Conversion errors when using 1.4 checkpoint

[rhulse-in]

i am trying to train on the 1.4 checkpoint, downloaded from huggingface, then shared from my gdrive. I am getting conversion errors regardless of whether I use "compatibility mode". Any idea what I'm doing wrong?

Traceback (most recent call last):
File "/content/convert_original_stable_diffusion_to_diffusers.py", line 628, in
checkpoint = torch.load(args.checkpoint_path)["state_dict"]
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 712, in load
return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 1049, in _load
result = unpickler.load()
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 1042, in find_class
return super().find_class(mod_name, name)
ModuleNotFoundError: No module named 'pytorch_lightning'
Conversion error, Insufficient RAM or corrupt CKPT, use a 4.7GB CKPT instead of 7GB
Conversion error, Insufficient RAM or corrupt CKPT, use a 4.7GB CKPT instead of 7GB

---

KeyboardInterrupt Traceback (most recent call last)

in
119 while not os.path.exists('/content/stable-diffusion-v1-5/unet/diffusion_pytorch_model.bin'):
120 print('Conversion error, Insufficient RAM or corrupt CKPT, use a 4.7GB CKPT instead of 7GB')
--> 121 time.sleep(5)
122 else:
123 while os.path.getsize('/content/model.ckpt') < 1810671599:

KeyboardInterrupt:

[TheLastBen]

Try
!pip install pytorch_lightning