Model reference · open weights
transformer is an open-weight language model from fla-hub. transformer-1.3B-100B (BF16) weighs 2.7 GB; the smallest configuration that runs it is RTX 3060 12 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
What it is
| Released by | fla-hub |
|---|---|
| Released | 2024-04-01 |
| Parameters | 1.4B |
| VRAM | 2.7 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory |
|---|---|---|---|
| 2K each | |||
| RTX 3060 12 GB | 20 | all 2K | 11.6 GB |
| RTX 4060 Ti 16 GB | 30 | all 2K | 15.4 GB |
| RTX 3090 24 GB | 49 | all 2K | 23.4 GB |
| RTX 4090 24 GB | 49 | all 2K | 23.4 GB |
| RTX 5090 32 GB | 68 | all 2K | 31.0 GB |
| L40S 48 GB | 100 | all 2K | 44.0 GB |
| A100 80 GB | 186 | all 2K | 78.2 GB |
| H100 80 GB | 178 | all 2K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 217 | all 2K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 251 | all 2K | 107 GB |
| H200 141 GB | 327 | all 2K | 138 GB |
| B200 180 GB | 421 | all 2K | 176 GB |
| Requests at once | 2K tokens each |
|---|---|
| 1 | 3.7 GB |
| 5 | 5.3 GB |
| 8 | 6.5 GB |
| 16 | 9.7 GB |
| 32 | 16.2 GB |
| 64 | 29.1 GB |
One card, with vLLM's small-card settings.