Question 1

How much VRAM does Yarn-Llama-2-7b-64k need?

Accepted Answer

Yarn-Llama-2-7b-64k (7B) requires approximately 15.061532592773437 GB of VRAM at full precision (FP16), 7.5 GB at 8-bit quantization (Q8), and 3.8 GB at 4-bit quantization (Q4). The actual VRAM usage may be slightly higher due to KV cache and runtime overhead.

Question 2

What GPUs can run Yarn-Llama-2-7b-64k?

Accepted Answer

We have 10 compatible GPUs indexed for Yarn-Llama-2-7b-64k, including L4, A40, A100 40GB, A10G, RTX A6000 and 5 more. Compatibility is determined by whether the GPU's VRAM can fit the model's memory footprint, potentially using quantization to reduce memory usage.

Question 3

What is the cheapest GPU to run Yarn-Llama-2-7b-64k?

Accepted Answer

The cheapest GPU to run Yarn-Llama-2-7b-64k in the cloud is the L4 at $0.20/hr. Other affordable options include A40 ($0.44/hr) and A100 40GB ($0.93/hr). Prices reflect the lowest on-demand rate across cloud providers.

Question 4

Can Yarn-Llama-2-7b-64k run on consumer hardware?

Accepted Answer

Yes, Yarn-Llama-2-7b-64k can run on consumer GPUs like the RTX A6000 (48 GB VRAM) at full precision (FP16).

Question 5

What quantization options are available for Yarn-Llama-2-7b-64k?

Accepted Answer

Yarn-Llama-2-7b-64k supports the following quantization levels: FP16 (full precision): ~15.061532592773437 GB, Q8 (8-bit): ~7.5 GB, Q6 (6-bit): ~5.6 GB, Q4 (4-bit): ~3.8 GB. Lower quantization reduces VRAM usage at the cost of some quality loss. Q8 offers a good balance between quality and memory savings, while Q4 provides maximum compression.

GPU	VRAM	Best Quant	From	Rent
L4	24 GB	FP16	$0.20/hr
A40	48 GB	FP16	$0.44/hr	Rent
A100 40GB	40 GB	FP16	$0.93/hr	Rent
A10G	24 GB	FP16	$1.01/hr
RTX A6000	48 GB	FP16	$1.09/hr
A10	24 GB	FP16	$1.29/hr
A100 80GB PCIe	80 GB	FP16	$1.39/hr	Rent
A100 80GB	80 GB	FP16	$1.39/hr	Rent
H100 MEGA	80 GB	FP16	$1.80/hr	Rent
H100 SXM	80 GB	FP16	$1.80/hr	Rent

Yarn-Llama-2-7b-64k

VRAM Requirements

Compatible GPUs

Yarn-Llama-2-7b-64k FAQ

VRAM requirements for Yarn-Llama-2-7b-64k at different quantization levels
Quantization	VRAM Required
FP16	15.1 GB
Q8	7.5 GB
Q6	5.6 GB
Q4	3.8 GB