split tinybox benchmark into two (#3741)

* split tinybox benchmark into two

* symlinks

.github/workflows/benchmark.yml

@@ -131,8 +131,6 @@ jobs:
     steps:
     - name: Checkout Code
       uses: actions/checkout@v4
-    - name: Show off tinybox
-      run: /opt/rocm/bin/rocm-bandwidth-test
     - name: Symlink models and datasets
       run: |
         mkdir -p weights
@@ -143,6 +141,8 @@ jobs:
         ln -s /raid/weights/LLaMA-2 weights/LLaMA-2
         mkdir -p extra/datasets
         ln -s /raid/datasets/imagenet extra/datasets/imagenet
+    - name: Show off tinybox
+      run: /opt/rocm/bin/rocm-bandwidth-test
     - name: Run model inference benchmark
       run: LD_PRELOAD="/opt/rocm/lib/libhsa-runtime64.so" HSA=1 NOCLANG=1 python3 test/external/external_model_benchmark.py
     - name: Test speed vs torch
@@ -167,6 +167,44 @@ jobs:
       run: |
         HSA=1 JIT=0 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing | tee gpt2_unjitted.txt
         HSA=1 JIT=1 python3 examples/gpt2.py --prompt "Hello." --count 10 --temperature 0 --timing | tee gpt2_jitted.txt
+    - uses: actions/upload-artifact@v4
+      with:
+        name: Speed (AMD)
+        path: |
+          onnx_inference_speed.csv
+          torch_speed.txt
+          llama_unjitted.txt
+          llama_jitted.txt
+          llama_beam.txt
+          llama_2_70B.txt
+          gpt2_unjitted.txt
+          gpt2_jitted.txt
+          matmul.txt
+          sd.txt
+          mixtral.txt
+
+  testmoreamdbenchmark:
+    name: tinybox Training
+    runs-on: [self-hosted, Linux, tinybox]
+    defaults:
+      run:
+        shell: bash -o pipefail {0}
+    if: github.repository_owner == 'tinygrad'
+    env:
+      PYTHONPATH: .
+    steps:
+    - name: Checkout Code
+      uses: actions/checkout@v4
+    - name: Symlink models and datasets
+      run: |
+        mkdir -p weights
+        ln -s ~/tinygrad/weights/bpe_simple_vocab_16e6.txt.gz weights/bpe_simple_vocab_16e6.txt.gz
+        ln -s ~/tinygrad/weights/LLaMA weights/LLaMA
+        ln -s ~/tinygrad/extra/datasets/cifar-10-python.tar.gz extra/datasets/cifar-10-python.tar.gz
+        ln -s /raid/weights/mixtral-8x7b-32kseqlen weights/mixtral-8x7b-32kseqlen
+        ln -s /raid/weights/LLaMA-2 weights/LLaMA-2
+        mkdir -p extra/datasets
+        ln -s /raid/datasets/imagenet extra/datasets/imagenet
     - name: Run 10 CIFAR training steps
       run: HSA=1 STEPS=10 python3 examples/hlb_cifar10.py | tee train_cifar.txt
     - name: Run 10 CIFAR training steps w HALF
@@ -183,10 +221,8 @@ jobs:
       run: HIP=1 BENCHMARK=10 BS=624 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py | tee train_resnet.txt
     - uses: actions/upload-artifact@v4
       with:
-        name: Speed (AMD)
+        name: Speed (AMD Training)
         path: |
-          onnx_inference_speed.csv
-          torch_speed.txt
           train_cifar.txt
           train_cifar_half.txt
           train_cifar_wino.txt
@@ -194,12 +230,3 @@ jobs:
           train_resnet.txt
           train_resnet_one_gpu.txt
           train_cifar_six_gpu.txt
-          llama_unjitted.txt
-          llama_jitted.txt
-          llama_beam.txt
-          llama_2_70B.txt
-          gpt2_unjitted.txt
-          gpt2_jitted.txt
-          matmul.txt
-          sd.txt
-          mixtral.txt