| name: sage-tiny | |
| num_layers: 2 | |
| d_model: 64 | |
| num_attn_heads: 4 | |
| num_kv_heads: 2 | |
| head_dim: 16 | |
| ffn_hidden_dim: 256 | |
| vocab_size: 32000 | |
| context_length: 64 | |
| rope_base_frequency: 500000 | |
| rope_scaling_factor: 1.0 | |
| dropout: 0.0 | |
| tie_word_embeddings: true | |
| name: sage-tiny | |
| num_layers: 2 | |
| d_model: 64 | |
| num_attn_heads: 4 | |
| num_kv_heads: 2 | |
| head_dim: 16 | |
| ffn_hidden_dim: 256 | |
| vocab_size: 32000 | |
| context_length: 64 | |
| rope_base_frequency: 500000 | |
| rope_scaling_factor: 1.0 | |
| dropout: 0.0 | |
| tie_word_embeddings: true | |