<
   ir_version: 10,
   opset_import: ["" : 23],
   producer_name: "pytorch"
>
main_graph (int32[batch,77] text) => (float[batch,640] text_embedding) 
   <
      float[batch,77,640] add_1071
      float[batch,77,640] add_1092
      float[batch,77,640] add_119
      float[batch,77,640] add_1207
      float[batch,77,640] add_1228
      float[batch,77,640] add_1343
      float[batch,77,640] add_1364
      float[batch,77,640] add_140
      float[batch,77,640] add_1479
      float[batch,77,640] add_1500
      float[batch,1,640] add_1500_pooled
      float[batch,1,640] add_1615
      float[batch,1,640] add_1636
      float[batch,77,640] add_255
      float[batch,77,640] add_276
      float[batch,77,640] add_391
      float[batch,77,640] add_4
      float[batch,77,640] add_412
      float[batch,77,640] add_527
      float[batch,77,640] add_548
      float[batch,77,640] add_663
      float[batch,77,640] add_684
      float[batch,77,640] add_799
      float[batch,77,640] add_820
      float[batch,77,640] add_935
      float[batch,77,640] add_956
      int64[batch] argmax
      float[batch,1] clamp_min
      float[batch,77,640] embedding
      float[batch,77,2560] gelu
      float[batch,77,2560] gelu_1
      float[batch,77,2560] gelu_10
      float[batch,1,2560] gelu_11
      float[batch,77,2560] gelu_2
      float[batch,77,2560] gelu_3
      float[batch,77,2560] gelu_4
      float[batch,77,2560] gelu_5
      float[batch,77,2560] gelu_6
      float[batch,77,2560] gelu_7
      float[batch,77,2560] gelu_8
      float[batch,77,2560] gelu_9
      float[batch,77,640] layer_norm
      float[batch,77,640] layer_norm_1
      float[batch,77,640] layer_norm_10
      float[batch,77,640] layer_norm_11
      float[batch,77,640] layer_norm_12
      float[batch,77,640] layer_norm_13
      float[batch,77,640] layer_norm_14
      float[batch,77,640] layer_norm_15
      float[batch,77,640] layer_norm_16
      float[batch,77,640] layer_norm_17
      float[batch,77,640] layer_norm_18
      float[batch,77,640] layer_norm_19
      float[batch,77,640] layer_norm_2
      float[batch,77,640] layer_norm_20
      float[batch,77,640] layer_norm_21
      float[batch,77,640] layer_norm_22
      float[batch,1,640] layer_norm_23
      float[batch,77,640] layer_norm_3
      float[batch,77,640] layer_norm_4
      float[batch,77,640] layer_norm_5
      float[batch,77,640] layer_norm_6
      float[batch,77,640] layer_norm_7
      float[batch,77,640] layer_norm_8
      float[batch,77,640] layer_norm_9
      float[batch,1] linalg_vector_norm
      float[batch,77,2560] linear_10
      float[batch,77,640] linear_11
      float[batch,77,2560] linear_14
      float[batch,77,640] linear_15
      float[batch,77,2560] linear_18
      float[batch,77,640] linear_19
      float[batch,77,2560] linear_2
      float[batch,77,2560] linear_22
      float[batch,77,640] linear_23
      float[batch,77,2560] linear_26
      float[batch,77,640] linear_27
      float[batch,77,640] linear_3
      float[batch,77,2560] linear_30
      float[batch,77,640] linear_31
      float[batch,77,2560] linear_34
      float[batch,77,640] linear_35
      float[batch,77,2560] linear_38
      float[batch,77,640] linear_39
      float[batch,77,2560] linear_42
      float[batch,77,640] linear_43
      float[batch,1,2560] linear_46
      float[batch,1,640] linear_47
      float[batch,77,2560] linear_6
      float[batch,77,640] linear_7
      float[batch,640] matmul
      float[batch,77,640] node_scaled_dot_product_attention_10_k
      float[batch,77,640] node_scaled_dot_product_attention_10_out
      float[batch,77,640] node_scaled_dot_product_attention_10_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_10_q
      float[batch,77,1920] node_scaled_dot_product_attention_10_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_10_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_10_v
