News for hackers. Stuff that matters.
Re:The RWKV language model: An RNN with the advantages of a ... by mromanuk on Thursday March 30, 2023 @09:23AM (#35372306)
which one is the next incarnation of performance/price/goodness after the RTX 3090 for running a mini AI homelab?
Re:The RWKV language model: An RNN with the advantages of a ... by espadrine on Thursday March 30, 2023 @10:41AM (#35373396)
To increase performance, you need bigger foundation models, like LLaMA 65B. You will always be RAM-bound for that (~40GB for LLaMA 65B INT4). So the next step is to upgrade your motherboard to have multiple GPU slots. If you really wanted to stick to a single-GPU setup, you could upgrade to RTX A6000, but it is more expensive than two RTX 3090 while holding the same amount of RAM.
Re:The RWKV language model: An RNN with the advantages of a ... by yieldcrv on Thursday March 30, 2023 @12:03PM (#35374861)
Its a dual slot gpu

What benefit does that offer over 2 GPUs?

Re:The RWKV language model: An RNN with the advantages of a ... by espadrine on Thursday March 30, 2023 @12:33PM (#35375350)
The two GPUs will need to exchange information in order to complete inference, by having one GPU hold half of the network weights, and the other holding the other half. That transmission of information will be limited by the PCIe bandwidth; for instance, 30 GB/s with v4x16.

Meanwhile the A6000 VRAM bandwidth is 768 GB/s (= 16 Gb/s (GDDR6) × 384 bit-width ÷ 8 bits per byte).

Re:The RWKV language model: An RNN with the advantages of a ... by coolspot on Thursday March 30, 2023 @01:10PM (#35375958)
Two 3090 can be connected using NVLink bridge which is much faster than PCI-E.
Re:The RWKV language model: An RNN with the advantages of a ... by espadrine on Thursday March 30, 2023 @02:26PM (#35377025)
I couldn’t find precise information on what bandwidth you’d get with NVLink on the 3090. To be fair, though, if all we do is inference, using Huggingface pipeline parallelism, the amount of data transferred is pretty small: 8192×2×n_tokens bytes; for most uses, with a recent PCIe setup, that bottleneck will take less than 0.1 ms per token generated, which may not be the dominant latency.

(Also, the RTX 3090 has faster VRAM, >900 GB/s, than the A6000, because it is GDDR6X.)

Re:The RWKV language model: An RNN with the advantages of a ... by azeirah on Thursday March 30, 2023 @04:49PM (#35378643)
Not entirely certain about this, but I believe that because Apple M-series GPUs use system memory you could possible run the larger models on accessible (albeit expensive) consumer hardware.

Need 64GB for 30B

And 128GB for the 65B

I'm not sure about the performance, but I think it should be ok? Especially given how much Apple has been investing in what -- I believe they call -- neural cores?

Here's some more context: https://news.ycombinator.com/item?id=35105364