LL

llama3-40-layer-base

LlamaGGUF by mradermacherUpdated 2024-05-19
Download on Hugging Face
Parameters9.78B
Context8K
Layers40
Architecturellama
Smallest quant3.8 GB
Quick answer

llama3-40-layer-base needs about 7.8 GB of memory at Q4_K_M (a 6.0 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 14 quants, from 3.8 GB (Q2_K) to 10.4 GB (Q8_0).

llama3-40-layer-base is a 9.78B-parameter Llama model in GGUF format, quantized by mradermacher. Its trained context length is 8K tokens. The repository has 14 quant files, from 3.8 GB (Q2_K) to 10.4 GB (Q8_0).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K3.8 GB 5.7 GB
Fits Get file →
IQ3_XS4.2 GB 6.1 GB
Fits Get file →
Q3_K_S4.4 GB 6.3 GB
Fits Get file →
IQ3_S4.4 GB 6.3 GB
Fits Get file →
IQ3_M4.6 GB 6.4 GB
Fits Get file →
Q3_K_M4.9 GB 6.7 GB
Fits Get file →
Q3_K_L5.2 GB 7.1 GB
Fits Get file →
IQ4_XS5.4 GB 7.3 GB
Tight Get file →
Q4_K_S5.7 GB 7.5 GB
Tight Get file →
Q4_K_M6.0 GB 7.8 GB
Tight Get file →
Q5_K_S6.8 GB 8.6 GB
Too big Get file →
Q5_K_M7.0 GB 8.8 GB
Too big Get file →
Q6_K8.0 GB 9.9 GB
Too big Get file →
Q8_010.4 GB 12.2 GB
Too big Get file →

What GPU do you need for llama3-40-layer-base?

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 (3.8 GB file, about 5.7 GB in use).
  • 8 GB VRAM: the best quant that fits is Q4_K_M (6.0 GB file, about 7.8 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (8.0 GB file, about 9.9 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (10.4 GB file, about 12.2 GB in use).

Which llama3-40-layer-base quant should you choose?

Q4_K_M is the usual starting point for llama3-40-layer-base: it balances file size and quality well. Here it is a 6.0 GB file.

The Q8_0 file is 10.4 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is Q2_K at 3.8 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
Parameters9.78B
Layers40
Trained context8,192 tokens
KV cache (fp16)160 KB per token

Source

Quantized bymradermacher
FamilyLlama
Downloads (HF)197
Last updated2024-05-19

Frequently asked questions about llama3-40-layer-base

How much VRAM does llama3-40-layer-base need?

About 7.8 GB at Q4_K_M with an 8K context window. That is the 6.0 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 llama3-40-layer-base on an 8 GB GPU?

Yes, with Q4_K_M: the 6.0 GB file needs about 7.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of llama3-40-layer-base GGUF?

It depends on the quant: Q2_K is 3.8 GB, IQ3_XS is 4.2 GB, Q3_K_S is 4.4 GB, IQ3_S is 4.4 GB, IQ3_M is 4.6 GB, Q3_K_M is 4.9 GB and 8 more.

What context length does llama3-40-layer-base support?

The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 1.3 GB at 8K and 1.3 GB at 8K.

Where can I download llama3-40-layer-base GGUF?

The files are on Hugging Face in the mradermacher/llama3-40-layer-base-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 from mradermacher

All by mradermacher →

More Llama models

All Llama →

Other medium models

All medium models →