      float[batch,77,640] node_scaled_dot_product_attention_11_k
      float[batch,1,640] node_scaled_dot_product_attention_11_out
      float[batch,1,640] node_scaled_dot_product_attention_11_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_11_q
      float[batch,77,1920] node_scaled_dot_product_attention_11_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_11_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_11_v
      float[batch,77,640] node_scaled_dot_product_attention_1_k
      float[batch,77,640] node_scaled_dot_product_attention_1_out
      float[batch,77,640] node_scaled_dot_product_attention_1_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_1_q
      float[batch,77,1920] node_scaled_dot_product_attention_1_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_1_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_1_v
      float[batch,77,640] node_scaled_dot_product_attention_2_k
      float[batch,77,640] node_scaled_dot_product_attention_2_out
      float[batch,77,640] node_scaled_dot_product_attention_2_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_2_q
      float[batch,77,1920] node_scaled_dot_product_attention_2_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_2_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_2_v
      float[batch,77,640] node_scaled_dot_product_attention_3_k
      float[batch,77,640] node_scaled_dot_product_attention_3_out
      float[batch,77,640] node_scaled_dot_product_attention_3_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_3_q
      float[batch,77,1920] node_scaled_dot_product_attention_3_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_3_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_3_v
      float[batch,77,640] node_scaled_dot_product_attention_4_k
      float[batch,77,640] node_scaled_dot_product_attention_4_out
      float[batch,77,640] node_scaled_dot_product_attention_4_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_4_q
      float[batch,77,1920] node_scaled_dot_product_attention_4_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_4_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_4_v
      float[batch,77,640] node_scaled_dot_product_attention_5_k
      float[batch,77,640] node_scaled_dot_product_attention_5_out
      float[batch,77,640] node_scaled_dot_product_attention_5_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_5_q
      float[batch,77,1920] node_scaled_dot_product_attention_5_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_5_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_5_v
      float[batch,77,640] node_scaled_dot_product_attention_6_k
      float[batch,77,640] node_scaled_dot_product_attention_6_out
      float[batch,77,640] node_scaled_dot_product_attention_6_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_6_q
      float[batch,77,1920] node_scaled_dot_product_attention_6_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_6_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_6_v
      float[batch,77,640] node_scaled_dot_product_attention_7_k
      float[batch,77,640] node_scaled_dot_product_attention_7_out
      float[batch,77,640] node_scaled_dot_product_attention_7_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_7_q
      float[batch,77,1920] node_scaled_dot_product_attention_7_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_7_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_7_v
      float[batch,77,640] node_scaled_dot_product_attention_8_k
      float[batch,77,640] node_scaled_dot_product_attention_8_out
      float[batch,77,640] node_scaled_dot_product_attention_8_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_8_q
      float[batch,77,1920] node_scaled_dot_product_attention_8_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_8_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_8_v
