NE

nemolita 21b i1

NemolitaGGUF by mradermacherUpdated 2024-08-09
Download on Hugging Face
Parameters20.4B
Context1000K
Layers70
Architecturellama
Smallest quant4.7 GB
Quick answer

nemolita 21b i1 needs about 15.2 GB of memory at Q4_K_M (a 12.4 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 21 quants, from 4.7 GB (IQ1_S) to 16.8 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 9.4 GB.

Nemolita 21b i1 – 20.4B llama model with tool‑use capability

Nemolita 21b i1 is a 20.4B parameter llama model merged from Mistral‑Nemo‑Instruct‑2407 and Nemomix‑v1.0‑12B. It supports tool usage and was trained on 1,024,000 context tokens. The model is available in multiple GGUF quantizations.

Good at

  • Tool‑based instruction following
  • Large context handling
  • High‑parameter language modeling

Limits

  • No vision support
  • License not specified
  • Requires large VRAM for higher quants

Best for

  • Developing tool‑integrated assistants
  • Running long‑form text generation
  • Fine‑tuning on custom instruction datasets

Who it’s for. Engineers and researchers who need a large llama model with tool‑use support and are comfortable managing GGUF files. The model is suitable for local deployment using llama.cpp.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S4.7 GB 7.6 GB
Tight Get file →
IQ1_M5.1 GB 8.0 GB
Tight Get file →
IQ2_XXS5.8 GB 8.6 GB
Too big Get file →
IQ2_XS6.3 GB 9.2 GB
Too big Get file →
IQ2_S6.7 GB 9.5 GB
Too big Get file →
IQ2_M7.2 GB 10.0 GB
Too big Get file →
Q2_K7.8 GB 10.6 GB
Too big Get file →
IQ3_XXS8.1 GB 10.9 GB
Too big Get file →
IQ3_XS8.6 GB 11.5 GB
Too big Get file →
Q3_K_S9.0 GB 11.9 GB
Too big Get file →
IQ3_S9.1 GB 11.9 GB
Too big Get file →
IQ3_M9.4 GB 12.2 GB
Too big Get file →
Q3_K_M10.0 GB 12.9 GB
Too big Get file →
Q3_K_L10.8 GB 13.7 GB
Too big Get file →
IQ4_XS11.1 GB 14.0 GB
Too big Get file →
Q4_011.7 GB 14.6 GB
Too big Get file →
Q4_K_S11.7 GB 14.6 GB
Too big Get file →
Q4_K_M12.4 GB 15.2 GB
Too big Get file →
Q5_K_S14.1 GB 17.0 GB
Too big Get file →
Q5_K_M14.5 GB 17.3 GB
Too big Get file →
Q6_K16.8 GB 19.6 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (file size 4.72 GB, estimated 7.6 GB memory at 8 k context). For 12 GB VRAM, IQ2_M (file size 7.19 GB, estimated 10 GB). For 16 GB VRAM, Q4_K_M (file size 12.36 GB, estimated 15.2 GB).

What GPU do you need for nemolita 21b 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: the best quant that fits is IQ1_M (5.1 GB file, about 8.0 GB in use).
  • 12 GB VRAM: the best quant that fits is IQ3_S (9.1 GB file, about 11.9 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_K_M (12.4 GB file, about 15.2 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (16.8 GB file, about 19.6 GB in use).

Which nemolita 21b i1 quant should you choose?

Q4_K_M is the usual starting point for nemolita 21b i1: it balances file size and quality well. Here it is a 12.4 GB file.

If memory is tight, the smallest file is IQ1_S at 4.7 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
Parameters20.4B
Layers70
Trained context1,024,000 tokens
KV cache (fp16)280 KB per token

Source

Quantized bymradermacher
FamilyNemolita
Downloads (HF)745
Last updated2024-08-09

Frequently asked questions about nemolita 21b i1

How much VRAM does nemolita 21b i1 need?

About 15.2 GB at Q4_K_M with an 8K context window. That is the 12.4 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 nemolita 21b i1 on an 8 GB GPU?

Yes, with IQ1_M: the 5.1 GB file needs about 8.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of nemolita 21b i1 GGUF?

It depends on the quant: IQ1_S is 4.7 GB, IQ1_M is 5.1 GB, IQ2_XXS is 5.8 GB, IQ2_XS is 6.3 GB, IQ2_S is 6.7 GB, IQ2_M is 7.2 GB and 15 more.

What context length does nemolita 21b i1 support?

The GGUF header lists a trained context length of 1,024,000 tokens (1000K). The KV cache grows with the context you use: about 2.3 GB at 8K and 9.4 GB at 32K.

Where can I download nemolita 21b i1 GGUF?

The files are on Hugging Face in the mradermacher/nemolita-21b-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Can I run Nemolita 21b i1 on 8 GB VRAM?

Yes, using the IQ1_S quantization which requires approximately 7.6 GB memory at 8 k context.

Does this model support vision tasks?

No, the model does not include vision capabilities.

What context length can it handle?

It was trained on 1,024,000 context tokens and can be used with that length in llama.cpp.

Other GGUFs of nemolita-21b

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Nemolita models

All Nemolita →

Other medium models

All medium models →