ollamadiffuser 1.2.3__py3-none-any.whl → 2.0.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (36) hide show
  1. ollamadiffuser/__init__.py +1 -1
  2. ollamadiffuser/api/server.py +312 -312
  3. ollamadiffuser/cli/config_commands.py +119 -0
  4. ollamadiffuser/cli/lora_commands.py +169 -0
  5. ollamadiffuser/cli/main.py +85 -1233
  6. ollamadiffuser/cli/model_commands.py +664 -0
  7. ollamadiffuser/cli/recommend_command.py +205 -0
  8. ollamadiffuser/cli/registry_commands.py +197 -0
  9. ollamadiffuser/core/config/model_registry.py +562 -11
  10. ollamadiffuser/core/config/settings.py +24 -2
  11. ollamadiffuser/core/inference/__init__.py +5 -0
  12. ollamadiffuser/core/inference/base.py +182 -0
  13. ollamadiffuser/core/inference/engine.py +204 -1405
  14. ollamadiffuser/core/inference/strategies/__init__.py +1 -0
  15. ollamadiffuser/core/inference/strategies/controlnet_strategy.py +170 -0
  16. ollamadiffuser/core/inference/strategies/flux_strategy.py +136 -0
  17. ollamadiffuser/core/inference/strategies/generic_strategy.py +164 -0
  18. ollamadiffuser/core/inference/strategies/gguf_strategy.py +113 -0
  19. ollamadiffuser/core/inference/strategies/hidream_strategy.py +104 -0
  20. ollamadiffuser/core/inference/strategies/sd15_strategy.py +134 -0
  21. ollamadiffuser/core/inference/strategies/sd3_strategy.py +80 -0
  22. ollamadiffuser/core/inference/strategies/sdxl_strategy.py +131 -0
  23. ollamadiffuser/core/inference/strategies/video_strategy.py +108 -0
  24. ollamadiffuser/mcp/__init__.py +0 -0
  25. ollamadiffuser/mcp/server.py +184 -0
  26. ollamadiffuser/ui/templates/index.html +62 -1
  27. ollamadiffuser/ui/web.py +116 -54
  28. {ollamadiffuser-1.2.3.dist-info → ollamadiffuser-2.0.0.dist-info}/METADATA +321 -108
  29. ollamadiffuser-2.0.0.dist-info/RECORD +61 -0
  30. {ollamadiffuser-1.2.3.dist-info → ollamadiffuser-2.0.0.dist-info}/WHEEL +1 -1
  31. {ollamadiffuser-1.2.3.dist-info → ollamadiffuser-2.0.0.dist-info}/entry_points.txt +1 -0
  32. ollamadiffuser/core/models/registry.py +0 -384
  33. ollamadiffuser/ui/samples/.DS_Store +0 -0
  34. ollamadiffuser-1.2.3.dist-info/RECORD +0 -45
  35. {ollamadiffuser-1.2.3.dist-info → ollamadiffuser-2.0.0.dist-info}/licenses/LICENSE +0 -0
  36. {ollamadiffuser-1.2.3.dist-info → ollamadiffuser-2.0.0.dist-info}/top_level.txt +0 -0
ollamadiffuser/core/config/model_registry.py CHANGED
@@ -103,7 +103,7 @@ class ModelRegistry:
103
103
  "min_ram_gb": 8,
104
104
  "recommended_ram_gb": 16,
105
105
  "disk_space_gb": 3,
106
- "supported_devices": ["CUDA", "CPU"],
106
+ "supported_devices": ["CUDA", "MPS", "CPU"],
107
107
  "performance_notes": "Heavily quantized, lowest quality but very fast"
108
108
  },
109
109
  "license_info": {
@@ -128,7 +128,7 @@ class ModelRegistry:
128
128
  "min_ram_gb": 10,
129
129
  "recommended_ram_gb": 16,
130
130
  "disk_space_gb": 4,
131
- "supported_devices": ["CUDA", "CPU"],
131
+ "supported_devices": ["CUDA", "MPS", "CPU"],
132
132
  "performance_notes": "Light quantization, good speed/quality balance"
133
133
  },
134
134
  "license_info": {
@@ -153,7 +153,7 @@ class ModelRegistry:
153
153
  "min_ram_gb": 12,
154
154
  "recommended_ram_gb": 20,
155
155
  "disk_space_gb": 6,
156
- "supported_devices": ["CUDA", "CPU"],
