github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-09 23:18:04 -05:00
Code Issues Packages Projects Releases Wiki Activity

no amdgpu kernel driver (#10408)

Browse Source 
* no amdgpu kernel driver

* don't test hip

* lower req
This commit is contained in:
George Hotz
2025-05-18 20:52:39 -07:00
committed by GitHub
parent 4b1f1a47bb
commit b06291077c
2 changed files with 17 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
.github/workflows/benchmark.yml vendored
View File

@@ -352,8 +352,10 @@ jobs:
steps:
- name: Checkout Code
uses: actions/checkout@v4
- name: Insert amdgpu
run: sudo modprobe amdgpu
- name: Remove amdgpu
run: sudo rmmod amdgpu || true
#- name: Insert amdgpu
# run: sudo modprobe amdgpu
- name: Symlink models and datasets
run: |
mkdir -p weights
@@ -372,12 +374,12 @@ jobs:
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: reset process replay
run: test/external/process_replay/reset.py
- name: setup perflevel
run: |
examples/mlperf/training_submission_v4.1/tinycorp/benchmarks/bert/implementations/tinybox_red/setup.sh
rocm-smi
- name: Show off tinybox
run: /opt/rocm/bin/rocm-bandwidth-test
#- name: setup perflevel
# run: |
# examples/mlperf/training_submission_v4.1/tinycorp/benchmarks/bert/implementations/tinybox_red/setup.sh
# rocm-smi
#- name: Show off tinybox
# run: /opt/rocm/bin/rocm-bandwidth-test
# TODO: unstable on AMD
#- 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
@@ -397,13 +399,13 @@ jobs:
run: AMD=1 SHOULD_USE_TC=1 HALF=1 DEBUG=2 ATOL=2e-2 python3 extra/gemm/simple_matmul.py | tee matmul_amd.txt
- name: Test AMD=1
run: DEBUG=2 AMD=1 python -m pytest -rA test/test_tiny.py
- name: Test HIP=1
run: DEBUG=2 HIP=1 python -m pytest -rA test/test_tiny.py
#- name: Test HIP=1
# run: DEBUG=2 HIP=1 python -m pytest -rA test/test_tiny.py
# TODO: AMD compiler bug causes this to fail
#- 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: Remove amdgpu
run: sleep 10 && sudo rmmod amdgpu # sleep a bit to let the driver unload the prev pid.
#- name: Remove amdgpu
# run: sleep 10 && sudo rmmod amdgpu # sleep a bit to let the driver unload the prev pid.
- name: Test AM cold start time
run: time AMD=1 AM_RESET=1 python3 test/test_tiny.py TestTiny.test_plus
- name: Test AM warm start time
@@ -428,8 +430,8 @@ jobs:
run: BENCHMARK_LOG=llama3_beam_4gpu AMD=1 JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0 | tee llama3_four_gpu.txt
# - name: Run LLaMA-3 8B on 6 GPUs
# run: AMD=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 6 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0 | tee llama3_six_gpu.txt
- name: Restore amdgpu
run: sudo modprobe amdgpu
#- name: Restore amdgpu
# run: sudo modprobe amdgpu
# - name: Run LLaMA-2 70B
# run: AMD=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama.py --gen 2 --size 70B --shard 6 --prompt "Hello." --count 10 --temperature 0 --timing | tee llama_2_70B.txt
- name: Run Mixtral 8x7B

2
test/external/speed_v_theoretical.py vendored
View File

@@ -90,7 +90,7 @@ class TestKernelSpeed(unittest.TestCase):
def test_conv_3x3_256_32_32_256_256(self): self._test_conv_3x3(256, 32, 32, 256, 256, nv_tflops=27, amd_tflops=20)
# theoretical is nv_tflops=165, amd_tflops=123
def test_gemm_4096(self): self._test_matmul(4096, nv_tflops=115, amd_tflops=74)
def test_gemm_4096(self): self._test_matmul(4096, nv_tflops=115, amd_tflops=65)
def test_gemm_8192(self): self._test_matmul(8192, nv_tflops=125, amd_tflops=65)
# theoretical is nv_gbs=1008, amd_gbs=960