LL

llama2

LlamaGGUF by DawoodthouseefUpdated 2023-10-14
Download on Hugging Face
Parameters13B
Context32K
Layers32
Architecturellama
Smallest quant3.1 GB
Quick answer

llama2 needs about 7.0 GB of memory at Q2_K (a 5.4 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 7 quants, from 3.1 GB (Q2_K) to 13.8 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

llama2 is a 13B-parameter Llama model in GGUF format, quantized by Dawoodthouseef. Its trained context length is 32K tokens. The repository has 7 quant files, from 3.1 GB (Q2_K) to 13.8 GB (Q8_0).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K3.1 GB 4.7 GB
Fits Get file →
Q5_K_M4.8 GB 6.4 GB
Fits Get file →
Q5_K_M4.8 GB 6.4 GB
Fits Get file →
Q2_K5.4 GB 7.0 GB
Fits Get file →
—7.2 GB 8.7 GB
Too big Get file →
Q8_07.2 GB 8.7 GB
Too big Get file →
Q8_013.8 GB 15.4 GB
Too big Get file →

What GPU do you need for llama2?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: the best quant that fits is Q2_K (3.1 GB file, about 4.7 GB in use).
  • 8 GB VRAM: the best quant that fits is Q2_K (5.4 GB file, about 7.0 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (7.2 GB file, about 8.7 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (13.8 GB file, about 15.4 GB in use).

Which llama2 quant should you choose?

Q2_K (5.4 GB) is the middle option in this repository. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

If memory is tight, the smallest file is Q2_K at 3.1 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters13B
Layers32
Trained context32,768 tokens
KV cache (fp16)128 KB per token

Source

Quantized byDawoodthouseef
FamilyLlama
Downloads (HF)122
Last updated2023-10-14

Frequently asked questions about llama2

How much VRAM does llama2 need?

About 7.0 GB at Q2_K with an 8K context window. That is the 5.4 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run llama2 on an 8 GB GPU?

Yes, with Q2_K: the 5.4 GB file needs about 7.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of llama2 GGUF?

It depends on the quant: Q2_K is 3.1 GB, Q5_K_M is 4.8 GB, Q5_K_M is 4.8 GB, Q2_K is 5.4 GB, file is 7.2 GB, Q8_0 is 7.2 GB and 1 more.

What context length does llama2 support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download llama2 GGUF?

The files are on Hugging Face in the Dawoodthouseef/llama2-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Similar models

Browse all →

More Llama models

All Llama →

Other medium models

All medium models →