sync: update submodule

Author: william-murray1204

stable_diffusion_cpp/__init__.py

@@ -4,4 +4,4 @@
 
 # isort: on
 
-__version__ = "0.2.5"
+__version__ = "0.2.6"

stable_diffusion_cpp/stable_diffusion.py

@@ -187,6 +187,7 @@ def txt_to_img(
         clip_skip: int = -1,
         cfg_scale: float = 7.0,
         guidance: float = 3.5,
+        eta: float = 0.0,
         width: int = 512,
         height: int = 512,
         sample_method: Optional[Union[str, SampleMethod, int, float]] = "euler_a",
@@ -214,6 +215,7 @@ def txt_to_img(
             clip_skip: Ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer.
             cfg_scale: Unconditional guidance scale.
             guidance: Guidance scale.
+            eta: Eta in DDIM, only for DDIM and TCD.
             width: Image height, in pixel space.
             height: Image width, in pixel space.
             sample_method: Sampling method.
@@ -286,6 +288,7 @@ def sd_progress_callback(
                 clip_skip,
                 cfg_scale,
                 guidance,
+                eta,
                 width,
                 height,
                 sample_method,
@@ -320,6 +323,7 @@ def img_to_img(
         clip_skip: int = -1,
         cfg_scale: float = 7.0,
         guidance: float = 3.5,
+        eta: float = 0.0,
         width: int = 512,
         height: int = 512,
         sample_method: Optional[Union[str, SampleMethod, int, float]] = "euler_a",
@@ -350,6 +354,7 @@ def img_to_img(
             clip_skip: Ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer.
             cfg_scale: Unconditional guidance scale.
             guidance: Guidance scale.
+            eta: Eta in DDIM, only for DDIM and TCD.
             width: Image height, in pixel space.
             height: Image width, in pixel space.
             sample_method: Sampling method.
@@ -451,6 +456,7 @@ def _create_blank_mask_image(width: int, height: int):
                 clip_skip,
                 cfg_scale,
                 guidance,
+                eta,
                 width,
                 height,
                 sample_method,
@@ -941,6 +947,8 @@ def validate_and_set_input(user_input: Union[str, int, float], type_map: Dict, a
     "ipndm": SampleMethod.IPNDM,
     "ipndm_v": SampleMethod.IPNDM_V,
     "lcm": SampleMethod.LCM,
+    "ddim_trailing": SampleMethod.DDIM_TRAILING,
+    "tcd": SampleMethod.TCD,
     "n_sample_methods": SampleMethod.N_SAMPLE_METHODS,
 }
 
@@ -985,11 +993,14 @@ def validate_and_set_input(user_input: Union[str, int, float], type_map: Dict, a
     "f64": GGMLType.SD_TYPE_F64,
     "iq1_m": GGMLType.SD_TYPE_IQ1_M,
     "bf16": GGMLType.SD_TYPE_BF16,
-    "q4_0_4_4": GGMLType.SD_TYPE_Q4_0_4_4,
-    "q4_0_4_8": GGMLType.SD_TYPE_Q4_0_4_8,
-    "q4_0_8_8": GGMLType.SD_TYPE_Q4_0_8_8,
+    # "q4_0_4_4": GGMLType.SD_TYPE_Q4_0_4_4,
+    # "q4_0_4_8": GGMLType.SD_TYPE_Q4_0_4_8,
+    # "q4_0_8_8": GGMLType.SD_TYPE_Q4_0_8_8,
     "tq1_0": GGMLType.SD_TYPE_TQ1_0,
     "tq2_0": GGMLType.SD_TYPE_TQ2_0,
+    # "iq4_nl_4_4": GGMLType.SD_TYPE_IQ4_NL_4_4,
+    # "iq4_nl_4_8": GGMLType.SD_TYPE_IQ4_NL_4_8,
+    # "iq4_nl_8_8": GGMLType.SD_TYPE_IQ4_NL_8_8,
     # Default
     "default": GGMLType.SD_TYPE_COUNT,
 }

stable_diffusion_cpp/stable_diffusion_cpp.py

@@ -173,6 +173,8 @@ class RNGType(IntEnum):
 #     IPNDM,
 #     IPNDM_V,
 #     LCM,
+#     DDIM_TRAILING,
+#     TCD,
 #     N_SAMPLE_METHODS
 # };
 class SampleMethod(IntEnum):
@@ -186,7 +188,9 @@ class SampleMethod(IntEnum):
     IPNDM = 7
     IPNDM_V = 8
     LCM = 9
-    N_SAMPLE_METHODS = 10
+    DDIM_TRAILING = 10
+    TCD = 11
+    N_SAMPLE_METHODS = 12
 
