xXFiEsTaDeAmOnXx's picture
Improve language tag (#1)
132bbba verified
|
raw
history blame
2.6 kB
metadata
base_model:
  - Qwen/Qwen2.5-3B-Instruct
library_name: transformers
tags:
  - mergekit
  - merge
language:
  - zho
  - eng
  - fra
  - spa
  - por
  - deu
  - ita
  - rus
  - jpn
  - kor
  - vie
  - tha
  - ara

merge

This is a merge of pre-trained language models created using mergekit.

Merge Details

The model was merged using the passthrough merge method, with a sliding window approach.

Duplicated Layers

Due to the sliding window method, layers overlap between slices. The following layers are duplicated across multiple slices:

  • Layers 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32

This means each layer is used in multiple slices, with each slice overlapping with the previous and next slice.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

slices:
  - sources:
      - model:  Qwen/Qwen2.5-3B-Instruct
        layer_range: [0, 4]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [2, 6]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [4, 8]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [6, 10]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [8, 12]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [10, 14]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [12, 16]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [14, 18]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [16, 20]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [18, 22]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [20, 24]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [22, 26]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [24, 28]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [26, 30]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [28, 32]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [30, 34]
  - sources:
      - model: Qwen/Qwen2.5-3B-Instruct
        layer_range: [32, 36]
tokenizer_source:  Qwen/Qwen2.5-3B-Instruct
merge_method: passthrough
dtype: bfloat16