      float[batch,77,640] node_scaled_dot_product_attention_9_k
      float[batch,77,640] node_scaled_dot_product_attention_9_out
      float[batch,77,640] node_scaled_dot_product_attention_9_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_9_q
      float[batch,77,1920] node_scaled_dot_product_attention_9_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_9_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_9_v
      float[batch,77,640] node_scaled_dot_product_attention_k
      float[batch,77,640] node_scaled_dot_product_attention_out
      float[batch,77,640] node_scaled_dot_product_attention_out_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_q
      float[batch,77,1920] node_scaled_dot_product_attention_qkv
      float[batch,77,1920] node_scaled_dot_product_attention_qkv_mm_out
      float[batch,77,640] node_scaled_dot_product_attention_v
      float[batch,77,640] scaled_dot_product_attention
      float[batch,77,640] scaled_dot_product_attention_1
      float[batch,77,640] scaled_dot_product_attention_10
      float[batch,77,640] scaled_dot_product_attention_11
      float[batch,1,640] scaled_dot_product_attention_11_pooled
      float[batch,77,640] scaled_dot_product_attention_2
      float[batch,77,640] scaled_dot_product_attention_3
      float[batch,77,640] scaled_dot_product_attention_4
      float[batch,77,640] scaled_dot_product_attention_5
      float[batch,77,640] scaled_dot_product_attention_6
      float[batch,77,640] scaled_dot_product_attention_7
      float[batch,77,640] scaled_dot_product_attention_8
      float[batch,77,640] scaled_dot_product_attention_9
      float[batch,77,2560] val_39
      float[batch,77,640] val_40
      float[batch,77,2560] val_41
      float[batch,77,640] val_42
      float[batch,77,2560] val_43
      float[batch,77,640] val_44
      float[batch,77,2560] val_45
      float[batch,77,640] val_46
      float[batch,77,2560] val_47
      float[batch,77,640] val_48
      float[batch,77,2560] val_49
      float[batch,77,640] val_50
      float[batch,77,2560] val_51
      float[batch,77,640] val_52
      float[batch,77,2560] val_53
      float[batch,77,640] val_54
      float[batch,77,2560] val_55
      float[batch,77,640] val_56
      float[batch,77,2560] val_57
      float[batch,77,640] val_58
      float[batch,77,2560] val_59
      float[batch,77,640] val_60
      float[batch,77] val_61
      float[batch,1,77] val_62
      float[batch,1,2560] val_63
      float[batch,1,640] val_64
      float[batch,1,640] val_65
      float[batch,640] val_66
   >
{
   val_38 = Gather <axis: int = 0> ("token_embedding.weight_fp16", text)
   embedding = Cast <to: int = 1> (val_38)
   add_4 = Add (embedding, positional_embedding)
   [node_layer_norm] layer_norm = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_4, "transformer.resblocks.0.ln_1.weight", "transformer.resblocks.0.ln_1.bias")
   [node_scaled_dot_product_attention_qkv_mm] node_scaled_dot_product_attention_qkv_mm_out = MatMul (layer_norm, val_2)
   [node_scaled_dot_product_attention_qkv_bias] node_scaled_dot_product_attention_qkv = Add (node_scaled_dot_product_attention_qkv_mm_out, "transformer.resblocks.0.attn.in_proj_bias")
   [node_scaled_dot_product_attention_qkv_split] node_scaled_dot_product_attention_q, node_scaled_dot_product_attention_k, node_scaled_dot_product_attention_v = Split <axis: int = -1> (node_scaled_dot_product_attention_qkv, attn3d_split_3x640)
   scaled_dot_product_attention = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_q, node_scaled_dot_product_attention_k, node_scaled_dot_product_attention_v)
   [node_scaled_dot_product_attention_out_mm] node_scaled_dot_product_attention_out_mm_out = MatMul (scaled_dot_product_attention, node_scaled_dot_product_attention_wo_t)
