README.md

metadata

base_model:
  - Aratako/Antler-7B-RP-v2
  - NTQAI/chatntq-ja-7b-v1.0
  - TFMC/Japanese-Starling-ChatV-7B
library_name: transformers
tags:
  - mergekit
  - merge
license: apache-2.0
language:
  - ja

japanese-chat-umievo-itr004-7b

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

Merge Details

Merge Method

This model was merged using the linear merge method.

Models Merged

The following models were included in the merge:

• Aratako/Antler-7B-RP-v2
• NTQAI/chatntq-ja-7b-v1.0
• TFMC/Japanese-Starling-ChatV-7B

Configuration

The following YAML configuration was used to produce this model:

dtype: bfloat16
merge_method: linear
parameters:
  int8_mask: 1.0
  normalize: 1.0
slices:
- sources:
  - layer_range: [0, 4]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.3117821548077149
  - layer_range: [0, 4]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.06622915649259134
  - layer_range: [0, 4]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.1456932001643398
- sources:
  - layer_range: [4, 8]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.235386369865962
  - layer_range: [4, 8]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.4897808973741158
  - layer_range: [4, 8]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: -0.13610409063514917
- sources:
  - layer_range: [8, 12]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: -0.02904316547026939
  - layer_range: [8, 12]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.27984511027305725
  - layer_range: [8, 12]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.17613114842003957
- sources:
  - layer_range: [12, 16]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.30708851106133633
  - layer_range: [12, 16]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.633483361515513
  - layer_range: [12, 16]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.23184742233331504
- sources:
  - layer_range: [16, 20]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.4261576863911635
  - layer_range: [16, 20]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.448790034700236
  - layer_range: [16, 20]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.908083073839566
- sources:
  - layer_range: [20, 24]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.49909163132229717
  - layer_range: [20, 24]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.46200102252131314
  - layer_range: [20, 24]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.3537781762598641
- sources:
  - layer_range: [24, 28]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.3053835052195379
  - layer_range: [24, 28]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: 0.554162215544036
  - layer_range: [24, 28]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.3526529317468468
- sources:
  - layer_range: [28, 32]
    model: NTQAI/chatntq-ja-7b-v1.0
    parameters:
      weight: 0.458122927784649
  - layer_range: [28, 32]
    model: TFMC/Japanese-Starling-ChatV-7B
    parameters:
      weight: -0.20730388433466773
  - layer_range: [28, 32]
    model: Aratako/Antler-7B-RP-v2
    parameters:
      weight: 0.5077362589077673