156
+ "supported_devices": ["CUDA", "MPS", "CPU"],
157
157
  "performance_notes": "Recommended quantization level - good quality and speed"
158
158
  },
159
159
  "license_info": {
@@ -178,7 +178,7 @@ class ModelRegistry:
178
178
  "min_ram_gb": 12,
179
179
  "recommended_ram_gb": 20,
180
180
  "disk_space_gb": 6,
181
- "supported_devices": ["CUDA", "CPU"],
181
+ "supported_devices": ["CUDA", "MPS", "CPU"],
182
182
  "performance_notes": "Q4_0 quantization - fast inference with good quality"
183
183
  },
184
184
  "license_info": {
@@ -203,7 +203,7 @@ class ModelRegistry:
203
203
  "min_ram_gb": 12,
204
204
  "recommended_ram_gb": 20,
205
205
  "disk_space_gb": 6,
206
- "supported_devices": ["CUDA", "CPU"],
206
+ "supported_devices": ["CUDA", "MPS", "CPU"],
207
207
  "performance_notes": "Q4_1 quantization - improved Q4_0 with better accuracy"
208
208
  },
209
209
  "license_info": {
@@ -228,7 +228,7 @@ class ModelRegistry:
228
228
  "min_ram_gb": 16,
229
229
  "recommended_ram_gb": 24,
230
230
  "disk_space_gb": 8,
231
- "supported_devices": ["CUDA", "CPU"],
231
+ "supported_devices": ["CUDA", "MPS", "CPU"],
232
232
  "performance_notes": "Higher quality quantization, slower but better results"
233
233
  },
234
234
  "license_info": {
@@ -253,7 +253,7 @@ class ModelRegistry:
253
253
  "min_ram_gb": 16,
254
254
  "recommended_ram_gb": 24,
255
255
  "disk_space_gb": 8,
256
- "supported_devices": ["CUDA", "CPU"],
256
+ "supported_devices": ["CUDA", "MPS", "CPU"],
257
257
  "performance_notes": "Q5_0 quantization - good balance of size and quality"
258
258
  },
259
259
  "license_info": {
@@ -278,7 +278,7 @@ class ModelRegistry:
278
278
  "min_ram_gb": 16,
279
279
  "recommended_ram_gb": 24,
280
280
  "disk_space_gb": 8,
281
- "supported_devices": ["CUDA", "CPU"],
281
+ "supported_devices": ["CUDA", "MPS", "CPU"],
282
282
  "performance_notes": "Q5_1 quantization - improved Q5_0 with better accuracy"
283
283
  },
284
284
  "license_info": {
@@ -303,7 +303,7 @@ class ModelRegistry:
303
303
  "min_ram_gb": 20,
304
304
  "recommended_ram_gb": 32,
305
305
  "disk_space_gb": 10,
306
- "supported_devices": ["CUDA", "CPU"],
306
+ "supported_devices": ["CUDA", "MPS", "CPU"],
307
307
  "performance_notes": "High quality quantization, close to original"
308
308
  },
309
309
  "license_info": {
@@ -328,7 +328,7 @@ class ModelRegistry:
328
328
  "min_ram_gb": 24,
329
329
  "recommended_ram_gb": 36,
330
330
  "disk_space_gb": 12,
331
- "supported_devices": ["CUDA", "CPU"],
331
+ "supported_devices": ["CUDA", "MPS", "CPU"],
332
332
  "performance_notes": "Very high quality, minimal quantization loss"
333
333
  },
334
334
  "license_info": {
@@ -353,7 +353,7 @@ class ModelRegistry:
353
353
  "min_ram_gb": 32,
354
354
  "recommended_ram_gb": 48,
355
355
  "disk_space_gb": 16,
356
- "supported_devices": ["CUDA", "CPU"],
356
+ "supported_devices": ["CUDA", "MPS", "CPU"],
357
357
  "performance_notes": "Full precision, best quality but largest size"
358
358
  },
359
359
  "license_info": {
@@ -550,6 +550,557 @@ class ModelRegistry:
550
550
  "supported_devices": ["CUDA", "MPS", "CPU"],
551
551
  "performance_notes": "Requires base SDXL model + ControlNet model. Good for depth-based control with SDXL quality."
