34.51it/s on A40
1 step That's like 29ms?
COMPUTING
-
A40 GPU achieves 34.51 iterations per second performance
By
–
-
When Can We Use Pure BF16 in AI Models?
By
–
Of course. But lots of people have observed the same thing in other contexts, and it's what we'd expect based on first principles. To me, the question isn't: when can't we use pure bf16, but when can we!
-
FP32 vs BF16 vs FP16 Mixed Precision Training Options
By
–
The main options are fp32 (or tf32), bf16, or fp16+mixed prec afaik.
-
Microsoft Cloud Revenue Beats Expectations While Google Underperforms
By
–
Tale of two clouds: Microsoft tops, Google flops Microsoft’s cloud computing revenue topped expectations, while Alphabet’s Google missed. Microsoft CEO Satya Nadella offered some anecdotal evidence to suggest progress in AI is helping, notably, over a million people paying to
-
Science and Engineering: Unlocking Universe Secrets and Building Tomorrow
By
–
the beauty of science & engineering is that there will always be more secrets of the universe to unlock, and then new things to build
-
BF16 Benefits Beyond Loss Scaling in Mixed Precision
By
–
Other than avoiding loss scaling, are there other benefits to using bf16? It still seems to require mixed precision afaict, right?
-
Cloud Provider Simplifies Oracle Database Migration Services
By
–
and another 'We are the only other cloud provider to run services making it simpler for customers to migrate their Oracle databases to our cloud.'
-
Distributed GPU Processing: File Synchronization and Aggregation
By
–
because in distributed mode, each GPU runs its own process. so this code will run simultaneously multiple times, once per GPU. so in this setting, each GPU writes a file, waits for everyone, then aggregates all the files, waits again, then deletes its own file 🙂
-
RISC-V Space Technology Event in Denver Next Week
By
–
We will be in Denver next week talking all things #riscv in #space, alongside our partners @BrekerSystems
, @Codasip
, @arteris_noc
, RISC-AI,
@ImperasSoftware, @cycuity and @Synopsys for an event exploring the intersection of #riscv, electronics design and space. Sign up here -
AI Dependency on Human Infrastructure and Maintenance
By
–
AI require electrical power, hardware maintenance, software updates, and many other things that they can't get without humans.