Workload
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Gradient computation | Workload 32,768 tokens | Gradient Computation Time (s)7.1 | 48 | |
| EEG Classification | Workload | Balanced Accuracy73.91 | 31 | |
| Brain Decoding | Workload | Balanced Accuracy78.67 | 9 | |
| Decoding | workload 64 KiB | Throughput (MiB/s)669.1 | 4 | |
| Encoding | workload 64 KiB | Throughput (MiB/s)236.7 | 4 | |
| Decoding | workload 1 KiB | Throughput (MiB/s)640.5 | 4 | |
| Encoding | workload 1 KiB | Throughput (MiB/s)218.1 | 4 | |
| Hybrid Mamba-Transformer Inference | workload uniform_short | Goodput (req/s)434.24 | 2 | |
| DNN Training | Workload Single-GPU (train) | Training Energy Savings15 | 1 |
Showing 9 of 9 rows