LL

Llama 3.1 8b Pruned 6 Layers

LlamaGGUF by mradermacherUpdated 2024-08-04
Download on Hugging Face
Parameters6.5B
Context128K
Layers25
Architecturellama
Smallest quant2.6 GB
Quick answer

Llama 3.1 8b Pruned 6 Layers needs about 5.4 GB of memory at Q4_K_M (a 4.0 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 15 quants, from 2.6 GB (Q2_K) to 13.0 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 3.4 GB.

Llama 3.1 8b Pruned 6 Layers – lightweight 6.5B parameter model

Llama 3.1 8b Pruned 6 Layers is a 6.5B parameter model with a 131072 token context window. It is available in multiple GGUF quantizations for efficient inference. The model is released by mradermacher and is not a mixture of experts.

Good at

  • Fast inference on modest hardware
  • Low memory footprint with Q2_K
  • Supports 8k token context

Limits

  • No vision support
  • License not specified
  • Requires 8k context for estimates

Best for

  • Local chatbot deployments
  • Rapid prototyping
  • Resource constrained inference

Who it’s for. Developers and researchers needing a compact Llama 3.1 variant for local deployment. The model is suitable for environments with limited GPU memory and for quick experimentation.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K2.6 GB 4.0 GB
Fits Get file →
IQ3_XS2.9 GB 4.2 GB
Fits Get file →
Q3_K_S3.0 GB 4.3 GB
Fits Get file →
IQ3_S3.0 GB 4.4 GB
Fits Get file →
IQ3_M3.1 GB 4.4 GB
Fits Get file →
Q3_K_M3.3 GB 4.6 GB
Fits Get file →
Q3_K_L3.5 GB 4.9 GB
Fits Get file →
IQ4_XS3.7 GB 5.0 GB
Fits Get file →
Q4_K_S3.8 GB 5.2 GB
Fits Get file →
Q4_K_M4.0 GB 5.4 GB
Fits Get file →
Q5_K_S4.5 GB 5.9 GB
Fits Get file →
Q5_K_M4.7 GB 6.0 GB
Fits Get file →
Q6_K5.3 GB 6.7 GB
Fits Get file →
Q8_06.9 GB 8.3 GB
Too big Get file →
F1613.0 GB 14.4 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (4.01 GB, ~5.4 GB memory). For 4 GB VRAM, Q2_K (2.62 GB, ~4 GB memory) is recommended. Higher precision quants like Q6_K or F16 require more than 8 GB VRAM.

What GPU do you need for Llama 3.1 8b Pruned 6 Layers?

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: the best quant that fits is Q2_K (2.6 GB file, about 4.0 GB in use).
  • 6 GB VRAM: the best quant that fits is Q5_K_M (4.7 GB file, about 6.0 GB in use).
  • 8 GB VRAM: the best quant that fits is Q6_K (5.3 GB file, about 6.7 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (6.9 GB file, about 8.3 GB in use).
  • 16 GB VRAM: the best quant that fits is F16 (13.0 GB file, about 14.4 GB in use).

Which Llama 3.1 8b Pruned 6 Layers quant should you choose?

Q4_K_M is the usual starting point for Llama 3.1 8b Pruned 6 Layers: it balances file size and quality well. Here it is a 4.0 GB file.

The Q8_0 file is 6.9 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 2.6 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
Parameters6.5B
Layers25
Trained context131,072 tokens
KV cache (fp16)100 KB per token

Source

Quantized bymradermacher
FamilyLlama
Downloads (HF)921
Last updated2024-08-04

Frequently asked questions about Llama 3.1 8b Pruned 6 Layers

How much VRAM does Llama 3.1 8b Pruned 6 Layers need?

About 5.4 GB at Q4_K_M with an 8K context window. That is the 4.0 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 Llama 3.1 8b Pruned 6 Layers on an 8 GB GPU?

Yes, with Q6_K: the 5.3 GB file needs about 6.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Llama 3.1 8b Pruned 6 Layers GGUF?

It depends on the quant: Q2_K is 2.6 GB, IQ3_XS is 2.9 GB, Q3_K_S is 3.0 GB, IQ3_S is 3.0 GB, IQ3_M is 3.1 GB, Q3_K_M is 3.3 GB and 9 more.

What context length does Llama 3.1 8b Pruned 6 Layers support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.8 GB at 8K and 3.4 GB at 32K.

Where can I download Llama 3.1 8b Pruned 6 Layers GGUF?

The files are on Hugging Face in the mradermacher/Llama-3.1-8b-Pruned-6-Layers-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Can I run this model on 8 GB VRAM?

Yes, using Q4_K_M or Q4_K_S will fit within 8 GB VRAM with estimated memory usage around 5.4 GB.

What is the context length of this model?

The model supports a 131072 token context window.

Does this model support vision tasks?

No, the model does not include vision capabilities.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Llama models

All Llama →

Other small models

All small models →