34 yesterday

c488475e0986 ยท 40GB
    Metadata
  • split.count
    3
  • split.no
    2
  • split.tensors.count
    1203
  • Tensor
    blk.6
  • blk.6.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.6.attn_k.weight
    BF16
    [5120, 1024]
  • blk.6.attn_k_norm.weight
    F32
    [128]
  • blk.6.attn_norm.weight
    F32
    [5120]
  • blk.6.attn_output.weight
    BF16
    [10240, 5120]
  • blk.6.attn_q.weight
    BF16
    [5120, 10240]
  • blk.6.attn_q_norm.weight
    F32
    [128]
  • blk.6.attn_v.weight
    BF16
    [5120, 1024]
  • blk.6.attn_vres_lambda.weight
    F32
    [1]
  • blk.6.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.6.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.6.ffn_norm.weight
    F32
    [5120]
  • blk.6.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.6.post_attention_norm.weight
    F32
    [5120]
  • blk.6.post_ffw_norm.weight
    F32
    [5120]
  • blk.7
  • blk.7.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.7.attn_k.weight
    BF16
    [5120, 1024]
  • blk.7.attn_k_norm.weight
    F32
    [128]
  • blk.7.attn_norm.weight
    F32
    [5120]
  • blk.7.attn_output.weight
    BF16
    [10240, 5120]
  • blk.7.attn_q.weight
    BF16
    [5120, 10240]
  • blk.7.attn_q_norm.weight
    F32
    [128]
  • blk.7.attn_v.weight
    BF16
    [5120, 1024]
  • blk.7.attn_vres_lambda.weight
    F32
    [1]
  • blk.7.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.7.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.7.ffn_norm.weight
    F32
    [5120]
  • blk.7.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.7.post_attention_norm.weight
    F32
    [5120]
  • blk.7.post_ffw_norm.weight
    F32
    [5120]
  • blk.8
  • blk.8.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.8.attn_k.weight
    BF16
    [5120, 1024]
  • blk.8.attn_k_norm.weight
    F32
    [128]
  • blk.8.attn_norm.weight
    F32
    [5120]
  • blk.8.attn_output.weight
    BF16
    [10240, 5120]
  • blk.8.attn_q.weight
    BF16
    [5120, 10240]
  • blk.8.attn_q_norm.weight
    F32
    [128]
  • blk.8.attn_v.weight
    BF16
    [5120, 1024]
  • blk.8.attn_vres_lambda.weight
    F32
    [1]
  • blk.8.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.8.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.8.ffn_norm.weight
    F32
    [5120]
  • blk.8.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.8.post_attention_norm.weight
    F32
    [5120]
  • blk.8.post_ffw_norm.weight
    F32
    [5120]
  • blk.9
  • blk.9.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.9.attn_k.weight
    BF16
    [5120, 1024]
  • blk.9.attn_k_norm.weight
    F32
    [128]
  • blk.9.attn_norm.weight
    F32
    [5120]
  • blk.9.attn_output.weight
    BF16
    [10240, 5120]
  • blk.9.attn_q.weight
    BF16
    [5120, 10240]
  • blk.9.attn_q_norm.weight
    F32
    [128]
  • blk.9.attn_v.weight
    BF16
    [5120, 1024]
  • blk.9.attn_vres_lambda.weight
    F32
    [1]
  • blk.9.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.9.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.9.ffn_norm.weight
    F32
    [5120]
  • blk.9.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.9.post_attention_norm.weight
    F32
    [5120]
  • blk.9.post_ffw_norm.weight
    F32
    [5120]
  • blk.59
  • blk.59.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.59.attn_k.weight
    BF16
    [5120, 1024]
  • blk.59.attn_k_norm.weight
    F32
    [128]
  • blk.59.attn_output.weight
    BF16
    [10240, 5120]
  • blk.59.attn_q.weight
    BF16
    [5120, 10240]
  • blk.59.attn_q_norm.weight
    F32
    [128]
  • blk.59.attn_v.weight
    BF16
    [5120, 1024]
  • blk.59.attn_vres_lambda.weight
    F32
    [1]
  • blk.59.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.59.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.59.ffn_norm.weight
    F32
    [5120]
  • blk.59.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.59.post_attention_norm.weight
    F32
    [5120]
  • blk.59.post_ffw_norm.weight
    F32
