Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

colab t4 #33

Open
werruww opened this issue Nov 22, 2024 · 4 comments
Open

colab t4 #33

werruww opened this issue Nov 22, 2024 · 4 comments
Labels
fixed fix a bug

Comments

@werruww
Copy link

werruww commented Nov 22, 2024

!git clone https://github.com/newgenai79/Sana_win

%cd Sana_win

!git clone https://github.com/woct0rdho/triton-windows/releases

!/content/Sana_win/environment_setup.bat sana

!chmod +x /content/Sana_win/environment_setup.bat

!/content/Sana_win/environment_setup.bat sana

@echo off
setlocal enabledelayedexpansion

set CONDA_ENV=%1
if "%CONDA_ENV%"=="" (
echo Please provide a conda environment name
exit /b 1
)

echo Creating Conda Environment: %CONDA_ENV%
call conda create -n %CONDA_ENV% python=3.10.0 -y
echo Conda Environment Created Successfully

echo Activating Conda Environment: %CONDA_ENV%
call conda activate %CONDA_ENV%
echo Conda Environment Activated

echo Installing CUDA Toolkit
call conda install -c nvidia cuda-toolkit -y
echo CUDA Toolkit Installed Successfully

echo Activating Conda Environment: %CONDA_ENV%
call conda activate %CONDA_ENV%
echo Conda Environment Activated

echo Updating pip to latest version
python -m pip install --upgrade pip
echo Pip Updated Successfully

echo Activating Conda Environment: %CONDA_ENV%
call conda activate %CONDA_ENV%
echo Conda Environment Activated

echo Installing xformers
call pip install -U xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121
echo Xformers Installed Successfully

echo Activating Conda Environment: %CONDA_ENV%
call conda activate %CONDA_ENV%
echo Conda Environment Activated

echo Installing project in editable mode
call pip install -e .
echo Project Installed Successfully

echo Activating Conda Environment: %CONDA_ENV%
call conda activate %CONDA_ENV%
echo Conda Environment Activated

echo Installing triton from wheel
pip install triton-3.0.0-cp310-cp310-win_amd64.whl
echo Installing triton complete

echo Environment Setup Complete!

#!/bin/bash

CONDA_ENV=$1
if [ -z "$CONDA_ENV" ]; then
echo "Please provide a conda environment name"
exit 1
fi

echo "Creating Conda Environment: $CONDA_ENV"
conda create -n "$CONDA_ENV" python=3.10.0 -y
echo "Conda Environment Created Successfully"

echo "Activating Conda Environment: $CONDA_ENV"
conda activate "$CONDA_ENV"
echo "Conda Environment Activated"

echo "Installing CUDA Toolkit"
conda install -c nvidia cuda-toolkit -y
echo "CUDA Toolkit Installed Successfully"

echo "Updating pip to latest version"
python -m pip install --upgrade pip
echo "Pip Updated Successfully"

echo "Installing xformers"
pip install -U xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121
echo "Xformers Installed Successful"

%%writefile environment_setup.sh
#!/bin/bash

... (paste the bash script here) ...

#!/bin/bash

CONDA_ENV=$1
if [ -z "$CONDA_ENV" ]; then
echo "Please provide a conda environment name"
exit 1
fi

echo "Creating Conda Environment: $CONDA_ENV"
conda create -n "$CONDA_ENV" python=3.10.0 -y
echo "Conda Environment Created Successfully"

echo "Activating Conda Environment: $CONDA_ENV"
conda activate "$CONDA_ENV"
echo "Conda Environment Activated"

echo "Installing CUDA Toolkit"
conda install -c nvidia cuda-toolkit -y
echo "CUDA Toolkit Installed Successfully"

echo "Updating pip to latest version"
python -m pip install --upgrade pip
echo "Pip Updated Successfully"

echo "Installing xformers"
pip install -U xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121
echo "Xformers Installed Successful"