 
 # enum schedule_t {
@@ -210,43 +214,46 @@ class Schedule(IntEnum):
 
 # // same as enum ggml_type
 # enum sd_type_t {
-#     SD_TYPE_F32  = 0,
-#     SD_TYPE_F16  = 1,
-#     SD_TYPE_Q4_0 = 2,
-#     SD_TYPE_Q4_1 = 3,
+#     SD_TYPE_F32     = 0,
+#     SD_TYPE_F16     = 1,
+#     SD_TYPE_Q4_0    = 2,
+#     SD_TYPE_Q4_1    = 3,
 #     // SD_TYPE_Q4_2 = 4, support has been removed
 #     // SD_TYPE_Q4_3 = 5, support has been removed
-#     SD_TYPE_Q5_0     = 6,
-#     SD_TYPE_Q5_1     = 7,
-#     SD_TYPE_Q8_0     = 8,
-#     SD_TYPE_Q8_1     = 9,
-#     SD_TYPE_Q2_K     = 10,
-#     SD_TYPE_Q3_K     = 11,
-#     SD_TYPE_Q4_K     = 12,
-#     SD_TYPE_Q5_K     = 13,
-#     SD_TYPE_Q6_K     = 14,
-#     SD_TYPE_Q8_K     = 15,
-#     SD_TYPE_IQ2_XXS  = 16,
-#     SD_TYPE_IQ2_XS   = 17,
-#     SD_TYPE_IQ3_XXS  = 18,
-#     SD_TYPE_IQ1_S    = 19,
-#     SD_TYPE_IQ4_NL   = 20,
-#     SD_TYPE_IQ3_S    = 21,
-#     SD_TYPE_IQ2_S    = 22,
-#     SD_TYPE_IQ4_XS   = 23,
-#     SD_TYPE_I8       = 24,
-#     SD_TYPE_I16      = 25,
-#     SD_TYPE_I32      = 26,
-#     SD_TYPE_I64      = 27,
-#     SD_TYPE_F64      = 28,
-#     SD_TYPE_IQ1_M    = 29,
-#     SD_TYPE_BF16     = 30,
-#     SD_TYPE_Q4_0_4_4 = 31,
-#     SD_TYPE_Q4_0_4_8 = 32,
-#     SD_TYPE_Q4_0_8_8 = 33,
-#     SD_TYPE_TQ1_0    = 34,
-#     SD_TYPE_TQ2_0    = 35,
-#     SD_TYPE_COUNT,
+#     SD_TYPE_Q5_0    = 6,
+#     SD_TYPE_Q5_1    = 7,
+#     SD_TYPE_Q8_0    = 8,
+#     SD_TYPE_Q8_1    = 9,
+#     SD_TYPE_Q2_K    = 10,
+#     SD_TYPE_Q3_K    = 11,
+#     SD_TYPE_Q4_K    = 12,
+#     SD_TYPE_Q5_K    = 13,
+#     SD_TYPE_Q6_K    = 14,
+#     SD_TYPE_Q8_K    = 15,
+#     SD_TYPE_IQ2_XXS = 16,
+#     SD_TYPE_IQ2_XS  = 17,
+#     SD_TYPE_IQ3_XXS = 18,
+#     SD_TYPE_IQ1_S   = 19,
+#     SD_TYPE_IQ4_NL  = 20,
+#     SD_TYPE_IQ3_S   = 21,
+#     SD_TYPE_IQ2_S   = 22,
+#     SD_TYPE_IQ4_XS  = 23,
+#     SD_TYPE_I8      = 24,
+#     SD_TYPE_I16     = 25,
+#     SD_TYPE_I32     = 26,
+#     SD_TYPE_I64     = 27,
+#     SD_TYPE_F64     = 28,
+#     SD_TYPE_IQ1_M   = 29,
+#     SD_TYPE_BF16    = 30,
+#     // SD_TYPE_Q4_0_4_4 = 31, support has been removed from gguf files
+#     // SD_TYPE_Q4_0_4_8 = 32,
+#     // SD_TYPE_Q4_0_8_8 = 33,
+#     SD_TYPE_TQ1_0   = 34,
+#     SD_TYPE_TQ2_0   = 35,
+#     // SD_TYPE_IQ4_NL_4_4 = 36,
+#     // SD_TYPE_IQ4_NL_4_8 = 37,
+#     // SD_TYPE_IQ4_NL_8_8 = 38,
+#     SD_TYPE_COUNT   = 39,
 # };
 class GGMLType(IntEnum):
     SD_TYPE_F32 = 0
@@ -281,12 +288,15 @@ class GGMLType(IntEnum):
     SD_TYPE_F64 = 28
     SD_TYPE_IQ1_M = 29
     SD_TYPE_BF16 = 30
-    SD_TYPE_Q4_0_4_4 = 31
-    SD_TYPE_Q4_0_4_8 = 32
-    SD_TYPE_Q4_0_8_8 = 33
+    # SD_TYPE_Q4_0_4_4 = 31 # support has been removed from gguf files
+    # SD_TYPE_Q4_0_4_8 = 32
+    # SD_TYPE_Q4_0_8_8 = 33
     SD_TYPE_TQ1_0 = 34
     SD_TYPE_TQ2_0 = 35
-    SD_TYPE_COUNT = 36
+    # SD_TYPE_IQ4_NL_4_4 = 36,
+    # SD_TYPE_IQ4_NL_4_8 = 37,
+    # SD_TYPE_IQ4_NL_8_8 = 38,
+    SD_TYPE_COUNT = 39
 