    [5120]
  • blk.60
  • blk.60.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.60.attn_k.weight
    BF16
    [5120, 1024]
  • blk.60.attn_k_norm.weight
    F32
    [128]
  • blk.60.attn_norm.weight
    F32
    [5120]
  • blk.60.attn_output.weight
    BF16
    [10240, 5120]
  • blk.60.attn_q.weight
    BF16
    [5120, 10240]
  • blk.60.attn_q_norm.weight
    F32
    [128]
  • blk.60.attn_v.weight
    BF16
    [5120, 1024]
  • blk.60.attn_vres_lambda.weight
    F32
    [1]
  • blk.60.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.60.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.60.ffn_norm.weight
    F32
    [5120]
  • blk.60.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.60.post_attention_norm.weight
    F32
    [5120]
  • blk.60.post_ffw_norm.weight
    F32
    [5120]
  • blk.61
  • blk.61.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.61.attn_k.weight
    BF16
    [5120, 1024]
  • blk.61.attn_k_norm.weight
    F32
    [128]
  • blk.61.attn_norm.weight
    F32
    [5120]
  • blk.61.attn_output.weight
    BF16
    [10240, 5120]
  • blk.61.attn_q.weight
    BF16
    [5120, 10240]
  • blk.61.attn_q_norm.weight
    F32
    [128]
  • blk.61.attn_v.weight
    BF16
    [5120, 1024]
  • blk.61.attn_vres_lambda.weight
    F32
    [1]
  • blk.61.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.61.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.61.ffn_norm.weight
    F32
    [5120]
  • blk.61.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.61.post_attention_norm.weight
    F32
    [5120]
  • blk.61.post_ffw_norm.weight
    F32
    [5120]
  • blk.62
  • blk.62.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.62.attn_k.weight
    BF16
    [5120, 1024]
  • blk.62.attn_k_norm.weight
    F32
    [128]
  • blk.62.attn_norm.weight
    F32
    [5120]
  • blk.62.attn_output.weight
    BF16
    [10240, 5120]
  • blk.62.attn_q.weight
    BF16
    [5120, 10240]
  • blk.62.attn_q_norm.weight
    F32
    [128]
  • blk.62.attn_v.weight
    BF16
    [5120, 1024]
  • blk.62.attn_vres_lambda.weight
    F32
    [1]
  • blk.62.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.62.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.62.ffn_norm.weight
    F32
    [5120]
  • blk.62.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.62.post_attention_norm.weight
    F32
    [5120]
  • blk.62.post_ffw_norm.weight
    F32
    [5120]
  • blk.63
  • blk.63.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.63.attn_k.weight
    BF16
    [5120, 1024]
  • blk.63.attn_k_norm.weight
    F32
    [128]
  • blk.63.attn_norm.weight
    F32
    [5120]
  • blk.63.attn_output.weight
    BF16
    [10240, 5120]
  • blk.63.attn_q.weight
    BF16
    [5120, 10240]
  • blk.63.attn_q_norm.weight
    F32
    [128]
  • blk.63.attn_v.weight
    BF16
    [5120, 1024]
  • blk.63.attn_vres_lambda.weight
    F32
    [1]
  • blk.63.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.63.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.63.ffn_norm.weight
    F32
    [5120]
  • blk.63.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.63.post_attention_norm.weight
    F32
    [5120]
  • blk.63.post_ffw_norm.weight
    F32
    [5120]
  • blk.64
  • blk.64.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.64.attn_k.weight
    BF16
    [5120, 1024]
  • blk.64.attn_k_norm.weight
    F32
    [128]
  • blk.64.attn_norm.weight
    F32
    [5120]
  • blk.64.attn_output.weight
    BF16
    [10240, 5120]
  • blk.64.attn_q.weight
    BF16
    [5120, 10240]
  • blk.64.attn_q_norm.weight
    F32
    [128]
  • blk.64.attn_v.weight
    BF16
    [5120, 1024]
  • blk.64.attn_vres_lambda.weight
    F32
    [1]
  • blk.64.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.64.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.64.ffn_norm.weight
    F32
    [5120]
  • blk.64.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.64.post_attention_norm.weight
    F32
    [5120]
  • blk.64.post_ffw_norm.weight
    F32
    [5120]
