5px;">
296
  </p>
297
 
298
  ---
pipeline.py ADDED
@@ -0,0 +1,440 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from diffusers import FluxControlPipeline, FluxTransformer2DModel
2
+ from typing import Any, Callable, Dict, List, Optional, Union
3
+ import torch
4
+
5
+ from diffusers.image_processor import PipelineImageInput
6
+ import numpy as np
7
+ import torch.nn.functional as F
8
+ from diffusers.pipelines.flux.pipeline_output import FluxPipelineOutput
9
+ from diffusers.pipelines.flux.pipeline_flux import calculate_shift, retrieve_timesteps, XLA_AVAILABLE
10
+
11
+
12
+ class Flex2Pipeline(FluxControlPipeline):
13
+ def __init__(
14
+ self,
15
+ scheduler,
16
+ vae,
17
+ text_encoder,
18
+ tokenizer,
19
+ text_encoder_2,
20
+ tokenizer_2,
21
+ transformer,
22
+ ):
23
+ super().__init__(scheduler, vae, text_encoder, tokenizer, text_encoder_2, tokenizer_2, transformer)
24
+
25
+ def check_inputs(
26
+ self,
27
+ prompt,
28
+ prompt_2,
29
+ height,
30
+ width,
31
+ prompt_embeds=None,
32
+ pooled_prompt_embeds=None,
33
+ callback_on_step_end_tensor_inputs=None,
34
+ max_sequence_length=None,
35
+ inpaint_image=None,
36
+ inpaint_mask=None,
37
+ control_image=None,
38
+ ):
39
+ super().check_inputs(
40
+ prompt,
41
+ prompt_2,
42
+ height,
43
+ width,
44
+ prompt_embeds=prompt_embeds,
45
+ pooled_prompt_embeds=pooled_prompt_embeds,
46
+ callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
47
+ max_sequence_length=max_sequence_length,
48
+ )
49
+ if inpaint_image is not None and inpaint_mask is None:
50
+ raise ValueError(
51
+ "If `inpaint_image` is passed, `inpaint_mask` must be passed as well. "
52
+ "Please make sure to pass both `inpaint_image` and `inpaint_mask`."
53
+ )
54
+ if inpaint_mask is not None and inpaint_image is None:
55
+ raise ValueError(
56
+ "If `inpaint_mask` is passed, `inpaint_image` must be passed as well. "
57
+ "Please make sure to pass both `inpaint_image` and `inpaint_mask`."
58
+ )
59
+
60
+ @torch.no_grad()
61
+ def __call__(
62
+ self,
63
+ prompt: Union[str, List[str]] = None,
64
+ prompt_2: Optional[Union[str, List[str]]] = None,
65
+ inpaint_image: Optional[PipelineImageInput] = None,
66
+ inpaint_mask: Optional[PipelineImageInput] = None,
67
+ control_image: Optional[PipelineImageInput] = None,
68
+ control_strength: Optional[float] = 1.0,
69
+ control_stop: Optional[float] = 1.0,
70
+ height: Optional[int] = None,
71
+ width: Optional[int] = None,
72
+ num_inference_steps: int = 28,
73
+ sigmas: Optional[List[float]] = None,
74
+ guidance_scale: float = 3.5,
75
+ num_images_per_prompt: Optional[int] = 1,
76
+ generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
77
+ latents: Optional[torch.FloatTensor] = None,
78
+ prompt_embeds: Optional[torch.FloatTensor] = None,
79
+ pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
80
+ output_type: Optional[str] = "pil",
81
+ return_dict: bool = True,
82
+ joint_attention_kwargs: Optional[Dict[str, Any]] = None,
83
+ callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
84
+ callback_on_step_end_tensor_inputs: List[str] = ["latents"],
85
+ max_sequence_length: int = 512,
86
+ **kwargs,
87
+ ):
88
+ r"""
89
+ Function invoked when calling the pipeline for generation.
90
+
91
+ Args:
92
+ prompt (`str` or `List[str]`, *optional*):
93
+ The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
94
+ instead.
95
+ prompt_2 (`str` or `List[str]`, *optional*):
96
+ The prompt or prompts to be sent to `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
97
+ will be used instead
98
+ inpaint_image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, `List[np.ndarray]`,:
99
+ `List[List[torch.Tensor]]`, `List[List[np.ndarray]]` or `List[List[PIL.Image.Image]]`):
100
+ The image to be inpainted.
101
+ inpaint_mask (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, `List[np.ndarray]`,:
102
+ `List[List[torch.Tensor]]`, `List[List[np.ndarray]]` or `List[List[PIL.Image.Image]]`):
103
+ A black and white mask to be used for inpainting. The white pixels are the areas to be inpainted, while the
104
+ black pixels are the areas to be kept.
