mirrors/tinygrad
2
0
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-06-24 02:14:17 +00:00
Code Issues Projects Releases Packages Wiki Activity

remove the different rtol / atol for openpilot CUDA in benchmark (#2907)

Browse source 
not sure what the issue was but seems to be fixed on master
This commit is contained in:
chenyu 2023-12-21 22:23:39 -05:00 • committed by GitHub
commit 1bbeb3fe2f
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 0 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

1
test/external/external_model_benchmark.py vendored
View file

@ -99,7 +99,6 @@ def benchmark_model(m, devices, validate_outs=False):
if validate_outs:
rtol, atol = 2e-3, 2e-3 # tolerance for fp16 models
if m == "openpilot" and 'CUDA' in devices: rtol, atol = 0.1, 0.1 # TODO: why is this broken?
inputs = {k:Tensor(inp) for k,inp in np_inputs.items()}
tinygrad_model = get_run_onnx(onnx_model)
tinygrad_out = tinygrad_model(inputs)