SC

Schisandra

SchisandraGGUF by bartowskiUpdated 2024-10-31
Download on Hugging Face
Parameters22.2B
Context32K
Layers56
Architecturellama
Smallest quant6.6 GB
Quick answer

Schisandra needs about 15.7 GB of memory at Q4_K_M (a 13.3 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 26 quants, from 6.6 GB (IQ2_XS) to 44.5 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 7.5 GB.

22B‑parameter llama model fine‑tuned for role‑playing and story writing

Schisandra is a 22.2B‑parameter llama model. It was created by merging eight Mistral Small finetunes and then re‑merged into the base model. The model is tuned for role‑playing and story writing tasks.

Good at

  • Role‑playing dialogues
  • Story generation
  • Character consistency

Limits

  • No vision support
  • Requires large VRAM for full precision
  • License is other

Best for

  • Writing interactive fiction
  • Creating character backstories
  • Generating long narrative scenes

Who it’s for. Developers and writers who need a large language model for creative writing. The model is suitable for those who can allocate sufficient GPU memory and are comfortable with llama.cpp quantizations.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ2_XS6.6 GB 9.0 GB
Too big Get file →
IQ2_S7.0 GB 9.4 GB
Too big Get file →
IQ2_M7.6 GB 10.0 GB
Too big Get file →
Q2_K8.3 GB 10.7 GB
Too big Get file →
Q2_K_L8.5 GB 10.8 GB
Too big Get file →
IQ3_XS9.2 GB 11.6 GB
Too big Get file →
Q3_K_S9.6 GB 12.0 GB
Too big Get file →
IQ3_M10.1 GB 12.4 GB
Too big Get file →
Q3_K_M10.8 GB 13.1 GB
Too big Get file →
Q3_K_L11.7 GB 14.1 GB
Too big Get file →
Q3_K_XL11.9 GB 14.3 GB
Too big Get file →
IQ4_XS11.9 GB 14.3 GB
Too big Get file →
Q4_012.6 GB 14.9 GB
Too big Get file →
Q4_012.6 GB 14.9 GB
Too big Get file →
Q4_012.6 GB 14.9 GB
Too big Get file →
Q4_012.6 GB 15.0 GB
Too big Get file →
Q4_K_S12.7 GB 15.0 GB
Too big Get file →
Q4_K_M13.3 GB 15.7 GB
Too big Get file →
Q4_K_L13.5 GB 15.9 GB
Too big Get file →
Q5_K_S15.3 GB 17.7 GB
Too big Get file →
Q5_K_M15.7 GB 18.1 GB
Too big Get file →
Q5_K_L15.8 GB 18.2 GB
Too big Get file →
Q6_K18.3 GB 20.6 GB
Too big Get file →
Q6_K_L18.4 GB 20.7 GB
Too big Get file →
Q8_023.6 GB 26.0 GB
Too big Get file →
F1644.5 GB 46.9 GB
Too big Get file →

For 8 GB VRAM, use Q4_0 (file size 12.57 GB, estimated 14.9 GB memory). For 12 GB VRAM, Q4_K_M (13.34 GB, 15.7 GB memory) is a good balance. If you have 16 GB VRAM, Q5_K_S (15.32 GB, 17.7 GB memory) offers higher quality while staying within limits.

What GPU do you need for Schisandra?

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 Q3_K_S (9.6 GB file, about 12.0 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_K_L (13.5 GB file, about 15.9 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K_L (18.4 GB file, about 20.7 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (23.6 GB file, about 26.0 GB in use).
  • 48 GB VRAM: the best quant that fits is F16 (44.5 GB file, about 46.9 GB in use).

Which Schisandra quant should you choose?

Q4_K_M is the usual starting point for Schisandra: it balances file size and quality well. Here it is a 13.3 GB file.

The Q8_0 file is 23.6 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 IQ2_XS at 6.6 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
Parameters22.2B
Layers56
Trained context32,768 tokens
KV cache (fp16)224 KB per token

Source

Quantized bybartowski
Licenseother
Downloads (HF)1,734
Last updated2024-10-31

Frequently asked questions about Schisandra

How much VRAM does Schisandra need?

About 15.7 GB at Q4_K_M with an 8K context window. That is the 13.3 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 Schisandra on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 9.0 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Schisandra GGUF?

It depends on the quant: IQ2_XS is 6.6 GB, IQ2_S is 7.0 GB, IQ2_M is 7.6 GB, Q2_K is 8.3 GB, Q2_K_L is 8.5 GB, IQ3_XS is 9.2 GB and 20 more.

What context length does Schisandra support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.9 GB at 8K and 7.5 GB at 32K.

Where can I download Schisandra GGUF?

The files are on Hugging Face in the bartowski/MS-Schisandra-22B-v0.1-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 Schisandra 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.

Can I run Schisandra on 8 GB VRAM?

Yes, use the Q4_0 quant which requires about 14.9 GB of memory at 8 k context. It will fit on 8 GB VRAM with offloading.

Does Schisandra support image inputs?

No, the model has no vision capability.

What is the maximum context length?

The model was trained with 32,768 tokens and can handle that context length.

Other GGUFs of MS-Schisandra-22B-v0.1

Similar models

Browse all →

More from bartowski

All by bartowski →

More Schisandra models

All Schisandra →

Other medium models

All medium models →