SF

SF 72B V1

GGUFGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters72.3B
Context32K
Layers80
Architecturellama
Smallest quant28.5 GB
Quick answer

SF 72B V1 needs about 67.2 GB of memory at Q4_K_M (a 45.2 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 10 quants, from 28.5 GB (Q2_K) to 45.2 GB (Q4_K_M). Longer context needs more: at 32K tokens the KV cache alone adds about 85.9 GB.

SF 72B V1 is a 72.3B-parameter model in GGUF format, quantized by mradermacher. Its trained context length is 32K tokens. The repository has 10 quant files, from 28.5 GB (Q2_K) to 45.2 GB (Q4_K_M).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K28.5 GB 50.5 GB
Too big Get file →
IQ3_XS31.4 GB 53.4 GB
Too big Get file →
IQ3_S33.0 GB 55.0 GB
Too big Get file →
Q3_K_S33.0 GB 55.0 GB
Too big Get file →
IQ3_M34.7 GB 56.7 GB
Too big Get file →
Q3_K_M36.7 GB 58.7 GB
Too big Get file →
Q3_K_L39.9 GB 61.9 GB
Too big Get file →
IQ4_XS40.6 GB 62.5 GB
Too big Get file →
Q4_K_S42.7 GB 64.7 GB
Too big Get file →
Q4_K_M45.2 GB 67.2 GB
Too big Get file →

What GPU do you need for SF 72B V1?

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 SF 72B V1 quant should you choose?

Q4_K_M is the usual starting point for SF 72B V1: it balances file size and quality well. Here it is a 45.2 GB file.

If memory is tight, the smallest file is Q2_K at 28.5 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
Parameters72.3B
Layers80
Trained context32,768 tokens
KV cache (fp16)2560 KB per token

Source

Quantized bymradermacher
Licenseapache-2.0
Downloads (HF)85
Last updated2024-05-06

Frequently asked questions about SF 72B V1

How much VRAM does SF 72B V1 need?

About 67.2 GB at Q4_K_M with an 8K context window. That is the 45.2 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 SF 72B V1 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 50.5 GB, so you would need a larger GPU or CPU offloading.

What is the file size of SF 72B V1 GGUF?

It depends on the quant: Q2_K is 28.5 GB, IQ3_XS is 31.4 GB, IQ3_S is 33.0 GB, Q3_K_S is 33.0 GB, IQ3_M is 34.7 GB, Q3_K_M is 36.7 GB and 4 more.

What context length does SF 72B V1 support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 21.5 GB at 8K and 85.9 GB at 32K.

Where can I download SF 72B V1 GGUF?

The files are on Hugging Face in the mradermacher/SF-72B-V1-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 SF 72B V1 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.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

Other large models

All large models →