use threefry in stable diffusion benchmark (#4988)

also updated default steps to 10. easier to tell the image is following the prompt.

.github/workflows/benchmark.yml

@@ -38,7 +38,7 @@ jobs:
     # if: github.event.inputs.run_process_replay == 'true' || contains(github.event.head_commit.message, '[run_process_replay]') || contains(github.event.pull_request.title, '[run_process_replay]')
     # run: echo "RUN_PROCESS_REPLAY=1" >> $GITHUB_ENV
     - name: Run Stable Diffusion
-      run: JIT=2 python3 examples/stable_diffusion.py --seed 0 --noshow --timing | tee sd.txt
+      run: JIT=2 THREEFRY=1 python3 examples/stable_diffusion.py --seed 0 --noshow --timing | tee sd.txt
     - name: Run model inference benchmark
       run: METAL=1 python3 test/external/external_model_benchmark.py
     - name: Test speed vs torch
@@ -155,7 +155,7 @@ jobs:
     - name: Fuzz Padded Tensor Core GEMM(PTX)
       run: CUDA=1 PTX=1 M_START=12 M_STOP=20 M_STEP=1 N_START=6 N_STOP=10 N_STEP=1 K_START=28 K_STOP=36 K_STEP=1 HALF=1 TC_OPT=2 python3 ./extra/gemm/fuzz_matmul.py
     - name: Run Stable Diffusion
-      run: CUDA=1 python3 examples/stable_diffusion.py --seed 0 --noshow --timing | tee sd.txt
+      run: CUDA=1 THREEFRY=1 python3 examples/stable_diffusion.py --seed 0 --noshow --timing | tee sd.txt
     - name: Run LLaMA
       run: |
         CUDA=1 JIT=0 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing | tee llama_unjitted.txt
@@ -306,7 +306,7 @@ jobs:
     #- name: Fuzz Padded Tensor Core GEMM
     #  run: HSA=1 M_START=12 M_STOP=20 M_STEP=1 N_START=12 N_STOP=20 N_STEP=1 K_START=28 K_STOP=36 K_STEP=1 HALF=1 TC_OPT=2 DEBUG=2 python3 ./extra/gemm/fuzz_matmul.py
     - name: Run Stable Diffusion
-      run: AMD=1 python3 examples/stable_diffusion.py --seed 0 --noshow --timing | tee sd.txt
+      run: AMD=1 THREEFRY=1 python3 examples/stable_diffusion.py --seed 0 --noshow --timing | tee sd.txt
     - name: Run LLaMA 7B
       run: |
         AMD=1 JIT=0 python3 examples/llama.py --gen 1 --prompt "Hello." --count 10 --temperature 0 --timing | tee llama_unjitted.txt

examples/stable_diffusion.py

@@ -576,7 +576,7 @@ class StableDiffusion:
 if __name__ == "__main__":
   default_prompt = "a horse sized cat eating a bagel"
   parser = argparse.ArgumentParser(description='Run Stable Diffusion', formatter_class=argparse.ArgumentDefaultsHelpFormatter)
-  parser.add_argument('--steps', type=int, default=5, help="Number of steps in diffusion")
+  parser.add_argument('--steps', type=int, default=10, help="Number of steps in diffusion")
   parser.add_argument('--prompt', type=str, default=default_prompt, help="Phrase to render")
   parser.add_argument('--out', type=str, default=Path(tempfile.gettempdir()) / "rendered.png", help="Output filename")
   parser.add_argument('--noshow', action='store_true', help="Don't show the image")