 
 # ==================================
@@ -387,7 +397,7 @@ class sd_image_t(ctypes.Structure):
 # ------------ txt2img ------------
 
 
-# SD_API sd_image_t* txt2img(sd_ctx_t* sd_ctx, const char* prompt, const char* negative_prompt, int clip_skip, float cfg_scale, float guidance, int width, int height, enum sample_method_t sample_method, int sample_steps, int64_t seed, int batch_count, const sd_image_t* control_cond, float control_strength, float style_strength, bool normalize_input, const char* input_id_images_path, int* skip_layers, size_t skip_layers_count, float slg_scale, float skip_layer_start, float skip_layer_end);
+# SD_API sd_image_t* txt2img(sd_ctx_t* sd_ctx, const char* prompt, const char* negative_prompt, int clip_skip, float cfg_scale, float guidance, float eta, int width, int height, enum sample_method_t sample_method, int sample_steps, int64_t seed, int batch_count, const sd_image_t* control_cond, float control_strength, float style_strength, bool normalize_input, const char* input_id_images_path, int* skip_layers, size_t skip_layers_count, float slg_scale, float skip_layer_start, float skip_layer_end);
 @ctypes_function(
     "txt2img",
     [
@@ -397,6 +407,7 @@ class sd_image_t(ctypes.Structure):
         ctypes.c_int,  # clip_skip
         ctypes.c_float,  # cfg_scale
         ctypes.c_float,  # guidance
+        ctypes.c_float,  # eta
         ctypes.c_int,  # width
         ctypes.c_int,  # height
         ctypes.c_int,  # sample_method
@@ -423,6 +434,7 @@ def txt2img(
     clip_skip: int,
     cfg_scale: float,
     guidance: float,
+    eta: float,
     width: int,
     height: int,
     sample_method: int,  # SampleMethod
@@ -446,18 +458,19 @@ def txt2img(
 # ------------ img2img ------------
 