105
+ control_image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, `List[np.ndarray]`,:
106
+ `List[List[torch.Tensor]]`, `List[List[np.ndarray]]` or `List[List[PIL.Image.Image]]`):
107
+ The control image (line, depth, pose, etc.) to be used for the generation. The control image
108
+ control_strength (`float`, *optional*, defaults to 1.0):
109
+ The strength of the control image. The higher the value, the more the control image will be used to
110
+ guide the generation. The lower the value, the less the control image will be used to guide the
111
+ generation.
112
+ control_stop (`float`, *optional*, defaults to 1.0):
113
+ The percentage of the generation to drop out the control. 0.0 to 1.0. 0.5 mean the control will be dropped
114
+ out at 50% of the generation.
115
+ height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
116
+ The height in pixels of the generated image. This is set to 1024 by default for the best results.
117
+ width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
118
+ The width in pixels of the generated image. This is set to 1024 by default for the best results.
119
+ num_inference_steps (`int`, *optional*, defaults to 50):
120
+ The number of denoising steps. More denoising steps usually lead to a higher quality image at the
121
+ expense of slower inference.
122
+ sigmas (`List[float]`, *optional*):
123
+ Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
124
+ their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
125
+ will be used.
126
+ guidance_scale (`float`, *optional*, defaults to 3.5):
127
+ Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
128
+ `guidance_scale` is defined as `w` of equation 2. of [Imagen
129
+ Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
130
+ 1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
131
+ usually at the expense of lower image quality.
132
+ num_images_per_prompt (`int`, *optional*, defaults to 1):
133
+ The number of images to generate per prompt.
134
+ generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
135
+ One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
136
+ to make generation deterministic.
137
+ latents (`torch.FloatTensor`, *optional*):
138
+ Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
139
+ generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
140
+ tensor will ge generated by sampling using the supplied random `generator`.
141
+ prompt_embeds (`torch.FloatTensor`, *optional*):
142
+ Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
143
+ provided, text embeddings will be generated from `prompt` input argument.
144
+ pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
145
+ Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
146
+ If not provided, pooled text embeddings will be generated from `prompt` input argument.
147
+ output_type (`str`, *optional*, defaults to `"pil"`):
148
+ The output format of the generate image. Choose between
149
+ [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
150
+ return_dict (`bool`, *optional*, defaults to `True`):
151
+ Whether or not to return a [`~pipelines.flux.FluxPipelineOutput`] instead of a plain tuple.
152
+ joint_attention_kwargs (`dict`, *optional*):
153
+ A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
154
+ `self.processor` in
155
+ [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
156
+ callback_on_step_end (`Callable`, *optional*):
157
+ A function that calls at the end of each denoising steps during the inference. The function is called
158
+ with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,
159
+ callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by
160
+ `callback_on_step_end_tensor_inputs`.
161
+ callback_on_step_end_tensor_inputs (`List`, *optional*):
162
+ The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
163
+ will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
164
+ `._callback_tensor_inputs` attribute of your pipeline class.
165
+ max_sequence_length (`int` defaults to 512): Maximum sequence length to use with the `prompt`.
166
+
167
+ Examples:
168
+
169
+ Returns:
170
+ [`~pipelines.flux.FluxPipelineOutput`] or `tuple`: [`~pipelines.flux.FluxPipelineOutput`] if `return_dict`
171
+ is True, otherwise a `tuple`. When returning a tuple, the first element is a list with the generated
172
+ images.
