Skip to content

docs: Add comprehensive GPU hardware requirements and quantization details #464

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
samay2504 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

docs: Add comprehensive GPU hardware requirements and quantization details #464

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 24 additions & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,8 +86,30 @@ To download the model weights. See

### System Requirements

Gemma can run on a CPU, GPU and TPU. For GPU, we recommend 8GB+ RAM on GPU for
The 2B checkpoint and 24GB+ RAM on GPU are used for the 7B checkpoint.
Gemma can run on a CPU, GPU and TPU. Below are the recommended hardware specifications:

#### GPU Requirements

**Gemma 2B Checkpoint:**
- **Minimum VRAM:** 8GB+
- **Recommended GPUs:**
- NVIDIA: RTX 3060 (12GB), RTX 4060 (8GB), RTX 3070 (8GB), RTX 4060 Ti (8GB/16GB)
- AMD: RX 6600 XT (8GB), RX 6700 XT (12GB), RX 7600 (8GB)
- Professional: NVIDIA T4 (16GB), A4000 (16GB)

**Gemma 7B Checkpoint:**
- **Minimum VRAM:** 24GB+
- **Recommended GPUs:**
- NVIDIA: RTX 4090 (24GB), RTX 3090 (24GB), RTX 3090 Ti (24GB)
- AMD: Radeon RX 7900 XTX (24GB), Radeon RX 7900 XT (20GB)
- Professional: NVIDIA A5000 (24GB), A6000 (48GB), L40 (48GB), H100 (80GB)

**Quantization Benefits:**
- **Full Precision (Float16/BF16):** ~19GB VRAM for 7B models
- **INT8 Quantization:** ~10GB VRAM for 7B models
- **INT4 Quantization:** ~5GB VRAM for 7B models

> **Note:** Using quantization techniques (INT8/INT4) can significantly reduce VRAM requirements, making 7B models accessible on GPUs with 12-16GB VRAM.

### Contributing

Expand Down