LiquidAI/LFM2-1.2B config.json · inference-research

Fields

id

sources/mbl-cfg-lfm2-1-2b

URL

https://huggingface.co/LiquidAI/LFM2-1.2B/raw/main/config.json

Publisher

Hugging Face

Kind

repo

Accessed

2026-10-04

notes

LFM2 hybrid. num_hidden_layers 16, hidden_size 2048, num_attention_heads 8, num_key_value_heads 2, head_dim 64, rope_theta 1000000, rope_scaling null, max_position_embeddings 128000, full_attn_idxs [2,5,8,10,12,14] = 6 of 16 layers are full attention, the rest short-conv. Read raw 2026-10-04.

Notes

LFM2 hybrid. num_hidden_layers 16, hidden_size 2048, num_attention_heads 8, num_key_value_heads 2, head_dim 64, rope_theta 1000000, rope_scaling null, max_position_embeddings 128000, full_attn_idxs [2,5,8,10,12,14] = 6 of 16 layers are full attention, the rest short-conv. Read raw 2026-10-04.

Cross-references

This record references no other records.