34 yesterday

eefc34038489 ยท 45GB
    Metadata
  • split.count
    3
  • split.no
    1
  • split.tensors.count
    1203
  • Tensor
    blk.4
  • blk.4.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.4.attn_k.weight
    BF16
    [5120, 1024]
  • blk.4.attn_k_norm.weight
    F32
    [128]
  • blk.4.attn_norm.weight
    F32
    [5120]
  • blk.4.attn_output.weight
    BF16
    [10240, 5120]
  • blk.4.attn_q.weight
    BF16
    [5120, 10240]
  • blk.4.attn_q_norm.weight
    F32
    [128]
  • blk.4.attn_v.weight
    BF16
    [5120, 1024]
  • blk.4.attn_vres_lambda.weight
    F32
    [1]
  • blk.4.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.4.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.4.ffn_norm.weight
    F32
    [5120]
  • blk.4.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.4.post_attention_norm.weight
    F32
    [5120]
  • blk.4.post_ffw_norm.weight
    F32
    [5120]
  • blk.5
  • blk.5.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.5.attn_k.weight
    BF16
    [5120, 1024]
  • blk.5.attn_k_norm.weight
    F32
    [128]
  • blk.5.attn_norm.weight
    F32
    [5120]
  • blk.5.attn_output.weight
    BF16
    [10240, 5120]
  • blk.5.attn_q.weight
    BF16
    [5120, 10240]
  • blk.5.attn_q_norm.weight
    F32
    [128]
  • blk.5.attn_v.weight
    BF16
    [5120, 1024]
  • blk.5.attn_vres_lambda.weight
    F32
    [1]
  • blk.5.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.5.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.5.ffn_norm.weight
    F32
    [5120]
  • blk.5.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.5.post_attention_norm.weight
    F32
    [5120]
  • blk.5.post_ffw_norm.weight
    F32
    [5120]
  • blk.32
  • blk.32.attn_q.weight
    BF16
    [5120, 10240]
  • blk.32.attn_v.weight
    BF16
    [5120, 1024]
  • blk.32.attn_vres_lambda.weight
    F32
    [1]
  • blk.33
  • blk.33.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.33.attn_k.weight
    BF16
    [5120, 1024]
  • blk.33.attn_k_norm.weight
    F32
    [128]
  • blk.33.attn_norm.weight
    F32
    [5120]
  • blk.33.attn_output.weight
    BF16
    [10240, 5120]
  • blk.33.attn_q.weight
    BF16
    [5120, 10240]
  • blk.33.attn_q_norm.weight
    F32
    [128]
  • blk.33.attn_v.weight
    BF16
    [5120, 1024]
  • blk.33.attn_vres_lambda.weight
    F32
    [1]
  • blk.33.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.33.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.33.ffn_norm.weight
    F32
    [5120]
  • blk.33.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.33.post_attention_norm.weight
    F32
    [5120]
  • blk.33.post_ffw_norm.weight
    F32
    [5120]
  • blk.34
  • blk.34.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.34.attn_k.weight
    BF16
    [5120, 1024]
  • blk.34.attn_k_norm.weight
    F32
    [128]
  • blk.34.attn_norm.weight
    F32
    [5120]
  • blk.34.attn_output.weight
    BF16
    [10240, 5120]
  • blk.34.attn_q.weight
    BF16
    [5120, 10240]
  • blk.34.attn_q_norm.weight
    F32
    [128]
  • blk.34.attn_v.weight
    BF16
    [5120, 1024]
  • blk.34.attn_vres_lambda.weight
    F32
    [1]
  • blk.34.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.34.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.34.ffn_norm.weight
    F32
    [5120]
  • blk.34.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.34.post_attention_norm.weight
    F32
    [5120]
  • blk.34.post_ffw_norm.weight
    F32
    [5120]
  • blk.35
  • blk.35.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.35.attn_k.weight
    BF16
    [5120, 1024]
  • blk.35.attn_k_norm.weight
    F32
    [128]
  • blk.35.attn_norm.weight
    F32
    [5120]
  • blk.35.attn_output.weight
    BF16
    [10240, 5120]
  • blk.35.attn_q.weight
    BF16
    [5120, 10240]