173
+ """
174
+
175
+ height = height or self.default_sample_size * self.vae_scale_factor
176
+ width = width or self.default_sample_size * self.vae_scale_factor
177
+
178
+ # 1. Check inputs. Raise error if not correct
179
+ self.check_inputs(
180
+ prompt,
181
+ prompt_2,
182
+ height,
183
+ width,
184
+ prompt_embeds=prompt_embeds,
185
+ pooled_prompt_embeds=pooled_prompt_embeds,
186
+ callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
187
+ max_sequence_length=max_sequence_length,
188
+ )
189
+
190
+ self._guidance_scale = guidance_scale
191
+ self._joint_attention_kwargs = joint_attention_kwargs
192
+ self._interrupt = False
193
+
194
+ # 2. Define call parameters
195
+ if prompt is not None and isinstance(prompt, str):
196
+ batch_size = 1
197
+ elif prompt is not None and isinstance(prompt, list):
198
+ batch_size = len(prompt)
199
+ else:
200
+ batch_size = prompt_embeds.shape[0]
201
+
202
+ device = self._execution_device
203
+
204
+ # 3. Prepare text embeddings
205
+ lora_scale = (
206
+ self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None
207
+ )
208
+ (
209
+ prompt_embeds,
210
+ pooled_prompt_embeds,
211
+ text_ids,
212
+ ) = self.encode_prompt(
213
+ prompt=prompt,
214
+ prompt_2=prompt_2,
215
+ prompt_embeds=prompt_embeds,
216
+ pooled_prompt_embeds=pooled_prompt_embeds,
217
+ device=device,
218
+ num_images_per_prompt=num_images_per_prompt,
219
+ max_sequence_length=max_sequence_length,
220
+ lora_scale=lora_scale,
221
+ )
222
+
223
+ # 4. Prepare latent variables
224
+ num_channels_latents = self.transformer.config.in_channels // 4
225
+
226
+ # only prepare latents for non controls
227
+ # (16 + 1 + 16 )
228
+ num_control_channels = 33
229
+ num_channels_latents = num_channels_latents - num_control_channels
230
+
231
+ control_latents = None
232
+ inpaint_latents = None
233
+ inpaint_latents_mask = None
234
+
235
+ latent_height = height // self.vae_scale_factor
236
+ latent_width = width // self.vae_scale_factor
237
+
238
+ # process the control and inpaint channels
239
+
240
+ if control_image is None:
241
+ control_latents = torch.zeros(
242
+ batch_size * num_images_per_prompt,
243
+ 3,
244
+ latent_height,
245
+ latent_width,
246
+ device=device,
247
+ dtype=self.vae.dtype,
248
+ )
249
+ else:
250
+ control_image = self.prepare_image(
251
+ image=control_image,
252
+ width=width,
253
+ height=height,
254
+ batch_size=batch_size * num_images_per_prompt,
255
+ num_images_per_prompt=num_images_per_prompt,
256
+ device=device,
257
+ dtype=self.vae.dtype,
258
+ )
259
+ control_image = self.vae.encode(control_image).latent_dist.sample(generator=generator)
260
+ control_latents = (control_image - self.vae.config.shift_factor) * self.vae.config.scaling_factor
261
+
262
+ # apply control strength
263
+ control_latents = control_latents * control_strength
264
+
265
+ if inpaint_image is None and inpaint_mask is None:
266
+ inpaint_latents = torch.zeros(
267
+ batch_size * num_images_per_prompt,
268
+ 3,
269
+ latent_height,
270
+ latent_width,
271
+ device=device,
272
+ dtype=self.vae.dtype,
273
+ )
274
+ inpaint_latents_mask = torch.ones(
275
+ batch_size * num_images_per_prompt,
276
+ 1,
277
+ latent_height,
278
+ latent_width,
279
+ device=device,
280
+ dtype=self.vae.dtype,
281
+ )
282
+ else:
283
+ inpaint_image = self.prepare_image(
284
+ image=inpaint_image,
285
+ width=width,
286
+ height=height,
287
+ batch_size=batch_size * num_images_per_prompt,
288
+ num_images_per_prompt=num_images_per_prompt,
289
+ device=device,
290
+ dtype=self.vae.dtype,
291
+ )
292
+ inpaint_image = self.vae.encode(inpaint_image).latent_dist.sample(generator=generator)
293
+ inpaint_latents = (inpaint_image - self.vae.config.shift_factor) * self.vae.config.scaling_factor
294
+ height_inpaint_image, width_inpaint_image = control_image.shape[2:]
295
+
296
+ inpaint_mask = self.prepare_image(
297
+ image=inpaint_mask,
298
+ width=width,
299
+ height=height,
300
+ batch_size=batch_size * num_images_per_prompt,
301
+ num_images_per_prompt=num_images_per_prompt,
302
+ device=device,
303
+ dtype=self.vae.dtype,
304
+ )
305
+ # mask is 3 ch -1 to 1. make it 1ch, 0 to 1
306
+ inpaint_mask = inpaint_mask[:, 0:1, :, :] * 0.5 + 0.5
307
+ # resize to match height_inpaint_image and width_inpaint_image
308
+ inpaint_latents_mask = F.interpolate(inpaint_mask, size=(height_inpaint_image, width_inpaint_image), mode="bilinear", align_corners=False)
309
+
310
+ # apply inverted mask to inpaint latents
311
