MI

Midnight Rose 103B v2.0.3 i1

MidnightGGUF by mradermacherUpdated 2024-11-18
Download on Hugging Face
Parameters103B
Context4K
Layers120
Architecturellama
Smallest quant22.0 GB
Quick answer

Midnight Rose 103B v2.0.3 i1 needs about 42.8 GB of memory at Q2_K (a 38.2 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 13 quants, from 22.0 GB (IQ1_S) to 46.4 GB (IQ3_M).

Midnight Rose 103B v2.0.3 i1 is a 103B-parameter Midnight model in GGUF format, quantized by mradermacher. Its trained context length is 4K tokens. The repository has 13 quant files, from 22.0 GB (IQ1_S) to 46.4 GB (IQ3_M).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S22.0 GB 24.5 GB
Too big Get file →
IQ1_M23.8 GB 26.3 GB
Too big Get file →
IQ2_XXS27.6 GB 30.2 GB
Too big Get file →
IQ2_XS30.7 GB 33.2 GB
Too big Get file →
IQ2_S32.2 GB 34.7 GB
Too big Get file →
IQ2_M35.0 GB 37.6 GB
Too big Get file →
Q2_K38.2 GB 40.7 GB
Too big Get file →
IQ3_XXS40.5 GB 43.1 GB
Too big Get file →
Q3_K_XS42.3 GB 44.8 GB
Too big Get file →
IQ3_XS42.5 GB 45.0 GB
Too big Get file →
Q3_K_S44.8 GB 47.3 GB
Too big Get file →
IQ3_S44.9 GB 47.4 GB
Too big Get file →
IQ3_M46.4 GB 48.9 GB
Too big Get file →

What GPU do you need for Midnight Rose 103B v2.0.3 i1?

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: no quant fits at an 8K context.
  • 32 GB VRAM: the best quant that fits is IQ1_M (23.8 GB file, about 28.3 GB in use).
  • 48 GB VRAM: the best quant that fits is IQ3_XS (42.5 GB file, about 47.0 GB in use).

Which Midnight Rose 103B v2.0.3 i1 quant should you choose?

Q2_K (38.2 GB) is the middle option in this repository. 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.

If memory is tight, the smallest file is IQ1_S at 22.0 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
Parameters103B
Layers120
Trained context4,096 tokens
KV cache (fp16)480 KB per token

Source

Quantized bymradermacher
FamilyMidnight
Licensellama2
Downloads (HF)473
Last updated2024-11-18

Frequently asked questions about Midnight Rose 103B v2.0.3 i1

How much VRAM does Midnight Rose 103B v2.0.3 i1 need?

About 42.8 GB at Q2_K with an 8K context window. That is the 38.2 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 Midnight Rose 103B v2.0.3 i1 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 26.5 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Midnight Rose 103B v2.0.3 i1 GGUF?

It depends on the quant: IQ1_S is 22.0 GB, IQ1_M is 23.8 GB, IQ2_XXS is 27.6 GB, IQ2_XS is 30.7 GB, IQ2_S is 32.2 GB, IQ2_M is 35.0 GB and 7 more.

What context length does Midnight Rose 103B v2.0.3 i1 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 4.0 GB at 8K and 2.0 GB at 4K.

Where can I download Midnight Rose 103B v2.0.3 i1 GGUF?

The files are on Hugging Face in the mradermacher/Midnight-Rose-103B-v2.0.3-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 Midnight Rose 103B v2.0.3 i1 use?

Hugging Face lists the license as llama2. 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 Midnight-Rose-103B-v2.0.3

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Midnight models

All Midnight →

Other large models

All large models →