VE

Velara

VelaraGGUF by RichardErkhovUpdated 2024-08-28
Download on Hugging Face
Parameters11.4B
Context32K
Layers51
Architecturellama
Smallest quant4.2 GB
Quick answer

Velara needs about 9.1 GB of memory at Q4_K_M (a 6.8 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 22 quants, from 4.2 GB (Q2_K) to 12.1 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 6.8 GB.

Velara is a 11.4B-parameter Velara model in GGUF format, quantized by RichardErkhov. Its trained context length is 32K tokens. The repository has 22 quant files, from 4.2 GB (Q2_K) to 12.1 GB (Q8_0).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K4.2 GB 6.5 GB
Fits Get file →
IQ3_XS4.7 GB 6.9 GB
Fits Get file →
Q3_K_S4.9 GB 7.2 GB
Fits Get file →
IQ3_S5.0 GB 7.2 GB
Fits Get file →
IQ3_M5.1 GB 7.3 GB
Tight Get file →
Q3_K5.5 GB 7.7 GB
Tight Get file →
Q3_K_M5.5 GB 7.7 GB
Tight Get file →
Q3_K_L6.0 GB 8.2 GB
Too big Get file →
IQ4_XS6.2 GB 8.4 GB
Too big Get file →
Q4_06.4 GB 8.7 GB
Too big Get file →
Q4_K_S6.5 GB 8.7 GB
Too big Get file →
IQ4_NL6.5 GB 8.7 GB
Too big Get file →
Q4_K6.8 GB 9.1 GB
Too big Get file →
Q4_K_M6.8 GB 9.1 GB
Too big Get file →
Q4_17.0 GB 9.2 GB
Too big Get file →
Q5_07.8 GB 10.1 GB
Too big Get file →
Q5_K_S7.8 GB 10.1 GB
Too big Get file →
Q5_K8.1 GB 10.3 GB
Too big Get file →
Q5_K_M8.1 GB 10.3 GB
Too big Get file →
Q5_18.6 GB 10.8 GB
Too big Get file →
Q6_K9.3 GB 11.6 GB
Too big Get file →
Q8_012.1 GB 14.3 GB
Too big Get file →

What GPU do you need for Velara?

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: no quant fits at an 8K context.
  • 8 GB VRAM: the best quant that fits is Q3_K (5.5 GB file, about 7.7 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (9.3 GB file, about 11.6 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (12.1 GB file, about 14.3 GB in use).

Which Velara quant should you choose?

Q4_K_M is the usual starting point for Velara: it balances file size and quality well. Here it is a 6.8 GB file.

The Q8_0 file is 12.1 GB, about 1.8 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 4.2 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
Parameters11.4B
Layers51
Trained context32,768 tokens
KV cache (fp16)204 KB per token

Source

Quantized byRichardErkhov
FamilyVelara
Downloads (HF)26
Last updated2024-08-28

Frequently asked questions about Velara

How much VRAM does Velara need?

About 9.1 GB at Q4_K_M with an 8K context window. That is the 6.8 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 Velara on an 8 GB GPU?

Yes, with Q3_K: the 5.5 GB file needs about 7.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Velara GGUF?

It depends on the quant: Q2_K is 4.2 GB, IQ3_XS is 4.7 GB, Q3_K_S is 4.9 GB, IQ3_S is 5.0 GB, IQ3_M is 5.1 GB, Q3_K is 5.5 GB and 16 more.

What context length does Velara support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.7 GB at 8K and 6.8 GB at 32K.

Where can I download Velara GGUF?

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

What license does Velara use?

Hugging Face lists the license as cc-by-nc-nd-4.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Similar models

Browse all →

More from RichardErkhov

All by RichardErkhov →

More Velara models

All Velara →

Other medium models

All medium models →