+ inpaint_latents = inpaint_latents * (1 - inpaint_latents_mask)
312
+
313
+ # concat the latent controls on the channel dimension every step
314
+ latent_controls = torch.cat([inpaint_latents, inpaint_latents_mask, control_latents], dim=1)
315
+ latent_no_controls = torch.cat([inpaint_latents, inpaint_latents_mask, torch.zeros_like(control_latents)], dim=1)
316
+
317
+ # pack the controls
318
+ height_latent_controls, width_latent_controls = latent_controls.shape[2:]
319
+ packed_latent_controls = self._pack_latents(
320
+ latent_controls,
321
+ batch_size * num_images_per_prompt,
322
+ num_control_channels,
323
+ height_latent_controls,
324
+ width_latent_controls,
325
+ )
326
+ packed_latent_no_controls = self._pack_latents(
327
+ latent_no_controls,
328
+ batch_size * num_images_per_prompt,
329
+ num_control_channels,
330
+ height_latent_controls,
331
+ width_latent_controls,
332
+ )
333
+
334
+ latents, latent_image_ids = self.prepare_latents(
335
+ batch_size * num_images_per_prompt,
336
+ num_channels_latents,
337
+ height,
338
+ width,
339
+ prompt_embeds.dtype,
340
+ device,
341
+ generator,
342
+ latents,
343
+ )
344
+
345
+ # 5. Prepare timesteps
346
+ sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
347
+ image_seq_len = latents.shape[1]
348
+ mu = calculate_shift(
349
+ image_seq_len,
350
+ self.scheduler.config.get("base_image_seq_len", 256),
351
+ self.scheduler.config.get("max_image_seq_len", 4096),
352
+ self.scheduler.config.get("base_shift", 0.5),
353
+ self.scheduler.config.get("max_shift", 1.15),
354
+ )
355
+ timesteps, num_inference_steps = retrieve_timesteps(
356
+ self.scheduler,
357
+ num_inference_steps,
358
+ device,
359
+ sigmas=sigmas,
360
+ mu=mu,
361
+ )
362
+ num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
363
+ self._num_timesteps = len(timesteps)
364
+
365
+ # handle guidance
366
+ if self.transformer.config.guidance_embeds:
367
+ guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32)
368
+ guidance = guidance.expand(latents.shape[0])
369
+ else:
370
+ guidance = None
371
+
372
+ control_cutoff = int(len(timesteps) * control_stop)
373
+
374
+ # 6. Denoising loop
375
+ with self.progress_bar(total=num_inference_steps) as progress_bar:
376
+ for i, t in enumerate(timesteps):
377
+ if self.interrupt:
378
+ continue
379
+
380
+ control_latents = packed_latent_controls if i < control_cutoff else packed_latent_no_controls
381
+
382
+ latent_model_input = torch.cat([latents, control_latents], dim=2)
383
+
384
+ # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
385
+ timestep = t.expand(latents.shape[0]).to(latents.dtype)
386
+
387
+ noise_pred = self.transformer(
388
+ hidden_states=latent_model_input,
389
+ timestep=timestep / 1000,
390
+ guidance=guidance,
391
+ pooled_projections=pooled_prompt_embeds,
392
+ encoder_hidden_states=prompt_embeds,
393
+ txt_ids=text_ids,
394
+ img_ids=latent_image_ids,
395
+ joint_attention_kwargs=self.joint_attention_kwargs,
396
+ return_dict=False,
397
+ )[0]
398
+
399
+ # compute the previous noisy sample x_t -> x_t-1
400
+ latents_dtype = latents.dtype
401
+ latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
402
+
403
+ if latents.dtype != latents_dtype:
404
+ if torch.backends.mps.is_available():
405
+ # some platforms (eg. apple mps) misbehave due to a pytorch bug: https://github.com/pytorch/pytorch/pull/99272
406
+ latents = latents.to(latents_dtype)
407
+
408
+ if callback_on_step_end is not None:
409
+ callback_kwargs = {}
410
+ for k in callback_on_step_end_tensor_inputs:
411
+ callback_kwargs[k] = locals()[k]
412
+ callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
413
+
414
+ latents = callback_outputs.pop("latents", latents)
415
+ prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
416
+
417
+ # call the callback, if provided
418
+ if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
419
+ progress_bar.update()
420
+
421
+ if XLA_AVAILABLE:
422
+ xm.mark_step()
423
+
424
+ if output_type == "latent":
425
+ image = latents
426
+ else:
427
+ latents = self._unpack_latents(latents, height, width, self.vae_scale_factor)
428
+ latents = (latents / self.vae.config.scaling_factor) + self.vae.config.shift_factor
429
+ image = self.vae.decode(latents, return_dict=False)[0]
430
+ image = self.image_processor.postprocess(image, output_type=output_type)
431
+
432
+ # Offload all models
433
+ self.maybe_free_model_hooks()
434
+
435
+ if not return_dict:
436
+ return (image,)
437
+
438
+ return FluxPipelineOutput(images=image)
439
+
440
+