I'm quite happy this is here. Downloaded it already, but this just means more people will get to use Dev edition, you know the people with 8GB give or take a bit as well. I just hope you had permission... from Ill, no offense.

ELI5 Comfy Tutorial?

e.g What folder does this model go into?

(Sorry but my small brain cannot keep up with the speed of AI image generation developments!)

lllyasviel writes that RTX 20xx series GPUs might not benefit from NF4 but an NVIDIA post does seem to indicate it should. Should I download this for my RTX 2060 12GB?

how to install ComfyUI_bitsandbytes_NF4

i have error

0.0 seconds (IMPORT FAILED): F:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_bitsandbytes_NF4

lllyasviel is genius once again. Can't believe how good it works and how little VRAM it uses. And I haven't yet tried FORGE even.

is nf4 will just run n cpu ? and what is bnb is?

Getting an error with the 5gb schnell version involving AttributeError: 'NoneType' object has no attribute 'unet_key_prefix'

'NoneType' object has no attribute 'unet_key_prefix'

*** Error completing request

unsure what this is about, I also notice you're using 20 steps still on that model, wouldn't the point of making schnell nf4 to make it even faster

How to run 6GB version? Comfy and Forge don't support it:
"ERROR: Could not detect model type of"

"TypeError: 'NoneType'object is not iterable"

FYI to others, I couldn't get the 5gb schnell to work for whatever reason.

Then in the 10gb model, I was getting an error where forge was looking for the files forge installs for the backend for the Dev version but for schnell.

Follow the path it gives you

Stable Diffusion WebUI Forge\backend\huggingface\black-forest-labs/FLUX.1-schnell

go up one directory and copy the contents from the flux dev folder into a new folder with the name above FLUX.1-schnell, then it should run.

On a laptop 4060 8GB, I'm getting 25 seconds or so which is fucking insane...

Thanks OP and the original DEV of the Nf4 implementation (he has a weird username and i'll never remember it heh)

May i ask what BNB means?
What is the difference between them?

This model is insanely fast with really good quality. In the images section you can see some of my creations with old DALL-E 3 prompts i have saved.

For me, the speed is the same as fp8 in ComfyUI. Is it supposed to be faster?

it's a pity the 5.5gb version won't load in comfyui :'(

I'm downloading the 10.7gb version now, will report back to say if it's any faster on my subpar hardware :P

Something I don't understand: What happened to clip_L and T5 encoder? Are they merged into these files? Wouldn't that reduce the quality significantly? T5 fp16 alone was larger than these files

TypeError: 'NoneType' object is not iterable

Error occurred when executing CheckpointLoaderNF4: load_checkpoint_guess_config() got an unexpected keyword argument 'model_options' File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_bitsandbytes_NF4\__init__.py", line 178, in load_checkpoint out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, embedding_directory=folder_paths.get_folder_paths("embeddings"), model_options={"custom_operations": OPS})

any benefit of using this 5gb nf4 except it use less memory ? any speed up in steps ?

3080 10gb first run schnell 25s, 2nd+ run shows lowvram mode 1735 and it takes 60s. I'm using ComfyUI.

Is this RalFinger model, or just repost of models from lllyasviel?

File "E:\webui_forge_cu121_torch21\webui\modules\call_queue.py", line 74, in f

res = list(func(*args, **kwargs))

TypeError: 'NoneType' object is not iterable

lllyasviel's model on huggingface is 11.5GB, while your model is 10.7GB. Can you explain why is that?

NF4+t5xxl_fp8 will not give you any speed boost compared to FP8+t5xxl_fp16 if your videocard has 16+GB VRAM. Tested in ComfyUI.

upd. If you don't have enough RAM or the checkpoint is on the HDD, you will get a speed boost, but this has nothing to do with the performance of your GPU!

upd. I test pytorch version: 2.4.0+cu121 & pytorch version: 2.4.0+cu124, results below. Hardware: RTX4070TI Super (16gb), i7 12700, 64GB RAM cl16 3600 d.rank, nvme 7000mbps.

cu124, fp8 and fp16 t5xxl.

6 seconds first launch.

26 sec - 1024x1024, 20 steps, euler simple.

cu124, nf4 and fp8 t5xxl.

3 seconds first launch.

26 sec - 1024x1024, 20 steps, euler simple.

cu121, fp8 and fp16 t5xxl.

27 sec first launch.

28 sec - 1024x1024, 20 steps, euler simple.

cu121, nf4 and fp8 t5xxl.

9 sec first launch.

28 sec - 1024x1024, 20 steps, euler simple.

