[Deterministic torch randn] Allow tensors to be generated on CPU (#1902)

* [Deterministic torch randn] Allow tensors to be generated on CPU

* fix more

* up

* fix more

* up

* Update src/diffusers/utils/torch_utils.py
Co-authored-by: Anton Lozhkov <anton@huggingface.co>

* Apply suggestions from code review

* up

* up

* Apply suggestions from code review
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
Co-authored-by: Anton Lozhkov <anton@huggingface.co>
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

examples/research_projects/intel_opts/textual_inversion/textual_inversion_bf16.py

@@ -336,7 +336,7 @@ class TextualInversionDataset(Dataset):
 
         if self.center_crop:
             crop = min(img.shape[0], img.shape[1])
-            h, w, = (
+            (h, w,) = (
                 img.shape[0],
                 img.shape[1],
             )

examples/textual_inversion/textual_inversion.py

@@ -381,7 +381,7 @@ class TextualInversionDataset(Dataset):
 
         if self.center_crop:
             crop = min(img.shape[0], img.shape[1])
-            h, w, = (
+            (h, w,) = (
                 img.shape[0],
                 img.shape[1],
             )

examples/textual_inversion/textual_inversion_flax.py

@@ -306,7 +306,7 @@ class TextualInversionDataset(Dataset):
 
         if self.center_crop:
             crop = min(img.shape[0], img.shape[1])
-            h, w, = (
+            (h, w,) = (
                 img.shape[0],
                 img.shape[1],
             )

scripts/convert_kakao_brain_unclip_to_diffusers.py

@@ -564,6 +564,7 @@ def super_res_unet_last_step_original_checkpoint_to_diffusers_checkpoint(model,
 
 # unet utils
 
+
 # <original>.time_embed -> <diffusers>.time_embedding
 def unet_time_embeddings(checkpoint, original_unet_prefix):
     diffusers_checkpoint = {}

src/diffusers/models/modeling_flax_pytorch_utils.py

@@ -37,6 +37,7 @@ def rename_key(key):
 # PyTorch => Flax #
 #####################
 
+
 # Adapted from https://github.com/huggingface/transformers/blob/c603c80f46881ae18b2ca50770ef65fa4033eacd/src/transformers/modeling_flax_pytorch_utils.py#L69
 # and https://github.com/patil-suraj/stable-diffusion-jax/blob/main/stable_diffusion_jax/convert_diffusers_to_jax.py
 def rename_key_and_reshape_tensor(pt_tuple_key, pt_tensor, random_flax_state_dict):

src/diffusers/pipelines/unclip/pipeline_unclip.py

@@ -24,7 +24,7 @@ from transformers.models.clip.modeling_clip import CLIPTextModelOutput
 from ...models import PriorTransformer, UNet2DConditionModel, UNet2DModel
 from ...pipelines import DiffusionPipeline, ImagePipelineOutput
 from ...schedulers import UnCLIPScheduler
-from ...utils import is_accelerate_available, logging
+from ...utils import is_accelerate_available, logging, torch_randn
 from .text_proj import UnCLIPTextProjModel
 
 
@@ -105,11 +105,7 @@ class UnCLIPPipeline(DiffusionPipeline):
 
     def prepare_latents(self, shape, dtype, device, generator, latents, scheduler):
         if latents is None:
-            if device.type == "mps":
-                # randn does not work reproducibly on mps
-                latents = torch.randn(shape, generator=generator, device="cpu", dtype=dtype).to(device)
-            else:
-                latents = torch.randn(shape, generator=generator, device=device, dtype=dtype)
+            latents = torch_randn(shape, generator=generator, device=device, dtype=dtype)
         else:
             if latents.shape != shape:
                 raise ValueError(f"Unexpected latents shape, got {latents.shape}, expected {shape}")

src/diffusers/pipelines/unclip/pipeline_unclip_image_variation.py

@@ -29,7 +29,7 @@ from transformers import (
 from ...models import UNet2DConditionModel, UNet2DModel
 from ...pipelines import DiffusionPipeline, ImagePipelineOutput
 from ...schedulers import UnCLIPScheduler
-from ...utils import is_accelerate_available, logging
+from ...utils import is_accelerate_available, logging, torch_randn
 from .text_proj import UnCLIPTextProjModel
 
 
@@ -113,11 +113,7 @@ class UnCLIPImageVariationPipeline(DiffusionPipeline):
     # Copied from diffusers.pipelines.unclip.pipeline_unclip.UnCLIPPipeline.prepare_latents
     def prepare_latents(self, shape, dtype, device, generator, latents, scheduler):
         if latents is None:
-            if device.type == "mps":
-                # randn does not work reproducibly on mps
-                latents = torch.randn(shape, generator=generator, device="cpu", dtype=dtype).to(device)
-            else:
-                latents = torch.randn(shape, generator=generator, device=device, dtype=dtype)
+            latents = torch_randn(shape, generator=generator, device=device, dtype=dtype)
         else:
             if latents.shape != shape:
                 raise ValueError(f"Unexpected latents shape, got {latents.shape}, expected {shape}")

src/diffusers/schedulers/scheduling_unclip.py

@@ -20,7 +20,7 @@ import numpy as np
 import torch
 
 from ..configuration_utils import ConfigMixin, register_to_config
-from ..utils import BaseOutput
+from ..utils import BaseOutput, torch_randn
 from .scheduling_utils import SchedulerMixin
 
