SM

Fireplace 34b iMatrix

SmaugGGUF by MarsupialAIUpdated 2024-05-11
Download on Hugging Face
Parameters34.4B
Context195K
Layers60
Architecturellama
Smallest quant9.3 GB
Quick answer

Fireplace 34b iMatrix needs about 23.2 GB of memory at its main file (a 20.7 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 7 quants, from 9.3 GB to 36.5 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 8.1 GB.

Fireplace 34b iMatrix is a 34.4B-parameter Smaug model in GGUF format, quantized by MarsupialAI. Its trained context length is 195K tokens. The repository has 7 quant files, from 9.3 GB to 36.5 GB.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—9.3 GB 11.8 GB
Too big Get file →
—13.3 GB 15.8 GB
Too big Get file →
—18.5 GB 21.0 GB
Too big Get file →
—20.7 GB 23.2 GB
Too big Get file →
—24.3 GB 26.8 GB
Too big Get file →
—28.2 GB 30.7 GB
Too big Get file →
—36.5 GB 39.1 GB
Too big Get file →

What GPU do you need for Fireplace 34b iMatrix?

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: the best quant that fits is the main file (9.3 GB file, about 11.8 GB in use).
  • 16 GB VRAM: the best quant that fits is the main file (13.3 GB file, about 15.8 GB in use).
  • 24 GB VRAM: the best quant that fits is the main file (20.7 GB file, about 23.2 GB in use).
  • 32 GB VRAM: the best quant that fits is the main file (28.2 GB file, about 30.7 GB in use).
  • 48 GB VRAM: the best quant that fits is the main file (36.5 GB file, about 39.1 GB in use).

Which Fireplace 34b iMatrix quant should you choose?

If memory is tight, the smallest file is the first one in the table at 9.3 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters34.4B
Layers60
Trained context200,000 tokens
KV cache (fp16)240 KB per token

Source

Quantized byMarsupialAI
FamilySmaug
Licenseother
Downloads (HF)99
Last updated2024-05-11

Frequently asked questions about Fireplace 34b iMatrix

How much VRAM does Fireplace 34b iMatrix need?

About 23.2 GB at its main file with an 8K context window. That is the 20.7 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 Fireplace 34b iMatrix on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 11.8 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Fireplace 34b iMatrix GGUF?

It depends on the quant: file is 9.3 GB, file is 13.3 GB, file is 18.5 GB, file is 20.7 GB, file is 24.3 GB, file is 28.2 GB and 1 more.

What context length does Fireplace 34b iMatrix support?

The GGUF header lists a trained context length of 200,000 tokens (195K). The KV cache grows with the context you use: about 2.0 GB at 8K and 8.1 GB at 32K.

Where can I download Fireplace 34b iMatrix GGUF?

The files are on Hugging Face in the MarsupialAI/Fireplace-34b_iMatrix_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 Fireplace 34b iMatrix use?

Hugging Face lists the license as other. 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 Smaug-34B-v0.1

Similar models

Browse all →

More from MarsupialAI

All by MarsupialAI →

More Smaug models

All Smaug →

Other large models

All large models →