TO

tokyotech llm Swallow 7b instruct v0.1

TokyotechGGUF by mmngaUpdated 2024-05-03
Download on Hugging Face
Parameters6.83B
Context4K
Layers32
Architecturellama
Smallest quant1.6 GB
Quick answer

tokyotech llm Swallow 7b instruct v0.1 needs about 8.9 GB of memory at Q4_K_M (a 4.1 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 24 quants, from 1.6 GB (IQ1_S) to 7.3 GB (Q8_0).

Tokyotech Swallow 7b Instruct v0.1 – a 6.83B Japanese‑focused LLM

The Swallow 7b Instruct v0.1 is a 6.83B parameter model trained on 4096 token context. It is released under the llama2 license and supports Japanese and English. The model is available in multiple GGUF quantizations for different memory budgets.

Good at

  • Japanese instruction following
  • English text generation
  • Medium‑size context handling

Limits

  • No vision support
  • Requires 8k context for full performance
  • License restricts commercial use

Best for

  • Chatbot prototypes
  • Japanese content creation
  • Educational demos

Who it’s for. Developers building Japanese language applications who need a lightweight instruction‑tuned model. Researchers testing llama‑based architectures in a bilingual setting.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S1.6 GB 4.2 GB
Fits Get file →
IQ1_M1.7 GB 4.3 GB
Fits Get file →
IQ2_XXS1.9 GB 4.5 GB
Fits Get file →
IQ2_XS2.1 GB 4.7 GB
Fits Get file →
IQ2_S2.2 GB 4.9 GB
Fits Get file →
IQ2_M2.4 GB 5.1 GB
Fits Get file →
Q2_K2.6 GB 5.2 GB
Fits Get file →
IQ3_XXS2.6 GB 5.3 GB
Fits Get file →
IQ3_XS2.9 GB 5.5 GB
Fits Get file →
IQ3_S3.0 GB 5.7 GB
Fits Get file →
Q3_K_S3.0 GB 5.7 GB
Fits Get file →
IQ3_M3.2 GB 5.8 GB
Fits Get file →
Q3_K_M3.4 GB 6.0 GB
Fits Get file →
Q3_K_L3.7 GB 6.3 GB
Fits Get file →
IQ4_XS3.7 GB 6.3 GB
Fits Get file →
IQ4_NL3.9 GB 6.5 GB
Fits Get file →
Q4_03.9 GB 6.5 GB
Fits Get file →
Q4_K_S3.9 GB 6.6 GB
Fits Get file →
Q4_K_M4.1 GB 6.8 GB
Fits Get file →
Q5_04.7 GB 7.4 GB
Tight Get file →
Q5_K_S4.7 GB 7.4 GB
Tight Get file →
Q5_K_M4.9 GB 7.5 GB
Tight Get file →
Q6_K5.6 GB 8.3 GB
Too big Get file →
Q8_07.3 GB 9.9 GB
Too big Get file →

For 8 GB VRAM, use IQ2_S (2.25 GB file, ~7 GB memory). For 12 GB VRAM, Q6_K (5.6 GB file, ~10.4 GB memory) offers higher precision. For 16 GB VRAM, Q8_0 (7.26 GB file, ~12.1 GB memory) provides the best quality.

What GPU do you need for tokyotech llm Swallow 7b 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: the best quant that fits is IQ3_M (3.2 GB file, about 8.0 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (5.6 GB file, about 10.4 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (7.3 GB file, about 12.1 GB in use).

Which tokyotech llm Swallow 7b instruct v0.1 quant should you choose?

Q4_K_M is the usual starting point for tokyotech llm Swallow 7b instruct v0.1: it balances file size and quality well. Here it is a 4.1 GB file.

The Q8_0 file is 7.3 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 IQ1_S at 1.6 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
Parameters6.83B
Layers32
Trained context4,096 tokens
KV cache (fp16)512 KB per token

Source

Quantized bymmnga
FamilyTokyotech
Licensellama2
Downloads (HF)1,894
Last updated2024-05-03

Frequently asked questions about tokyotech llm Swallow 7b instruct v0.1

How much VRAM does tokyotech llm Swallow 7b instruct v0.1 need?

About 8.9 GB at Q4_K_M with an 8K context window. That is the 4.1 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 tokyotech llm Swallow 7b instruct v0.1 on an 8 GB GPU?

Yes, with IQ3_M: the 3.2 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 tokyotech llm Swallow 7b instruct v0.1 GGUF?

It depends on the quant: IQ1_S is 1.6 GB, IQ1_M is 1.7 GB, IQ2_XXS is 1.9 GB, IQ2_XS is 2.1 GB, IQ2_S is 2.2 GB, IQ2_M is 2.4 GB and 18 more.

What context length does tokyotech llm Swallow 7b 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 4.3 GB at 8K and 2.1 GB at 4K.

Where can I download tokyotech llm Swallow 7b instruct v0.1 GGUF?

The files are on Hugging Face in the mmnga/tokyotech-llm-Swallow-7b-instruct-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 tokyotech llm Swallow 7b 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 this model on 8 GB VRAM?

Yes, using the IQ2_S quantization which requires about 7 GB of VRAM at 8k context.

Does this model support image inputs?

No, the Swallow 7b Instruct v0.1 has no vision capabilities.

What license does it use?

The model is released under the llama2 license.

Similar models

Browse all →

More from mmnga

All by mmnga →

More Tokyotech models

All Tokyotech →

Other small models

All small models →