  • blk.65
  • blk.65.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.65.attn_k.weight
    BF16
    [5120, 1024]
  • blk.65.attn_k_norm.weight
    F32
    [128]
  • blk.65.attn_norm.weight
    F32
    [5120]
  • blk.65.attn_output.weight
    BF16
    [10240, 5120]
  • blk.65.attn_q.weight
    BF16
    [5120, 10240]
  • blk.65.attn_q_norm.weight
    F32
    [128]
  • blk.65.attn_v.weight
    BF16
    [5120, 1024]
  • blk.65.attn_vres_lambda.weight
    F32
    [1]
  • blk.65.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.65.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.65.ffn_norm.weight
    F32
    [5120]
  • blk.65.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.65.post_attention_norm.weight
    F32
    [5120]
  • blk.65.post_ffw_norm.weight
    F32
    [5120]
  • blk.66
  • blk.66.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.66.attn_k.weight
    BF16
    [5120, 1024]
  • blk.66.attn_k_norm.weight
    F32
    [128]
  • blk.66.attn_norm.weight
    F32
    [5120]
  • blk.66.attn_output.weight
    BF16
    [10240, 5120]
  • blk.66.attn_q.weight
    BF16
    [5120, 10240]
  • blk.66.attn_q_norm.weight
    F32
    [128]
  • blk.66.attn_v.weight
    BF16
    [5120, 1024]
  • blk.66.attn_vres_lambda.weight
    F32
    [1]
  • blk.66.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.66.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.66.ffn_norm.weight
    F32
    [5120]
  • blk.66.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.66.post_attention_norm.weight
    F32
    [5120]
  • blk.66.post_ffw_norm.weight
    F32
    [5120]
  • blk.67
  • blk.67.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.67.attn_k.weight
    BF16
    [5120, 1024]
  • blk.67.attn_k_norm.weight
    F32
    [128]
  • blk.67.attn_norm.weight
    F32
    [5120]
  • blk.67.attn_output.weight
    BF16
    [10240, 5120]
  • blk.67.attn_q.weight
    BF16
    [5120, 10240]
  • blk.67.attn_q_norm.weight
    F32
    [128]
  • blk.67.attn_v.weight
    BF16
    [5120, 1024]
  • blk.67.attn_vres_lambda.weight
    F32
    [1]
  • blk.67.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.67.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.67.ffn_norm.weight
    F32
    [5120]
  • blk.67.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.67.post_attention_norm.weight
    F32
    [5120]
  • blk.67.post_ffw_norm.weight
    F32
    [5120]
  • blk.68
  • blk.68.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.68.attn_k.weight
    BF16
    [5120, 1024]
  • blk.68.attn_k_norm.weight
    F32
    [128]
  • blk.68.attn_norm.weight
    F32
    [5120]
  • blk.68.attn_output.weight
    BF16
    [10240, 5120]
  • blk.68.attn_q.weight
    BF16
    [5120, 10240]
  • blk.68.attn_q_norm.weight
    F32
    [128]
  • blk.68.attn_v.weight
    BF16
    [5120, 1024]
  • blk.68.attn_vres_lambda.weight
    F32
    [1]
  • blk.68.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.68.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.68.ffn_norm.weight
    F32
    [5120]
  • blk.68.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.68.post_attention_norm.weight
    F32
    [5120]
  • blk.68.post_ffw_norm.weight
    F32
    [5120]
  • blk.69
  • blk.69.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.69.attn_k.weight
    BF16
    [5120, 1024]
  • blk.69.attn_k_norm.weight
    F32
    [128]
  • blk.69.attn_norm.weight
    F32
    [5120]
  • blk.69.attn_output.weight
    BF16
    [10240, 5120]
  • blk.69.attn_q.weight
    BF16
    [5120, 10240]
  • blk.69.attn_q_norm.weight
    F32
    [128]
  • blk.69.attn_v.weight
    BF16
    [5120, 1024]
  • blk.69.attn_vres_lambda.weight
    F32
    [1]
  • blk.69.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.69.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.69.ffn_norm.weight
    F32
    [5120]
  • blk.69.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.69.post_attention_norm.weight
    F32
    [5120]
  • blk.69.post_ffw_norm.weight
    F32
    [5120]
  • blk.70