!chmod +x environment_setup.sh

!./environment_setup.sh sana

Step 1: Clone the repository

git clone https://github.com/newgenai79/Sana_win

Step 2: Navigate inside the cloned repository

cd Sana_win

Step 3: Download triton wheel and place in Sana_win (triton-3.0.0-cp310-cp310-win_amd64.whl)

https://github.com/woct0rdho/triton-windows/releases

Step 4: Install requirements (make sure conda or miniconda is installed)

environment_setup.bat sana

Step 5: Download models (refer main repository https://github.com/NVlabs/Sana#-news)
mkdir Efficient-Large-Model\Sana_1600M_512px\checkpoints
mkdir Efficient-Large-Model\Sana_1600M_1024px\checkpoints

Sana
└───Efficient-Large-Model
	├───Sana_1600M_1024px
	│   └───checkpoints
	│           Sana_1600M_1024px.pth
	│
	└───Sana_1600M_512px
		└───checkpoints
				Sana_1600M_512px.pth

Step 6: Gradio UI (1.6B)
512x512 model
python app\app_sana.py --config configs\sana_config\512ms\Sana_1600M_img512.yaml --model_path Efficient-Large-Model\Sana_1600M_512px\checkpoints\Sana_1600M_512px.pth

1024x1024 model
python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth

!git clone https://github.com/newgenai79/Sana_win

%cd Sana_win

!wget -P Sana_win https://github.com/woct0rdho/triton-windows/releases/download/v3.0.0/triton-3.0.0-cp310-cp310-win_amd64.whl

!pip install tritonclient[http]

!chmod +x environment_setup.sh
!./environment_setup.sh sana

import os
from getpass import getpass

HF_TOKEN = getpass("Enter your Hugging Face access token:hR")
os.environ["HF_TOKEN"] = HF_TOKEN

!huggingface-cli login

!mkdir -p Efficient-Large-Model/Sana_1600M_512px/checkpoints
!wget -P Efficient-Large-Model/Sana_1600M_512px/checkpoints https://huggingface.co/newgenai/Sana/resolve/main/Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

cache_dir="./Efficient-Large-Model/Sana_1600M_512px/checkpoints"

#!pip install huggingface_hub

import os
from getpass import getpass # Import the getpass function

from huggingface_hub import hf_hub_download, login

Get the Hugging Face access token

HF_TOKEN = getpass("Enter your Hugging Face access token:")

Log in to Hugging Face using the access token

login(token=HF_TOKEN)

Download the file

file_path = hf_hub_download(
repo_id="Efficient-Large-Model/Sana_1600M_1024px",
filename="checkpoints/Sana_1600M_1024px.pth",
cache_dir="./Efficient-Large-Model/Sana_1600M_1024px/checkpoints", # local path
token=HF_TOKEN
)

print(f"File downloaded to: {file_path}")

!pip install gradio

!pip install spaces

!python /content/Sana_win/app/app_sana.py --config configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

import torch
import torchvision

print(f"PyTorch CUDA version: {torch.version.cuda}")
print(f"torchvision CUDA version: {torchvision.version}")

!pip uninstall torchvision -y

!pip install torchvision==0.16.0+cu118 -f https://download.pytorch.org/whl/torch_stable.html

!apt-get install cuda-toolkit-11-8 # example if CUDA 11.8 is needed

!python /content/Sana_win/app/app_sana.py --config configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

!apt-get install libjpeg-dev libpng-dev -y

!pip install torch torchvision --upgrade

Traceback (most recent call last):
File "/content/Sana_win/app/app_sana.py", line 34, in
from app.sana_pipeline import SanaPipeline
ModuleNotFoundError: No module named 'app'

/content/Sana_win/app/app_sana_512.py

!python /content/Sana_win/app/app_sana_512.py --config /content/Sana_win/configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

!python /content/Sana_win/app/app_sana_512.py

!pip install --upgrade torch

!pip install torch==1.13.0+cu117 # Adjust version and CUDA as needed

!python app/app_sana.py --config configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

!python /content/Sana_win/app/app_sana_512.py --config /content/Sana_win/configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

%cd /content/Sana_win/app

import sys
import os

sys.path.append(os.path.abspath(os.path.join(os.getcwd(), os.pardir)))
!python app_sana_512.py

import sys
import os

sys.path.append(os.path.abspath(os.path.join(os.getcwd(), os.pardir)))