  • blk.35.attn_q_norm.weight
    F32
    [128]
  • blk.35.attn_v.weight
    BF16
    [5120, 1024]
  • blk.35.attn_vres_lambda.weight
    F32
    [1]
  • blk.35.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.35.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.35.ffn_norm.weight
    F32
    [5120]
  • blk.35.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.35.post_attention_norm.weight
    F32
    [5120]
  • blk.35.post_ffw_norm.weight
    F32
    [5120]
  • blk.36
  • blk.36.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.36.attn_k.weight
    BF16
    [5120, 1024]
  • blk.36.attn_k_norm.weight
    F32
    [128]
  • blk.36.attn_norm.weight
    F32
    [5120]
  • blk.36.attn_output.weight
    BF16
    [10240, 5120]
  • blk.36.attn_q.weight
    BF16
    [5120, 10240]
  • blk.36.attn_q_norm.weight
    F32
    [128]
  • blk.36.attn_v.weight
    BF16
    [5120, 1024]
  • blk.36.attn_vres_lambda.weight
    F32
    [1]
  • blk.36.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.36.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.36.ffn_norm.weight
    F32
    [5120]
  • blk.36.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.36.post_attention_norm.weight
    F32
    [5120]
  • blk.36.post_ffw_norm.weight
    F32
    [5120]
  • blk.37
  • blk.37.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.37.attn_k.weight
    BF16
    [5120, 1024]
  • blk.37.attn_k_norm.weight
    F32
    [128]
  • blk.37.attn_norm.weight
    F32
    [5120]
  • blk.37.attn_output.weight
    BF16
    [10240, 5120]
  • blk.37.attn_q.weight
    BF16
    [5120, 10240]
  • blk.37.attn_q_norm.weight
    F32
    [128]
  • blk.37.attn_v.weight
    BF16
    [5120, 1024]
  • blk.37.attn_vres_lambda.weight
    F32
    [1]
  • blk.37.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.37.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.37.ffn_norm.weight
    F32
    [5120]
  • blk.37.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.37.post_attention_norm.weight
    F32
    [5120]
  • blk.37.post_ffw_norm.weight
    F32
    [5120]
  • blk.38
  • blk.38.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.38.attn_k.weight
    BF16
    [5120, 1024]
  • blk.38.attn_k_norm.weight
    F32
    [128]
  • blk.38.attn_norm.weight
    F32
    [5120]
  • blk.38.attn_output.weight
    BF16
    [10240, 5120]
  • blk.38.attn_q.weight
    BF16
    [5120, 10240]
  • blk.38.attn_q_norm.weight
    F32
    [128]
  • blk.38.attn_v.weight
    BF16
    [5120, 1024]
  • blk.38.attn_vres_lambda.weight
    F32
    [1]
  • blk.38.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.38.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.38.ffn_norm.weight
    F32
    [5120]
  • blk.38.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.38.post_attention_norm.weight
    F32
    [5120]
  • blk.38.post_ffw_norm.weight
    F32
    [5120]
  • blk.39
  • blk.39.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.39.attn_k.weight
    BF16
    [5120, 1024]
  • blk.39.attn_k_norm.weight
    F32
    [128]
  • blk.39.attn_norm.weight
    F32
    [5120]
  • blk.39.attn_output.weight
    BF16
    [10240, 5120]
  • blk.39.attn_q.weight
    BF16
    [5120, 10240]
  • blk.39.attn_q_norm.weight
    F32
    [128]
  • blk.39.attn_v.weight
    BF16
    [5120, 1024]
  • blk.39.attn_vres_lambda.weight
    F32
    [1]
  • blk.39.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.39.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.39.ffn_norm.weight
    F32
    [5120]
  • blk.39.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.39.post_attention_norm.weight
    F32
    [5120]
  • blk.39.post_ffw_norm.weight
    F32
    [5120]
  • blk.40
  • blk.40.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.40.attn_k.weight
    BF16
