How can I solve this issue on the colab implementation?
Opened this issue · 5 comments
Example usage
EXPORT_VIDEO=True
EXPORT_MESH=True
INFER_CONFIG="./configs/infer-b.yaml"
MODEL_NAME="zxhezexin/openlrm-mix-base-1.1"
IMAGE_INPUT="./assets/sample_input/owl.png"
!python -m openlrm.launch infer.lrm --infer $INFER_CONFIG model_name=$MODEL_NAME image_input=$IMAGE_INPUT export_video=$EXPORT_VIDEO export_mesh=$EXPORT_MESH
/usr/bin/python3: No module named openlrm
Hi,
No module named openlrm means your working dir is not the root dir of your cloned repo.
You can run !realpath . to see which is the working dir for your notebook.
Thanks! I fixed that issue, but now I'm receiving another problem, saying something about the xFormers
Example usage
%cd ../OpenLRM
EXPORT_VIDEO=True
EXPORT_MESH=True
INFER_CONFIG="./configs/infer-b.yaml"
MODEL_NAME="zxhezexin/openlrm-mix-base-1.1"
IMAGE_INPUT="./assets/sample_input/owl.png"
!python -m openlrm.launch infer.lrm --infer $INFER_CONFIG model_name=$MODEL_NAME image_input=$IMAGE_INPUT export_video=$EXPORT_VIDEO export_mesh=$EXPORT_MESH
........................................................................................................................................................................................................
/OpenLRM
[2024-03-21 13:24:10,947] openlrm.models.modeling_lrm: [INFO] Using DINOv2 as the encoder
/OpenLRM/openlrm/models/encoders/dinov2/layers/swiglu_ffn.py:51: UserWarning: xFormers is not available (SwiGLU)
warnings.warn("xFormers is not available (SwiGLU)")
/OpenLRM/openlrm/models/encoders/dinov2/layers/attention.py:33: UserWarning: xFormers is not available (Attention)
warnings.warn("xFormers is not available (Attention)")
/OpenLRM/openlrm/models/encoders/dinov2/layers/block.py:46: UserWarning: xFormers is not available (Block)
warnings.warn("xFormers is not available (Block)")
0% 0/1 [00:00<?, ?it/s]/OpenLRM/openlrm/datasets/cam_utils.py:153: UserWarning: Using torch.cross without specifying the dim arg is deprecated.
Please either pass the dim explicitly or simply use torch.linalg.cross.
The default value of dim will change to agree with that of linalg.cross in a future release. (Triggered internally at ../aten/src/ATen/native/Cross.cpp:63.)
x_axis = torch.cross(up_world, z_axis)
^C
I am still working on the colab implementation, for now, these are the script modifications on app.py & the errors on it
..............................................................................................................................................................................................
Copyright (c) 2023-2024, Zexin He
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
https://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
import os
from PIL import Image
import numpy as np
import gradio as gr
def assert_input_image(input_image):
if input_image is None:
raise gr.Error("No image selected or uploaded!")
def prepare_working_dir():
import tempfile
working_dir = tempfile.TemporaryDirectory()
return working_dir
def init_preprocessor():
from openlrm.utils.preprocess import Preprocessor
global preprocessor
preprocessor = Preprocessor()
def preprocess_fn(image_in: np.ndarray, remove_bg: bool, recenter: bool, working_dir):
image_raw = os.path.join(working_dir.name, "raw.png")
with Image.fromarray(image_in) as img:
img.save(image_raw)
image_out = os.path.join(working_dir.name, "rembg.png")
success = preprocessor.preprocess(image_path=image_raw, save_path=image_out, rmbg=remove_bg, recenter=recenter)
assert success, f"Failed under preprocess_fn!"
