SW

Swallow-13b-instruct-v0.1

SwallowGGUF by mradermacherUpdated 2024-08-02
Download on Hugging Face
Parameters13.1B
Context4K
Layers40
Architecturellama
Smallest quant3.0 GB
Quick answer

Swallow-13b-instruct-v0.1 needs about 15.2 GB of memory at Q4_K_M (a 7.9 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 21 quants, from 3.0 GB (IQ1_S) to 10.8 GB (Q6_K).

Swallow‑13b‑instruct‑v0.1 – 13.1B LLaMA‑2 based instruction model

Swallow‑13b‑instruct‑v0.1 is a 13.1B parameter LLaMA‑2 model fine‑tuned for instruction following. It supports Japanese and English and was released on April 26, 2024. The model is available in multiple GGUF quantizations.

Good at

  • Instruction following
  • Japanese‑English bilingual text
  • Fine‑tuned LLaMA‑2 architecture

Limits

  • No vision support
  • Requires 8 k context for estimates
  • License: llama2

Best for

  • Chatbots with bilingual support
  • Japanese content generation
  • Instruction‑driven research prototypes

Who it’s for. Developers and researchers needing a bilingual instruction model based on LLaMA‑2. The model is suitable for applications that require Japanese and English language handling without vision capabilities.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S3.0 GB 6.8 GB
Fits Get file →
IQ1_M3.2 GB 7.1 GB
Fits Get file →
IQ2_XXS3.6 GB 7.5 GB
Tight Get file →
IQ2_XS3.9 GB 7.8 GB
Tight Get file →
IQ2_S4.3 GB 8.1 GB
Too big Get file →
IQ2_M4.6 GB 8.4 GB
Too big Get file →
Q2_K4.9 GB 8.8 GB
Too big Get file →
IQ3_XXS5.0 GB 8.9 GB
Too big Get file →
IQ3_XS5.4 GB 9.3 GB
Too big Get file →
IQ3_S5.7 GB 9.6 GB
Too big Get file →
Q3_K_S5.7 GB 9.6 GB
Too big Get file →
IQ3_M6.1 GB 9.9 GB
Too big Get file →
Q3_K_M6.4 GB 10.3 GB
Too big Get file →
Q3_K_L7.0 GB 10.9 GB
Too big Get file →
IQ4_XS7.0 GB 10.9 GB
Too big Get file →
Q4_07.5 GB 11.3 GB
Too big Get file →
Q4_K_S7.5 GB 11.4 GB
Too big Get file →
Q4_K_M7.9 GB 11.8 GB
Too big Get file →
Q5_K_S9.1 GB 12.9 GB
Too big Get file →
Q5_K_M9.3 GB 13.2 GB
Too big Get file →
Q6_K10.8 GB 14.6 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (2.96 GB, 10.2 GB memory) or IQ1_M (3.2 GB, 10.4 GB memory). For 16 GB VRAM, Q4_K_M (7.95 GB, 15.2 GB memory) or Q5_K_M (9.32 GB, 16.5 GB memory) are good choices. Higher‑quality quants like Q6_K (10.77 GB, 18 GB memory) require more VRAM.

What GPU do you need for Swallow-13b-instruct-v0.1?

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 IQ2_M (4.6 GB file, about 11.8 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_K_M (7.9 GB file, about 15.2 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (10.8 GB file, about 18.0 GB in use).

Which Swallow-13b-instruct-v0.1 quant should you choose?

Q4_K_M is the usual starting point for Swallow-13b-instruct-v0.1: it balances file size and quality well. Here it is a 7.9 GB file.

If memory is tight, the smallest file is IQ1_S at 3.0 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
Parameters13.1B
Layers40
Trained context4,096 tokens
KV cache (fp16)800 KB per token

Source

Quantized bymradermacher
FamilySwallow
Licensellama2
Downloads (HF)619
Last updated2024-08-02

Frequently asked questions about Swallow-13b-instruct-v0.1

How much VRAM does Swallow-13b-instruct-v0.1 need?

About 15.2 GB at Q4_K_M with an 8K context window. That is the 7.9 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 Swallow-13b-instruct-v0.1 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 10.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Swallow-13b-instruct-v0.1 GGUF?

It depends on the quant: IQ1_S is 3.0 GB, IQ1_M is 3.2 GB, IQ2_XXS is 3.6 GB, IQ2_XS is 3.9 GB, IQ2_S is 4.3 GB, IQ2_M is 4.6 GB and 15 more.

What context length does Swallow-13b-instruct-v0.1 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 6.7 GB at 8K and 3.4 GB at 4K.

Where can I download Swallow-13b-instruct-v0.1 GGUF?

The files are on Hugging Face in the mradermacher/Swallow-13b-instruct-v0.1-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 Swallow-13b-instruct-v0.1 use?

Hugging Face lists the license as llama2. 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 Swallow‑13b‑instruct‑v0.1 on 8 GB VRAM?

Yes, using IQ1_S or IQ1_M quantizations will fit within 8 GB VRAM with an estimated memory usage of 10.2–10.4 GB at 8 k context.

Does the model support vision tasks?

No, Swallow‑13b‑instruct‑v0.1 does not include vision capabilities.

What languages does the model handle?

The model supports Japanese and English, with a tokenizer optimized for Japanese data.

Other GGUFs of Swallow-13b-instruct-v0.1

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Swallow models

All Swallow →

Other medium models

All medium models →