552
552
  }
553
+ },
554
+
555
+ # ===== Tier 1: Additional base models (use existing strategies) =====
556
+
557
+ "stable-diffusion-3.5-large": {
558
+ "repo_id": "stabilityai/stable-diffusion-3.5-large",
559
+ "model_type": "sd3",
560
+ "variant": "fp16",
561
+ "parameters": {
562
+ "num_inference_steps": 28,
563
+ "guidance_scale": 3.5
564
+ },
565
+ "hardware_requirements": {
566
+ "min_vram_gb": 12,
567
+ "recommended_vram_gb": 16,
568
+ "min_ram_gb": 24,
569
+ "recommended_ram_gb": 48,
570
+ "disk_space_gb": 16,
571
+ "supported_devices": ["CUDA", "MPS"],
572
+ "performance_notes": "Large SD3.5 model. Higher quality than medium variant."
573
+ },
574
+ "license_info": {
575
+ "type": "Stability AI Community License",
576
+ "requires_agreement": True,
577
+ "commercial_use": False
578
+ }
579
+ },
580
+
581
+ "stable-diffusion-3.5-large-turbo": {
582
+ "repo_id": "stabilityai/stable-diffusion-3.5-large-turbo",
583
+ "model_type": "sd3",
584
+ "variant": "fp16",
585
+ "parameters": {
586
+ "num_inference_steps": 4,
587
+ "guidance_scale": 0.0
588
+ },
589
+ "hardware_requirements": {
590
+ "min_vram_gb": 12,
591
+ "recommended_vram_gb": 16,
592
+ "min_ram_gb": 24,
593
+ "recommended_ram_gb": 48,
594
+ "disk_space_gb": 16,
595
+ "supported_devices": ["CUDA", "MPS"],
596
+ "performance_notes": "Turbo variant with 4-step generation. Fast inference."
597
+ },
598
+ "license_info": {
599
+ "type": "Stability AI Community License",
600
+ "requires_agreement": True,
601
+ "commercial_use": False
602
+ }
603
+ },
604
+
605
+ "realvisxl-v4": {
606
+ "repo_id": "SG161222/RealVisXL_V4.0",
607
+ "model_type": "sdxl",
608
+ "variant": "fp16",
609
+ "parameters": {
610
+ "num_inference_steps": 30,
611
+ "guidance_scale": 5.0
612
+ },
613
+ "hardware_requirements": {
614
+ "min_vram_gb": 6,
615
+ "recommended_vram_gb": 10,
616
+ "min_ram_gb": 12,
617
+ "recommended_ram_gb": 24,
618
+ "disk_space_gb": 7,
619
+ "supported_devices": ["CUDA", "MPS", "CPU"],
620
+ "performance_notes": "Photorealistic SDXL finetune. Very popular community model."