return image_out
def demo_openlrm(infer_impl):
def core_fn(image: str, source_cam_dist: float, working_dir):
dump_video_path = os.path.join(working_dir.name, "output.mp4")
dump_mesh_path = os.path.join(working_dir.name, "output.ply")
infer_impl(
image_path=image,
source_cam_dist=source_cam_dist,
export_video=True,
export_mesh=True,
dump_video_path=dump_video_path,
dump_mesh_path=dump_mesh_path,
)
return dump_video_path
return dump_mesh_path
def example_fn(image: np.ndarray):
from gradio.utils import get_cache_folder
working_dir = get_cache_folder()
image = preprocess_fn(
image_in=image,
remove_bg=True,
recenter=True,
working_dir=working_dir,
)
video = core_fn(
image=image,
source_cam_dist=2.0,
working_dir=working_dir,
)
mesh = core_fn(
mesh=mesh,
working_dir=working_dir,
)
return image, video, mesh
_TITLE = '''OpenLRM: Open-Source Large Reconstruction Models'''
_DESCRIPTION = '''
<div>
<a style="display:inline-block" href='https://github.com/3DTopia/OpenLRM'><img src='https://img.shields.io/github/stars/3DTopia/OpenLRM?style=social'/></a>
<a style="display:inline-block; margin-left: .5em" href="https://huggingface.co/zxhezexin"><img src='https://img.shields.io/badge/Model-Weights-blue'/></a>
</div>
OpenLRM is an open-source implementation of Large Reconstruction Models.
<strong>Image-to-3D in 10 seconds with A100!</strong>
<strong>Disclaimer:</strong> This demo uses `openlrm-mix-base-1.1` model with 288x288 rendering resolution here for a quick demonstration.
'''
with gr.Blocks(analytics_enabled=False) as demo:
# HEADERS
with gr.Row():
with gr.Column(scale=1):
gr.Markdown('# ' + _TITLE)
with gr.Row():
gr.Markdown(_DESCRIPTION)
# DISPLAY
with gr.Row():
with gr.Column(variant='panel', scale=1):
with gr.Tabs(elem_id="openlrm_input_image"):
with gr.TabItem('Input Image'):
with gr.Row():
input_image = gr.Image(label="Input Image", image_mode="RGBA", width="auto", sources="upload", type="numpy", elem_id="content_image")
with gr.Column(variant='panel', scale=1):
with gr.Tabs(elem_id="openlrm_processed_image"):
with gr.TabItem('Processed Image'):
with gr.Row():
processed_image = gr.Image(label="Processed Image", image_mode="RGBA", type="filepath", elem_id="processed_image", width="auto", interactive=False)
with gr.Column(variant='panel', scale=1):
with gr.Tabs(elem_id="openlrm_render_video"):
with gr.TabItem('Rendered Video'):
with gr.Row():
output_video = gr.Video(label="Rendered Video", format="mp4", width="auto", autoplay=True)
with gr.Column(variant='panel', scale=1):
with gr.Tabs(elem_id="openlrm_render_mesh"):
with gr.TabItem('Rendered Mesh'):
with gr.Row():
output_mesh = gr.Mesh(label="Rendered Mesh", format="ply", width="auto", autoplay=True)
# SETTING
with gr.Row():
with gr.Column(variant='panel', scale=1):
with gr.Tabs(elem_id="openlrm_attrs"):
with gr.TabItem('Settings'):
with gr.Column(variant='panel'):
gr.Markdown(
"""
<strong>Best Practice</strong>:
Centered objects in reasonable sizes. Try adjusting source camera distances.
