It stands for Importance Matrix, a technique used to improve the quality of quantized models. The Imatrix is calculated based on calibration data, and it helps determine the importance of different model activations during the quantization process. The idea is to preserve the most important information during quantization, which can help reduce the loss of model performance, especially when the calibration data is diverse. [1] [2]

Steps:

Base⇢ GGUF(F16)⇢ Imatrix-Data(F16)⇢ GGUF(Imatrix-Quants)

Quants:

python

    quantization_options = [
        "Q4_K_M", "IQ4_XS", "Q5_K_M", "Q5_K_S", "Q6_K",
        "Q8_0", "IQ3_M", "IQ3_S", "IQ3_XXS"
    ]

This is experimental.

For imatrix data generation, kalomaze's groups_merged.txt with added roleplay chats was used, you can find it here.

Alt-img:

Original model information:

So this should have been v1 but i done goofed, so heres "v2".

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

yaml

slices:
  - sources:
      - model: ChaoticNeutrals/Eris_Floramix_DPO_7B
        layer_range: [0, 32]
      - model: ResplendentAI/Datura_7B
        layer_range: [0, 32]
merge_method: slerp
base_model: ChaoticNeutrals/Eris_Floramix_DPO_7B
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

Lewdiculous/Eris-Daturamix-7b-v2-GGUF-IQ-Imatrix

Configuration

Lewdiculous/
Eris-Daturamix-7b-v2-GGUF-IQ-Imatrix