I am 3d Artist and my brother is React,html.css,java developer
I am 3d Artist and my brother is React,html.css,java developer
Mmy portfolio
SIs the art workflow based on stable diffusion?
F@Discord Admin I am having this problem, asked you some days ago in youtube comments, can you help me here please? 
FI was using Stable Difffusion normally like you demonstrated in the video, but today I got an error.
OutOfMemoryError Traceback (most recent call last)
<ipython-input-9-9817a4f7d2d0> in <module>
10
11 with autocast("cuda"), torch.inference_mode():
---> 12 images = pipe(
13 prompt,
14 height=height,
13 frames
/usr/local/lib/python3.8/dist-packages/diffusers/models/cross_attention.py in get_attention_scores(self, query, key, attention_mask)
221
222 attention_probs = attention_scores.softmax(dim=-1)
--> 223 attention_probs = attention_probs.to(dtype)
224
225 return attention_probs
OutOfMemoryError: CUDA out of memory. Tried to allocate 2.50 GiB (GPU 0; 14.76 GiB total capacity; 12.21 GiB already allocated; 1.22 GiB free; 12.26 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Now I'm only able to generate images with 256 x 256 px and lower.
Any idea how to fix this and generate images of 512 x 512 again?
Thank you!
OutOfMemoryError Traceback (most recent call last)
<ipython-input-9-9817a4f7d2d0> in <module>
10
11 with autocast("cuda"), torch.inference_mode():
---> 12 images = pipe(
13 prompt,
14 height=height,
13 frames
/usr/local/lib/python3.8/dist-packages/diffusers/models/cross_attention.py in get_attention_scores(self, query, key, attention_mask)
221
222 attention_probs = attention_scores.softmax(dim=-1)
--> 223 attention_probs = attention_probs.to(dtype)
224
225 return attention_probs
OutOfMemoryError: CUDA out of memory. Tried to allocate 2.50 GiB (GPU 0; 14.76 GiB total capacity; 12.21 GiB already allocated; 1.22 GiB free; 12.26 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Now I'm only able to generate images with 256 x 256 px and lower.
Any idea how to fix this and generate images of 512 x 512 again?
Thank you!
Fnot at all atm
Fthis is in google colab?
Fyes@
Fif so what is your batch size
Fyes!
Fshow me settings please
Fok
F!accelerate launch train_dreambooth.py \
--pretrained_model_name_or_path=$MODEL_NAME \
--pretrained_vae_name_or_path="stabilityai/sd-vae-ft-mse" \
--output_dir=$OUTPUT_DIR \
--revision="fp16" \
--with_prior_preservation --prior_loss_weight=1.0 \
--seed=1337 \
--resolution=512 \
--train_batch_size=1 \
--train_text_encoder \
--mixed_precision="fp16" \
--use_8bit_adam \
--gradient_accumulation_steps=1 \
--learning_rate=1e-6 \
--lr_scheduler="constant" \
--lr_warmup_steps=496 \
--num_class_images=744 \
--sample_batch_size=4 \
--max_train_steps=4960 \
--save_interval=10000 \
--save_sample_prompt="felipe" \
--concepts_list="concepts_list.json"
Reduce the
--pretrained_model_name_or_path=$MODEL_NAME \
--pretrained_vae_name_or_path="stabilityai/sd-vae-ft-mse" \
--output_dir=$OUTPUT_DIR \
--revision="fp16" \
--with_prior_preservation --prior_loss_weight=1.0 \
--seed=1337 \
--resolution=512 \
--train_batch_size=1 \
--train_text_encoder \
--mixed_precision="fp16" \
--use_8bit_adam \
--gradient_accumulation_steps=1 \
--learning_rate=1e-6 \
--lr_scheduler="constant" \
--lr_warmup_steps=496 \
--num_class_images=744 \
--sample_batch_size=4 \
--max_train_steps=4960 \
--save_interval=10000 \
--save_sample_prompt="felipe" \
--concepts_list="concepts_list.json"
Reduce the --save_interval to lower than --max_train_steps to save weights from intermediate steps.
--save_sample_prompt can be same as --instance_prompt to generate intermediate samples (saved along with weights in samples directory).
Fok this is training settings
Fyou get error during training?
Fno, I was using normally and suddenly one day it stopped working
Fworking only with 256 x 256 images or lower
Fok
Fshow me
Fwhere you get error
Fwhen you click generate images?
Fok, I'll make a video from scratch
Fjust 1 sec
Fok
Fby the way
Ffirst try
Fnum_samples = 1
Fah, ok!
MFelipe are u brazilian?
Fyep!
MShow
FTá funcionando pra vc ai direitinho?
Fworked with one sample
MEu não cheguei a usar esse colab ainda, utilizo geralmente o do automatic1111 mesmo
AGoogle Colab
Fbut why this time with only one sample?
Fand now it's not working even with 1 sample
GHey all! Trying to setup controlnet with Automatic as per the latest video
Gbut I'm hitting an error I'm not sure how to fix
Gonce I hit generate I get this in CMD
G
GSolved the first part of the problem, but now I have this:
DI can't connect to colab, it writes that I used limits of GPU usage((((
Fwait some hours and it will work again
Fhappened to me too
Fu have 10 samples not 1
Fdid you try with base sd 1.5
GHi! Yeah, I'm using v1-5-pruned.ckpt