    [5120, 1024]
  • blk.40.attn_k_norm.weight
    F32
    [128]
  • blk.40.attn_norm.weight
    F32
    [5120]
  • blk.40.attn_output.weight
    BF16
    [10240, 5120]
  • blk.40.attn_q.weight
    BF16
    [5120, 10240]
  • blk.40.attn_q_norm.weight
    F32
    [128]
  • blk.40.attn_v.weight
    BF16
    [5120, 1024]
  • blk.40.attn_vres_lambda.weight
    F32
    [1]
  • blk.40.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.40.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.40.ffn_norm.weight
    F32
    [5120]
  • blk.40.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.40.post_attention_norm.weight
    F32
    [5120]
  • blk.40.post_ffw_norm.weight
    F32
    [5120]
  • blk.41
  • blk.41.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.41.attn_k.weight
    BF16
    [5120, 1024]
  • blk.41.attn_k_norm.weight
    F32
    [128]
  • blk.41.attn_norm.weight
    F32
    [5120]
  • blk.41.attn_output.weight
    BF16
    [10240, 5120]
  • blk.41.attn_q.weight
    BF16
    [5120, 10240]
  • blk.41.attn_q_norm.weight
    F32
    [128]
  • blk.41.attn_v.weight
    BF16
    [5120, 1024]
  • blk.41.attn_vres_lambda.weight
    F32
    [1]
  • blk.41.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.41.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.41.ffn_norm.weight
    F32
    [5120]
  • blk.41.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.41.post_attention_norm.weight
    F32
    [5120]
  • blk.41.post_ffw_norm.weight
    F32
    [5120]
  • blk.42
  • blk.42.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.42.attn_k.weight
    BF16
    [5120, 1024]
  • blk.42.attn_k_norm.weight
    F32
    [128]
  • blk.42.attn_norm.weight
    F32
    [5120]
  • blk.42.attn_output.weight
    BF16
    [10240, 5120]
  • blk.42.attn_q.weight
    BF16
    [5120, 10240]
  • blk.42.attn_q_norm.weight
    F32
    [128]
  • blk.42.attn_v.weight
    BF16
    [5120, 1024]
  • blk.42.attn_vres_lambda.weight
    F32
    [1]
  • blk.42.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.42.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.42.ffn_norm.weight
    F32
    [5120]
  • blk.42.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.42.post_attention_norm.weight
    F32
    [5120]
  • blk.42.post_ffw_norm.weight
    F32
    [5120]
  • blk.43
  • blk.43.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.43.attn_k.weight
    BF16
    [5120, 1024]
  • blk.43.attn_k_norm.weight
    F32
    [128]
  • blk.43.attn_norm.weight
    F32
    [5120]
  • blk.43.attn_output.weight
    BF16
    [10240, 5120]
  • blk.43.attn_q.weight
    BF16
    [5120, 10240]
  • blk.43.attn_q_norm.weight
    F32
    [128]
  • blk.43.attn_v.weight
    BF16
    [5120, 1024]
  • blk.43.attn_vres_lambda.weight
    F32
    [1]
  • blk.43.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.43.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.43.ffn_norm.weight
    F32
    [5120]
  • blk.43.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.43.post_attention_norm.weight
    F32
    [5120]
  • blk.43.post_ffw_norm.weight
    F32
    [5120]
  • blk.44
  • blk.44.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.44.attn_k.weight
    BF16
    [5120, 1024]
  • blk.44.attn_k_norm.weight
    F32
    [128]
  • blk.44.attn_norm.weight
    F32
    [5120]
  • blk.44.attn_output.weight
    BF16
    [10240, 5120]
  • blk.44.attn_q.weight
    BF16
    [5120, 10240]
  • blk.44.attn_q_norm.weight
    F32
    [128]
  • blk.44.attn_v.weight
    BF16
    [5120, 1024]
  • blk.44.attn_vres_lambda.weight
    F32
    [1]
  • blk.44.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.44.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.44.ffn_norm.weight
    F32
    [5120]
  • blk.44.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.44.post_attention_norm.weight
