diff --git a/api/requirements/amd-linux.txt b/api/requirements/amd-linux.txt
index c44cd431..063d6d61 100644
--- a/api/requirements/amd-linux.txt
+++ b/api/requirements/amd-linux.txt
@@ -1,4 +1,4 @@
 --extra-index-url https://download.pytorch.org/whl/rocm5.4.2
 torch==2.0.0
 torchvision==0.15.1
-https://download.onnxruntime.ai/onnxruntime_training-1.14.1%2Brocm54-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
\ No newline at end of file
+https://download.onnxruntime.ai/onnxruntime_training-1.14.1%2Brocm54-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
\ No newline at end of file
diff --git a/docs/architecture.md b/docs/architecture.md
new file mode 100644
index 00000000..533b9be9
--- /dev/null
+++ b/docs/architecture.md
@@ -0,0 +1,12 @@
+# onnx-web Architecture
+
+## Contents
+
+## Chain Pipelines
+
+Chain pipelines provide a way to run any number and combination of other pipelines, automatically splitting them
+into the correct tile size and passing the output on to the next stage.
+
+## Worker Pool
+
+The worker pool is a process pool that manages one or more worker processes for each device (typically a GPU).