CH

Chaifighter-20B-v2

ChaifighterGGUF by backyardaiUpdated 2024-05-22
Download on Hugging Face
Parameters19.5B
Context4K
Layers88
Architecturellama
Smallest quant4.2 GB
Quick answer

Chaifighter-20B-v2 needs about 15.1 GB of memory at Q4_K_M (a 11.7 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 23 quants, from 4.2 GB (IQ1_S) to 38.9 GB (BF16).

Chaifighter‑20B‑v2: a 20B Llama‑based model for long‑form roleplay

Chaifighter‑20B‑v2 is a 19.5B‑parameter Llama model optimized for verbose, creative storytelling. It supports a 4096‑token context window and is designed for long‑form roleplay and essay writing. The model is a frankenmerge of several open‑source models.

Good at

  • Long‑form creative writing
  • Verbose roleplay
  • Reasoning and coherence

Limits

  • 4096‑token context window
  • No vision support
  • License not specified

Best for

  • Character‑driven storytelling
  • Essay drafting
  • Creative brainstorming

Who it’s for. Developers and hobbyists who need a medium‑size model for detailed, long‑form dialogue. It is suitable for local inference on GPUs with moderate VRAM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S4.2 GB 6.2 GB
Fits Get file →
IQ1_M4.6 GB 6.6 GB
Fits Get file →
IQ2_XXS5.2 GB 7.2 GB
Tight Get file →
IQ2_XS5.8 GB 7.8 GB
Tight Get file →
IQ2_S6.1 GB 8.1 GB
Too big Get file →
IQ2_M6.6 GB 8.6 GB
Too big Get file →
Q2_K_S6.7 GB 8.7 GB
Too big Get file →
Q2_K7.2 GB 9.2 GB
Too big Get file →
IQ3_XXS7.5 GB 9.5 GB
Too big Get file →
IQ3_XS8.0 GB 10.0 GB
Too big Get file →
Q3_K_S8.4 GB 10.4 GB
Too big Get file →
IQ3_S8.5 GB 10.4 GB
Too big Get file →
IQ3_M8.7 GB 10.7 GB
Too big Get file →
Q3_K_M9.4 GB 11.4 GB
Too big Get file →
Q3_K_L10.2 GB 12.2 GB
Too big Get file →
IQ4_XS10.4 GB 12.4 GB
Too big Get file →
Q4_K_S11.1 GB 13.0 GB
Too big Get file →
Q4_K_M11.7 GB 13.7 GB
Too big Get file →
Q5_K_S13.4 GB 15.4 GB
Too big Get file →
Q5_K_M13.8 GB 15.7 GB
Too big Get file →
Q6_K16.0 GB 17.9 GB
Too big Get file →
Q8_020.7 GB 22.7 GB
Too big Get file →
BF1638.9 GB 40.9 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (4.2 GB, 7.7 GB memory) or IQ1_M (4.59 GB, 8 GB memory). For 12 GB VRAM, IQ2_M (6.62 GB, 10.1 GB memory) or Q2_K (7.21 GB, 10.7 GB memory) are good choices. Higher‑quality Q3 or Q4 quantizations require 16 GB or more.

What GPU do you need for Chaifighter-20B-v2?

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 (4.6 GB file, about 8.0 GB in use).
  • 12 GB VRAM: the best quant that fits is IQ3_S (8.5 GB file, about 11.9 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_K_M (11.7 GB file, about 15.1 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (16.0 GB file, about 19.4 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (20.7 GB file, about 24.1 GB in use).
  • 48 GB VRAM: the best quant that fits is BF16 (38.9 GB file, about 42.4 GB in use).

Which Chaifighter-20B-v2 quant should you choose?

Q4_K_M is the usual starting point for Chaifighter-20B-v2: it balances file size and quality well. Here it is a 11.7 GB file.

The Q8_0 file is 20.7 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 4.2 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
Parameters19.5B
Layers88
Trained context4,096 tokens
KV cache (fp16)352 KB per token

Source

Quantized bybackyardai
Downloads (HF)964
Last updated2024-05-22

Frequently asked questions about Chaifighter-20B-v2

How much VRAM does Chaifighter-20B-v2 need?

About 15.1 GB at Q4_K_M with an 8K context window. That is the 11.7 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 Chaifighter-20B-v2 on an 8 GB GPU?

Yes, with IQ1_M: the 4.6 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 Chaifighter-20B-v2 GGUF?

It depends on the quant: IQ1_S is 4.2 GB, IQ1_M is 4.6 GB, IQ2_XXS is 5.2 GB, IQ2_XS is 5.8 GB, IQ2_S is 6.1 GB, IQ2_M is 6.6 GB and 17 more.

What context length does Chaifighter-20B-v2 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 3.0 GB at 8K and 1.5 GB at 4K.

Where can I download Chaifighter-20B-v2 GGUF?

The files are on Hugging Face in the backyardai/Chaifighter-20B-v2-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 Chaifighter‑20B‑v2 on 8 GB VRAM?

Yes, use the IQ1_S or IQ1_M quantization. They fit within 8 GB VRAM and use about 7.7–8 GB of memory at 8K context.

Does Chaifighter‑20B‑v2 support image input?

No, the model has no vision capabilities.

Other GGUFs of Chaifighter-20B-v2

Similar models

Browse all →

More from backyardai

All by backyardai →

More Chaifighter models

All Chaifighter →

Other medium models

All medium models →