!python app_sana_512.py

!python /content/Sana_win/app/app_sana_512.py --config /content/Sana_win/configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

!python app_sana_512.py --config configs/sana_config/512ms/Sana_1600M_img512.yaml --model_path Efficient-Large-Model/Sana_1600M_512px/checkpoints/Sana_1600M_512px.pth

!cat /content/Sana_win/app/app_sana_512.py

with open('/content/Sana_win/app/app_sana_512.py', 'r') as f:
print(f.read())

!python /content/Sana_win/app/app_sana_512.py

@werruww
Copy link
Author

werruww commented Nov 22, 2024

!git clone https://github.com/NVlabs/Sana.git
%cd Sana

%cd /content/Sana

!pip install torch
!pip install -U xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121
!pip install -e .

%cd /content/Sana

!git config --global credential.helper store

!huggingface-cli login

%env DEMO_PORT=15432
!python app/app_sana.py
--config=/content/Sana/configs/sana_config/1024ms/Sana_1600M_img1024.yaml
--model_path=hf://Efficient-Large-Model/Sana_1600M_1024px/checkpoints/Sana_1600M_1024px.pth

env: DEMO_PORT=15432
2024-11-22 22:36:27.683067: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-11-22 22:36:27.901396: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-11-22 22:36:27.964233: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2024-11-22 22:36:28.314640: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-11-22 22:36:30.456657: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
2024-11-23 06:36:35 - [Sana] - INFO - Sampler flow_dpm-solver, flow_shift: 3.0
2024-11-23 06:36:35 - [Sana] - INFO - Inference with torch.float16, PAG guidance layer: [8]
[DC-AE] Loading model from mit-han-lab/dc-ae-f32c32-sana-1.0
Loading checkpoint shards: 100% 2/2 [00:01<00:00, 1.17it/s]
2024-11-23 06:37:39 - [Sana] - WARNING - use pe: False, position embed interpolation: 1.0, base size: 32
2024-11-23 06:37:39 - [Sana] - WARNING - attention type: linear; ffn type: glumbconv; autocast linear attn: False
^C

@werruww
Copy link
Author

werruww commented Nov 22, 2024

@werruww
Copy link
Author

werruww commented Nov 22, 2024

import torch
from app.sana_pipeline import SanaPipeline
from torchvision.utils import save_image

Import the huggingface_hub library

from huggingface_hub import hf_hub_download

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
generator = torch.Generator(device=device).manual_seed(42)

sana = SanaPipeline("/content/Sana/configs/sana_config/512ms/Sana_600M_img512.yaml")

Use hf_hub_download to download the model weights

repo_id = "Efficient-Large-Model/Sana_1600M_512px"
filename = "checkpoints/Sana_1600M_512px.pth"
model_path = hf_hub_download(repo_id=repo_id, filename=filename)
sana.from_pretrained(model_path)
prompt = 'a cyberpunk cat with a neon sign that says "Sana"'

image = sana(
prompt=prompt,
height=128,
width=128,
guidance_scale=1.0,
pag_guidance_scale=1.0,
num_inference_steps=5,
generator=generator,
)
save_image(image, '/content/sana.png', nrow=1, normalize=True, value_range=(-1, 1))

2024-11-23 06:52:38 - [Sana] - INFO - Sampler flow_dpm-solver, flow_shift: 1.0
2024-11-23 06:52:38 - [Sana] - INFO - Inference with torch.float16, PAG guidance layer: [14]
[DC-AE] Loading model from mit-han-lab/dc-ae-f32c32-sana-1.0
Loading checkpoint shards: 100%
 2/2 [00:01<00:00,  1.51it/s]
2024-11-23 06:53:23 - [Sana] - WARNING - use pe: False, position embed interpolation: 1.0, base size: 16
2024-11-23 06:53:23 - [Sana] - WARNING - attention type: linear; ffn type: glumbconv; autocast linear attn: False
2024-11-23 06:53:31 - [Sana] - INFO - use_fp32_attention: True
2024-11-23 06:53:31 - [Sana] - INFO - SanaMS:SanaMS_600M_P1_D28,Model Parameters: 591,747,872
Sana_1600M_512px.pth: 100%
 6.42G/6.42G [04:13<00:00, 25.7MB/s]
[Sana] Loading model from /root/.cache/huggingface/hub/models--Efficient-Large-Model--Sana_1600M_512px/snapshots/b17b080d9f3b6c4fb71ba7b0d8384d445d500dd1/checkpoints/Sana_1600M_512px.pth

