split metal model tests (#12119)

* split metal model tests

* llama too

.github/workflows/test.yml

@@ -411,6 +411,8 @@ jobs:
       - name: Run process replay tests
         uses: ./.github/actions/process-replay
 
+# ****** ONNX Tests ******
+
   testonnxcpu:
     name: 'ONNX (CPU) Tests'
     runs-on: ubuntu-22.04
@@ -488,6 +490,8 @@ jobs:
       - name: Test 1B LLM
         run: echo "What's a male chicken called? Answer with only one word." | MAX_BUFFER_SIZE=0 python3 -m tinygrad.apps.llm | grep -i rooster
 
+# ****** Models Tests ******
+
   testmodels:
     name: Models (llvm+cpu+gpu)
     runs-on: ubuntu-22.04
@@ -513,6 +517,29 @@ jobs:
       - name: Run process replay tests
         uses: ./.github/actions/process-replay
 
+  testmetalmodels:
+    name: Models (metal)
+    runs-on: macos-14
+    timeout-minutes: 20
+    env:
+      IGNORE_OOB: 0
+
+    steps:
+    - name: Checkout Code
+      uses: actions/checkout@v4
+    - name: Setup Environment
+      uses: ./.github/actions/setup-tinygrad
+      with:
+        key: metal
+        deps: testing
+        python-version: '3.11'
+    - name: Test models (Metal)
+      run: METAL=1 python -m pytest -n=auto test/models -v --durations=20
+    - name: Test LLaMA compile speed
+      run: METAL=1 python test/external/external_test_speed_llama.py
+
+# ****** Feature Tests ******
+
   testrangeify:
     name: Linux (rangeify)
     runs-on: ubuntu-24.04
@@ -807,7 +834,7 @@ jobs:
 
 # ****** OSX Tests ******
 
-  testmetal2:
+  testmetal:
     name: MacOS (unit)
     runs-on: macos-14
     timeout-minutes: 20
@@ -820,23 +847,19 @@ jobs:
     - name: Setup Environment
       uses: ./.github/actions/setup-tinygrad
       with:
-        key: metal2
+        key: metal
         deps: testing
         python-version: '3.11'
         amd: 'true'
         cuda: 'true'
         ocelot: 'true'
         llvm: 'true'
-    - name: Test models (Metal)
-      run: METAL=1 python -m pytest -n=auto test/models -v --durations=20
     - name: Run ONNX
       run: METAL=1 python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
     - name: Test tensor core ops (fake)
       run: TC=2 METAL=1 DEBUG=3 python test/test_ops.py TestOps.test_gemm
     - name: Test tensor core ops (real)
       run: METAL=1 DEBUG=3 python test/test_ops.py TestOps.test_big_gemm
-    - name: Test LLaMA compile speed
-      run: METAL=1 python test/external/external_test_speed_llama.py
     - name: Test Beam Search
       run: METAL=1 IGNORE_BEAM_CACHE=1 python3 -m pytest extra/optimization/test_beam_search.py
     #- name: Fuzz Test linearizer