AL

ALMA 13B

AlmaGGUF by s3nhUpdated 2024-01-18
Download on Hugging Face
Parameters13B
Context4K
Layers40
Architecturellama
Smallest quant5.7 GB
Quick answer

ALMA 13B needs about 15.1 GB of memory at Q4_K_M (a 7.9 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 5 quants, from 5.7 GB (Q3_K_S) to 13.8 GB (Q8_0).

ALMA 13B is a 13B-parameter Alma model in GGUF format, quantized by s3nh. Its trained context length is 4K tokens. The repository has 5 quant files, from 5.7 GB (Q3_K_S) to 13.8 GB (Q8_0).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q3_K_S5.7 GB 9.5 GB
Too big Get file →
Q4_K_M7.9 GB 11.7 GB
Too big Get file →
Q5_K_M9.2 GB 13.1 GB
Too big Get file →
Q6_K10.7 GB 14.5 GB
Too big Get file →
Q8_013.8 GB 17.7 GB
Too big Get file →

What GPU do you need for ALMA 13B?

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: the best quant that fits is Q4_K_M (7.9 GB file, about 15.1 GB in use).
  • 24 GB VRAM: the best quant that fits is Q8_0 (13.8 GB file, about 21.0 GB in use).

Which ALMA 13B quant should you choose?

Q4_K_M is the usual starting point for ALMA 13B: it balances file size and quality well. Here it is a 7.9 GB file.

The Q8_0 file is 13.8 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 Q3_K_S at 5.7 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters13B
Layers40
Trained context4,096 tokens
KV cache (fp16)800 KB per token

Source

Quantized bys3nh
FamilyAlma
Licenseopenrail
Downloads (HF)20
Last updated2024-01-18

Frequently asked questions about ALMA 13B

How much VRAM does ALMA 13B need?

About 15.1 GB at Q4_K_M with an 8K context window. That is the 7.9 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 ALMA 13B on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 12.9 GB, so you would need a larger GPU or CPU offloading.

What is the file size of ALMA 13B GGUF?

It depends on the quant: Q3_K_S is 5.7 GB, Q4_K_M is 7.9 GB, Q5_K_M is 9.2 GB, Q6_K is 10.7 GB, Q8_0 is 13.8 GB.

What context length does ALMA 13B support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 6.7 GB at 8K and 3.4 GB at 4K.

Where can I download ALMA 13B GGUF?

The files are on Hugging Face in the s3nh/ALMA-13B-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 ALMA 13B use?

Hugging Face lists the license as openrail. 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 s3nh

All by s3nh →

More Alma models

All Alma →

Other medium models

All medium models →