From 2ac3f92d4e3b662414971b67bf9ab77b92e50362 Mon Sep 17 00:00:00 2001 From: Patrick von Platen Date: Sun, 2 Jul 2023 14:23:59 +0000 Subject: [PATCH 1/2] add gitignore --- .gitignore | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/.gitignore b/.gitignore index 0c70c5e90..eb43a0340 100644 --- a/.gitignore +++ b/.gitignore @@ -4,4 +4,5 @@ *.egg-info build /outputs -/checkpoints \ No newline at end of file +/checkpoints +__pycache__ From dea0f4c57b0053eb7ff547da1acdd21188154353 Mon Sep 17 00:00:00 2001 From: Patrick von Platen Date: Mon, 3 Jul 2023 12:41:01 +0000 Subject: [PATCH 2/2] finish test setup --- scripts/demo/sampling.py | 6 ++++-- scripts/demo/streamlit_helpers.py | 5 ++++- 2 files changed, 8 insertions(+), 3 deletions(-) diff --git a/scripts/demo/sampling.py b/scripts/demo/sampling.py index 0aa1a424b..2aac31f80 100644 --- a/scripts/demo/sampling.py +++ b/scripts/demo/sampling.py @@ -126,9 +126,10 @@ def run_txt2img( use_identity_guider=not version_dict["is_guided"] ) - num_samples = num_rows * num_cols + # num_samples = num_rows * num_cols + num_samples = 1 - if st.button("Sample"): + if True: st.write(f"**Model I:** {version}") out = do_sample( state["model"], @@ -307,6 +308,7 @@ def apply_refiner( ) else: raise ValueError(f"unknown mode {mode}") + if isinstance(out, (tuple, list)): samples, samples_z = out else: diff --git a/scripts/demo/streamlit_helpers.py b/scripts/demo/streamlit_helpers.py index ddc9c6ba5..39f326194 100644 --- a/scripts/demo/streamlit_helpers.py +++ b/scripts/demo/streamlit_helpers.py @@ -13,6 +13,7 @@ from torchvision import transforms from torchvision.utils import make_grid from safetensors.torch import load_file as load_safetensors +from pytorch_lightning import seed_everything from sgm.modules.diffusionmodules.sampling import ( EulerEDMSampler, @@ -206,6 +207,7 @@ def init_save_locally(_dir, init_value: bool = False): else: save_path = None + return True, "/home/patrick/images/sgm_test" return save_locally, save_path @@ -513,7 +515,8 @@ def do_sample( additional_model_inputs[k] = batch[k] shape = (math.prod(num_samples), C, H // F, W // F) - randn = torch.randn(shape).to("cuda") + seed_everything(0) + randn = torch.randn(shape, device="cuda") def denoiser(input, sigma, c): return model.denoiser(