621
+ },
622
+ "license_info": {
623
+ "type": "CreativeML Open RAIL++-M",
624
+ "requires_agreement": False,
625
+ "commercial_use": True
626
+ }
627
+ },
628
+
629
+ "dreamshaper": {
630
+ "repo_id": "Lykon/DreamShaper",
631
+ "model_type": "sd15",
632
+ "variant": "fp16",
633
+ "parameters": {
634
+ "num_inference_steps": 30,
635
+ "guidance_scale": 7.0
636
+ },
637
+ "hardware_requirements": {
638
+ "min_vram_gb": 4,
639
+ "recommended_vram_gb": 6,
640
+ "min_ram_gb": 8,
641
+ "recommended_ram_gb": 16,
642
+ "disk_space_gb": 5,
643
+ "supported_devices": ["CUDA", "MPS", "CPU"],
644
+ "performance_notes": "Popular community SD 1.5 model. Runs on most GPUs."
645
+ },
646
+ "license_info": {
647
+ "type": "CreativeML Open RAIL-M",
648
+ "requires_agreement": False,
649
+ "commercial_use": True
650
+ }
651
+ },
652
+
653
+ "realistic-vision-v6": {
654
+ "repo_id": "SG161222/Realistic_Vision_V6.0_B1_noVAE",
655
+ "model_type": "sd15",
656
+ "variant": "fp16",
657
+ "parameters": {
658
+ "num_inference_steps": 30,
659
+ "guidance_scale": 5.0
660
+ },
661
+ "hardware_requirements": {
662
+ "min_vram_gb": 4,
663
+ "recommended_vram_gb": 6,
664
+ "min_ram_gb": 8,
665
+ "recommended_ram_gb": 16,
666
+ "disk_space_gb": 5,
667
+ "supported_devices": ["CUDA", "MPS", "CPU"],
668
+ "performance_notes": "Photorealistic SD 1.5 model. Great for portraits."
669
+ },
670
+ "license_info": {
671
+ "type": "CreativeML Open RAIL-M",
672
+ "requires_agreement": False,
673
+ "commercial_use": True
674
+ }
675
+ },
676
+
677
+ "sdxl-turbo": {
678
+ "repo_id": "stabilityai/sdxl-turbo",
679
+ "model_type": "sdxl",
680
+ "variant": "fp16",
681
+ "parameters": {
682
+ "num_inference_steps": 1,
683
+ "guidance_scale": 0.0
684
+ },
685
+ "hardware_requirements": {
686
+ "min_vram_gb": 6,
687
+ "recommended_vram_gb": 10,
688
+ "min_ram_gb": 12,
689
+ "recommended_ram_gb": 24,
690
+ "disk_space_gb": 7,
691
+ "supported_devices": ["CUDA", "MPS", "CPU"],
692
+ "performance_notes": "Distilled SDXL. Single-step inference, extremely fast."
693
+ },
694
+ "license_info": {
695
+ "type": "Stability AI Non-Commercial Research Community License",
696
+ "requires_agreement": True,
697
+ "commercial_use": False
698
+ }
699
+ },
700
+
701
+ # ===== Tier 2: Scheduler override models =====
702
+
703
+ "sdxl-lightning-4step": {
704
+ "repo_id": "ByteDance/SDXL-Lightning",
705
+ "model_type": "sdxl",
706
+ "variant": "fp16",
707
+ "parameters": {
708
+ "num_inference_steps": 4,
709
+ "guidance_scale": 0.0,
710
+ "scheduler_class": "EulerDiscreteScheduler",
711
+ "scheduler_kwargs": {
712
+ "timestep_spacing": "trailing"
713
+ }
714
+ },
715
+ "hardware_requirements": {
716
+ "min_vram_gb": 6,
717
+ "recommended_vram_gb": 10,
718
+ "min_ram_gb": 12,
719
+ "recommended_ram_gb": 24,
720
+ "disk_space_gb": 7,
721
+ "supported_devices": ["CUDA", "MPS", "CPU"],
722
+ "performance_notes": "Lightning-fast SDXL with 4 steps."