 
-# SD_API sd_image_t* img2img(sd_ctx_t* sd_ctx, sd_image_t init_image, sd_image_t mask_image, const char* prompt, const char* negative_prompt, int clip_skip, float cfg_scale, float guidance, int width, int height, enum sample_method_t sample_method, int sample_steps, float strength, int64_t seed, int batch_count, const sd_image_t* control_cond, float control_strength, float style_strength, bool normalize_input, const char* input_id_images_path, int* skip_layers, size_t skip_layers_count, float slg_scale, float skip_layer_start, float skip_layer_end);
+# SD_API sd_image_t* img2img(sd_ctx_t* sd_ctx, sd_image_t init_image, sd_image_t mask_image, const char* prompt, const char* negative_prompt, int clip_skip, float cfg_scale, float guidance, float eta, int width, int height, enum sample_method_t sample_method, int sample_steps, float strength, int64_t seed, int batch_count, const sd_image_t* control_cond, float control_strength, float style_strength, bool normalize_input, const char* input_id_images_path, int* skip_layers, size_t skip_layers_count, float slg_scale, float skip_layer_start, float skip_layer_end);
 @ctypes_function(
     "img2img",
     [
         sd_ctx_t_p_ctypes,  # sd_ctx
         sd_image_t,  # init_image
-        sd_image_t, # mask_image
+        sd_image_t,  # mask_image
         ctypes.c_char_p,  # prompt
         ctypes.c_char_p,  # negative_prompt
         ctypes.c_int,  # clip_skip
         ctypes.c_float,  # cfg_scale
         ctypes.c_float,  # guidance
+        ctypes.c_float,  # eta
         ctypes.c_int,  # width
         ctypes.c_int,  # height
         ctypes.c_int,  # sample_method
@@ -487,6 +500,7 @@ def img2img(
     clip_skip: int,
     cfg_scale: float,
     guidance: float,
+    eta: float,
     width: int,
     height: int,
     sample_method: int,  # SampleMethod

tests/test_txt2img.py

@@ -4,7 +4,8 @@
 from stable_diffusion_cpp import StableDiffusion
 
 # MODEL_PATH = "C:\\stable-diffusion\\turbovisionxlSuperFastXLBasedOnNew_tvxlV431Bakedvae.q8_0.gguf"  # GGUF model wont work for LORAs (GGML_ASSERT error)
-MODEL_PATH = "C:\\stable-diffusion\\turbovisionxlSuperFastXLBasedOnNew_tvxlV431Bakedvae.safetensors"
+# MODEL_PATH = "C:\\stable-diffusion\\turbovisionxlSuperFastXLBasedOnNew_tvxlV431Bakedvae.safetensors"
+MODEL_PATH = "C:\\stable-diffusion\\catCitronAnimeTreasure_v10.safetensors"
 
 LORA_DIR = "C:\\stable-diffusion\\loras"
 
@@ -20,8 +21,10 @@ def callback(step: int, steps: int, time: float):
 
 try:
     prompts = [
-        {"add": "_lora", "prompt": "a lovely cat <lora:pixel_art:1>"},  # With LORA
-        {"add": "", "prompt": "a lovely cat"},  # Without LORA
+        # {"add": "_lora", "prompt": "a lovely cat <lora:realism_lora:1>"},  # With LORA
+        # {"add": "", "prompt": "a lovely cat"},  # Without LORA
+        {"add": "_lora", "prompt": "a cute cat glass statue <lora:glass_statue_v1:1>"},  # With LORA
+        {"add": "2", "prompt": "a cute cat glass statue"},  # Without LORA
     ]
 
     OUTPUT_DIR = "tests/outputs"

tests/tests.md

@@ -1,5 +1,5 @@
 ```bash
-python tests\test_controlnet.py; python tests\test_convert_model.py; python tests\test_flux.py; python tests\test_img2img.py; python tests\test_preprocess_canny.py; python tests\test_system_info.py; python tests\test_txt2img.py; python tests\test_upscale.py; python tests\test_photomaker.py; python tests\test_inpainting.py
+python tests\test_txt2img.py; python tests\test_controlnet.py; python tests\test_convert_model.py; python tests\test_flux.py; python tests\test_img2img.py; python tests\test_preprocess_canny.py; python tests\test_system_info.py; python tests\test_upscale.py; python tests\test_photomaker.py; python tests\test_inpainting.py
 ```
 
 ```bash