   [node_scaled_dot_product_attention_out_bias] node_scaled_dot_product_attention_out = Add (node_scaled_dot_product_attention_out_mm_out, "transformer.resblocks.0.attn.out_proj.bias")
   add_119 = Add (add_4, node_scaled_dot_product_attention_out)
   layer_norm_1 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_119, "transformer.resblocks.0.ln_2.weight", "transformer.resblocks.0.ln_2.bias")
   val_39 = MatMul (layer_norm_1, val_3)
   linear_2 = Add (val_39, "transformer.resblocks.0.mlp.c_fc.bias")
   [node_gelu] gelu = Gelu <approximate: string = "none"> (linear_2)
   val_40 = MatMul (gelu, val_4)
   linear_3 = Add (val_40, "transformer.resblocks.0.mlp.c_proj.bias")
   add_140 = Add (add_119, linear_3)
   layer_norm_2 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_140, "transformer.resblocks.1.ln_1.weight", "transformer.resblocks.1.ln_1.bias")
   [node_scaled_dot_product_attention_1_qkv_mm] node_scaled_dot_product_attention_1_qkv_mm_out = MatMul (layer_norm_2, val_5)
   [node_scaled_dot_product_attention_1_qkv_bias] node_scaled_dot_product_attention_1_qkv = Add (node_scaled_dot_product_attention_1_qkv_mm_out, "transformer.resblocks.1.attn.in_proj_bias")
   [node_scaled_dot_product_attention_1_qkv_split] node_scaled_dot_product_attention_1_q, node_scaled_dot_product_attention_1_k, node_scaled_dot_product_attention_1_v = Split <axis: int = -1> (node_scaled_dot_product_attention_1_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_1 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_1_q, node_scaled_dot_product_attention_1_k, node_scaled_dot_product_attention_1_v)
   [node_scaled_dot_product_attention_1_out_mm] node_scaled_dot_product_attention_1_out_mm_out = MatMul (scaled_dot_product_attention_1, node_scaled_dot_product_attention_1_wo_t)
   [node_scaled_dot_product_attention_1_out_bias] node_scaled_dot_product_attention_1_out = Add (node_scaled_dot_product_attention_1_out_mm_out, "transformer.resblocks.1.attn.out_proj.bias")
   add_255 = Add (add_140, node_scaled_dot_product_attention_1_out)
   layer_norm_3 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_255, "transformer.resblocks.1.ln_2.weight", "transformer.resblocks.1.ln_2.bias")
   val_41 = MatMul (layer_norm_3, val_6)
   linear_6 = Add (val_41, "transformer.resblocks.1.mlp.c_fc.bias")
   gelu_1 = Gelu <approximate: string = "none"> (linear_6)
   val_42 = MatMul (gelu_1, val_7)
   linear_7 = Add (val_42, "transformer.resblocks.1.mlp.c_proj.bias")
   add_276 = Add (add_255, linear_7)
   layer_norm_4 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_276, "transformer.resblocks.2.ln_1.weight", "transformer.resblocks.2.ln_1.bias")
   [node_scaled_dot_product_attention_2_qkv_mm] node_scaled_dot_product_attention_2_qkv_mm_out = MatMul (layer_norm_4, val_8)
   [node_scaled_dot_product_attention_2_qkv_bias] node_scaled_dot_product_attention_2_qkv = Add (node_scaled_dot_product_attention_2_qkv_mm_out, "transformer.resblocks.2.attn.in_proj_bias")
   [node_scaled_dot_product_attention_2_qkv_split] node_scaled_dot_product_attention_2_q, node_scaled_dot_product_attention_2_k, node_scaled_dot_product_attention_2_v = Split <axis: int = -1> (node_scaled_dot_product_attention_2_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_2 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_2_q, node_scaled_dot_product_attention_2_k, node_scaled_dot_product_attention_2_v)
   [node_scaled_dot_product_attention_2_out_mm] node_scaled_dot_product_attention_2_out_mm_out = MatMul (scaled_dot_product_attention_2, node_scaled_dot_product_attention_2_wo_t)
   [node_scaled_dot_product_attention_2_out_bias] node_scaled_dot_product_attention_2_out = Add (node_scaled_dot_product_attention_2_out_mm_out, "transformer.resblocks.2.attn.out_proj.bias")
