face-parsing
by jonathandinu
85M params · image-segmentation · 205 likes · 474.4k downloads
face-parsing is a 85M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (15518 tok/s), GeForce RTX 5090 (23271 tok/s), M4 Max 128GB (5678 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 15518 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 23271 tok/s | comfortable |
| M4 Max 128GB | 128GB | 5678 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 2839 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 2839 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 28431 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 57185 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 13720 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 11346 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 11013 tok/s | comfortable |