AE

aegolius acadicus 34b v3

AegoliusGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters35.4B
Context32K
Layers32
Architecturellama
Smallest quant12.9 GB
Quick answer

aegolius acadicus 34b v3 needs about 23.0 GB of memory at Q4_K_M (a 21.4 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 14 quants, from 12.9 GB (Q2_K) to 37.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

Aegolius Acadicus 34b v3 – 35.4B parameter MOE Llama model

Aegolius Acadicus 34b v3 is a 35.4B parameter mixture of experts model built on Llama architecture. It was trained with a context length of 32768 tokens and is released under the apache-2.0 license. The model is tailored for knowledge‑centric tasks.

Good at

  • Knowledge‑heavy dialogue
  • Multi‑disciplinary reasoning
  • Fine‑tuned Llama2 and Mistral integration

Limits

  • No vision support
  • Requires large VRAM for higher quants
  • License restricts commercial use

Best for

  • Research prototypes
  • Academic knowledge bases
  • Custom Lora adapter development

Who it’s for. Developers and researchers who need a large MOE model for language understanding and want to experiment with Llama2 and Mistral fine‑tuned adapters. The model is suitable for those who can manage the memory demands and comply with the apache‑2.0 license.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K12.9 GB 14.5 GB
Too big Get file →
IQ3_XS14.5 GB 16.0 GB
Too big Get file →
Q3_K_S15.3 GB 16.9 GB
Too big Get file →
IQ3_S15.3 GB 16.9 GB
Too big Get file →
IQ3_M15.6 GB 17.1 GB
Too big Get file →
Q3_K_M17.0 GB 18.5 GB
Too big Get file →
Q3_K_L18.4 GB 19.9 GB
Too big Get file →
IQ4_XS19.1 GB 20.7 GB
Too big Get file →
Q4_K_S20.1 GB 21.7 GB
Too big Get file →
Q4_K_M21.4 GB 23.0 GB
Too big Get file →
Q5_K_S24.4 GB 26.0 GB
Too big Get file →
Q5_K_M25.1 GB 26.7 GB
Too big Get file →
Q6_K29.1 GB 30.6 GB
Too big Get file →
Q8_037.6 GB 39.2 GB
Too big Get file →

For 8 GB VRAM, use Q2_K (file size 12.93 GB, estimated 14.5 GB memory). For 12 GB VRAM, Q3_K_S (15.28 GB, 16.9 GB memory) or IQ3_S (15.3 GB, 16.9 GB memory). For 16 GB VRAM, Q4_K_S (20.14 GB, 21.7 GB memory) or Q4_K_M (21.44 GB, 23 GB memory).

What GPU do you need for aegolius acadicus 34b v3?

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 IQ3_XS (14.5 GB file, about 16.0 GB in use).
  • 24 GB VRAM: the best quant that fits is Q4_K_M (21.4 GB file, about 23.0 GB in use).
  • 32 GB VRAM: the best quant that fits is Q6_K (29.1 GB file, about 30.6 GB in use).
  • 48 GB VRAM: the best quant that fits is Q8_0 (37.6 GB file, about 39.2 GB in use).

Which aegolius acadicus 34b v3 quant should you choose?

Q4_K_M is the usual starting point for aegolius acadicus 34b v3: it balances file size and quality well. Here it is a 21.4 GB file.

The Q8_0 file is 37.6 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 Q2_K at 12.9 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
Parameters35.4B (approximate, mixture of experts)
Layers32
Trained context32,768 tokens
KV cache (fp16)128 KB per token

Source

Quantized bymradermacher
FamilyAegolius
Licenseapache-2.0
Downloads (HF)781
Last updated2024-05-06

Frequently asked questions about aegolius acadicus 34b v3

How much VRAM does aegolius acadicus 34b v3 need?

About 23.0 GB at Q4_K_M with an 8K context window. That is the 21.4 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 aegolius acadicus 34b v3 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 14.5 GB, so you would need a larger GPU or CPU offloading.

What is the file size of aegolius acadicus 34b v3 GGUF?

It depends on the quant: Q2_K is 12.9 GB, IQ3_XS is 14.5 GB, Q3_K_S is 15.3 GB, IQ3_S is 15.3 GB, IQ3_M is 15.6 GB, Q3_K_M is 17.0 GB and 8 more.

What context length does aegolius acadicus 34b v3 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 aegolius acadicus 34b v3 GGUF?

The files are on Hugging Face in the mradermacher/aegolius-acadicus-34b-v3-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 aegolius acadicus 34b v3 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.

Can I run this model on 8 GB VRAM?

Only the Q2_K quant is feasible, using about 14.5 GB of memory at 8 k context.

Does the model support vision tasks?

No, the model has no vision capabilities.

What license applies to this model?

The model is released under the apache-2.0 license.

Other GGUFs of aegolius-acadicus-34b-v3

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Aegolius models

All Aegolius →

Other large models

All large models →