AI-generated art is getting better and better. A new open-source text-to-image ai model has come out called LAION-400M. The image below was generated by that AI program. This is Linux desktop 2030. A strange boxy PC form factor and a monitor so big it needs 2 feet to hold it up. This is amazing. The only problem is the program was used a lot last night and now it is broken. But it could generate photorealistic images.
Look at the tool here, this can generate amazing artwork.
https://huggingface.co/spaces/multimodalart/latentdiffusion.
When it loads I get this Python error.
Downloading: 98%|█████████▊| 6.05G/6.15G [01:18<00:01, 71.7MB/s] Downloading: 99%|█████████▊| 6.06G/6.15G [01:18<00:01, 70.5MB/s] Downloading: 99%|█████████▊| 6.07G/6.15G [01:18<00:01, 73.3MB/s] Downloading: 99%|█████████▉| 6.08G/6.15G [01:18<00:01, 73.8MB/s] Downloading: 99%|█████████▉| 6.09G/6.15G [01:18<00:00, 76.6MB/s] Downloading: 99%|█████████▉| 6.09G/6.15G [01:18<00:00, 75.0MB/s] Downloading: 99%|█████████▉| 6.10G/6.15G [01:18<00:00, 72.4MB/s] Downloading: 99%|█████████▉| 6.11G/6.15G [01:18<00:00, 73.5MB/s] Downloading: 99%|█████████▉| 6.12G/6.15G [01:18<00:00, 74.7MB/s] Downloading: 100%|█████████▉| 6.12G/6.15G [01:19<00:00, 73.6MB/s] Downloading: 100%|█████████▉| 6.13G/6.15G [01:19<00:00, 74.2MB/s] Downloading: 100%|█████████▉| 6.14G/6.15G [01:19<00:00, 72.6MB/s] Downloading: 100%|█████████▉| 6.15G/6.15G [01:19<00:00, 75.4MB/s] Downloading: 100%|██████████| 6.15G/6.15G [01:19<00:00, 77.5MB/s] Loading model from /home/user/.cache/huggingface/hub/e8bab807983d189ffbc03ea923f84ebe9558b3f33943aee8f6c12da5fad85e11.9b73e76dc7071bdcee7b70e04cd611a77122fdd3b0a17dd2db82bb30a1534c0d Traceback (most recent call last): File "app.py", line 95, in <module> model = load_model_from_config(config,model_path_e) File "app.py", line 31, in load_model_from_config pl_sd = torch.load(ckpt, map_location="cuda") File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 712, in load return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args) File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 1046, in _load result = unpickler.load() File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 1016, in persistent_load load_tensor(dtype, nbytes, key, _maybe_decode_ascii(location)) File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 1001, in load_tensor wrap_storage=restore_location(storage, location), File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 970, in restore_location return default_restore_location(storage, map_location) File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 176, in default_restore_location result = fn(storage, location) File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/serialization.py", line 158, in _cuda_deserialize return obj.cuda(device) File "/home/user/.pyenv/versions/3.8.9/lib/python3.8/site-packages/torch/_utils.py", line 79, in _cuda return new_type(self.size()).copy_(self, non_blocking) RuntimeError: CUDA error: out of memory CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. |
I hope this is fixed eventually. This would be a great tool to create interesting artwork and research trends with AI. Here is a link to the AI artwork generator. Maybe the service needs to be restarted regularly, it is a very interesting tool. It uses the LAION-400-Million Open Dataset. This is a 10TB dataset of 256*256 pixel images used to generate the AI artwork. The information about this dataset is here: https://laion.ai/laion-400-open-dataset/. Each image has metadata associated with it. This makes it easy to find the certain image you are looking for. Eventually, computers will be able to generate any high-quality image you desire from text input. Will they be able to generate a movie as well? Imagine an AI written movie that uses AI to create all of the footage and even inserts famous actors with Deepfake technology. That would be incredible. An AI-generated Star Wars movie would be better than the latest movies for sure. At least AI would not kill off all the popular characters and ruin the franchise.