    F32
    [5120]
  • blk.44.post_ffw_norm.weight
    F32
    [5120]
  • blk.45
  • blk.45.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.45.attn_k.weight
    BF16
    [5120, 1024]
  • blk.45.attn_k_norm.weight
    F32
    [128]
  • blk.45.attn_norm.weight
    F32
    [5120]
  • blk.45.attn_output.weight
    BF16
    [10240, 5120]
  • blk.45.attn_q.weight
    BF16
    [5120, 10240]
  • blk.45.attn_q_norm.weight
    F32
    [128]
  • blk.45.attn_v.weight
    BF16
    [5120, 1024]
  • blk.45.attn_vres_lambda.weight
    F32
    [1]
  • blk.45.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.45.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.45.ffn_norm.weight
    F32
    [5120]
  • blk.45.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.45.post_attention_norm.weight
    F32
    [5120]
  • blk.45.post_ffw_norm.weight
    F32
    [5120]
  • blk.46
  • blk.46.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.46.attn_k.weight
    BF16
    [5120, 1024]
  • blk.46.attn_k_norm.weight
    F32
    [128]
  • blk.46.attn_norm.weight
    F32
    [5120]
  • blk.46.attn_output.weight
    BF16
    [10240, 5120]
  • blk.46.attn_q.weight
    BF16
    [5120, 10240]
  • blk.46.attn_q_norm.weight
    F32
    [128]
  • blk.46.attn_v.weight
    BF16
    [5120, 1024]
  • blk.46.attn_vres_lambda.weight
    F32
    [1]
  • blk.46.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.46.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.46.ffn_norm.weight
    F32
    [5120]
  • blk.46.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.46.post_attention_norm.weight
    F32
    [5120]
  • blk.46.post_ffw_norm.weight
    F32
    [5120]
  • blk.47
  • blk.47.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.47.attn_k.weight
    BF16
    [5120, 1024]
  • blk.47.attn_k_norm.weight
    F32
    [128]
  • blk.47.attn_norm.weight
    F32
    [5120]
  • blk.47.attn_output.weight
    BF16
    [10240, 5120]
  • blk.47.attn_q.weight
    BF16
    [5120, 10240]
  • blk.47.attn_q_norm.weight
    F32
    [128]
  • blk.47.attn_v.weight
    BF16
    [5120, 1024]
  • blk.47.attn_vres_lambda.weight
    F32
    [1]
  • blk.47.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.47.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.47.ffn_norm.weight
    F32
    [5120]
  • blk.47.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.47.post_attention_norm.weight
    F32
    [5120]
  • blk.47.post_ffw_norm.weight
    F32
    [5120]
  • blk.48
  • blk.48.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.48.attn_k.weight
    BF16
    [5120, 1024]
  • blk.48.attn_k_norm.weight
    F32
    [128]
  • blk.48.attn_norm.weight
    F32
    [5120]
  • blk.48.attn_output.weight
    BF16
    [10240, 5120]
  • blk.48.attn_q.weight
    BF16
    [5120, 10240]
  • blk.48.attn_q_norm.weight
    F32
    [128]
  • blk.48.attn_v.weight
    BF16
    [5120, 1024]
  • blk.48.attn_vres_lambda.weight
    F32
    [1]
  • blk.48.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.48.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.48.ffn_norm.weight
    F32
    [5120]
  • blk.48.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.48.post_attention_norm.weight
    F32
    [5120]
  • blk.48.post_ffw_norm.weight
    F32
    [5120]
  • blk.49
  • blk.49.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.49.attn_k.weight
    BF16
    [5120, 1024]
  • blk.49.attn_k_norm.weight
    F32
    [128]
  • blk.49.attn_norm.weight
    F32
    [5120]
  • blk.49.attn_output.weight
    BF16
    [10240, 5120]
  • blk.49.attn_q.weight
    BF16
    [5120, 10240]
  • blk.49.attn_q_norm.weight
    F32
    [128]
  • blk.49.attn_v.weight
    BF16
    [5120, 1024]
  • blk.49.attn_vres_lambda.weight
    F32
    [1]