723
+ },
724
+ "license_info": {
725
+ "type": "OpenRAIL++",
726
+ "requires_agreement": False,
727
+ "commercial_use": True
728
+ }
729
+ },
730
+
731
+ # ===== Tier 3: FLUX pipeline variants =====
732
+
733
+ "flux.1-fill-dev": {
734
+ "repo_id": "black-forest-labs/FLUX.1-Fill-dev",
735
+ "model_type": "flux",
736
+ "variant": "fp16",
737
+ "parameters": {
738
+ "pipeline_class": "FluxFillPipeline",
739
+ "num_inference_steps": 28,
740
+ "guidance_scale": 30.0,
741
+ "max_sequence_length": 512
742
+ },
743
+ "hardware_requirements": {
744
+ "min_vram_gb": 20,
745
+ "recommended_vram_gb": 24,
746
+ "min_ram_gb": 32,
747
+ "recommended_ram_gb": 64,
748
+ "disk_space_gb": 24,
749
+ "supported_devices": ["CUDA"],
750
+ "performance_notes": "FLUX inpainting/outpainting model."
751
+ },
752
+ "license_info": {
753
+ "type": "FLUX.1 Non-Commercial License",
754
+ "requires_agreement": True,
755
+ "commercial_use": False
756
+ }
757
+ },
758
+
759
+ "flux.1-canny-dev": {
760
+ "repo_id": "black-forest-labs/FLUX.1-Canny-dev",
761
+ "model_type": "flux",
762
+ "variant": "fp16",
763
+ "parameters": {
764
+ "pipeline_class": "FluxControlPipeline",
765
+ "num_inference_steps": 28,
766
+ "guidance_scale": 30.0,
767
+ "max_sequence_length": 512
768
+ },
769
+ "hardware_requirements": {
770
+ "min_vram_gb": 20,
771
+ "recommended_vram_gb": 24,
772
+ "min_ram_gb": 32,
773
+ "recommended_ram_gb": 64,
774
+ "disk_space_gb": 24,
775
+ "supported_devices": ["CUDA"],
776
+ "performance_notes": "FLUX canny edge control model."
777
+ },
778
+ "license_info": {
779
+ "type": "FLUX.1 Non-Commercial License",
780
+ "requires_agreement": True,
781
+ "commercial_use": False
782
+ }
783
+ },
784
+
785
+ "flux.1-depth-dev": {
786
+ "repo_id": "black-forest-labs/FLUX.1-Depth-dev",
787
+ "model_type": "flux",
788
+ "variant": "fp16",
789
+ "parameters": {
790
+ "pipeline_class": "FluxControlPipeline",
791
+ "num_inference_steps": 28,
792
+ "guidance_scale": 10.0,
793
+ "max_sequence_length": 512
794
+ },
795
+ "hardware_requirements": {
796
+ "min_vram_gb": 20,
797
+ "recommended_vram_gb": 24,
798
+ "min_ram_gb": 32,
799
+ "recommended_ram_gb": 64,
800
+ "disk_space_gb": 24,
801
+ "supported_devices": ["CUDA"],
802
+ "performance_notes": "FLUX depth control model."
803
+ },
804
+ "license_info": {
805
+ "type": "FLUX.1 Non-Commercial License",
806
+ "requires_agreement": True,
807
+ "commercial_use": False
808
+ }
809
+ },
810
+
811
+ # ===== Tier 4: New-generation models (GenericPipelineStrategy) =====
812
+
813
+ "flux.2-dev": {
814
+ "repo_id": "black-forest-labs/FLUX.2-dev",
815
+ "model_type": "generic",
816
+ "variant": "fp16",
817
+ "parameters": {
818
+ "pipeline_class": "Flux2Pipeline",
819
+ "num_inference_steps": 28,
820
+ "guidance_scale": 3.5,
821
+ "torch_dtype": "bfloat16",
822
+ "enable_cpu_offload": True
823
+ },
824
+ "hardware_requirements": {
825
+ "min_vram_gb": 14,
826
+ "recommended_vram_gb": 24,
827
+ "min_ram_gb": 32,
828
+ "recommended_ram_gb": 64,
829
+ "disk_space_gb": 36,
830
+ "supported_devices": ["CUDA"],
831
+ "performance_notes": "FLUX.2 32B parameter model. Requires CPU offloading on consumer GPUs. Requires diffusers from source."
