6 ms·
Every Flop Counts: Scaling a 300B LLM Without Premium GPUs
- osti 1y agoI think this is the one where they train LLM without NVIDIA GPU's.
- flowerthoughts 1y agoThey never mention what hardware they're on. Table 1 is the closest thing. Device specs for six devices: 120-989 TFLOPS and 64-96 GB RAM. An RTX 5090 is about 105 TFLOPS. https://www.techpowerup.com/gpu-specs/geforce-rtx-5090.c4216 https://www.techpowerup.com/gpu-specs/geforce-rtx-5090.c4216
- bshark 1y agoThe 96GB (HBM2e) SKU is named PPU from T-head semiconductor (basically a subsidiary of Alibaba). The spec is very similar to H20. Other chips they were using include Huawei Ascend 910B (64GB) and maybe other domestic designed chips.
- boulos 1y agoI was surprised not to see a Kunlun P800 there.
- rahen 1y agoI'm pretty surprised by the claimed memory usage for 300B parameters (table 1). If we compare similar models: - Llama 3.1 with 405B parameters: 2 TB of memory (FP32), 500 GB (FP8) - DeepSeek R1 with 671B parameters: 1.3 TB (scaling linearly, around 600 GB for 300B parameters) Ling claims no more than 96 GB of memory, most likely for inference. That's far more than a 20% reduction. Am I missing something?
- fxtentacle 1y agoSome of these models still produce great results with something low like 2.7 bits per variable.
- cavisne 1y agoI think they only claim their "Ling-Lite" 17B model can fit on a single 96GB GPU, their 300B model needs 8 of them (768GB of HBM)
- vednig 1y agoThey've shared some interesting optimization techniques for bigger LLMs that's all, not exactly low powered devices as in power consumption. Still a good read.