SH

shqiponja 90

ShqiponjaGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters89.7B
Context32K
Layers100
Architecturellama
Smallest quant35.0 GB
Quick answer

shqiponja 90 needs about 67.9 GB of memory at Q3_K_S (a 40.5 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 7 quants, from 35.0 GB (Q2_K) to 49.2 GB (Q3_K_L). Longer context needs more: at 32K tokens the KV cache alone adds about 107.4 GB.

shqiponja 90 is a 89.7B-parameter Shqiponja model in GGUF format, quantized by mradermacher. Its trained context length is 32K tokens. The repository has 7 quant files, from 35.0 GB (Q2_K) to 49.2 GB (Q3_K_L).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K35.0 GB 62.3 GB
Too big Get file →
IQ3_XS38.5 GB 65.8 GB
Too big Get file →
IQ3_S40.5 GB 67.9 GB
Too big Get file →
Q3_K_S40.5 GB 67.9 GB
Too big Get file →
IQ3_M42.6 GB 70.0 GB
Too big Get file →
Q3_K_M45.2 GB 72.5 GB
Too big Get file →
Q3_K_L49.2 GB 76.5 GB
Too big Get file →

What GPU do you need for shqiponja 90?

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: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: no quant fits at an 8K context.
  • 32 GB VRAM: no quant fits at an 8K context.
  • 48 GB VRAM: no quant fits at an 8K context.

Which shqiponja 90 quant should you choose?

Q3_K_S (40.5 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

If memory is tight, the smallest file is Q2_K at 35.0 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
Parameters89.7B
Layers100
Trained context32,768 tokens
KV cache (fp16)3200 KB per token

Source

Quantized bymradermacher
FamilyShqiponja
Licensemit
Downloads (HF)64
Last updated2024-05-06

Frequently asked questions about shqiponja 90

How much VRAM does shqiponja 90 need?

About 67.9 GB at Q3_K_S with an 8K context window. That is the 40.5 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 shqiponja 90 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 62.3 GB, so you would need a larger GPU or CPU offloading.

What is the file size of shqiponja 90 GGUF?

It depends on the quant: Q2_K is 35.0 GB, IQ3_XS is 38.5 GB, IQ3_S is 40.5 GB, Q3_K_S is 40.5 GB, IQ3_M is 42.6 GB, Q3_K_M is 45.2 GB and 1 more.

What context length does shqiponja 90 support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 26.8 GB at 8K and 107.4 GB at 32K.

Where can I download shqiponja 90 GGUF?

The files are on Hugging Face in the mradermacher/shqiponja-90-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 shqiponja 90 use?

Hugging Face lists the license as mit. 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 shqiponja-90

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Shqiponja models

All Shqiponja →

Other large models

All large models →