GO

GodziLLa 30B i1

GodzillaGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters32.5B
Context2K
Layers60
Architecturellama
Smallest quant7.4 GB
Quick answer

GodziLLa 30B i1 needs about 33.5 GB of memory at Q4_K_M (a 19.9 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 21 quants, from 7.4 GB (IQ1_S) to 26.9 GB (Q6_K).

32.5B‑parameter LLaMA‑based model for research and non‑commercial use

GodziLLa 30B i1 is a 32.5B‑parameter LLaMA model derived from Maya LoRAs and CalderaAI’s Lazarus‑30B. It is released under a CC‑BY‑NC‑4.0 license and is intended for research on composite LLMs. The model supports a 2048‑token context window.

Good at

  • Research on LoRA adapters
  • Non‑commercial text generation
  • Stress‑testing composite LLM limits

Limits

  • Requires attribution and non‑commercial use
  • No vision or multimodal support
  • Context limited to 2048 tokens

Best for

  • Academic studies of LoRA behavior
  • Testing LLaMA‑based composites
  • Benchmarking against open‑source LLMs

Who it’s for. Researchers and developers working on LLaMA‑based models who need a large‑parameter composite for non‑commercial experimentation. The model is not intended for commercial deployment.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S7.4 GB 11.2 GB
Too big Get file →
IQ1_M8.0 GB 11.8 GB
Too big Get file →
IQ2_XXS9.0 GB 12.8 GB
Too big Get file →
IQ2_XS9.9 GB 13.7 GB
Too big Get file →
IQ2_S10.7 GB 14.4 GB
Too big Get file →
IQ2_M11.5 GB 15.2 GB
Too big Get file →
Q2_K12.3 GB 16.1 GB
Too big Get file →
IQ3_XXS12.6 GB 16.4 GB
Too big Get file →
IQ3_XS13.6 GB 17.3 GB
Too big Get file →
IQ3_S14.3 GB 18.1 GB
Too big Get file →
Q3_K_S14.3 GB 18.1 GB
Too big Get file →
IQ3_M15.1 GB 18.9 GB
Too big Get file →
Q3_K_M16.0 GB 19.8 GB
Too big Get file →
Q3_K_L17.5 GB 21.3 GB
Too big Get file →
IQ4_XS17.6 GB 21.4 GB
Too big Get file →
Q4_018.7 GB 22.4 GB
Too big Get file →
Q4_K_S18.7 GB 22.5 GB
Too big Get file →
Q4_K_M19.9 GB 23.6 GB
Too big Get file →
Q5_K_S22.6 GB 26.4 GB
Too big Get file →
Q5_K_M23.3 GB 27.1 GB
Too big Get file →
Q6_K26.9 GB 30.7 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (7.4 GB) or IQ1_M (8.01 GB). For 12 GB VRAM, IQ2_XS (9.92 GB) or IQ2_S (10.67 GB) are suitable. For 16 GB VRAM, IQ3_XS (13.56 GB) or Q3_K_S (14.32 GB) provide a balance of size and memory usage. Estimates are approximate.

What GPU do you need for GodziLLa 30B i1?

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: the best quant that fits is IQ2_XS (9.9 GB file, about 23.5 GB in use).
  • 32 GB VRAM: the best quant that fits is IQ4_XS (17.6 GB file, about 31.2 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (26.9 GB file, about 40.5 GB in use).

Which GodziLLa 30B i1 quant should you choose?

Q4_K_M is the usual starting point for GodziLLa 30B i1: it balances file size and quality well. Here it is a 19.9 GB file.

If memory is tight, the smallest file is IQ1_S at 7.4 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
Parameters32.5B
Layers60
Trained context2,048 tokens
KV cache (fp16)1560 KB per token

Source

Quantized bymradermacher
FamilyGodzilla
Downloads (HF)2,798
Last updated2024-05-06

Frequently asked questions about GodziLLa 30B i1

How much VRAM does GodziLLa 30B i1 need?

About 33.5 GB at Q4_K_M with an 8K context window. That is the 19.9 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 GodziLLa 30B i1 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 21.0 GB, so you would need a larger GPU or CPU offloading.

What is the file size of GodziLLa 30B i1 GGUF?

It depends on the quant: IQ1_S is 7.4 GB, IQ1_M is 8.0 GB, IQ2_XXS is 9.0 GB, IQ2_XS is 9.9 GB, IQ2_S is 10.7 GB, IQ2_M is 11.5 GB and 15 more.

What context length does GodziLLa 30B i1 support?

The GGUF header lists a trained context length of 2,048 tokens (2K). The KV cache grows with the context you use: about 13.1 GB at 8K and 3.3 GB at 2K.

Where can I download GodziLLa 30B i1 GGUF?

The files are on Hugging Face in the mradermacher/GodziLLa-30B-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 GodziLLa 30B i1 use?

Hugging Face lists the license as cc-by-nc-4.0. 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 GodziLLa 30B i1 on 8 GB VRAM?

Yes, the IQ1_S and IQ1_M quantizations fit within 8 GB VRAM with an estimated 21–21.6 GB memory usage at 8k context.

Is commercial use allowed?

No. The model is licensed under CC‑BY‑NC‑4.0 and may only be used for non‑commercial purposes with attribution.

What context length does the model support?

The model is trained with a 2048‑token context window and does not support longer contexts.

Other GGUFs of GodziLLa-30B

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Godzilla models

All Godzilla →

Other medium models

All medium models →