Merge branch 'main' into main

Author: chaowenguo0

docs/source/en/api/pipelines/sana.md

@@ -59,10 +59,10 @@ Refer to the [Quantization](../../quantization/overview) overview to learn more
 ```py
 import torch
 from diffusers import BitsAndBytesConfig as DiffusersBitsAndBytesConfig, SanaTransformer2DModel, SanaPipeline
-from transformers import BitsAndBytesConfig as BitsAndBytesConfig, AutoModelForCausalLM
+from transformers import BitsAndBytesConfig as BitsAndBytesConfig, AutoModel
 
 quant_config = BitsAndBytesConfig(load_in_8bit=True)
-text_encoder_8bit = AutoModelForCausalLM.from_pretrained(
+text_encoder_8bit = AutoModel.from_pretrained(
     "Efficient-Large-Model/Sana_1600M_1024px_diffusers",
     subfolder="text_encoder",
     quantization_config=quant_config,

tests/models/transformers/test_models_prior.py

@@ -132,7 +132,6 @@ def test_output_pretrained(self):
             output = model(**input)[0]
 
         output_slice = output[0, :5].flatten().cpu()
-        print(output_slice)
 
         # Since the VAE Gaussian prior's generator is seeded on the appropriate device,
         # the expected output slices are not the same for CPU and GPU.
@@ -182,7 +181,6 @@ def test_kandinsky_prior(self, seed, expected_slice):
         assert list(sample.shape) == [1, 768]
 
         output_slice = sample[0, :8].flatten().cpu()
-        print(output_slice)
         expected_output_slice = torch.tensor(expected_slice)
 
         assert torch_all_close(output_slice, expected_output_slice, atol=1e-3)

tests/models/unets/test_models_unet_2d_condition.py

@@ -175,8 +175,7 @@ def create_ip_adapter_plus_state_dict(model):
     )
 
     ip_image_projection_state_dict = OrderedDict()
-    keys = [k for k in image_projection.state_dict() if "layers." in k]
-    print(keys)
+
     for k, v in image_projection.state_dict().items():
         if "2.to" in k:
             k = k.replace("2.to", "0.to")

tests/pipelines/controlnet/test_flax_controlnet.py

@@ -78,7 +78,7 @@ def test_canny(self):
         expected_slice = jnp.array(
             [0.167969, 0.116699, 0.081543, 0.154297, 0.132812, 0.108887, 0.169922, 0.169922, 0.205078]
         )
-        print(f"output_slice: {output_slice}")
+
         assert jnp.abs(output_slice - expected_slice).max() < 1e-2
 
     def test_pose(self):
@@ -123,5 +123,5 @@ def test_pose(self):
         expected_slice = jnp.array(
             [[0.271484, 0.261719, 0.275391, 0.277344, 0.279297, 0.291016, 0.294922, 0.302734, 0.302734]]
         )
-        print(f"output_slice: {output_slice}")
+
         assert jnp.abs(output_slice - expected_slice).max() < 1e-2

tests/pipelines/kandinsky/test_kandinsky_combined.py

@@ -308,8 +308,6 @@ def test_kandinsky(self):
 
         image_from_tuple_slice = image_from_tuple[0, -3:, -3:, -1]
 
-        print(image_from_tuple_slice)
-
         assert image.shape == (1, 64, 64, 3)
 
         expected_slice = np.array([0.0320, 0.0860, 0.4013, 0.0518, 0.2484, 0.5847, 0.4411, 0.2321, 0.4593])

tests/pipelines/ledits_pp/test_ledits_pp_stable_diffusion.py

@@ -146,7 +146,7 @@ def test_ledits_pp_inversion(self):
         )
 
         latent_slice = sd_pipe.init_latents[0, -1, -3:, -3:].to(device)
-        print(latent_slice.flatten())
+
         expected_slice = np.array([-0.9084, -0.0367, 0.2940, 0.0839, 0.6890, 0.2651, -0.7104, 2.1090, -0.7822])
         assert np.abs(latent_slice.flatten() - expected_slice).max() < 1e-3
 
@@ -167,12 +167,12 @@ def test_ledits_pp_inversion_batch(self):