  • blk.49.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.49.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.49.ffn_norm.weight
    F32
    [5120]
  • blk.49.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.49.post_attention_norm.weight
    F32
    [5120]
  • blk.49.post_ffw_norm.weight
    F32
    [5120]
  • blk.50
  • blk.50.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.50.attn_k.weight
    BF16
    [5120, 1024]
  • blk.50.attn_k_norm.weight
    F32
    [128]
  • blk.50.attn_norm.weight
    F32
    [5120]
  • blk.50.attn_output.weight
    BF16
    [10240, 5120]
  • blk.50.attn_q.weight
    BF16
    [5120, 10240]
  • blk.50.attn_q_norm.weight
    F32
    [128]
  • blk.50.attn_v.weight
    BF16
    [5120, 1024]
  • blk.50.attn_vres_lambda.weight
    F32
    [1]
  • blk.50.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.50.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.50.ffn_norm.weight
    F32
    [5120]
  • blk.50.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.50.post_attention_norm.weight
    F32
    [5120]
  • blk.50.post_ffw_norm.weight
    F32
    [5120]
  • blk.51
  • blk.51.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.51.attn_k.weight
    BF16
    [5120, 1024]
  • blk.51.attn_k_norm.weight
    F32
    [128]
  • blk.51.attn_norm.weight
    F32
    [5120]
  • blk.51.attn_output.weight
    BF16
    [10240, 5120]
  • blk.51.attn_q.weight
    BF16
    [5120, 10240]
  • blk.51.attn_q_norm.weight
    F32
    [128]
  • blk.51.attn_v.weight
    BF16
    [5120, 1024]
  • blk.51.attn_vres_lambda.weight
    F32
    [1]
  • blk.51.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.51.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.51.ffn_norm.weight
    F32
    [5120]
  • blk.51.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.51.post_attention_norm.weight
    F32
    [5120]
  • blk.51.post_ffw_norm.weight
    F32
    [5120]
  • blk.52
  • blk.52.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.52.attn_k.weight
    BF16
    [5120, 1024]
  • blk.52.attn_k_norm.weight
    F32
    [128]
  • blk.52.attn_norm.weight
    F32
    [5120]
  • blk.52.attn_output.weight
    BF16
    [10240, 5120]
  • blk.52.attn_q.weight
    BF16
    [5120, 10240]
  • blk.52.attn_q_norm.weight
    F32
    [128]
  • blk.52.attn_v.weight
    BF16
    [5120, 1024]
  • blk.52.attn_vres_lambda.weight
    F32
    [1]
  • blk.52.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.52.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.52.ffn_norm.weight
    F32
    [5120]
  • blk.52.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.52.post_attention_norm.weight
    F32
    [5120]
  • blk.52.post_ffw_norm.weight
    F32
    [5120]
  • blk.53
  • blk.53.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.53.attn_k.weight
    BF16
    [5120, 1024]
  • blk.53.attn_k_norm.weight
    F32
    [128]
  • blk.53.attn_norm.weight
    F32
    [5120]
  • blk.53.attn_output.weight
    BF16
    [10240, 5120]
  • blk.53.attn_q.weight
    BF16
    [5120, 10240]
  • blk.53.attn_q_norm.weight
    F32
    [128]
  • blk.53.attn_v.weight
    BF16
    [5120, 1024]
  • blk.53.attn_vres_lambda.weight
    F32
    [1]
  • blk.53.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.53.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.53.ffn_norm.weight
    F32
    [5120]
  • blk.53.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.53.post_attention_norm.weight
    F32
    [5120]
  • blk.53.post_ffw_norm.weight
    F32
    [5120]
  • blk.54
  • blk.54.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.54.attn_k.weight
    BF16
    [5120, 1024]
  • blk.54.attn_k_norm.weight
    F32
    [128]
  • blk.54.attn_norm.weight
    F32
    [5120]
  • blk.54.attn_output.weight
    BF16
    [10240, 5120]
