moves what happens where around a bit and updates readme

.github/workflows/main.yml

@@ -109,6 +109,9 @@ jobs:
     name: GPU CI
     runs-on: ubuntu-latest
     timeout-minutes: 15
+    # TODO: @jakep
+    # once you fix the base checks, uncomment the following line for expense reasons
+    # needs: [checks]
     env:
       BEAKER_TOKEN: ${{ secrets.BEAKER_TOKEN }}
       BEAKER_IMAGE: chrisw/olmocr-gpu-ci

README.md

@@ -56,13 +56,12 @@ conda activate olmocr
 
 git clone https://github.com/allenai/olmocr.git
 cd olmocr
-pip install -e .
-```
 
-Install sglang with [flashinfer](https://github.com/flashinfer-ai/flashinfer) if you want to run inference on GPU.
-```bash
-pip install sgl-kernel==0.0.3.post1 --force-reinstall --no-deps
-pip install "sglang[all]==0.4.2" --find-links https://flashinfer.ai/whl/cu124/torch2.4/flashinfer/
+# If running on CPU, run this:
+pip install -e .
+
+# If running on GPU, run this instead:
+pip install -e .[gpu] --find-links https://flashinfer.ai/whl/cu124/torch2.4/flashinfer/
 ```
 
 ### Local Usage Example

scripts/beaker/gpu-ci-script.sh

@@ -9,8 +9,7 @@ git clone https://github.com/allenai/olmocr.git olmocr \
     .[gpu] \
     pytest \
     --find-links https://flashinfer.ai/whl/cu124/torch2.4/flashinfer/ \
-  && python -m olmocr.pipeline ./localworkspace --pdfs tests/gnarly_pdfs/*.pdf \
-  && pytest tests/gnarly_pdfs/test_gnarly_pdfs.py
+  && bash tests/gnarly_pdfs/test_gnarly_pdfs.sh
 
 
 

tests/gnarly_pdfs/test_gnarly_pdfs.sh

@@ -0,0 +1,6 @@
+#/usr/bin/bash
+
+set -ex
+
+python -m olmocr.pipeline ./localworkspace --pdfs tests/gnarly_pdfs/*.pdf \
+  && pytest tests/gnarly_pdfs/test_gnarly_pdfs.py