Mtrini-27B-Tellus-IQ2_XS-Imatrix

This repository contains the importance matrix and calibration data used for the Mtrini-27B-Tellus IQ2_XS quantization.

CompiwerAI — Building AI For Everyone.

Files

Importance Matrix

Mtrini-27B-Tellus-IQ2_XS.imatrix

Importance information used during IQ2_XS quantization.

Calibration Dataset

mtrini-calibration-large.txt

Calibration text used to generate the importance matrix.

Calibration Configuration

Property	Value
Context size	512
Threads	20
Chunks	32
Calibration PPL	~2.4629 ± 0.04545
Importance entries	496

Quantization Pipeline

1. Mtrini-27B-Tellus merged model
2. F16 GGUF conversion
3. Calibration dataset preparation
4. Importance-matrix generation
5. IQ2_XS quantization
6. GGUF validation

Related Models

• IQ2_XS: CompiwerAI/Mtrini-27B-Tellus-IQ2_XS
• F16: CompiwerAI/Mtrini-27B-Tellus-GGUF
• Merged: CompiwerAI/Mtrini-27B-Tellus-Merged