  • blk.54.attn_q.weight
    BF16
    [5120, 10240]
  • blk.54.attn_q_norm.weight
    F32
    [128]
  • blk.54.attn_v.weight
    BF16
    [5120, 1024]
  • blk.54.attn_vres_lambda.weight
    F32
    [1]
  • blk.54.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.54.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.54.ffn_norm.weight
    F32
    [5120]
  • blk.54.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.54.post_attention_norm.weight
    F32
    [5120]
  • blk.54.post_ffw_norm.weight
    F32
    [5120]
  • blk.55
  • blk.55.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.55.attn_k.weight
    BF16
    [5120, 1024]
  • blk.55.attn_k_norm.weight
    F32
    [128]
  • blk.55.attn_norm.weight
    F32
    [5120]
  • blk.55.attn_output.weight
    BF16
    [10240, 5120]
  • blk.55.attn_q.weight
    BF16
    [5120, 10240]
  • blk.55.attn_q_norm.weight
    F32
    [128]
  • blk.55.attn_v.weight
    BF16
    [5120, 1024]
  • blk.55.attn_vres_lambda.weight
    F32
    [1]
  • blk.55.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.55.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.55.ffn_norm.weight
    F32
    [5120]
  • blk.55.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.55.post_attention_norm.weight
    F32
    [5120]
  • blk.55.post_ffw_norm.weight
    F32
    [5120]
  • blk.56
  • blk.56.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.56.attn_k.weight
    BF16
    [5120, 1024]
  • blk.56.attn_k_norm.weight
    F32
    [128]
  • blk.56.attn_norm.weight
    F32
    [5120]
  • blk.56.attn_output.weight
    BF16
    [10240, 5120]
  • blk.56.attn_q.weight
    BF16
    [5120, 10240]
  • blk.56.attn_q_norm.weight
    F32
    [128]
  • blk.56.attn_v.weight
    BF16
    [5120, 1024]
  • blk.56.attn_vres_lambda.weight
    F32
    [1]
  • blk.56.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.56.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.56.ffn_norm.weight
    F32
    [5120]
  • blk.56.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.56.post_attention_norm.weight
    F32
    [5120]
  • blk.56.post_ffw_norm.weight
    F32
    [5120]
  • blk.57
  • blk.57.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.57.attn_k.weight
    BF16
    [5120, 1024]
  • blk.57.attn_k_norm.weight
    F32
    [128]
  • blk.57.attn_norm.weight
    F32
    [5120]
  • blk.57.attn_output.weight
    BF16
    [10240, 5120]
  • blk.57.attn_q.weight
    BF16
    [5120, 10240]
  • blk.57.attn_q_norm.weight
    F32
    [128]
  • blk.57.attn_v.weight
    BF16
    [5120, 1024]
  • blk.57.attn_vres_lambda.weight
    F32
    [1]
  • blk.57.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.57.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.57.ffn_norm.weight
    F32
    [5120]
  • blk.57.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.57.post_attention_norm.weight
    F32
    [5120]
  • blk.57.post_ffw_norm.weight
    F32
    [5120]
  • blk.58
  • blk.58.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.58.attn_k.weight
    BF16
    [5120, 1024]
  • blk.58.attn_k_norm.weight
    F32
    [128]
  • blk.58.attn_norm.weight
    F32
    [5120]
  • blk.58.attn_output.weight
    BF16
    [10240, 5120]
  • blk.58.attn_q.weight
    BF16
    [5120, 10240]
  • blk.58.attn_q_norm.weight
    F32
    [128]
  • blk.58.attn_v.weight
    BF16
    [5120, 1024]
  • blk.58.attn_vres_lambda.weight
    F32
    [1]
  • blk.58.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.58.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.58.ffn_norm.weight
    F32
    [5120]
  • blk.58.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.58.post_attention_norm.weight
    F32
    [5120]
  • blk.58.post_ffw_norm.weight
    F32
    [5120]
  • blk.59
  • blk.59.attn_norm.weight
    F32
    [5120]