Lfm3ForCausalLM(
  (model): Lfm3Model(
    (embed_tokens): Embedding(128000, 64, padding_idx=124893)
    (layers): ModuleList(
      (0-1): 2 x Lfm3DecoderLayer(
        (conv): Lfm3ShortConv(
          (conv): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(6,), groups=64, bias=False)
          (in_proj): Linear(in_features=64, out_features=192, bias=False)
          (out_proj): Linear(in_features=64, out_features=64, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (2): Lfm3DecoderLayer(
        (self_attn): Lfm3Attention(
          (q_proj): Linear(in_features=64, out_features=128, bias=False)
          (k_proj): Linear(in_features=64, out_features=32, bias=False)
          (v_proj): Linear(in_features=64, out_features=32, bias=False)
          (o_proj): Linear(in_features=128, out_features=64, bias=False)
          (q_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (k_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (gate_proj): Linear(in_features=64, out_features=4, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (3): Lfm3DecoderLayer(
        (conv): Lfm3ShortConv(
          (conv): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(6,), groups=64, bias=False)
          (in_proj): Linear(in_features=64, out_features=192, bias=False)
          (out_proj): Linear(in_features=64, out_features=64, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (4): Lfm3DecoderLayer(
        (self_attn): Lfm3Attention(
          (q_proj): Linear(in_features=64, out_features=128, bias=False)
          (k_proj): Linear(in_features=64, out_features=32, bias=False)
          (v_proj): Linear(in_features=64, out_features=32, bias=False)
          (o_proj): Linear(in_features=128, out_features=64, bias=False)
          (q_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (k_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (gate_proj): Linear(in_features=64, out_features=4, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (5): Lfm3DecoderLayer(
        (conv): Lfm3ShortConv(
          (conv): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(6,), groups=64, bias=False)
          (in_proj): Linear(in_features=64, out_features=192, bias=False)
          (out_proj): Linear(in_features=64, out_features=64, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (6): Lfm3DecoderLayer(
        (self_attn): Lfm3Attention(
          (q_proj): Linear(in_features=64, out_features=128, bias=False)
          (k_proj): Linear(in_features=64, out_features=32, bias=False)
          (v_proj): Linear(in_features=64, out_features=32, bias=False)
          (o_proj): Linear(in_features=128, out_features=64, bias=False)
          (q_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (k_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (gate_proj): Linear(in_features=64, out_features=4, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (7): Lfm3DecoderLayer(
        (conv): Lfm3ShortConv(
          (conv): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(6,), groups=64, bias=False)
          (in_proj): Linear(in_features=64, out_features=192, bias=False)
          (out_proj): Linear(in_features=64, out_features=64, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (8): Lfm3DecoderLayer(
        (self_attn): Lfm3Attention(
          (q_proj): Linear(in_features=64, out_features=128, bias=False)
          (k_proj): Linear(in_features=64, out_features=32, bias=False)
          (v_proj): Linear(in_features=64, out_features=32, bias=False)
          (o_proj): Linear(in_features=128, out_features=64, bias=False)
          (q_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (k_norm): Lfm3RMSNorm((32,), eps=1e-05)
          (gate_proj): Linear(in_features=64, out_features=4, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
      (9): Lfm3DecoderLayer(
        (conv): Lfm3ShortConv(
          (conv): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(6,), groups=64, bias=False)
          (in_proj): Linear(in_features=64, out_features=192, bias=False)
          (out_proj): Linear(in_features=64, out_features=64, bias=False)
        )
        (feed_forward): Lfm3MLP(
          (w1): Linear(in_features=64, out_features=128, bias=False)
          (w3): Linear(in_features=64, out_features=128, bias=False)
          (w2): Linear(in_features=128, out_features=64, bias=False)
        )
        (operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_operator_norm): Lfm3RMSNorm((64,), eps=1e-05)
        (post_ffn_norm): Lfm3RMSNorm((64,), eps=1e-05)
      )
    )
    (rotary_emb): Lfm3RotaryEmbedding()
    (embedding_norm): Lfm3RMSNorm((64,), eps=1e-05)
  )
  (lm_head): Linear(in_features=64, out_features=128000, bias=False)
)