ci: more parallelism, less duplication (#16509)

.github/workflows/test.yml

@@ -43,31 +43,21 @@ jobs:
         cd $HOME/test_external_dir
         uv venv venv
         uv pip install --python venv $GITHUB_WORKSPACE mypy
+        cp $GITHUB_WORKSPACE/examples/beautiful_mnist.py .
         venv/bin/python -c "from tinygrad.tensor import Tensor; print(Tensor([1,2,3,4,5]))"
         venv/bin/mypy -c "from tinygrad.tensor import Tensor; print(Tensor([1,2,3,4,5]))"
-    - name: Run beautiful_mnist with tinygrad only
-      run: |
-        mkdir $GITHUB_WORKSPACE/test_dir
-        cd $GITHUB_WORKSPACE/test_dir
-        uv venv venv
-        uv pip install --python venv $GITHUB_WORKSPACE
-        cp $GITHUB_WORKSPACE/examples/beautiful_mnist.py .
         BS=2 STEPS=10 MAX_BUFFER_SIZE=0 venv/bin/python beautiful_mnist.py
-    - name: Test Docs Build
-      run: python -m mkdocs build --strict
     - name: Test Docs
-      run: python docs/abstractions3.py
-    - name: Test README
-      run: awk '/```python/{flag=1;next}/```/{flag=0}flag' README.md > README.py &&  python README.py
-    - name: Test Quickstart
-      run: awk '/```python/{flag=1;next}/```/{flag=0}flag' docs/quickstart.md > quickstart.py &&  python quickstart.py
+      run: |
+        parallel --link --tagstring '[{1}]' '{2}' \
+          ::: mkdocs abstractions3 readme quickstart export \
+          ::: 'mkdocs build --strict' \
+              'python docs/abstractions3.py' \
+              $'awk \'/```python/{flag=1;next}/```/{flag=0}flag\' README.md | python' \
+              $'awk \'/```python/{flag=1;next}/```/{flag=0}flag\' docs/quickstart.md | python' \
+              'DEV=CPU python examples/compile_efficientnet.py > recognize.c && clang -O2 recognize.c -lm -o recognize && cat test/models/efficientnet/Chicken.jpg | ./recognize | grep cock'
     - name: Test DEBUG
       run: DEBUG=100 python3 -c "from tinygrad import Tensor; N = 1024; a, b = Tensor.rand(N, N), Tensor.rand(N, N); c = (a.reshape(N, 1, N) * b.T.reshape(1, N, N)).sum(axis=2); print((c.numpy() - (a.numpy() @ b.numpy())).mean())"
-    - name: Compile EfficientNet to C and test it
-      run: |
-        DEV=CPU python examples/compile_efficientnet.py > recognize.c
-        clang -O2 recognize.c -lm -o recognize
-        cat test/models/efficientnet/Chicken.jpg | ./recognize | grep cock
 
   torchbackend:
     name: Torch Backend Tests
@@ -202,7 +192,7 @@ jobs:
     - name: Lint bad-indentation and trailing-whitespace with pylint
       run: python -m pylint --disable=all -e W0311 -e C0303 --jobs=0 --indent-string='  ' --recursive=y .
     - name: Run pre-commit linting hooks
-      run: SKIP=tiny,tests,example pre-commit run --all-files
+      run: SKIP=tiny,tests,example,mypy pre-commit run --all-files
     - name: Lint additional files with ruff
       run: |
         python3 -m ruff check examples/mlperf/ --ignore E501
@@ -263,7 +253,7 @@ jobs:
         deps: testing_unit
         llvm: 'true'
     - name: Run pre-commit test hooks
-      run: SKIP=ruff,mypy pre-commit run --all-files
+      run: SKIP=ruff,mypy,tests pre-commit run --all-files
     - name: Check Device.DEFAULT
       run: python -c "from tinygrad import Device; assert Device.DEFAULT == 'CPU', Device.DEFAULT"
     - name: Run unit tests
@@ -463,15 +453,17 @@ jobs:
         uses: ./.github/actions/setup-tinygrad
         with:
           key: apps_llm
-      - name: Test 1B LLM (llama)
-        run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.llm --model llama3.2:1b | tee /dev/stderr | grep -i rooster
-      - name: Test 1B LLM (llama q4)
-        run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.llm --model llama3.2:1b-q4 | tee /dev/stderr | grep -i rooster
-      - name: Test 1B LLM (qwen3.5)
-        run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.llm --model qwen3.5:0.8b | tee /dev/stderr | grep -i rooster
-      - name: Test 1B LLM (qwen)
-        # NOTE: qwen is dumb and only knows about female chickens
-        run: echo "What's a female chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.llm --model qwen3:0.6b | tee /dev/stderr | grep -i hen
+      - name: Test LLMs
+        env:
+          MAX_BUFFER_SIZE: 0
+        run: |
+          parallel --link --tagstring '[{1}]' '{2}' \
+            ::: llama 'llama q4' qwen3.5 qwen \
+            ::: $'echo "What\'s a male chicken called? Answer with only one word." | python3 -m tinygrad.llm --model llama3.2:1b | tee /dev/stderr | grep -i rooster' \
+                $'echo "What\'s a male chicken called? Answer with only one word." | python3 -m tinygrad.llm --model llama3.2:1b-q4 | tee /dev/stderr | grep -i rooster' \
+                $'echo "What\'s a male chicken called? Answer with only one word." | python3 -m tinygrad.llm --model qwen3.5:0.8b | tee /dev/stderr | grep -i rooster' \
+                $'echo "What\'s a female chicken called? Answer with only one word." | python3 -m tinygrad.llm --model qwen3:0.6b | tee /dev/stderr | grep -i hen'
+                # NOTE: qwen is dumb and only knows about female chickens
 
 # ****** Models Tests ******
 
@@ -531,14 +523,8 @@ jobs:
         deps: testing
         llvm: "true"
         qemu: "true"
-    - name: Set MOCKDSP env
-      run: printf "MOCKDSP=1" >> $GITHUB_ENV
-    - name: Run test_tiny on DSP
-      run: DEBUG=2 DEV=DSP python test/test_tiny.py
-    - name: Test transcendentals
-      run: DEBUG=2 DEV=DSP python test/backend/test_transcendental.py TestTranscendentalVectorized
-    - name: Test quantize onnx
-      run: DEBUG=2 DEV=DSP python3 test/backend/test_quantize_onnx.py
+    - name: Run tests
+      run: MOCKDSP=1 DEV=DSP python -m pytest -n=auto test/test_tiny.py test/backend/test_transcendental.py::TestTranscendentalVectorized test/backend/test_quantize_onnx.py
 
   testlinux:
     strategy: