KO

KoSOLAR v2.0

KosolarGGUF by mradermacherUpdated 2024-12-07
Download on Hugging Face
Parameters10.7B
Context4K
Layers48
Architecturellama
Smallest quant4.0 GB
Quick answer

KoSOLAR v2.0 needs about 8.6 GB of memory at Q4_K_M (a 6.5 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 13 quants, from 4.0 GB (Q2_K) to 21.5 GB (F16).

KoSOLAR v2.0 is a 10.7B-parameter Kosolar model in GGUF format, quantized by mradermacher. Its trained context length is 4K tokens. The repository has 13 quant files, from 4.0 GB (Q2_K) to 21.5 GB (F16).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K4.0 GB 5.3 GB
Fits Get file →
Q3_K_S4.7 GB 6.0 GB
Fits Get file →
Q3_K_M5.2 GB 6.5 GB
Fits Get file →
Q3_K_L5.7 GB 7.0 GB
Fits Get file →
IQ4_XS5.8 GB 7.1 GB
Fits Get file →
Q4_06.1 GB 7.4 GB
Tight Get file →
Q4_K_S6.1 GB 7.4 GB
Tight Get file →
Q4_K_M6.5 GB 7.8 GB
Tight Get file →
Q5_K_S7.4 GB 8.7 GB
Too big Get file →
Q5_K_M7.6 GB 8.9 GB
Too big Get file →
Q6_K8.8 GB 10.1 GB
Too big Get file →
Q8_011.4 GB 12.7 GB
Too big Get file →
F1621.5 GB 22.8 GB
Too big Get file →

What GPU do you need for KoSOLAR v2.0?

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 IQ4_XS (5.8 GB file, about 7.9 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (8.8 GB file, about 10.9 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (11.4 GB file, about 13.5 GB in use).
  • 24 GB VRAM: the best quant that fits is F16 (21.5 GB file, about 23.6 GB in use).

Which KoSOLAR v2.0 quant should you choose?

Q4_K_M is the usual starting point for KoSOLAR v2.0: it balances file size and quality well. Here it is a 6.5 GB file.

The Q8_0 file is 11.4 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 4.0 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
Parameters10.7B
Layers48
Trained context4,096 tokens
KV cache (fp16)192 KB per token

Source

Quantized bymradermacher
FamilyKosolar
Licenseapache-2.0
Downloads (HF)82
Last updated2024-12-07

Frequently asked questions about KoSOLAR v2.0

How much VRAM does KoSOLAR v2.0 need?

About 8.6 GB at Q4_K_M with an 8K context window. That is the 6.5 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 KoSOLAR v2.0 on an 8 GB GPU?

Yes, with IQ4_XS: the 5.8 GB file needs about 7.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of KoSOLAR v2.0 GGUF?

It depends on the quant: Q2_K is 4.0 GB, Q3_K_S is 4.7 GB, Q3_K_M is 5.2 GB, Q3_K_L is 5.7 GB, IQ4_XS is 5.8 GB, Q4_0 is 6.1 GB and 7 more.

What context length does KoSOLAR v2.0 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 1.6 GB at 8K and 0.8 GB at 4K.

Where can I download KoSOLAR v2.0 GGUF?

The files are on Hugging Face in the mradermacher/KoSOLAR-v2.0-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 KoSOLAR v2.0 use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Other GGUFs of KoSOLAR-v2.0

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Kosolar models

All Kosolar →

Other medium models

All medium models →