3 ms·
Not sure about the benchmark, but the code includes the option for mixed precision training via Apex/AMP.
by joshvm 6y ago
Not sure about the benchmark, but the code includes the option for mixed precision training via Apex/AMP.
- bcatanzaro 6y agoIf you click around enough you’ll see they benchmarked in 32-bit FP. Glad they have a mixed precision training option but I really think it’s a mistake in 2020 to do work related to efficient inference using 32-but FP. The problem is that your conclusions aren’t independent of this choice. A different network might be far better in terms of accuracy/speed tradeoffs when evaluated at a lower precision. But there is no reason to use 32-but precision for inference, so this is just a big mistake.