Instructions to use MLXBits/10eros-experimental-mlx-q8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use MLXBits/10eros-experimental-mlx-q8 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("MLXBits/10eros-experimental-mlx-q8", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - MLX
How to use MLXBits/10eros-experimental-mlx-q8 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download MLXBits/10eros-experimental-mlx-q8 --local-dir 10eros-experimental-mlx-q8
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
curious for a 4bit just in case..
I saw your sulphur mlx quant goes down to q4. curious to try eros too and see if either does anything on my m4 / 24gb..
Have you used that one? I honestly didn't think a 24GB Mac could handle this model at all. My 32GB Mac barely handled the Q4 quant of Sulphur, unless I was doing low-res or very short clips <10 sec.
Have you used that one? I honestly didn't think a 24GB Mac could handle this model at all. My 32GB Mac barely handled the Q4 quant of Sulphur, unless I was doing low-res or very short clips <10 sec.
Haven’t yet, but I am used to low res / short clips regardless from the WAN days. Mac video gen is painful in general. Candidly I’m trying to figure how to run it, ha ha, I am used to swarmUI frontend over comfy (I run my Mac mini headless and comfy is too hard to manipulate via a Remote Desktop on my phone) so working through your directions from sulphur with Gemini assistance tonight
Give that one a try first, if you could. I just don't want to sink the time into a quant for a machine I'm pretty sure can't handle it. I was doing 10 sec clips at 384x480, on a 32GB Macbook Air, and that was pretty much the ceiling before performance and heat caused it to slow down a lot.