Update transformers

[Breefield]

Getting:

ImportError: `diffusers` requires transformers >= 4.25.1 to function correctly, but 4.21.3 was found in your environment. You can upgrade it with pip: `pip install transformers --upgrade`

System:

Dockerfile:		 docker-build/Dockerfile
requirements:		 requirements-lin-cuda.txt
volumename:		 invokeai_data
arch:			 x86_64
platform:		 Linux/x86_64
invokeai_tag:		 invokeai:x86_64

[blessedcoolant]

@lstein diffusers released a new patch today that addresses the issue with the versioning on transformers. https://github.com/huggingface/diffusers/releases/tag/v0.10.2

[lstein]

Here's the problem I originally encountered when I try to use diffusers 0.10.0. It occurs when the NSFW checker (which is part of diffusers) loads:

Installing model for NSFW content detection...

A problem occurred during initialization.
The error was: "cannot import name 'CompVisVDenoiser' from 'k_diffusion.external' (/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/k_diffusion/external.py)"
Traceback (most recent call last):
  File "/data/lstein/InvokeAI/scripts/configure_invokeai.py", line 757, in main
    download_safety_checker()
  File "/data/lstein/InvokeAI/scripts/configure_invokeai.py", line 518, in download_safety_checker
    from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
  File "/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/diffusers/__init__.py", line 39, in <module>
    from .pipelines import (
  File "/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/diffusers/pipelines/__init__.py", line 42, in <module>
    from .alt_diffusion import AltDiffusionImg2ImgPipeline, AltDiffusionPipeline
  File "/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/diffusers/pipelines/alt_diffusion/__init__.py", line 33, in <module>
    from .pipeline_alt_diffusion import AltDiffusionPipeline
  File "/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion.py", line 36, in <module>
    from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
  File "/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/diffusers/pipelines/stable_diffusion/__init__.py", line 72, in <module>
    from .pipeline_stable_diffusion_k_diffusion import StableDiffusionKDiffusionPipeline
  File "/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_k_diffusion.py", line 20, in <module>
    from k_diffusion.external import CompVisDenoiser, CompVisVDenoiser
ImportError: cannot import name 'CompVisVDenoiser' from 'k_diffusion.external' (/data/lstein/InvokeAI/.venv/lib/python3.9/site-packages/k_diffusion/external.py)

I was able to fix this by updating the link to Birch-San's repository to pull from HEAD rather than a release tag. I'm going to modify this PR accordingly.