  • blk.70.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.70.attn_k.weight
    BF16
    [5120, 1024]
  • blk.70.attn_k_norm.weight
    F32
    [128]
  • blk.70.attn_norm.weight
    F32
    [5120]
  • blk.70.attn_output.weight
    BF16
    [10240, 5120]
  • blk.70.attn_q.weight
    BF16
    [5120, 10240]
  • blk.70.attn_q_norm.weight
    F32
    [128]
  • blk.70.attn_v.weight
    BF16
    [5120, 1024]
  • blk.70.attn_vres_lambda.weight
    F32
    [1]
  • blk.70.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.70.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.70.ffn_norm.weight
    F32
    [5120]
  • blk.70.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.70.post_attention_norm.weight
    F32
    [5120]
  • blk.70.post_ffw_norm.weight
    F32
    [5120]
  • blk.71
  • blk.71.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.71.attn_k.weight
    BF16
    [5120, 1024]
  • blk.71.attn_k_norm.weight
    F32
    [128]
  • blk.71.attn_norm.weight
    F32
    [5120]
  • blk.71.attn_output.weight
    BF16
    [10240, 5120]
  • blk.71.attn_q.weight
    BF16
    [5120, 10240]
  • blk.71.attn_q_norm.weight
    F32
    [128]
  • blk.71.attn_v.weight
    BF16
    [5120, 1024]
  • blk.71.attn_vres_lambda.weight
    F32
    [1]
  • blk.71.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.71.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.71.ffn_norm.weight
    F32
    [5120]
  • blk.71.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.71.post_attention_norm.weight
    F32
    [5120]
  • blk.71.post_ffw_norm.weight
    F32
    [5120]
  • blk.72
  • blk.72.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.72.attn_k.weight
    BF16
    [5120, 1024]
  • blk.72.attn_k_norm.weight
    F32
    [128]
  • blk.72.attn_norm.weight
    F32
    [5120]
  • blk.72.attn_output.weight
    BF16
    [10240, 5120]
  • blk.72.attn_q.weight
    BF16
    [5120, 10240]
  • blk.72.attn_q_norm.weight
    F32
    [128]
  • blk.72.attn_v.weight
    BF16
    [5120, 1024]
  • blk.72.attn_vres_lambda.weight
    F32
    [1]
  • blk.72.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.72.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.72.ffn_norm.weight
    F32
    [5120]
  • blk.72.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.72.post_attention_norm.weight
    F32
    [5120]
  • blk.72.post_ffw_norm.weight
    F32
    [5120]
  • blk.73
  • blk.73.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.73.attn_k.weight
    BF16
    [5120, 1024]
  • blk.73.attn_k_norm.weight
    F32
    [128]
  • blk.73.attn_norm.weight
    F32
    [5120]
  • blk.73.attn_output.weight
    BF16
    [10240, 5120]
  • blk.73.attn_q.weight
    BF16
    [5120, 10240]
  • blk.73.attn_q_norm.weight
    F32
    [128]
  • blk.73.attn_v.weight
    BF16
    [5120, 1024]
  • blk.73.attn_vres_lambda.weight
    F32
    [1]
  • blk.73.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.73.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.73.ffn_norm.weight
    F32
    [5120]
  • blk.73.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.73.post_attention_norm.weight
    F32
    [5120]
  • blk.73.post_ffw_norm.weight
    F32
    [5120]
  • blk.74
  • blk.74.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.74.attn_k.weight
    BF16
    [5120, 1024]
  • blk.74.attn_k_norm.weight
    F32
    [128]
  • blk.74.attn_norm.weight
    F32
    [5120]
  • blk.74.attn_output.weight
    BF16
    [10240, 5120]
  • blk.74.attn_q.weight
    BF16
    [5120, 10240]
  • blk.74.attn_q_norm.weight
    F32
    [128]
  • blk.74.attn_v.weight
    BF16
    [5120, 1024]
  • blk.74.attn_vres_lambda.weight
    F32
    [1]
  • blk.74.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.74.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.74.ffn_norm.weight
    F32
    [5120]
  • blk.74.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.74.post_attention_norm.weight
    F32
    [5120]
  • blk.74.post_ffw_norm.weight
    F32
    [5120]
  • blk.75