 
@@ -273,15 +273,9 @@ class UnCLIPScheduler(SchedulerMixin, ConfigMixin):
         # 6. Add noise
         variance = 0
         if t > 0:
-            device = model_output.device
-            if device.type == "mps":
-                # randn does not work reproducibly on mps
-                variance_noise = torch.randn(model_output.shape, dtype=model_output.dtype, generator=generator)
-                variance_noise = variance_noise.to(device)
-            else:
-                variance_noise = torch.randn(
-                    model_output.shape, generator=generator, device=device, dtype=model_output.dtype
-                )
+            variance_noise = torch_randn(
+                model_output.shape, dtype=model_output.dtype, generator=generator, device=model_output.device
+            )
 
             variance = self._get_variance(
                 t,

src/diffusers/utils/__init__.py

@@ -64,6 +64,7 @@ from .import_utils import (
 from .logging import get_logger
 from .outputs import BaseOutput
 from .pil_utils import PIL_INTERPOLATION
+from .torch_utils import torch_randn
 
 
 if is_torch_available():

src/diffusers/utils/torch_utils.py

+# Copyright 2022 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+PyTorch utilities: Utilities related to PyTorch
+"""
+from typing import List, Optional, Tuple, Union
+
+from . import logging
+from .import_utils import is_torch_available
+
+
+if is_torch_available():
+    import torch
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+def torch_randn(
+    shape: Union[Tuple, List],
+    generator: Optional[Union[List["torch.Generator"], "torch.Generator"]] = None,
+    device: Optional["torch.device"] = None,
+    dtype: Optional["torch.dtype"] = None,
+):
+    """This is a helper function that allows to create random tensors on the desired `device` with the desired `dtype`. When
+    passing a list of generators one can seed each batched size individually. If CPU generators are passed the tensor
+    will always be created on CPU.
+    """
+    # device on which tensor is created defaults to device
+    rand_device = device
+    batch_size = shape[0]
+
+    if generator is not None:
+        if generator.device != device and generator.device.type == "cpu":
+            rand_device = "cpu"
+            if device != "mps":
+                logger.info(
+                    f"The passed generator was created on 'cpu' even though a tensor on {device} was expected."
+                    f" Tensors will be created on 'cpu' and then moved to {device}. Note that one can probably"
+                    f" slighly speed up this function by passing a generator that was created on the {device} device."
+                )
+        elif generator.device.type != device.type and generator.device.type == "cuda":
+            raise ValueError(f"Cannot generate a {device} tensor from a generator of type {generator.device.type}.")
+
+    if isinstance(generator, list):
+        shape = (1,) + shape[1:]
+        latents = [
+            torch.randn(shape, generator=generator[i], device=rand_device, dtype=dtype) for i in range(batch_size)
+        ]
+        latents = torch.cat(latents, dim=0).to(device)
+    else:
+        latents = torch.randn(shape, generator=generator, device=rand_device, dtype=dtype).to(device)
+
+    return latents

tests/pipelines/unclip/test_unclip.py

@@ -382,7 +382,7 @@ class UnCLIPPipelineIntegrationTests(unittest.TestCase):
         pipeline = pipeline.to(torch_device)
         pipeline.set_progress_bar_config(disable=None)
 
-        generator = torch.Generator(device=torch_device).manual_seed(0)
+        generator = torch.Generator(device="cpu").manual_seed(0)
         output = pipeline(
             "horse",
             num_images_per_prompt=1,

tests/pipelines/unclip/test_unclip_image_variation.py

@@ -480,7 +480,7 @@ class UnCLIPImageVariationPipelineIntegrationTests(unittest.TestCase):
         pipeline.set_progress_bar_config(disable=None)
         pipeline.enable_sequential_cpu_offload()
 
-        generator = torch.Generator(device=torch_device).manual_seed(0)
+        generator = torch.Generator(device="cpu").manual_seed(0)
         output = pipeline(
             input_image,
             num_images_per_prompt=1,

utils/custom_init_isort.py

@@ -96,6 +96,7 @@ def ignore_underscore(key):
 
 def sort_objects(objects, key=None):
     "Sort a list of `objects` following the rules of isort. `key` optionally maps an object to a str."
+
     # If no key is provided, we use a noop.
     def noop(x):
         return x
@@ -117,6 +118,7 @@ def sort_objects_in_import(import_statement):
     """
     Return the same `import_statement` but with objects properly sorted.
     """
+
     # This inner function sort imports between [ ].
     def _replace(match):
         imports = match.groups()[0]