Model reference · open weights
gemma is an open-weight language model from unsloth. gemma-2b-it (BF16) weighs 5.0 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 | unsloth |
|---|---|
| Released | 2024-02-21 |
| Parameters | 2.5B |
| VRAM | 5.0 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory |
|---|---|---|---|
| 8K each | |||
| RTX 3060 12 GB | 38 | all 8K | 11.6 GB |
| RTX 4060 Ti 16 GB | 63 | all 8K | 15.4 GB |
| RTX 3090 24 GB | 116 | all 8K | 23.4 GB |
| RTX 4090 24 GB | 115 | all 8K | 23.4 GB |
| RTX 5090 32 GB | 166 | all 8K | 31.0 GB |
| L40S 48 GB | 252 | all 8K | 44.0 GB |
| A100 80 GB | 478 | all 8K | 78.2 GB |
| H100 80 GB | 441 | all 8K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 545 | all 8K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 635 | all 8K | 107 GB |
| H200 141 GB | 838 | all 8K | 138 GB |
| B200 180 GB | 1000+ | all 8K | 176 GB |
| Requests at once | 8K tokens each |
|---|---|
| 1 | 6.0 GB |
| 5 | 6.6 GB |
| 8 | 7.1 GB |
| 16 | 8.3 GB |
| 32 | 10.7 GB |
| 64 | 15.5 GB |
One card, with vLLM's small-card settings.
From the model card
All notebooks are beginner friendly! Add your dataset, click "Run All", and you'll get a 2x faster finetuned model which can be exported to GGUF, vLLM or uploaded to Hugging Face.
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.