Open gozi1123 opened 2 years ago
Not sure if you are still having this issue or not but here is what worked for me.
You need to change your end of line (EOL) from CRLF to LF in notepad ++. Here is a quote from google on how to do that: In Notepad++ go to the View > Show Symbol menu and select Show End of Line. Once you select View > Show Symbol > Show End of Line you can see the CR LF characters visually. You can then use the menu item Edit > EOL Conversion and select Unix (LF).
As soon as I did that and saved the file it was able to download the model for me.
Also my pretrained model path looks like this:
--pretrained_model_name_or_path=$MODEL_NAME \
and at the top of the file
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
Hope that helps.
Describe the bug
When running the training script via terminal (ubuntu 18.04) I get an error
train_booth.sh: line 5: --pretrained_model_name_or_path=CompVis/stable-diffusion-v1-4: No such file or directory
This doesn't happen when running the script in Pycharm
Reproduction
accelerate launch train_dreambooth.py \ --pretrained_model_name_or_path="CompVis/stable-diffusion-v1-4" \ --instance_data_dir="PATH" \ --class_data_dir="woman" \ --output_dir="PATH" \ --with_prior_preservation --prior_loss_weight=1.0 \ --instance_prompt="propt" \ --class_prompt="photo of a woman" \ --seed=1337 \ --resolution=512 \ --train_batch_size=1 \ --mixed_precision="fp16" \ --use_8bit_adam \ --gradient_accumulation_steps=1 \ --learning_rate=5e-6 \ --lr_scheduler="constant" \ --lr_warmup_steps=0 \ --num_class_images=50 \ --sample_batch_size=4 \ --max_train_steps=1000 \ --gradient_checkpointing
Logs
System Info
diffusers
version: 0.6.0.dev0