Model reference · open weights
OpenDalleV1.1 is an open-weight image model from dataautogpt3. OpenDalleV1.1 (FP32) weighs 6.9 GB; the smallest configuration that runs it is RTX 3060 12 GB.
What it is
| Released by | dataautogpt3 |
|---|---|
| Type | Image models |
| Task | Text→image |
| Parameters (lead) | 2.6B |
| Runs with | diffusers |
| Released | 2023-12-21 |
| Popularity | 1k downloads / month |
| Weights | 6.9 GB (OpenDalleV1.1 (FP32), file size) |
| Licence | Non-commercial |
What it runs on
Weights 6.9 GB (file size) · its biggest part 5.1 GB · working memory for one 1024×1024 image about 5.0 GB · overhead about 537 MB.
| Card | One 1024×1024 image | Counted memory |
|---|---|---|
| RTX 3060 12 GB | tight (encoders offloaded) | 11.6 GB |
| RTX 4060 Ti 16 GB | fits | 15.4 GB |
| RTX 3090 24 GB | fits | 23.4 GB |
| RTX 4090 24 GB | fits | 23.4 GB |
| RTX 5090 32 GB | fits | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
Estimates, not measurements: the weights are the build's file size; one 1024×1024 image needs about 5 GB of working memory (larger images more); "encoders offloaded" means only the biggest part is on the card at once — diffusers' model offload, or ComfyUI unloading the text encoder. diffusers can also place a pipeline's parts on separate cards (device_map) — not estimated here. Counted memory is 92 % of what CUDA reports for the card.
From the model card
my newest model and best current model is located here: https://huggingface.co/dataautogpt3/ProteusV0.2
OpenDalle v1.1 on Hugging Face - It's Here!
Realism & Style: improved
We're talking about a major glow-up in the realism and style department. Expect images that not only hit the bullseye with your prompts but also bring that extra zing of artistic flair. It's like your prompts went to art school!
Prompt Loyalty: Our Heartbeat
The soul of OpenDalle? Sticking to your prompts like glue. v1.1 takes your words and turns them into visual masterpieces that are just what you pictured – maybe even better.
Where We Stand: The Cool Middle Kid
Here's the scoop: OpenDalle v1.1 is proudly strutting a notch above SDXL. While DALLE-3 is still the big cheese, we're hot on its heels. Think of us as the cool, savvy middle sibling, rocking both brains and beauty.
Use these settings for the best results with OpenDalle v1.1:
CFG Scale: Use a CFG scale of 8 to 7
Steps: 60 to 70 steps for more detail, 35 steps for faster results.
Sampler: DPM2
Scheduler: Normal or Karras
from diffusers import AutoPipelineForText2Image
import torch
pipeline = AutoPipelineForText2Image.from_pretrained('dataautogpt3/OpenDalleV1.1', torch_dtype=torch.float16).to('cuda')
image = pipeline('black fluffy gorgeous dangerous cat animal creature, large orange eyes, big fluffy ears, piercing gaze, full moon, dark ambiance, best quality, extremely detailed').images[0]
Non-Commercial Personal Use License Agreement
For dataautogpt3/OpenDalleV1.1
This Non-Commercial Personal Use License Agreement ("Agreement") is between Alexander Izquierdo ("Licensor") and the individual or entity ("Licensee") using the Stable Diffusion model with unique merging method and tuning ("Model") hosted on the Hugging Face repository named OpenDalleV1.1.
a. Licensor hereby grants to Licensee a non-exclusive, non-transferable, non-sublicensable license to use the Model for personal, non-commercial purposes.
b. "Personal, non-commercial purposes" are defined as use that does not involve any form of compensation or monetary gain. This includes, but is not limited to, academic research, educational use, and hobbyist projects.
c. The Licensee is permitted to modify, merge, and use the Model for personal projects, provided that such use adheres to the terms of this Agreement.
a. The Licensor explicitly retains all rights, title, and interest in and to the unique merging method used in the Model. This merging method is the proprietary creation and intellectual property of the Licensor.
b. The Licensee shall not claim ownership, reverse engineer, or attempt to recreate the merging method for any purpose.
c. The Licensor retains all rights, title, and interest in and to the Model, including any modifications or improvements made by the Licensee.
d. The Licensee agrees to attribute the Licensor in any academic or public display of the Model or derivative works.
a. The Licensee shall not use the Model or the merging method for any commercial purposes.
b. The Licensee shall not distribute, sublicense, lease, or lend the Model or the merging method to any third party.
c. The Licensee shall not publicly display, perform, or communicate the Model, the merging method, or any derivative works thereof without the prior written consent of the Licensor.
This Agreement will terminate automatically if the Licensee breaches any of its terms and conditions.
The Model and the merging method are provided "as is," and the Licensor makes no warranties, express or implied, regarding their performance, reliability, or suitability for any purpose.
The Licensor shall not be liable for any damages arising out of or related to the use or inability to use the Model or the merging method.
a. This Agreement constitutes the entire agreement between the parties and supersedes all prior agreements and understandings, whether written or oral, relating to its subject matter.
b. Any amendment to this Agreement must be in writing and signed by both parties.
c. This Agreement shall be governed by the laws of Maryland.
IN WITNESS WHEREOF, the parties have executed this Agreement as of the Effective Date.
Quoted from the model card on Hugging Face — the full card is behind the Hugging Face link above.
Running it yourself
Rent a machine by the hour — ComfyUI is installed on it. Open ComfyUI through the tunnel: its default workflow loads a checkpoint — choose this model's file in Load Checkpoint, with the settings its model card gives.
# on your rented machine: pip install diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline
pipe = DiffusionPipeline.from_pretrained("dataautogpt3/OpenDalleV1.1", torch_dtype=torch.bfloat16).to("cuda")
image = pipe(prompt="a red bicycle on a cobbled street").images[0]
image.save("/workspace/out.png")# on your rented machine (the ssh line is on its page in the console)
# get REPO FILE FOLDER: one file into /workspace/models/FOLDER, where ComfyUI loads it from
get() { hf download "$1" "$2" --local-dir /workspace/hf-files && mkdir -p "/workspace/models/$3" && mv "/workspace/hf-files/$2" "/workspace/models/$3/$4"; }
# the model (6.5 GB)
get dataautogpt3/OpenDalleV1.1 OpenDalleV1.1.safetensors checkpoints
start-comfyui# on your computer, in a second terminal: ComfyUI in your browser at http://localhost:8188
# HOST and PORT are your machine's, from its page in the console
ssh -L 8188:localhost:8188 dev@HOST -p PORT