From d2ac95fa7b2a8d0bcc5361ee16dba9cbb81ff8b2 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 27 Jan 2023 11:28:12 +0300 Subject: remove the need to place configs near models --- modules/sd_models_config.py | 65 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 65 insertions(+) create mode 100644 modules/sd_models_config.py (limited to 'modules/sd_models_config.py') diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py new file mode 100644 index 00000000..ea773a10 --- /dev/null +++ b/modules/sd_models_config.py @@ -0,0 +1,65 @@ +import re +import os + +from modules import shared, paths + +sd_configs_path = shared.sd_configs_path +sd_repo_configs_path = os.path.join(paths.paths['Stable Diffusion'], "configs", "stable-diffusion") + + +config_default = shared.sd_default_config +config_sd2 = os.path.join(sd_repo_configs_path, "v2-inference.yaml") +config_sd2v = os.path.join(sd_repo_configs_path, "v2-inference-v.yaml") +config_inpainting = os.path.join(sd_configs_path, "v1-inpainting-inference.yaml") +config_instruct_pix2pix = os.path.join(sd_configs_path, "instruct-pix2pix.yaml") +config_alt_diffusion = os.path.join(sd_configs_path, "alt-diffusion-inference.yaml") + +re_parametrization_v = re.compile(r'-v\b') + + +def guess_model_config_from_state_dict(sd, filename): + fn = os.path.basename(filename) + + sd2_cond_proj_weight = sd.get('cond_stage_model.model.transformer.resblocks.0.attn.in_proj_weight', None) + diffusion_model_input = sd.get('model.diffusion_model.input_blocks.0.0.weight', None) + roberta_weight = sd.get('cond_stage_model.roberta.embeddings.word_embeddings.weight', None) + + if sd2_cond_proj_weight is not None and sd2_cond_proj_weight.shape[1] == 1024: + if re.search(re_parametrization_v, fn) or "v2-1_768" in fn: + return config_sd2v + else: + return config_sd2 + + if diffusion_model_input is not None: + if diffusion_model_input.shape[1] == 9: + return config_inpainting + if diffusion_model_input.shape[1] == 8: + return config_instruct_pix2pix + + if roberta_weight is not None: + return config_alt_diffusion + + return config_default + + +def find_checkpoint_config(state_dict, info): + if info is None: + return guess_model_config_from_state_dict(state_dict, "") + + config = find_checkpoint_config_near_filename(info) + if config is not None: + return config + + return guess_model_config_from_state_dict(state_dict, info.filename) + + +def find_checkpoint_config_near_filename(info): + if info is None: + return None + + config = os.path.splitext(info.filename)[0] + ".yaml" + if os.path.exists(config): + return config + + return None + -- cgit v1.2.1 From 6f31d2210c189f8db118e6f95add7ba2a64f0238 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 27 Jan 2023 11:54:19 +0300 Subject: support detecting midas model fix broken api for checkpoint list --- modules/sd_models_config.py | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'modules/sd_models_config.py') diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py index ea773a10..4d1e92e1 100644 --- a/modules/sd_models_config.py +++ b/modules/sd_models_config.py @@ -10,6 +10,7 @@ sd_repo_configs_path = os.path.join(paths.paths['Stable Diffusion'], "configs", config_default = shared.sd_default_config config_sd2 = os.path.join(sd_repo_configs_path, "v2-inference.yaml") config_sd2v = os.path.join(sd_repo_configs_path, "v2-inference-v.yaml") +config_depth_model = os.path.join(sd_repo_configs_path, "v2-midas-inference.yaml") config_inpainting = os.path.join(sd_configs_path, "v1-inpainting-inference.yaml") config_instruct_pix2pix = os.path.join(sd_configs_path, "instruct-pix2pix.yaml") config_alt_diffusion = os.path.join(sd_configs_path, "alt-diffusion-inference.yaml") @@ -22,7 +23,9 @@ def guess_model_config_from_state_dict(sd, filename): sd2_cond_proj_weight = sd.get('cond_stage_model.model.transformer.resblocks.0.attn.in_proj_weight', None) diffusion_model_input = sd.get('model.diffusion_model.input_blocks.0.0.weight', None) - roberta_weight = sd.get('cond_stage_model.roberta.embeddings.word_embeddings.weight', None) + + if sd.get('depth_model.model.pretrained.act_postprocess3.0.project.0.bias', None) is not None: + return config_depth_model if sd2_cond_proj_weight is not None and sd2_cond_proj_weight.shape[1] == 1024: if re.search(re_parametrization_v, fn) or "v2-1_768" in fn: @@ -36,7 +39,7 @@ def guess_model_config_from_state_dict(sd, filename): if diffusion_model_input.shape[1] == 8: return config_instruct_pix2pix - if roberta_weight is not None: + if sd.get('cond_stage_model.roberta.embeddings.word_embeddings.weight', None) is not None: return config_alt_diffusion return config_default -- cgit v1.2.1 From 6b82efd737827bbeef202f04ff5a8faec9b64ef8 Mon Sep 17 00:00:00 2001 From: MrCheeze Date: Fri, 27 Jan 2023 20:06:19 -0500 Subject: add v2-inpainting model detection, and broaden v-model detection to include anything with 768 in the name --- modules/sd_models_config.