RuntimeError Traceback (most recent call last)
in <cell line: 15>()
13 filename = "checkpoints/Sana_1600M_512px.pth"
14 model_path = hf_hub_download(repo_id=repo_id, filename=filename)
---> 15 sana.from_pretrained(model_path)
16 prompt = 'a cyberpunk cat with a neon sign that says "Sana"'
17

1 frames
/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py in load_state_dict(self, state_dict, strict, assign)
2213
2214 if len(error_msgs) > 0:
-> 2215 raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
2216 self.class.name, "\n\t".join(error_msgs)))
2217 return _IncompatibleKeys(missing_keys, unexpected_keys)

RuntimeError: Error(s) in loading state_dict for SanaMS:
size mismatch for x_embedder.proj.weight: copying a param with shape torch.Size([2240, 32, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 32, 1, 1]).
size mismatch for x_embedder.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for t_embedder.mlp.0.weight: copying a param with shape torch.Size([2240, 256]) from checkpoint, the shape in current model is torch.Size([1152, 256]).
size mismatch for t_embedder.mlp.0.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for t_embedder.mlp.2.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for t_embedder.mlp.2.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for t_block.1.weight: copying a param with shape torch.Size([13440, 2240]) from checkpoint, the shape in current model is torch.Size([6912, 1152]).
size mismatch for t_block.1.bias: copying a param with shape torch.Size([13440]) from checkpoint, the shape in current model is torch.Size([6912]).
size mismatch for y_embedder.y_proj.fc1.weight: copying a param with shape torch.Size([2240, 2304]) from checkpoint, the shape in current model is torch.Size([1152, 2304]).
size mismatch for y_embedder.y_proj.fc1.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for y_embedder.y_proj.fc2.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for y_embedder.y_proj.fc2.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for attention_y_norm.weight: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.0.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.0.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.0.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.0.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.0.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.0.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.0.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.0.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.0.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.0.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.0.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.0.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.0.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.0.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.0.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.1.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.1.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.1.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.1.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.1.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.1.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.1.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.1.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.1.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.1.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.1.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.1.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.1.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.1.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.1.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.2.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.2.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.2.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.2.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.2.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.2.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.2.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.2.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.2.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.2.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.2.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.2.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.2.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.2.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.2.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.3.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.3.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.3.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.3.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.3.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.3.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.3.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.3.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.3.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.3.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.3.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.3.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.3.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.3.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.3.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.4.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.4.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.4.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.4.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.4.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.4.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.4.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.4.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.4.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.4.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.4.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.4.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.4.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.4.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.4.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.5.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.5.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.5.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.5.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.5.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.5.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.5.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.5.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.5.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.5.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.5.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.5.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.5.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.5.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.5.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.6.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.6.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.6.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.6.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.6.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.6.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.6.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.6.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.6.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.6.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.6.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.6.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.6.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.6.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.6.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.7.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.7.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.7.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.7.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.7.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.7.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.7.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.7.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.7.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.7.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.7.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.7.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.7.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.7.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.7.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.8.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.8.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.8.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.8.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.8.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.8.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.8.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.8.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.8.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.8.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.8.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.8.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.8.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.8.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.8.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.9.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.9.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.9.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.9.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.9.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.9.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.9.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.9.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.9.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.9.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.9.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.9.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.9.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.9.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.9.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.10.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.10.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.10.