We first merge 14 models to produce
EmbeddedLLM/Mistral-7B-Merge-14-v0.3,
which is then merged again with
Weyaxi/OpenHermes-2.5-neural-chat-v3-3-openchat-3.5-1210-Slerp using Gradient SLERP.
The result is a model that performs quite well but may require further instruction fine-tuning.
Either ChatML or Llama-2 chat template.
1slices:
2 - sources:
3 - model: Weyaxi/OpenHermes-2.5-neural-chat-v3-3-openchat-3.5-1210-Slerp
4 layer_range: [0, 32]
5 - model: EmbeddedLLM/Mistral-7B-Merge-14-v0.3
6 layer_range: [0, 32]
7
8merge_method: slerp
9base_model: Weyaxi/OpenHermes-2.5-neural-chat-v3-3-openchat-3.5-1210-Slerp
10
11parameters:
12 t:
13 - filter: self_attn
14 value: [0, 0.5, 0.3, 0.7, 1]
15 - filter: mlp
16 value: [1, 0.5, 0.7, 0.3, 0]
17 - value: 0.5 # fallback for rest of tensors
18tokenizer_source: base
19embed_slerp: true
20
21dtype: bfloat16
