Merge branch 'main' into benchmarking-overhaul

Author: sayakpaul

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

@@ -113,7 +113,7 @@ image = pipe(
     height=832, 
     width=480
 ).images[0]
-image[0].save("output.png")
+image.save("output.png")
 ```
 
 ## SanaSprintPipeline

docs/source/en/quantization/torchao.md

@@ -91,7 +91,7 @@ The quantization methods supported are as follows:
 
 Some quantization methods are aliases (for example, `int8wo` is the commonly used shorthand for `int8_weight_only`). This allows using the quantization methods described in the torchao docs as-is, while also making it convenient to remember their shorthand notations.
 
-Refer to the official torchao documentation for a better understanding of the available quantization methods and the exhaustive list of configuration options available.
+Refer to the [official torchao documentation](https://docs.pytorch.org/ao/stable/index.html) for a better understanding of the available quantization methods and the exhaustive list of configuration options available.
 
 ## Serializing and Deserializing quantized models
 
@@ -155,5 +155,5 @@ transformer.load_state_dict(state_dict, strict=True, assign=True)
 
 ## Resources
 
-- [TorchAO Quantization API](https://github.com/pytorch/ao/blob/main/torchao/quantization/README.md)
+- [TorchAO Quantization API](https://docs.pytorch.org/ao/stable/index.html)
 - [Diffusers-TorchAO examples](https://github.com/sayakpaul/diffusers-torchao)

examples/cogvideo/train_cogvideox_image_to_video_lora.py

@@ -555,7 +555,7 @@ def _load_dataset_from_local_path(self):
 
         if any(not path.is_file() for path in instance_videos):
             raise ValueError(
-                "Expected '--video_column' to be a path to a file in `--instance_data_root` containing line-separated paths to video data but found atleast one path that is not a valid file."
+                "Expected '--video_column' to be a path to a file in `--instance_data_root` containing line-separated paths to video data but found at least one path that is not a valid file."
             )
 
         return instance_prompts, instance_videos

examples/cogvideo/train_cogvideox_lora.py

@@ -539,7 +539,7 @@ def _load_dataset_from_local_path(self):
 
         if any(not path.is_file() for path in instance_videos):
             raise ValueError(
-                "Expected '--video_column' to be a path to a file in `--instance_data_root` containing line-separated paths to video data but found atleast one path that is not a valid file."
+                "Expected '--video_column' to be a path to a file in `--instance_data_root` containing line-separated paths to video data but found at least one path that is not a valid file."
             )
 
         return instance_prompts, instance_videos

examples/research_projects/multi_subject_dreambooth_inpainting/README.md

@@ -73,7 +73,7 @@ accelerate launch train_multi_subject_dreambooth_inpaint.py \
 
 ## 3. Results
 
-A [![Weights & Biases](https://img.shields.io/badge/Weights%20&%20Biases-Report-blue)](https://wandb.ai/gzguevara/uncategorized/reports/Multi-Subject-Dreambooth-for-Inpainting--Vmlldzo2MzY5NDQ4?accessToken=y0nya2d7baguhbryxaikbfr1203amvn1jsmyl07vk122mrs7tnph037u1nqgse8t) is provided showing the training progress by every 50 steps. Note, the reported  weights & baises run was performed on a A100 GPU with the following stetting:
+A [![Weights & Biases](https://img.shields.io/badge/Weights%20&%20Biases-Report-blue)](https://wandb.ai/gzguevara/uncategorized/reports/Multi-Subject-Dreambooth-for-Inpainting--Vmlldzo2MzY5NDQ4?accessToken=y0nya2d7baguhbryxaikbfr1203amvn1jsmyl07vk122mrs7tnph037u1nqgse8t) is provided showing the training progress by every 50 steps. Note, the reported weights & biases run was performed on a A100 GPU with the following stetting:
 
 ```bash
 accelerate launch train_multi_subject_dreambooth_inpaint.py \

src/diffusers/hooks/faster_cache.py

@@ -146,7 +146,7 @@ class FasterCacheConfig:
     alpha_low_frequency: float = 1.1
     alpha_high_frequency: float = 1.1
 
-    # n as described in CFG-Cache explanation in the paper - dependant on the model
+    # n as described in CFG-Cache explanation in the paper - dependent on the model
     unconditional_batch_skip_range: int = 5
     unconditional_batch_timestep_skip_range: Tuple[int, int] = (-1, 641)
 

src/diffusers/hooks/hooks.py

@@ -45,7 +45,7 @@ def initialize_hook(self, module: torch.nn.Module) -> torch.nn.Module:
 
     def deinitalize_hook(self, module: torch.nn.Module) -> torch.nn.Module:
         r"""
-        Hook that is executed when a model is deinitalized.
+        Hook that is executed when a model is deinitialized.
 
         Args:
             module (`torch.nn.Module`):

src/diffusers/hooks/layerwise_casting.py

@@ -62,7 +62,7 @@ def initialize_hook(self, module: torch.nn.Module):
 
     def deinitalize_hook(self, module: torch.nn.Module):
         raise NotImplementedError(
-            "LayerwiseCastingHook does not support deinitalization. A model once enabled with layerwise casting will "
+            "LayerwiseCastingHook does not support deinitialization. A model once enabled with layerwise casting will "
             "have casted its weights to a lower precision dtype for storage. Casting this back to the original dtype "
             "will lead to precision loss, which might have an impact on the model's generation quality. The model should "
             "be re-initialized and loaded in the original dtype."

src/diffusers/loaders/peft.py

@@ -251,7 +251,7 @@ def load_lora_adapter(
 
             rank = {}
             for key, val in state_dict.items():
-                # Cannot figure out rank from lora layers that don't have atleast 2 dimensions.
+                # Cannot figure out rank from lora layers that don't have at least 2 dimensions.
                 # Bias layers in LoRA only have a single dimension
                 if "lora_B" in key and val.ndim > 1:
                     # Check out https://github.com/huggingface/peft/pull/2419 for the `^` symbol.

src/diffusers/models/autoencoders/autoencoder_kl.py

@@ -63,8 +63,8 @@ class AutoencoderKL(ModelMixin, ConfigMixin, FromOriginalModelMixin, PeftAdapter
             Synthesis with Latent Diffusion Models](https://huggingface.co/papers/2112.10752) paper.
         force_upcast (`bool`, *optional*, default to `True`):
             If enabled it will force the VAE to run in float32 for high image resolution pipelines, such as SD-XL. VAE
-            can be fine-tuned / trained to a lower range without loosing too much precision in which case
-            `force_upcast` can be set to `False` - see: https://huggingface.co/madebyollin/sdxl-vae-fp16-fix
+            can be fine-tuned / trained to a lower range without losing too much precision in which case `force_upcast`
+            can be set to `False` - see: https://huggingface.co/madebyollin/sdxl-vae-fp16-fix
         mid_block_add_attention (`bool`, *optional*, default to `True`):
             If enabled, the mid_block of the Encoder and Decoder will have attention blocks. If set to false, the
             mid_block will only have resnet blocks