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'modules/sd_models_config.py') diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py index 4d1e92e1..73854a45 100644 --- a/modules/sd_models_config.py +++ b/modules/sd_models_config.py @@ -10,6 +10,7 @@ sd_repo_configs_path = os.path.join(paths.paths['Stable Diffusion'], "configs", config_default = shared.sd_default_config config_sd2 = os.path.join(sd_repo_configs_path, "v2-inference.yaml") config_sd2v = os.path.join(sd_repo_configs_path, "v2-inference-v.yaml") +config_sd2_inpainting = os.path.join(sd_repo_configs_path, "v2-inpainting-inference.yaml") config_depth_model = os.path.join(sd_repo_configs_path, "v2-midas-inference.yaml") config_inpainting = os.path.join(sd_configs_path, "v1-inpainting-inference.yaml") config_instruct_pix2pix = os.path.join(sd_configs_path, "instruct-pix2pix.yaml") @@ -28,7 +29,9 @@ def guess_model_config_from_state_dict(sd, filename): return config_depth_model if sd2_cond_proj_weight is not None and sd2_cond_proj_weight.shape[1] == 1024: - if re.search(re_parametrization_v, fn) or "v2-1_768" in fn: + if diffusion_model_input.shape[1] == 9: + return config_sd2_inpainting + elif re.search(re_parametrization_v, fn) or "768" in fn: return config_sd2v else: return config_sd2 -- cgit v1.2.1 From 4c52dfe4ac98c53431ecd267d59f27391d3a63e7 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 28 Jan 2023 08:30:17 +0300 Subject: make the detection for -v models less broad --- modules/sd_models_config.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/sd_models_config.py') diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py index 73854a45..00217990 100644 --- a/modules/sd_models_config.py +++ b/modules/sd_models_config.py @@ -31,7 +31,7 @@ def guess_model_config_from_state_dict(sd, filename): if sd2_cond_proj_weight is not None and sd2_cond_proj_weight.shape[1] == 1024: if diffusion_model_input.shape[1] == 9: return config_sd2_inpainting - elif re.search(re_parametrization_v, fn) or "768" in fn: + elif re.search(re_parametrization_v, fn): return config_sd2v else: return config_sd2 -- cgit v1.2.1 From d04e3e921e8ee71442a1f4a1d6e91c05b8238007 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 28 Jan 2023 15:24:29 +0300 Subject: automatically detect v-parameterization for SD2 checkpoints --- modules/sd_models_config.py | 51 ++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 46 insertions(+), 5 deletions(-) (limited to 'modules/sd_models_config.py') diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py index 00217990..91c21700 100644 --- a/modules/sd_models_config.py +++ b/modules/sd_models_config.py @@ -1,7 +1,9 @@ import re import os -from modules import shared, paths +import torch + +from modules import shared, paths, sd_disable_initialization sd_configs_path = shared.sd_configs_path sd_repo_configs_path = os.path.join(paths.paths['Stable Diffusion'], "configs", "stable-diffusion") @@ -16,12 +18,51 @@ config_inpainting = os.path.join(sd_configs_path, "v1-inpainting-inference.yaml" config_instruct_pix2pix = os.path.join(sd_configs_path, "instruct-pix2pix.yaml") config_alt_diffusion = os.path.join(sd_configs_path, "alt-diffusion-inference.yaml") -re_parametrization_v = re.compile(r'-v\b') +def is_using_v_parameterization_for_sd2(state_dict): + """ + Detects whether unet in state_dict is using v-parameterization. Returns True if it is. You're welcome. + """ -def guess_model_config_from_state_dict(sd, filename): - fn = os.path.basename(filename) + import ldm.modules.diffusionmodules.openaimodel + from modules import devices + + device = devices.cpu + + with sd_disable_initialization.DisableInitialization(): + unet = ldm.modules.diffusionmodules.openaimodel.UNetModel( + use_checkpoint=True, + use_fp16=False, + image_size=32, + in_channels=4, + out_channels=4, + model_channels=320, + attention_resolutions=[4, 2, 1], + num_res_blocks=2, + channel_mult=[1, 2, 4, 4], + num_head_channels=64, + use_spatial_transformer=True, + use_linear_in_transformer=True, + transformer_depth=1, + context_dim=1024, + legacy=False + ) + unet.eval() + + with torch.no_grad(): + unet_sd = {k.replace("model.diffusion_model.", ""): v for k, v in state_dict.items() if "model.diffusion_model." in k} + unet.load_state_dict(unet_sd, strict=True) + unet.to(device=device, dtype=torch.float) + test_cond = torch.ones((1, 2, 1024), device=device) * 0.5 + x_test = torch.ones((1, 4, 8, 8), device=device) * 0.5 + + out = (unet(x_test, torch.asarray([999], device=device), context=test_cond) - x_test).mean().item() + + return out < -1 + + +def guess_model_config_from_state_dict(sd, filename): sd2_cond_proj_weight = sd.get('cond_stage_model.model.transformer.resblocks.0.attn.in_proj_weight', None) diffusion_model_input = sd.get('model.diffusion_model.input_blocks.0.0.weight', None) @@ -31,7 +72,7 @@ def guess_model_config_from_state_dict(sd, filename): if sd2_cond_proj_weight is not None and sd2_cond_proj_weight.shape[1] == 1024: if diffusion_model_input.shape[1] == 9: return config_sd2_inpainting - elif re.search(re_parametrization_v, fn): + elif is_using_v_parameterization_for_sd2(sd): return config_sd2v else: return config_sd2 -- cgit v1.2.1