"""
)
checkbox_rembg = gr.Checkbox(True, label='Remove background')
checkbox_recenter = gr.Checkbox(True, label='Recenter the object')
slider_cam_dist = gr.Slider(1.0, 3.5, value=2.0, step=0.1, label="Source Camera Distance")
submit = gr.Button('Generate', elem_id="openlrm_generate", variant='primary')
# EXAMPLES
with gr.Row():
examples = [
['assets/sample_input/owl.png'],
['assets/sample_input/building.png'],
['assets/sample_input/mailbox.png'],
['assets/sample_input/fire.png'],
['assets/sample_input/girl.png'],
['assets/sample_input/lamp.png'],
['assets/sample_input/hydrant.png'],
['assets/sample_input/hotdogs.png'],
['assets/sample_input/traffic.png'],
['assets/sample_input/ceramic.png'],
]
gr.Examples(
examples=examples,
inputs=[input_image],
outputs=[processed_image, output_video],
fn=example_fn,
cache_examples=bool(os.getenv('SPACE_ID')),
examples_per_page=20,
)
working_dir = gr.State()
submit.click(
fn=assert_input_image,
inputs=[input_image],
queue=False,
).success(
fn=prepare_working_dir,
outputs=[working_dir],
queue=False,
).success(
fn=preprocess_fn,
inputs=[input_image, checkbox_rembg, checkbox_recenter, working_dir],
outputs=[processed_image],
).success(
fn=core_fn,
inputs=[processed_image, slider_cam_dist, working_dir],
outputs=[output_video],
).success(
fn=core_fn,
inputs=[processed_mesh, working_dir],
outputs=[output_mesh],
)
demo.queue()
demo.launch(share = True)
def launch_gradio_app():
os.environ.update({
"APP_ENABLED": "1",
"APP_MODEL_NAME": "zxhezexin/openlrm-mix-base-1.1",
"APP_INFER": "./configs/infer-gradio.yaml",
"APP_TYPE": "infer.lrm",
"NUMBA_THREADING_LAYER": 'omp',
})
from openlrm.runners import REGISTRY_RUNNERS
from openlrm.runners.infer.base_inferrer import Inferrer
InferrerClass : Inferrer = REGISTRY_RUNNERS[os.getenv("APP_TYPE")]
with InferrerClass() as inferrer:
init_preprocessor()
if not bool(os.getenv('SPACE_ID')):
from openlrm.utils.proxy import no_proxy
demo = no_proxy(demo_openlrm)
else:
demo = demo_openlrm
demo(infer_impl=inferrer.infer_single)
if name == 'main':
launch_gradio_app()
..............................................................................................................................................................................................
config.json: 100% 322/322 [00:00<00:00, 1.49MB/s]
pytorch_model.bin: 100% 1.04G/1.04G [00:08<00:00, 116MB/s]
[2024-03-22 19:07:46,354] openlrm.models.modeling_lrm: [INFO] Using DINOv2 as the encoder
/content/OpenLRM/openlrm/models/encoders/dinov2/layers/swiglu_ffn.py:43: UserWarning: xFormers is available (SwiGLU)
warnings.warn("xFormers is available (SwiGLU)")
/content/OpenLRM/openlrm/models/encoders/dinov2/layers/attention.py:27: UserWarning: xFormers is available (Attention)
warnings.warn("xFormers is available (Attention)")
/content/OpenLRM/openlrm/models/encoders/dinov2/layers/block.py:39: UserWarning: xFormers is available (Block)
warnings.warn("xFormers is available (Block)")
Downloading: "https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_reg4_pretrain.pth" to /root/.cache/torch/hub/checkpoints/dinov2_vitb14_reg4_pretrain.pth
100% 330M/330M [00:01<00:00, 174MB/s]
Downloading data from 'https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnx' to file '/root/.u2net/u2net.onnx'.
100%|████████████████████████████████████████| 176M/176M [00:00<00:00, 861GB/s]
Traceback (most recent call last):
File "/content/OpenLRM/openlrm/utils/proxy.py", line 34, in wrapper
return func(*args, **kwargs)
File "/content/OpenLRM/app.py", line 132, in demo_openlrm
output_mesh = gr.Mesh(label="Rendered Mesh", format="ply", width="auto", autoplay=True)
AttributeError: module 'gradio' has no attribute 'Mesh'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/content/OpenLRM/app.py", line 226, in
launch_gradio_app()
File "/content/OpenLRM/app.py", line 221, in launch_gradio_app
demo(infer_impl=inferrer.infer_single)
File "/content/OpenLRM/openlrm/utils/proxy.py", line 36, in wrapper
os.environ['http_proxy'] = http_proxy
File "/usr/lib/python3.10/os.py", line 685, in setitem
value = self.encodevalue(value)
File "/usr/lib/python3.10/os.py", line 757, in encode
raise TypeError("str expected, not %s" % type(value).name)
TypeError: str expected, not NoneType
..............................................................................................................................................................................................
Thanks!
Hi,
Plz refer to this issue here. #24 (comment)
The newest commit has solved this problem.