832
+ },
833
+ "license_info": {
834
+ "type": "FLUX.2 Dev Non-Commercial License",
835
+ "requires_agreement": True,
836
+ "commercial_use": False
837
+ }
838
+ },
839
+
840
+ "flux.2-klein-4b": {
841
+ "repo_id": "black-forest-labs/FLUX.2-klein-4B",
842
+ "model_type": "generic",
843
+ "variant": "fp16",
844
+ "parameters": {
845
+ "pipeline_class": "Flux2KleinPipeline",
846
+ "num_inference_steps": 28,
847
+ "guidance_scale": 3.5,
848
+ "torch_dtype": "bfloat16"
849
+ },
850
+ "hardware_requirements": {
851
+ "min_vram_gb": 10,
852
+ "recommended_vram_gb": 16,
853
+ "min_ram_gb": 16,
854
+ "recommended_ram_gb": 32,
855
+ "disk_space_gb": 10,
856
+ "supported_devices": ["CUDA", "MPS"],
857
+ "performance_notes": "Compact 4B FLUX.2 model. Apache 2.0 licensed. Requires diffusers from source."
858
+ },
859
+ "license_info": {
860
+ "type": "Apache 2.0",
861
+ "requires_agreement": False,
862
+ "commercial_use": True
863
+ }
864
+ },
865
+
866
+ "z-image-turbo": {
867
+ "repo_id": "Tongyi-MAI/Z-Image-Turbo",
868
+ "model_type": "generic",
869
+ "variant": "fp16",
870
+ "parameters": {
871
+ "pipeline_class": "ZImagePipeline",
872
+ "num_inference_steps": 8,
873
+ "guidance_scale": 5.0,
874
+ "torch_dtype": "bfloat16"
875
+ },
876
+ "hardware_requirements": {
877
+ "min_vram_gb": 10,
878
+ "recommended_vram_gb": 16,
879
+ "min_ram_gb": 16,
880
+ "recommended_ram_gb": 32,
881
+ "disk_space_gb": 14,
882
+ "supported_devices": ["CUDA"],
883
+ "performance_notes": "Alibaba 6B model. 8-step turbo generation. Bilingual Chinese/English. Requires diffusers from source."
884
+ },
885
+ "license_info": {
886
+ "type": "Apache 2.0",
887
+ "requires_agreement": False,
888
+ "commercial_use": True
889
+ }
890
+ },
891
+
892
+ "sana-1.5": {
893
+ "repo_id": "Efficient-Large-Model/SANA1.5_1.6B_1024px_diffusers",
894
+ "model_type": "generic",
895
+ "variant": "fp16",
896
+ "parameters": {
897
+ "pipeline_class": "SanaPipeline",
898
+ "num_inference_steps": 20,
899
+ "guidance_scale": 5.0,
900
+ "torch_dtype": "float16"
901
+ },
902
+ "hardware_requirements": {
903
+ "min_vram_gb": 8,
904
+ "recommended_vram_gb": 12,
905
+ "min_ram_gb": 16,
906
+ "recommended_ram_gb": 32,
907
+ "disk_space_gb": 4,
908
+ "supported_devices": ["CUDA", "MPS", "CPU"],
909
+ "performance_notes": "NVIDIA 1.6B model. Very efficient, competitive with 12B models."
