Fix some of the tests on GPU-enabled systems #22
Merged
+10
−16
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The tests for steps 5+ place the test tensors on the CPU, but on GPU-enabled systems the graph is by default placed on the GPU. This causes a mismatch on compilation, leading to these tests failing with cryptic errors on systems with NVIDIA or AMD GPUs.
Additionally, once the graph is run on the GPU, the default datatype is bfloat16 but that datatype isn't supported in NumPy. When results are read back into the NumPy array, these reads fail without an appropriate cast to float32 values.
This fixes these tests for steps 5, 6, 7, 9, 10, and 11 and enables them to run on GPU-equipped machines. Steps 8 and 12 have additional issues to fix, even when this casting is handled. They will be fixed later.
Fixes issue #12