LL

Llama3merge7-15B-MoE

LlamaGGUF by RichardErkhovUpdated 2024-07-21
Download on Hugging Face
Parameters15B
Context8K
Layers32
Architecturellama
Smallest quant4.9 GB
Quick answer

Llama3merge7-15B-MoE needs about 9.9 GB of memory at Q4_K_M (a 8.3 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 14 quants, from 4.9 GB (IQ3_M) to 8.3 GB (Q4_K_M).

Llama3merge7-15B-MoE is a 15B-parameter Llama model in GGUF format, quantized by RichardErkhov. Its trained context length is 8K tokens. The repository has 14 quant files, from 4.9 GB (IQ3_M) to 8.3 GB (Q4_K_M).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ3_M4.9 GB 6.5 GB
Fits Get file →
Q2_K5.2 GB 6.8 GB
Fits Get file →
Q3_K_M5.7 GB 7.3 GB
Tight Get file →
IQ3_XS5.8 GB 7.4 GB
Tight Get file →
Q3_K_S6.1 GB 7.7 GB
Tight Get file →
IQ3_S6.1 GB 7.7 GB
Tight Get file →
Q3_K6.7 GB 8.3 GB
Too big Get file →
Q3_K_L7.2 GB 8.8 GB
Too big Get file →
IQ4_XS7.5 GB 9.1 GB
Too big Get file →
Q4_07.8 GB 9.4 GB
Too big Get file →
Q4_K_S7.9 GB 9.5 GB
Too big Get file →
IQ4_NL7.9 GB 9.5 GB
Too big Get file →
Q4_K8.3 GB 9.9 GB
Too big Get file →
Q4_K_M8.3 GB 9.9 GB
Too big Get file →

What GPU do you need for Llama3merge7-15B-MoE?

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_S (6.1 GB file, about 7.7 GB in use).
  • 12 GB VRAM: the best quant that fits is Q4_K (8.3 GB file, about 9.9 GB in use).

Which Llama3merge7-15B-MoE quant should you choose?

Q4_K_M is the usual starting point for Llama3merge7-15B-MoE: it balances file size and quality well. Here it is a 8.3 GB file.

If memory is tight, the smallest file is IQ3_M at 4.9 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters15B (approximate, mixture of experts)
Layers32
Trained context8,192 tokens
KV cache (fp16)128 KB per token

Source

Quantized byRichardErkhov
FamilyLlama
Downloads (HF)4
Last updated2024-07-21

Frequently asked questions about Llama3merge7-15B-MoE

How much VRAM does Llama3merge7-15B-MoE need?

About 9.9 GB at Q4_K_M with an 8K context window. That is the 8.3 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 Llama3merge7-15B-MoE on an 8 GB GPU?

Yes, with IQ3_S: the 6.1 GB file needs about 7.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Llama3merge7-15B-MoE GGUF?

It depends on the quant: IQ3_M is 4.9 GB, Q2_K is 5.2 GB, Q3_K_M is 5.7 GB, IQ3_XS is 5.8 GB, Q3_K_S is 6.1 GB, IQ3_S is 6.1 GB and 8 more.

What context length does Llama3merge7-15B-MoE support?

The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 1.1 GB at 8K and 1.1 GB at 8K.

Where can I download Llama3merge7-15B-MoE GGUF?

The files are on Hugging Face in the RichardErkhov/allknowingroger_-_Llama3merge7-15B-MoE-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Similar models

Browse all →

More from RichardErkhov

All by RichardErkhov →

More Llama models

All Llama →

Other medium models

All medium models →