Yarn Llama 2 70b 32k needs about 31.4 GB of memory at Q3_K_XS (a 28.3 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 10 quants, from 18.3 GB (IQ2_XXS) to 47.5 GB (Q5_K_S). Longer context needs more: at 32K tokens the KV cache alone adds about 10.7 GB.
Yarn Llama 2 70b 32k is a 69B-parameter Llama model in GGUF format, quantized by Artefact2. Its trained context length is 32K tokens. The repository has 10 quant files, from 18.3 GB (IQ2_XXS) to 47.5 GB (Q5_K_S).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XXS | 18.3 GB | 21.5 GB | Too big | Get file → | |
| IQ2_XS | 20.3 GB | 23.5 GB | Too big | Get file → | |
| Q2_K_S | 23.6 GB | 26.7 GB | Too big | Get file → | |
| Q2_K | 25.5 GB | 28.6 GB | Too big | Get file → | |
| Q3_K_XS | 28.3 GB | 31.4 GB | Too big | Get file → | |
| Q3_K_S | 29.9 GB | 33.1 GB | Too big | Get file → | |
| Q3_K_M | 33.3 GB | 36.5 GB | Too big | Get file → | |
| Q3_K_L | 36.1 GB | 39.3 GB | Too big | Get file → | |
| Q4_K_S | 39.2 GB | 42.4 GB | Too big | Get file → | |
| Q5_K_S | 47.5 GB | 50.6 GB | Too big | Get file → |
What GPU do you need for Yarn Llama 2 70b 32k?
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: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is IQ2_XS (20.3 GB file, about 23.5 GB in use).
- 32 GB VRAM: the best quant that fits is Q3_K_XS (28.3 GB file, about 31.4 GB in use).
- 48 GB VRAM: the best quant that fits is Q4_K_S (39.2 GB file, about 42.4 GB in use).
Which Yarn Llama 2 70b 32k quant should you choose?
Q3_K_XS (28.3 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
If memory is tight, the smallest file is IQ2_XXS at 18.3 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
Source
Frequently asked questions about Yarn Llama 2 70b 32k
How much VRAM does Yarn Llama 2 70b 32k need?
About 31.4 GB at Q3_K_XS with an 8K context window. That is the 28.3 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 Yarn Llama 2 70b 32k on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 21.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Yarn Llama 2 70b 32k GGUF?
It depends on the quant: IQ2_XXS is 18.3 GB, IQ2_XS is 20.3 GB, Q2_K_S is 23.6 GB, Q2_K is 25.5 GB, Q3_K_XS is 28.3 GB, Q3_K_S is 29.9 GB and 4 more.
What context length does Yarn Llama 2 70b 32k support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 2.7 GB at 8K and 10.7 GB at 32K.
Where can I download Yarn Llama 2 70b 32k GGUF?
The files are on Hugging Face in the Artefact2/Yarn-Llama-2-70b-32k-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 Yarn Llama 2 70b 32k use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.