   add_391 = Add (add_276, node_scaled_dot_product_attention_2_out)
   layer_norm_5 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_391, "transformer.resblocks.2.ln_2.weight", "transformer.resblocks.2.ln_2.bias")
   val_43 = MatMul (layer_norm_5, val_9)
   linear_10 = Add (val_43, "transformer.resblocks.2.mlp.c_fc.bias")
   gelu_2 = Gelu <approximate: string = "none"> (linear_10)
   val_44 = MatMul (gelu_2, val_10)
   linear_11 = Add (val_44, "transformer.resblocks.2.mlp.c_proj.bias")
   add_412 = Add (add_391, linear_11)
   layer_norm_6 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_412, "transformer.resblocks.3.ln_1.weight", "transformer.resblocks.3.ln_1.bias")
   [node_scaled_dot_product_attention_3_qkv_mm] node_scaled_dot_product_attention_3_qkv_mm_out = MatMul (layer_norm_6, val_11)
   [node_scaled_dot_product_attention_3_qkv_bias] node_scaled_dot_product_attention_3_qkv = Add (node_scaled_dot_product_attention_3_qkv_mm_out, "transformer.resblocks.3.attn.in_proj_bias")
   [node_scaled_dot_product_attention_3_qkv_split] node_scaled_dot_product_attention_3_q, node_scaled_dot_product_attention_3_k, node_scaled_dot_product_attention_3_v = Split <axis: int = -1> (node_scaled_dot_product_attention_3_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_3 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_3_q, node_scaled_dot_product_attention_3_k, node_scaled_dot_product_attention_3_v)
   [node_scaled_dot_product_attention_3_out_mm] node_scaled_dot_product_attention_3_out_mm_out = MatMul (scaled_dot_product_attention_3, node_scaled_dot_product_attention_3_wo_t)
   [node_scaled_dot_product_attention_3_out_bias] node_scaled_dot_product_attention_3_out = Add (node_scaled_dot_product_attention_3_out_mm_out, "transformer.resblocks.3.attn.out_proj.bias")
   add_527 = Add (add_412, node_scaled_dot_product_attention_3_out)
   layer_norm_7 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_527, "transformer.resblocks.3.ln_2.weight", "transformer.resblocks.3.ln_2.bias")
   val_45 = MatMul (layer_norm_7, val_12)
   linear_14 = Add (val_45, "transformer.resblocks.3.mlp.c_fc.bias")
   gelu_3 = Gelu <approximate: string = "none"> (linear_14)
   val_46 = MatMul (gelu_3, val_13)
   linear_15 = Add (val_46, "transformer.resblocks.3.mlp.c_proj.bias")
   add_548 = Add (add_527, linear_15)
   layer_norm_8 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_548, "transformer.resblocks.4.ln_1.weight", "transformer.resblocks.4.ln_1.bias")
   [node_scaled_dot_product_attention_4_qkv_mm] node_scaled_dot_product_attention_4_qkv_mm_out = MatMul (layer_norm_8, val_14)
   [node_scaled_dot_product_attention_4_qkv_bias] node_scaled_dot_product_attention_4_qkv = Add (node_scaled_dot_product_attention_4_qkv_mm_out, "transformer.resblocks.4.attn.in_proj_bias")
   [node_scaled_dot_product_attention_4_qkv_split] node_scaled_dot_product_attention_4_q, node_scaled_dot_product_attention_4_k, node_scaled_dot_product_attention_4_v = Split <axis: int = -1> (node_scaled_dot_product_attention_4_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_4 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_4_q, node_scaled_dot_product_attention_4_k, node_scaled_dot_product_attention_4_v)
   [node_scaled_dot_product_attention_4_out_mm] node_scaled_dot_product_attention_4_out_mm_out = MatMul (scaled_dot_product_attention_4, node_scaled_dot_product_attention_4_wo_t)
   [node_scaled_dot_product_attention_4_out_bias] node_scaled_dot_product_attention_4_out = Add (node_scaled_dot_product_attention_4_out_mm_out, "transformer.resblocks.4.attn.out_proj.bias")
   add_663 = Add (add_548, node_scaled_dot_product_attention_4_out)
   layer_norm_9 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_663, "transformer.resblocks.4.ln_2.weight", "transformer.resblocks.4.ln_2.bias")
   val_47 = MatMul (layer_norm_9, val_15)