910
+ },
911
+ "license_info": {
912
+ "type": "Apache 2.0",
913
+ "requires_agreement": False,
914
+ "commercial_use": True
915
+ }
916
+ },
917
+
918
+ "cogview4": {
919
+ "repo_id": "THUDM/CogView4-6B",
920
+ "model_type": "generic",
921
+ "variant": "fp16",
922
+ "parameters": {
923
+ "pipeline_class": "CogView4Pipeline",
924
+ "num_inference_steps": 50,
925
+ "guidance_scale": 3.5,
926
+ "torch_dtype": "bfloat16"
927
+ },
928
+ "hardware_requirements": {
929
+ "min_vram_gb": 12,
930
+ "recommended_vram_gb": 18,
931
+ "min_ram_gb": 16,
932
+ "recommended_ram_gb": 32,
933
+ "disk_space_gb": 14,
934
+ "supported_devices": ["CUDA", "MPS"],
935
+ "performance_notes": "Zhipu AI 6B model with GLM-4 text encoder. Bilingual Chinese/English. Tight fit on 16GB Apple Silicon."
936
+ },
937
+ "license_info": {
938
+ "type": "Apache 2.0",
939
+ "requires_agreement": False,
940
+ "commercial_use": True
941
+ }
942
+ },
943
+
944
+ "kolors": {
945
+ "repo_id": "Kwai-Kolors/Kolors-diffusers",
946
+ "model_type": "generic",
947
+ "variant": "fp16",
948
+ "parameters": {
949
+ "pipeline_class": "KolorsPipeline",
950
+ "num_inference_steps": 50,
951
+ "guidance_scale": 5.0,
952
+ "torch_dtype": "float16",
953
+ "enable_cpu_offload": True
954
+ },
955
+ "hardware_requirements": {
956
+ "min_vram_gb": 8,
957
+ "recommended_vram_gb": 18,
958
+ "min_ram_gb": 16,
959
+ "recommended_ram_gb": 32,
960
+ "disk_space_gb": 18,
961
+ "supported_devices": ["CUDA", "MPS"],
962
+ "performance_notes": "Kuaishou 8.6B model with ChatGLM3 encoder. Bilingual Chinese/English. Midjourney-v6 level quality."
963
+ },
964
+ "license_info": {
965
+ "type": "Kolors License",
966
+ "requires_agreement": False,
967
+ "commercial_use": True
968
+ }
969
+ },
970
+
971
+ "hunyuan-dit": {
972
+ "repo_id": "Tencent-Hunyuan/HunyuanDiT-v1.2-Diffusers",
973
+ "model_type": "generic",
974
+ "variant": "fp16",
975
+ "parameters": {
976
+ "pipeline_class": "HunyuanDiTPipeline",
977
+ "num_inference_steps": 50,
978
+ "guidance_scale": 5.0,
979
+ "torch_dtype": "float16",
980
+ "enable_cpu_offload": True
981
+ },
982
+ "hardware_requirements": {
983
+ "min_vram_gb": 6,
984
+ "recommended_vram_gb": 16,
985
+ "min_ram_gb": 16,
986
+ "recommended_ram_gb": 32,
987
+ "disk_space_gb": 12,
988
+ "supported_devices": ["CUDA", "MPS"],
989
+ "performance_notes": "Tencent 1.5B model with dual text encoders (mT5 + bilingual CLIP). Fine-grained Chinese understanding."
990
+ },
991
+ "license_info": {
992
+ "type": "Tencent Hunyuan Community License",
993
+ "requires_agreement": False,
994
+ "commercial_use": True
995
+ }
996
+ },
997
+
998
+ "lumina-2": {
999
+ "repo_id": "Alpha-VLLM/Lumina-Image-2.0",
1000
+ "model_type": "generic",
1001
+ "variant": "fp16",
1002
+ "parameters": {
1003
+ "pipeline_class": "Lumina2Pipeline",
1004
+ "num_inference_steps": 30,
1005
+ "guidance_scale": 4.0,
1006
+ "torch_dtype": "bfloat16",
1007
+ "enable_cpu_offload": True
1008
+ },
1009
+ "hardware_requirements": {
1010
+ "min_vram_gb": 8,
1011
+ "recommended_vram_gb": 16,
1012
+ "min_ram_gb": 16,
1013
+ "recommended_ram_gb": 32,
1014
+ "disk_space_gb": 6,
1015
+ "supported_devices": ["CUDA", "MPS"],
1016
+ "performance_notes": "Alpha-VLLM 2B model. Unified text+image token architecture. Works on Apple Silicon (tight on 16GB)."
