segformer-b0-finetuned-ade-512-512
by nvidia
4M params · image-segmentation · 183 likes · 474.9k downloads
segformer-b0-finetuned-ade-512-512 is a 4M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (349834 tok/s), GeForce RTX 5090 (524618 tok/s), M4 Max 128GB (128018 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 349834 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 524618 tok/s | comfortable |
| M4 Max 128GB | 128GB | 128018 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 64009 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 64009 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 640945 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 1289137 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 309305 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 255781 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 248278 tok/s | comfortable |