I'm back and doing well! I've got a job in the field now, so we'll see in the long run how that effects my open source output.
Here we have a 11b Llama 3 instruct model for future work.
EDIT: Made a yaml mistake with part funnel, but it still works well.
This is a merge stock of 3 models:
-
Part Wave
-
Part Block
-
Part Funnel
With Part Funnel as the base.
Part Wave:
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [0, 12]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [8, 18]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [13, 23]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [18, 32]
Part Block:
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [0, 15]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [8, 23]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [16, 32]
Part Funnel:
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [0, 15]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [14, 14]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [13, 13]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [12, 12]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [11, 11]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [10, 10]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [9, 9]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [8, 23]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [22, 22]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [21, 21]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [20, 20]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [19, 19]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [18, 18]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [17, 17]
-
sources:
-
model: NousResearch/Meta-Llama-3-8B-Instruct
layer_range: [16, 32]