1017
+ },
1018
+ "license_info": {
1019
+ "type": "Apache 2.0",
1020
+ "requires_agreement": False,
1021
+ "commercial_use": True
1022
+ }
1023
+ },
1024
+
1025
+ "pixart-sigma": {
1026
+ "repo_id": "PixArt-alpha/PixArt-Sigma-XL-2-1024-MS",
1027
+ "model_type": "generic",
1028
+ "variant": "fp16",
1029
+ "parameters": {
1030
+ "pipeline_class": "PixArtSigmaPipeline",
1031
+ "num_inference_steps": 20,
1032
+ "guidance_scale": 4.5,
1033
+ "torch_dtype": "float16"
1034
+ },
1035
+ "hardware_requirements": {
1036
+ "min_vram_gb": 6,
1037
+ "recommended_vram_gb": 10,
1038
+ "min_ram_gb": 8,
1039
+ "recommended_ram_gb": 16,
1040
+ "disk_space_gb": 3,
1041
+ "supported_devices": ["CUDA", "MPS", "CPU"],
1042
+ "performance_notes": "Very lightweight 0.6B model. 4K capable. Outperforms much larger models."
1043
+ },
1044
+ "license_info": {
1045
+ "type": "PixArt Open License",
1046
+ "requires_agreement": False,
1047
+ "commercial_use": True
1048
+ }
1049
+ },
1050
+
1051
+ "auraflow": {
1052
+ "repo_id": "fal/AuraFlow-v0.3",
1053
+ "model_type": "generic",
1054
+ "variant": "fp16",
1055
+ "parameters": {
1056
+ "pipeline_class": "AuraFlowPipeline",
1057
+ "num_inference_steps": 50,
1058
+ "guidance_scale": 3.5,
1059
+ "torch_dtype": "float16",
1060
+ "enable_cpu_offload": True
1061
+ },
1062
+ "hardware_requirements": {
1063
+ "min_vram_gb": 12,
1064
+ "recommended_vram_gb": 20,
1065
+ "min_ram_gb": 16,
1066
+ "recommended_ram_gb": 32,
1067
+ "disk_space_gb": 14,
1068
+ "supported_devices": ["CUDA"],
1069
+ "performance_notes": "Fal 6.8B model. Largest Apache 2.0 licensed text-to-image model."
1070
+ },
1071
+ "license_info": {
1072
+ "type": "Apache 2.0",
1073
+ "requires_agreement": False,
1074
+ "commercial_use": True
1075
+ }
1076
+ },
1077
+
1078
+ "omnigen": {
1079
+ "repo_id": "Shitao/OmniGen-v1-diffusers",
1080
+ "model_type": "generic",
1081
+ "variant": "fp16",
1082
+ "parameters": {
1083
+ "pipeline_class": "OmniGenPipeline",
1084
+ "num_inference_steps": 50,
1085
+ "guidance_scale": 2.5,
1086
+ "torch_dtype": "bfloat16",
1087
+ "enable_cpu_offload": True,
1088
+ "supports_negative_prompt": False
1089
+ },
1090
+ "hardware_requirements": {
1091
+ "min_vram_gb": 12,
1092
+ "recommended_vram_gb": 18,
1093
+ "min_ram_gb": 16,
1094
+ "recommended_ram_gb": 32,
1095
+ "disk_space_gb": 10,
1096
+ "supported_devices": ["CUDA"],
1097
+ "performance_notes": "BAAI 3.8B unified model. Text-to-image, editing, subject-driven generation without extra plugins."
1098
+ },
1099
+ "license_info": {
1100
+ "type": "MIT",
1101
+ "requires_agreement": False,
1102
+ "commercial_use": True
1103
+ }
553
1104
  }
554
1105
  }
555
1106