v0.4.3
v0.4.3 (2024-04-28)
Fix
- fix: 🐛 do not generate inf values if elements has infinity disabled for bfloat16 tensors.
Before, if a float strategy that prevented infinities was provided to as elements
to tensor_strategy
, the resultant tensor could still occasionally have infinities because the strategy would generate float32 values that exceeded the maximium/minimum values for bfloat16. (cda8577
)
Unknown
- Merge pull request #21 from qthequartermasterman/fix-bfloat16-infinities
fix: 🐛 do not generate inf values if elements has infinity disabled for bfloat16 tensors. (8398018
)