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.10.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.10.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.10.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.10.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.10.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.10.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.10.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.10.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.10.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.10.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.10.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.10.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.11.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.11.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.11.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.11.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.11.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.11.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.11.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.11.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.11.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.11.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.11.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.11.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.11.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.11.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.11.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.12.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.12.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.12.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.12.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.12.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.12.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.12.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.12.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.12.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.12.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.12.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.12.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.12.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.12.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.12.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.13.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.13.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.13.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.13.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.13.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.13.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.13.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.13.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.13.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.13.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.13.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.13.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.13.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.13.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.13.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.14.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.14.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.14.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.14.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.14.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.14.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.14.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.14.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.14.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.14.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.14.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.14.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.14.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.14.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.14.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.15.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.15.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.15.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.15.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.15.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.15.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.15.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.15.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.15.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.15.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.15.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.15.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.15.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.15.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.15.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.16.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.16.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.16.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.16.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.16.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.16.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.16.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.16.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.16.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.16.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.16.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.16.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.16.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.16.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.16.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.17.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.17.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.17.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.17.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.17.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.17.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.17.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.17.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.17.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.17.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.17.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.17.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.17.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.17.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.17.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.18.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.18.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.18.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.18.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.18.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.18.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.18.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.18.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.18.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.18.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.18.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.18.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.18.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.18.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.18.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for blocks.19.scale_shift_table: copying a param with shape torch.Size([6, 2240]) from checkpoint, the shape in current model is torch.Size([6, 1152]).
size mismatch for blocks.19.attn.qkv.weight: copying a param with shape torch.Size([6720, 2240]) from checkpoint, the shape in current model is torch.Size([3456, 1152]).
size mismatch for blocks.19.attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.19.attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.19.cross_attn.q_linear.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.19.cross_attn.q_linear.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.19.cross_attn.kv_linear.weight: copying a param with shape torch.Size([4480, 2240]) from checkpoint, the shape in current model is torch.Size([2304, 1152]).
size mismatch for blocks.19.cross_attn.kv_linear.bias: copying a param with shape torch.Size([4480]) from checkpoint, the shape in current model is torch.Size([2304]).
size mismatch for blocks.19.cross_attn.proj.weight: copying a param with shape torch.Size([2240, 2240]) from checkpoint, the shape in current model is torch.Size([1152, 1152]).
size mismatch for blocks.19.cross_attn.proj.bias: copying a param with shape torch.Size([2240]) from checkpoint, the shape in current model is torch.Size([1152]).
size mismatch for blocks.19.mlp.inverted_conv.conv.weight: copying a param with shape torch.Size([11200, 2240, 1, 1]) from checkpoint, the shape in current model is torch.Size([5760, 1152, 1, 1]).
size mismatch for blocks.19.mlp.inverted_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.19.mlp.depth_conv.conv.weight: copying a param with shape torch.Size([11200, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([5760, 1, 3, 3]).
size mismatch for blocks.19.mlp.depth_conv.conv.bias: copying a param with shape torch.Size([11200]) from checkpoint, the shape in current model is torch.Size([5760]).
size mismatch for blocks.19.mlp.point_conv.conv.weight: copying a param with shape torch.Size([2240, 5600, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 2880, 1, 1]).
size mismatch for final_layer.scale_shift_table: copying a param with shape torch.Size([2, 2240]) from checkpoint, the shape in current model is torch.Size([2, 1152]).
size mismatch for final_layer.linear.weight: copying a param with shape torch.Size([32, 2240]) from checkpoint, the shape in current model is torch.Size([32, 1152]).

@lawrence-cj
Copy link
Collaborator

lawrence-cj commented Nov 24, 2024

sana = SanaPipeline("/content/Sana/configs/sana_config/512ms/Sana_600M_img512.yaml")
repo_id = "Efficient-Large-Model/Sana_1600M_512px"
filename = "checkpoints/Sana_1600M_512px.pth"

The config file you are using is for 600M model, but the checkpoint is for 1600M model.

@lawrence-cj lawrence-cj mentioned this issue Nov 24, 2024
@lawrence-cj lawrence-cj added the fixed fix a bug label Nov 24, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
fixed fix a bug
Projects
None yet
Development

No branches or pull requests

2 participants