please help guys i can't understand this err

"Error occurred when executing UNETLoader: Error(s) in loading state_dict for Flux: size mismatch for img_in.weight: copying a param with shape torch.Size([98304, 1]) from checkpoint, the shape in current model is torch.Size([3072, 64]). size mismatch for time_in.in_layer.weight: copying a param with shape torch.Size([393216, 1]) from checkpoint, the shape in current model is torch.Size([3072, 256]). size mismatch for time_in.out_layer.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for vector_in.in_layer.weight: copying a param with shape torch.Size([1179648, 1]) from checkpoint, the shape in current model is torch.Size([3072, 768]). size mismatch for vector_in.out_layer.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for guidance_in.in_layer.weight: copying a param with shape torch.Size([393216, 1]) from checkpoint, the shape in current model is torch.Size([3072, 256]). size mismatch for guidance_in.out_layer.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for txt_in.weight: copying a param with shape torch.Size([6291456, 1]) from checkpoint, the shape in current model is torch.Size([3072, 4096]). size mismatch for double_blocks.0.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.0.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.0.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.0.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.0.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.0.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.0.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.0.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.0.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.0.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.1.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.1.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.1.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.1.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.1.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.1.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.1.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.1.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.1.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.1.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.2.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.2.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.2.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.2.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.2.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.2.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.2.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.2.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.2.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.2.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.3.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.3.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.3.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.3.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.3.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.3.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.3.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.3.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.3.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.3.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.4.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.4.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.4.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.4.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.4.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.4.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.4.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.4.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.4.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.4.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.5.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.5.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.5.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.5.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.5.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.5.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.5.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.5.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.5.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.5.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.6.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.6.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.6.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.6.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.6.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.6.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.6.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.6.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.6.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.6.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.7.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.7.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.7.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.7.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.7.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.7.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.7.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.7.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.7.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.7.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.8.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.8.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.8.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.8.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.8.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.8.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.8.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.8.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.8.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.8.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.9.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.9.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.9.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.9.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.9.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.9.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.9.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.9.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.9.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.9.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.10.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.10.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.10.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.10.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.10.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.10.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.10.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.10.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.10.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.10.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.11.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.11.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.11.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.11.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.11.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.11.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.11.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.11.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.11.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.11.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.12.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.12.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.12.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.12.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.12.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.12.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.12.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.12.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.12.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.12.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.13.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.13.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.13.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.13.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.13.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.13.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.13.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.13.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.13.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.13.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.14.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.14.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.14.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.14.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.14.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.14.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.14.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.14.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.14.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.14.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.15.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.15.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.15.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.15.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.15.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.15.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.15.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.15.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.15.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.15.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.16.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.16.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.16.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.16.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.16.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.16.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.16.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.16.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.16.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.16.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.17.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.17.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.17.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.17.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.17.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.17.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.17.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.17.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.17.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.17.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.18.img_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.18.img_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.18.img_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.18.img_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.18.img_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for double_blocks.18.txt_mod.lin.weight: copying a param with shape torch.Size([28311552, 1]) from checkpoint, the shape in current model is torch.Size([18432, 3072]). size mismatch for double_blocks.18.txt_attn.qkv.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for double_blocks.18.txt_attn.proj.weight: copying a param with shape torch.Size([4718592, 1]) from checkpoint, the shape in current model is torch.Size([3072, 3072]). size mismatch for double_blocks.18.txt_mlp.0.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([12288, 3072]). size mismatch for double_blocks.18.txt_mlp.2.weight: copying a param with shape torch.Size([18874368, 1]) from checkpoint, the shape in current model is torch.Size([3072, 12288]). size mismatch for single_blocks.0.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.0.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.0.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.1.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.1.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.1.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.2.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.2.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.2.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.3.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.3.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.3.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.4.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.4.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.4.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.5.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.5.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.5.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.6.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.6.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.6.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.7.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.7.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.7.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.8.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.8.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.8.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.9.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.9.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.9.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.10.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.10.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.10.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.11.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.11.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.11.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.12.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.12.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.12.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.13.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.13.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.13.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.14.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.14.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.14.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.15.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.15.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.15.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.16.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.16.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.16.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.17.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.17.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.17.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.18.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.18.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.18.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.19.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.19.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.19.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.20.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.20.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.20.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.21.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.21.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.21.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.22.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.22.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.22.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.23.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.23.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.23.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.24.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.24.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.24.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.25.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.25.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.25.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.26.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.26.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.26.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.27.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.27.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.27.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.28.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.28.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.28.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.29.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.29.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.29.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.30.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.30.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.30.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.31.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.31.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.31.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.32.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.32.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.32.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.33.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.33.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.33.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.34.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.34.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.34.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.35.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.35.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.35.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.36.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.36.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.36.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for single_blocks.37.linear1.weight: copying a param with shape torch.Size([33030144, 1]) from checkpoint, the shape in current model is torch.Size([21504, 3072]). size mismatch for single_blocks.37.linear2.weight: copying a param with shape torch.Size([23592960, 1]) from checkpoint, the shape in current model is torch.Size([3072, 15360]). size mismatch for single_blocks.37.modulation.lin.weight: copying a param with shape torch.Size([14155776, 1]) from checkpoint, the shape in current model is torch.Size([9216, 3072]). size mismatch for final_layer.linear.weight: copying a param with shape torch.Size([98304, 1]) from checkpoint, the shape in current model is torch.Size([64, 3072]). size mismatch for final_layer.adaLN_modulation.1.weight: copying a param with shape torch.Size([9437184, 1]) from checkpoint, the shape in current model is torch.Size([6144, 3072]). File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\nodes.py", line 836, in load_unet model = comfy.sd.load_unet(unet_path, dtype=dtype) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 645, in load_unet model = load_unet_state_dict(sd, dtype=dtype) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 637, in load_unet_state_dict model.load_model_weights(new_sd, "") File "E:\cumfy ui\ComfyUI_windows_portable\ComfyUI\comfy\model_base.py", line 225, in load_model_weights m, u = self.diffusion_model.load_state_dict(to_load, strict=False) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\cumfy ui\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 2189, in load_state_dict raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(

Close"

In this comment I say thank you to lllyasviel, join in the comments)

Very fast in Forge

Once you understand how to install everything NF4 is so amazing. Quality and speed at the same time, love it! 💕

Is anyone mange to run a lora with this version yet?

Is there any chance of this working on a 6GB of VRAM?

what vae?

@RalFinger, plan on updating to this version... https://huggingface.co/lllyasviel/flux1-dev-bnb-nf4/blob/main/flux1-dev-bnb-nf4-v2.safetensors ?

It's sooooooooooo much faster. If you'd believe that. I know it's not Comfy yet, but maybe you could make it so. Works in Forge, of course.

I made some comparisons between different sampling steps

https://civitai.com/articles/6690