mlp_mixer_b16_224_miil

cv/classification/mlp_mixer_b16_224_miil/README.md

@@ -0,0 +1,66 @@
+## mlp_mixer_b16_224_miil
+
+### Installation
+- Install the latest version of OneFlow
+```bash
+python3 -m pip install oneflow -f https://staging.oneflow.info/branch/master/[PLATFORM]
+```
+Find more information on [install oneflow](https://github.com/Oneflow-Inc/oneflow#install-oneflow)
+
+- Install flowvision
+
+Then install the latest stable release of flowvision
+
+```bash
+pip install flowvision==0.2.1
+```
+
+- Install other requirements
+```bash
+python3 -m pip install -r requirements.txt
+```
+
+### Dataset
+#### ImageNet
+For ImageNet dataset, you can download it from http://image-net.org/. We provide the following two ways to load data:
+
+- For standard folder dataset, move validation images to labeled sub-folders. The file structure should look like:
+  ```bash
+  $ tree data
+  imagenet
+  ├── train
+  │   ├── class1
+  │   │   ├── img1.jpeg
+  │   │   ├── img2.jpeg
+  │   │   └── ...
+  │   ├── class2
+  │   │   ├── img3.jpeg
+  │   │   └── ...
+  │   └── ...
+  └── val
+      ├── class1
+      │   ├── img4.jpeg
+      │   ├── img5.jpeg
+      │   └── ...
+      ├── class2
+      │   ├── img6.jpeg
+      │   └── ...
+      └── ...
+
+
+### Training
+
+You can use bash script `train.sh` to train this model.
+
+```bash
+sh train.sh
+```
+
+### inference
+
+Bash script `infer.sh` is used to infer the trained model.
+
+```bash
+sh infer.sh
+```
+

cv/classification/mlp_mixer_b16_224_miil/config.py

@@ -0,0 +1 @@
+../config.py

cv/classification/mlp_mixer_b16_224_miil/configs

@@ -0,0 +1 @@
+../configs

cv/classification/mlp_mixer_b16_224_miil/data

@@ -0,0 +1 @@
+../data

cv/classification/mlp_mixer_b16_224_miil/infer.sh

@@ -0,0 +1,16 @@
+export PYTHONPATH=$PWD:$PYTHONPATH
+set -aux
+
+GPU_NUMS=1
+PORT=12346
+MODEL_ARCH="mlp_mixer_b16_224_miil"
+
+python3 -m oneflow.distributed.launch \
+        --nproc_per_node $GPU_NUMS \
+        --master_addr 127.0.0.1 \
+        --master_port $PORT \
+        main.py \
+        --cfg configs/default_settings.yaml \
+        --model_arch $MODEL_ARCH \
+        --throughput
+

cv/classification/mlp_mixer_b16_224_miil/logger.py

@@ -0,0 +1 @@
+../logger.py

cv/classification/mlp_mixer_b16_224_miil/lr_scheduler.py

@@ -0,0 +1 @@
+../lr_scheduler.py

cv/classification/mlp_mixer_b16_224_miil/main.py

@@ -0,0 +1 @@
+../main.py

cv/classification/mlp_mixer_b16_224_miil/optimizer.py

@@ -0,0 +1 @@
+../optimizer.py

cv/classification/mlp_mixer_b16_224_miil/requirements.txt

@@ -0,0 +1 @@
+../requirements.txt

cv/classification/mlp_mixer_b16_224_miil/train.sh

@@ -0,0 +1,16 @@
+export PYTHONPATH=$PWD:$PYTHONPATH
+set -aux
+
+GPU_NUMS=8
+PORT=12346
+MODEL_ARCH="mlp_mixer_b16_224_miil"
+
+python3 -m oneflow.distributed.launch \
+        --nproc_per_node $GPU_NUMS \
+        --master_addr 127.0.0.1 \
+        --master_port $PORT \
+        main.py \
+        --cfg configs/default_settings.yaml \
+        --model_arch $MODEL_ARCH \
+        --lr 0.001
+

cv/classification/mlp_mixer_b16_224_miil/utils.py

@@ -0,0 +1 @@
+../utils.py