  • blk.75.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.75.attn_k.weight
    BF16
    [5120, 1024]
  • blk.75.attn_k_norm.weight
    F32
    [128]
  • blk.75.attn_norm.weight
    F32
    [5120]
  • blk.75.attn_output.weight
    BF16
    [10240, 5120]
  • blk.75.attn_q.weight
    BF16
    [5120, 10240]
  • blk.75.attn_q_norm.weight
    F32
    [128]
  • blk.75.attn_v.weight
    BF16
    [5120, 1024]
  • blk.75.attn_vres_lambda.weight
    F32
    [1]
  • blk.75.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.75.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.75.ffn_norm.weight
    F32
    [5120]
  • blk.75.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.75.post_attention_norm.weight
    F32
    [5120]
  • blk.75.post_ffw_norm.weight
    F32
    [5120]
  • blk.76
  • blk.76.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.76.attn_k.weight
    BF16
    [5120, 1024]
  • blk.76.attn_k_norm.weight
    F32
    [128]
  • blk.76.attn_norm.weight
    F32
    [5120]
  • blk.76.attn_output.weight
    BF16
    [10240, 5120]
  • blk.76.attn_q.weight
    BF16
    [5120, 10240]
  • blk.76.attn_q_norm.weight
    F32
    [128]
  • blk.76.attn_v.weight
    BF16
    [5120, 1024]
  • blk.76.attn_vres_lambda.weight
    F32
    [1]
  • blk.76.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.76.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.76.ffn_norm.weight
    F32
    [5120]
  • blk.76.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.76.post_attention_norm.weight
    F32
    [5120]
  • blk.76.post_ffw_norm.weight
    F32
    [5120]
  • blk.77
  • blk.77.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.77.attn_k.weight
    BF16
    [5120, 1024]
  • blk.77.attn_k_norm.weight
    F32
    [128]
  • blk.77.attn_norm.weight
    F32
    [5120]
  • blk.77.attn_output.weight
    BF16
    [10240, 5120]
  • blk.77.attn_q.weight
    BF16
    [5120, 10240]
  • blk.77.attn_q_norm.weight
    F32
    [128]
  • blk.77.attn_v.weight
    BF16
    [5120, 1024]
  • blk.77.attn_vres_lambda.weight
    F32
    [1]
  • blk.77.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.77.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.77.ffn_norm.weight
    F32
    [5120]
  • blk.77.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.77.post_attention_norm.weight
    F32
    [5120]
  • blk.77.post_ffw_norm.weight
    F32
    [5120]
  • blk.78
  • blk.78.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.78.attn_k.weight
    BF16
    [5120, 1024]
  • blk.78.attn_k_norm.weight
    F32
    [128]
  • blk.78.attn_norm.weight
    F32
    [5120]
  • blk.78.attn_output.weight
    BF16
    [10240, 5120]
  • blk.78.attn_q.weight
    BF16
    [5120, 10240]
  • blk.78.attn_q_norm.weight
    F32
    [128]
  • blk.78.attn_v.weight
    BF16
    [5120, 1024]
  • blk.78.attn_vres_lambda.weight
    F32
    [1]
  • blk.78.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.78.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.78.ffn_norm.weight
    F32
    [5120]
  • blk.78.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.78.post_attention_norm.weight
    F32
    [5120]
  • blk.78.post_ffw_norm.weight
    F32
    [5120]
  • blk.79
  • blk.79.attn_gate.weight
    BF16
    [5120, 10240]
  • blk.79.attn_k.weight
    BF16
    [5120, 1024]
  • blk.79.attn_k_norm.weight
    F32
    [128]
  • blk.79.attn_norm.weight
    F32
    [5120]
  • blk.79.attn_output.weight
    BF16
    [10240, 5120]
  • blk.79.attn_q.weight
    BF16
    [5120, 10240]
  • blk.79.attn_q_norm.weight
    F32
    [128]
  • blk.79.attn_v.weight
    BF16
    [5120, 1024]
  • blk.79.attn_vres_lambda.weight
    F32
    [1]
  • blk.79.ffn_down.weight
    BF16
    [40960, 5120]
  • blk.79.ffn_gate.weight
    BF16
    [5120, 40960]
  • blk.79.ffn_norm.weight
    F32
    [5120]
  • blk.79.ffn_up.weight
    BF16
    [5120, 40960]
  • blk.79.post_attention_norm.weight
    F32
    [5120]
  • blk.79.post_ffw_norm.weight
    F32
    [5120]
  • output_norm.weight
    F32
    [5120]