KeyError: "EncoderDecoder: 'VIT_MLA is not in the models registry'"

[APeiZou]

hello，why is it this KeyError: "EncoderDecoder: 'VIT_MLA is not in the models registry'". I should do what to registry VIT_MLA

[XiongweiWu]

@APeiZou Hi, you may just follow the installation instructions, and the VIT_MLA backbone will be automatically registered into the software.

I think the code should be ok to use but anyway I will double-check it soon since I am a bit busy with the coming ddl now. Once I finish the checking I will update here.

[EricBizet]

I am currently encountering the same issue (KeyError: "EncoderDecoder: 'VIT_MLA is not in the models registry'") while trying to run an inference and get_started.md does not provide information on setting up the VIT_MLA backbone.

Current code:

from mmseg.apis import init_segmentor, inference_segmentor, show_result_pyplot
from mmseg.core.evaluation import get_palette
config_file = 'configs/SETR_MLA_768x768_80k_base.py'
checkpoint_file = 'models/model.pth'
# build the model from a config file and a checkpoint file
model = init_segmentor(config_file, checkpoint_file, device='cuda:0')

Any updates on this one ? Thank you

[XiongweiWu]

@EricBizet Sorry for replying late. I will check it this weekend. Sorry again I am super busy with other 2 projects and I fail to answer you on time.

[EricBizet]

No worries, I managed to get it to work be using this Dockerfile I made based off yours:

FROM pytorch/pytorch:1.6.0-cuda10.1-cudnn7-devel

WORKDIR /usr/src/app

RUN apt-get update && apt-get install -y git
RUN apt-get install ffmpeg libsm6 libxext6  -y

RUN pip install mmcv-full==1.3.0 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/index.html

RUN git clone https://github.com/LARC-CMU-SMU/FoodSeg103-Benchmark-v1.git mmsegmentation
WORKDIR /usr/src/app/mmsegmentation
RUN pip install -e .

RUN pip install jupyterlab==2.2.0 timm

ENTRYPOINT ["/bin/bash", "-c", "jupyter notebook --no-browser --allow-root --ip 0.0.0.0"]

The installation instructions were about pulling the official mmseg repo which does not include the same backbone files as yours (and if doing so, your need to copy and paste files manually in the backbone folder).

Cheers,

Eric

[XiongweiWu]

@EricBizet Thx, I also find this issue. Current install instruction is not comprehensive, and I will refine the code as well as the install instruction soon.