FU

FusionNet_7Bx2_MoE_14B

FusionnetGGUF by mradermacherUpdated 2024-08-02
Download on Hugging Face
Parameters12.9B
Context32K
Layers32
Architecturellama
Smallest quant2.7 GB
Quick answer

FusionNet_7Bx2_MoE_14B needs about 9.4 GB of memory at Q4_K_M (a 7.8 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 21 quants, from 2.7 GB (IQ1_S) to 10.6 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

FusionNet_7Bx2_MoE_14B – 12.9B‑parameter MoE Llama model

FusionNet_7Bx2_MoE_14B is a 12.9B‑parameter Llama‑based model fine‑tuned with a mixture of experts. It supports a 32768‑token context window and is released under the MIT license. The model is available in multiple GGUF quantizations.

Good at

  • Large context handling
  • Mixture of experts for specialized inference
  • Fine‑tuned English language performance

Limits

  • Requires 8‑k context for memory estimates
  • No vision or multimodal support
  • Large memory footprint for higher‑quality quants

Best for

  • Chatbots needing long context
  • Research prototyping with MoE models
  • Fine‑tuning experiments on local hardware

Who it’s for. Developers and researchers who need a large‑parameter Llama model with MoE capabilities and a flexible quantization set. The model is suitable for local deployment using llama.cpp.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S2.7 GB 4.3 GB
Fits Get file →
IQ1_M3.0 GB 4.6 GB
Fits Get file →
IQ2_XXS3.5 GB 5.0 GB
Fits Get file →
IQ2_XS3.8 GB 5.4 GB
Fits Get file →
IQ2_S4.0 GB 5.5 GB
Fits Get file →
IQ2_M4.3 GB 5.9 GB
Fits Get file →
Q2_K4.8 GB 6.3 GB
Fits Get file →
IQ3_XXS5.0 GB 6.6 GB
Fits Get file →
IQ3_XS5.3 GB 6.9 GB
Fits Get file →
Q3_K_S5.6 GB 7.2 GB
Fits Get file →
IQ3_S5.6 GB 7.2 GB
Fits Get file →
IQ3_M5.7 GB 7.3 GB
Tight Get file →
Q3_K_M6.2 GB 7.8 GB
Tight Get file →
Q3_K_L6.7 GB 8.3 GB
Too big Get file →
IQ4_XS6.9 GB 8.5 GB
Too big Get file →
Q4_07.3 GB 8.9 GB
Too big Get file →
Q4_K_S7.3 GB 8.9 GB
Too big Get file →
Q4_K_M7.8 GB 9.4 GB
Too big Get file →
Q5_K_S8.9 GB 10.4 GB
Too big Get file →
Q5_K_M9.1 GB 10.7 GB
Too big Get file →
Q6_K10.6 GB 12.1 GB
Too big Get file →

For 8‑GB VRAM, use IQ1_S (4.3 GB memory) or IQ1_M (4.6 GB). With 12‑GB VRAM, Q4_K_M (9.4 GB) or Q5_K_S (10.4 GB) are viable. For 16‑GB VRAM, Q5_K_M (10.7 GB) or Q6_K (12.1 GB) provide higher quality. Memory estimates are approximate.

What GPU do you need for FusionNet_7Bx2_MoE_14B?

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: the best quant that fits is IQ2_M (4.3 GB file, about 5.9 GB in use).
  • 8 GB VRAM: the best quant that fits is Q3_K_M (6.2 GB file, about 7.8 GB in use).
  • 12 GB VRAM: the best quant that fits is Q5_K_M (9.1 GB file, about 10.7 GB in use).
  • 16 GB VRAM: the best quant that fits is Q6_K (10.6 GB file, about 12.1 GB in use).

Which FusionNet_7Bx2_MoE_14B quant should you choose?

Q4_K_M is the usual starting point for FusionNet_7Bx2_MoE_14B: it balances file size and quality well. Here it is a 7.8 GB file.

If memory is tight, the smallest file is IQ1_S at 2.7 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
Parameters12.9B (approximate, mixture of experts)
Layers32
Trained context32,768 tokens
KV cache (fp16)128 KB per token

Source

Quantized bymradermacher
FamilyFusionnet
Licensemit
Downloads (HF)2,657
Last updated2024-08-02

Frequently asked questions about FusionNet_7Bx2_MoE_14B

How much VRAM does FusionNet_7Bx2_MoE_14B need?

About 9.4 GB at Q4_K_M with an 8K context window. That is the 7.8 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 FusionNet_7Bx2_MoE_14B on an 8 GB GPU?

Yes, with Q3_K_M: the 6.2 GB file needs about 7.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of FusionNet_7Bx2_MoE_14B GGUF?

It depends on the quant: IQ1_S is 2.7 GB, IQ1_M is 3.0 GB, IQ2_XXS is 3.5 GB, IQ2_XS is 3.8 GB, IQ2_S is 4.0 GB, IQ2_M is 4.3 GB and 15 more.

What context length does FusionNet_7Bx2_MoE_14B support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download FusionNet_7Bx2_MoE_14B GGUF?

The files are on Hugging Face in the mradermacher/FusionNet_7Bx2_MoE_14B-i1-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 FusionNet_7Bx2_MoE_14B 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.

Can I run FusionNet_7Bx2_MoE_14B on 8 GB VRAM?

Yes, use IQ1_S or IQ1_M quantizations, which require about 4.3–4.6 GB of VRAM at 8‑k context.

Does this model support vision or multimodal inputs?

No, the model only handles text and has no vision capabilities.

What license applies to this model?

The model is released under the MIT license, allowing free use and modification.

Other GGUFs of FusionNet_7Bx2_MoE_14B

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Fusionnet models

All Fusionnet →

Other medium models

All medium models →