QW

Qwen2-57B-A14B

QwenGGUF by mradermacherUpdated 2024-08-02
Download on Hugging Face
Parameters57.4B
Context128K
Layers28
Architectureqwen2moe
Smallest quant21.1 GB
Quick answer

Qwen2-57B-A14B needs about 35.8 GB of memory at Q4_K_M (a 34.9 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 15 quants, from 21.1 GB (Q2_K) to 47.1 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.

Qwen2-57B-A14B is a 57.4B-parameter Qwen model in GGUF format, quantized by mradermacher. Its trained context length is 128K tokens. The repository has 15 quant files, from 21.1 GB (Q2_K) to 47.1 GB (Q6_K).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K21.1 GB 22.0 GB
Too big Get file →
IQ3_XXS22.2 GB 23.2 GB
Too big Get file →
IQ3_XS23.6 GB 24.6 GB
Too big Get file →
Q3_K_S24.9 GB 25.9 GB
Too big Get file →
IQ3_S24.9 GB 25.9 GB
Too big Get file →
IQ3_M25.2 GB 26.2 GB
Too big Get file →
Q3_K_M27.5 GB 28.5 GB
Too big Get file →
Q3_K_L29.8 GB 30.8 GB
Too big Get file →
IQ4_XS30.7 GB 31.7 GB
Too big Get file →
Q4_032.6 GB 33.6 GB
Too big Get file →
Q4_K_S32.7 GB 33.7 GB
Too big Get file →
Q4_K_M34.9 GB 35.8 GB
Too big Get file →
Q5_K_S39.6 GB 40.5 GB
Too big Get file →
Q5_K_M40.8 GB 41.8 GB
Too big Get file →
Q6_K47.1 GB 48.1 GB
Too big Get file →

What GPU do you need for Qwen2-57B-A14B?

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 IQ3_XXS (22.2 GB file, about 23.2 GB in use).
  • 32 GB VRAM: the best quant that fits is IQ4_XS (30.7 GB file, about 31.7 GB in use).
  • 48 GB VRAM: the best quant that fits is Q5_K_M (40.8 GB file, about 41.8 GB in use).

Which Qwen2-57B-A14B quant should you choose?

Q4_K_M is the usual starting point for Qwen2-57B-A14B: it balances file size and quality well. Here it is a 34.9 GB file.

If memory is tight, the smallest file is Q2_K at 21.1 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

Architectureqwen2moe
Parameters57.4B (approximate, mixture of experts)
Active parameters14B
Layers28
Trained context131,072 tokens
KV cache (fp16)56 KB per token

Source

Quantized bymradermacher
FamilyQwen
Licenseapache-2.0
Downloads (HF)547
Last updated2024-08-02

Frequently asked questions about Qwen2-57B-A14B

How much VRAM does Qwen2-57B-A14B need?

About 35.8 GB at Q4_K_M with an 8K context window. That is the 34.9 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 Qwen2-57B-A14B on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 22.0 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Qwen2-57B-A14B GGUF?

It depends on the quant: Q2_K is 21.1 GB, IQ3_XXS is 22.2 GB, IQ3_XS is 23.6 GB, Q3_K_S is 24.9 GB, IQ3_S is 24.9 GB, IQ3_M is 25.2 GB and 9 more.

What context length does Qwen2-57B-A14B support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.

Where can I download Qwen2-57B-A14B GGUF?

The files are on Hugging Face in the mradermacher/Qwen2-57B-A14B-i1-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 Qwen2-57B-A14B 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.

Other GGUFs of Qwen2-57B-A14B

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Qwen models

All Qwen →

Other large models

All large models →