base.yml

defs:
    model: &base_config
        n_positions: 141
        n_layer: 12
        n_embd: 256
        n_head: 8
    
    model: &small_config
        <<: *base_config
        n_layer: 6
        n_head: 4
        n_embd: 64

    model: &retrieval_config
        <<: *base_config
        n_layer: 2
        n_embd: 128

model: &std_gpt2
    <<: *base_config
    type: gpt2

model: &std_llama
    <<: *base_config
    type: llama

model: &std_mamba
    <<: *base_config
    type: mamba

model: &small_gpt2
    <<: *small_config
    type: gpt2

model: &small_llama
    <<: *small_config
    type: llama

model: &small_mamba
    <<: *small_config
    type: mamba

model: &retrieval_gpt2
    <<: *retrieval_config
    type: gpt2

model: &retrieval_llama
    <<: *retrieval_config
    type: llama

model: &retrieval_mamba
    <<: *retrieval_config
    type: mamba