Mistral-v0.3-13B-32k-Base

Mistral-v0.3-13B-32k-Base is a merge of the following models using mergekit:

🧩 Configuration

dtype: bfloat16
merge_method: linear # use linear so we can include multiple models, albeit at a zero weight
parameters:
  weight: 1.0
slices:
- sources:
  - layer_range: [0, 8]
    model: win10/Breeze-13B-32k-Base-v1_0
  - layer_range: [0, 8]
    model: win10/Mistral-v0.3-13B-32k-Base
    parameters:
      weight: 0
- sources:
  - layer_range: [4, 12]
    model: win10/Mistral-v0.3-13B-32k-Base
  - layer_range: [4, 12]
    model: win10/Breeze-13B-32k-Base-v1_0
    parameters:
      weight: 0
- sources:
  - layer_range: [8, 16]
    model: win10/Breeze-13B-32k-Base-v1_0
    parameters:
        weight: 0
  - layer_range: [8, 16]
    model: win10/Mistral-v0.3-13B-32k-Base
    parameters:
      weight: 0
- sources:
  - layer_range: [12, 20]
    model: win10/Mistral-v0.3-13B-32k-Base
  - layer_range: [12, 20]
    model: win10/Breeze-13B-32k-Base-v1_0
    parameters:
      weight: 0
- sources:
  - layer_range: [16, 24]
    model: win10/Breeze-13B-32k-Base-v1_0
  - layer_range: [16, 24]
    model: win10/Mistral-v0.3-13B-32k-Base
    parameters:
      weight: 0
- sources:
  - layer_range: [20, 28]
    model: win10/Mistral-v0.3-13B-32k-Base
  - layer_range: [20, 28]
    model: win10/Breeze-13B-32k-Base-v1_0
    parameters:
      weight: 0
- sources:
  - layer_range: [24, 32]
    model: win10/Breeze-13B-32k-Base-v1_0
  - layer_range: [24, 32]
    model: win10/Mistral-v0.3-13B-32k-Base
    parameters:
      weight: 0
- sources:
  - model: win10/Mistral-v0.3-13B-32k-Base
    layer_range: [28, 42]
  - model: win10/Breeze-13B-32k-Base-v1_0
    layer_range: [28, 42]
    parameters:
      weight: 0
- sources:
  - model: win10/Breeze-13B-32k-Base-v1_0
    layer_range: [32, 48]
  - model: win10/Mistral-v0.3-13B-32k-Base
    layer_range: [32, 48]
    parameters:
      weight: 0
- sources:
  - model: win10/Mistral-v0.3-13B-32k-Base
    layer_range: [38, 52]
  - model: win10/Breeze-13B-32k-Base-v1_0
    layer_range: [38, 52]
    parameters:
      weight: 0
- sources:
  - model: win10/Breeze-13B-32k-Base-v1_0
    layer_range: [48, 56]
  - model: win10/Mistral-v0.3-13B-32k-Base
    layer_range: [48, 56]
    parameters:
      weight: 0
tokenizer_source: union
Downloads last month
9
Safetensors
Model size
12B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support