| 5 Nov 2025 |
 apyh | In reply to @connorbaker:matrix.org
Why are you building for so many CUDA capabilities? I can’t really think of a reason you’d need that range in particular. 's a distributed ml training application that needs to run on everything from gtx 10xx gpus to modern data center GH/GB200s :/ | 03:27:37 |
| apyh | most common hardware is gonna be 30xx 40xx 50xx, h100, a100, b200 | 03:27:56 |
| apyh | though.. i could just see what pytorch precompiled wherls runs on and limit to that | 03:28:54 |
| apyh | should be fine | 03:28:56 |
 connor (burnt/out) (UTC-8) | Flash attention doesn’t support anything older than Ampere I thought | 03:29:07 |
 Robbie Buxton | V2 does | 03:29:19 |
| Robbie Buxton | V3 is hopper only | 03:29:24 |
| apyh | ya its only v3 iirc | 03:29:26 |
| Robbie Buxton | V4 (cute) is Blackwell | 03:29:33 |
| Robbie Buxton | But that’s a wip | 03:29:38 |
| apyh | and yeah fair enough I could drop fa for older gpus, ig i can provide cuda capabilities different per package | 03:30:06 |
| Robbie Buxton | Requires cutlass-dsl | 03:30:07 |
| connor (burnt/out) (UTC-8) | V2 doesn’t support older than Ampere per their readme, unless they forgot to update it | 03:30:25 |
| apyh | In reply to @connorbaker:matrix.org
V2 doesn’t support older than Ampere per their readme, unless they forgot to update it yeah makes sense then maybe I'll just drop and see if anyone complains lol | 03:30:43 |
| Robbie Buxton | In reply to @connorbaker:matrix.org
V2 doesn’t support older than Ampere per their readme, unless they forgot to update it It’s not optimized but it runs | 03:30:46 |
