Erno Hopearuoho

Is there an easy way to get the width of the vertical axis units? ```python import pandas_datareader.data as web import datetime import requests_cache import os from time import sleep from...

@Gonzih thanks! That does indeed solve the issue with gradient checkpointing and actually gets really close to being able to train with AMD gpus. Dreambooth example actually starts this time,...

408x408 with https://github.com/ShivamShrirao/diffusers/tree/main/examples/dreambooth using `--output_dir=$OUTPUT_DIR --with_prior_preservation --prior_loss_weight=1.0 --train_batch_size=1 --gradient_accumulation_steps=1 --learning_rate=5e-6 --lr_scheduler="constant" --lr_warmup_steps=0 --num_class_images=1500 --max_train_steps=800 --gradient_checkpointing --mixed_precision="fp16" --sample_batch_size=4` 1.78it/s.

> Probably the latter, assuming bitsandbytes isn't doing anything weird. In theory one might be able to just compile a rocm version with hipcc, but the python code has some...

Duplicate of this issue that's gaining some momentum: https://github.com/huggingface/diffusers/issues/672 (should eventually at least solve the initial problem this ticket was about)

+1 Would be great to get this working on AMD hardware if there's no hard technical limitation blocking compatibility. Newer Radeon cards often have quite high VRAM and work really...

Thanks for the answer, the new additions to the documentation are really helpful 👍 I have a bit busy schedule for now, but I'll try to at least get a...

Can replicate this with HIPBLAS. 768x768 works, 768x1024 works, 1024x1024 fails, 1280x1280 fails. Interesting also how 1024x1024 and 1280x1280 fail in different ways. EDIT: Actually that seems to be only...

Yup, we'll have to add this at some point. It's mostly about figuring out the interface. One way is to have each of them separate, but it's possible that something...

Hmm, so this would be the item rarity (grey, green, blue, purple and orange)?