         )
 
         latent_slice = sd_pipe.init_latents[0, -1, -3:, -3:].to(device)
-        print(latent_slice.flatten())
+
         expected_slice = np.array([0.2528, 0.1458, -0.2166, 0.4565, -0.5657, -1.0286, -0.9961, 0.5933, 1.1173])
         assert np.abs(latent_slice.flatten() - expected_slice).max() < 1e-3
 
         latent_slice = sd_pipe.init_latents[1, -1, -3:, -3:].to(device)
-        print(latent_slice.flatten())
+
         expected_slice = np.array([-0.0796, 2.0583, 0.5501, 0.5358, 0.0282, -0.2803, -1.0470, 0.7023, -0.0072])
         assert np.abs(latent_slice.flatten() - expected_slice).max() < 1e-3
 

tests/pipelines/ledits_pp/test_ledits_pp_stable_diffusion_xl.py

@@ -216,14 +216,14 @@ def test_ledits_pp_inversion_batch(self):
         )
 
         latent_slice = sd_pipe.init_latents[0, -1, -3:, -3:].to(device)
-        print(latent_slice.flatten())
+
         expected_slice = np.array([0.2528, 0.1458, -0.2166, 0.4565, -0.5656, -1.0286, -0.9961, 0.5933, 1.1172])
         assert np.abs(latent_slice.flatten() - expected_slice).max() < 1e-3
 
         latent_slice = sd_pipe.init_latents[1, -1, -3:, -3:].to(device)
-        print(latent_slice.flatten())
+
         expected_slice = np.array([-0.0796, 2.0583, 0.5500, 0.5358, 0.0282, -0.2803, -1.0470, 0.7024, -0.0072])
-        print(latent_slice.flatten())
+
         assert np.abs(latent_slice.flatten() - expected_slice).max() < 1e-3
 
     def test_ledits_pp_warmup_steps(self):

tests/pipelines/pag/test_pag_sd.py

@@ -318,7 +318,7 @@ def test_pag_cfg(self):
 
         image_slice = image[0, -3:, -3:, -1].flatten()
         assert image.shape == (1, 512, 512, 3)
-        print(image_slice.flatten())
+
         expected_slice = np.array(
             [0.58251953, 0.5722656, 0.5683594, 0.55029297, 0.52001953, 0.52001953, 0.49951172, 0.45410156, 0.50146484]
         )
@@ -339,7 +339,6 @@ def test_pag_uncond(self):
         expected_slice = np.array(
             [0.5986328, 0.52441406, 0.3972168, 0.4741211, 0.34985352, 0.22705078, 0.4128418, 0.2866211, 0.31713867]
         )
-        print(image_slice.flatten())
         assert (
             np.abs(image_slice.flatten() - expected_slice).max() < 1e-3
         ), f"output is different from expected, {image_slice.flatten()}"

tests/pipelines/pag/test_pag_sd_img2img.py

@@ -255,7 +255,7 @@ def test_pag_cfg(self):
 
         image_slice = image[0, -3:, -3:, -1].flatten()
         assert image.shape == (1, 512, 512, 3)
-        print(image_slice.flatten())
+
         expected_slice = np.array(
             [0.58251953, 0.5722656, 0.5683594, 0.55029297, 0.52001953, 0.52001953, 0.49951172, 0.45410156, 0.50146484]
         )
@@ -276,7 +276,7 @@ def test_pag_uncond(self):
         expected_slice = np.array(
             [0.5986328, 0.52441406, 0.3972168, 0.4741211, 0.34985352, 0.22705078, 0.4128418, 0.2866211, 0.31713867]
         )
-        print(image_slice.flatten())
+
         assert (
             np.abs(image_slice.flatten() - expected_slice).max() < 1e-3
         ), f"output is different from expected, {image_slice.flatten()}"

tests/pipelines/pag/test_pag_sd_inpaint.py

@@ -292,7 +292,7 @@ def test_pag_cfg(self):
 
         image_slice = image[0, -3:, -3:, -1].flatten()
         assert image.shape == (1, 512, 512, 3)
-        print(image_slice.flatten())
+
         expected_slice = np.array(
             [0.38793945, 0.4111328, 0.47924805, 0.39208984, 0.4165039, 0.41674805, 0.37060547, 0.36791992, 0.40625]
         )