[QDP] fix test bug #734
[QDP] fix test bug #734
Conversation
Signed-off-by: Hsien-Cheng Huang <[email protected]>
ryankert01
left a comment
ryankert01
left a comment
There was a problem hiding this comment.
sorry for my vscode rust extension is changing a lot of our codespace, I've highlighted the changes.
PTAL @guan404ming @rich7420 @400Ping @machichima
| assert_eq!( | ||
| tensor.dtype.bits, 128, | ||
| "Should be 128 bits (2x64-bit floats)" | ||
| ); |
There was a problem hiding this comment.
this is the failed test
There was a problem hiding this comment.
I am thinking as we change the precision to float32 by default, should we set the expected value to 2x32 rather than change the precision back to float64?
There was a problem hiding this comment.
see this, working on float32 right now.
| println!("Testing DLPack tensor metadata..."); | ||
|
|
||
| let engine = match QdpEngine::new(0) { | ||
| let engine = match QdpEngine::new_with_precision(0, Precision::Float64) { |
There was a problem hiding this comment.
This is what I changed
|
Can you provide the reproduction scrip you run for the error? |
cargo test |
|
strange, my local test passed. |
|
@400Ping As I'm concerned, there's a PR change default from fp64 to fp32 |
|
I can reproduce the error on the newest 
|
Yes, I am also working on the follow up as well. |
|
local test all passed. |
4 participants
Purpose of PR
default has been changed to f32, so the bits result would be 32*2=64
solution: Make the test not init the executor to default, instead init to a predetermined f64
Related Issues or PRs
Closes #733
Changes Made
Breaking Changes
Checklist
TODO
new()fp32, bring back the coverage