   linear_18 = Add (val_47, "transformer.resblocks.4.mlp.c_fc.bias")
   gelu_4 = Gelu <approximate: string = "none"> (linear_18)
   val_48 = MatMul (gelu_4, val_16)
   linear_19 = Add (val_48, "transformer.resblocks.4.mlp.c_proj.bias")
   add_684 = Add (add_663, linear_19)
   layer_norm_10 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_684, "transformer.resblocks.5.ln_1.weight", "transformer.resblocks.5.ln_1.bias")
   [node_scaled_dot_product_attention_5_qkv_mm] node_scaled_dot_product_attention_5_qkv_mm_out = MatMul (layer_norm_10, val_17)
   [node_scaled_dot_product_attention_5_qkv_bias] node_scaled_dot_product_attention_5_qkv = Add (node_scaled_dot_product_attention_5_qkv_mm_out, "transformer.resblocks.5.attn.in_proj_bias")
   [node_scaled_dot_product_attention_5_qkv_split] node_scaled_dot_product_attention_5_q, node_scaled_dot_product_attention_5_k, node_scaled_dot_product_attention_5_v = Split <axis: int = -1> (node_scaled_dot_product_attention_5_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_5 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_5_q, node_scaled_dot_product_attention_5_k, node_scaled_dot_product_attention_5_v)
   [node_scaled_dot_product_attention_5_out_mm] node_scaled_dot_product_attention_5_out_mm_out = MatMul (scaled_dot_product_attention_5, node_scaled_dot_product_attention_5_wo_t)
   [node_scaled_dot_product_attention_5_out_bias] node_scaled_dot_product_attention_5_out = Add (node_scaled_dot_product_attention_5_out_mm_out, "transformer.resblocks.5.attn.out_proj.bias")
   add_799 = Add (add_684, node_scaled_dot_product_attention_5_out)
   layer_norm_11 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_799, "transformer.resblocks.5.ln_2.weight", "transformer.resblocks.5.ln_2.bias")
   val_49 = MatMul (layer_norm_11, val_18)
   linear_22 = Add (val_49, "transformer.resblocks.5.mlp.c_fc.bias")
   gelu_5 = Gelu <approximate: string = "none"> (linear_22)
   val_50 = MatMul (gelu_5, val_19)
   linear_23 = Add (val_50, "transformer.resblocks.5.mlp.c_proj.bias")
   add_820 = Add (add_799, linear_23)
   layer_norm_12 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_820, "transformer.resblocks.6.ln_1.weight", "transformer.resblocks.6.ln_1.bias")
   [node_scaled_dot_product_attention_6_qkv_mm] node_scaled_dot_product_attention_6_qkv_mm_out = MatMul (layer_norm_12, val_20)
   [node_scaled_dot_product_attention_6_qkv_bias] node_scaled_dot_product_attention_6_qkv = Add (node_scaled_dot_product_attention_6_qkv_mm_out, "transformer.resblocks.6.attn.in_proj_bias")
   [node_scaled_dot_product_attention_6_qkv_split] node_scaled_dot_product_attention_6_q, node_scaled_dot_product_attention_6_k, node_scaled_dot_product_attention_6_v = Split <axis: int = -1> (node_scaled_dot_product_attention_6_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_6 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_6_q, node_scaled_dot_product_attention_6_k, node_scaled_dot_product_attention_6_v)
   [node_scaled_dot_product_attention_6_out_mm] node_scaled_dot_product_attention_6_out_mm_out = MatMul (scaled_dot_product_attention_6, node_scaled_dot_product_attention_6_wo_t)
   [node_scaled_dot_product_attention_6_out_bias] node_scaled_dot_product_attention_6_out = Add (node_scaled_dot_product_attention_6_out_mm_out, "transformer.resblocks.6.attn.out_proj.bias")
   add_935 = Add (add_820, node_scaled_dot_product_attention_6_out)
   layer_norm_13 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_935, "transformer.resblocks.6.ln_2.weight", "transformer.resblocks.6.ln_2.bias")
   val_51 = MatMul (layer_norm_13, val_21)
   linear_26 = Add (val_51, "transformer.resblocks.6.mlp.c_fc.bias")
   gelu_6 = Gelu <approximate: string = "none"> (linear_26)
   val_52 = MatMul (gelu_6, val_22)
   linear_27 = Add (val_52, "transformer.resblocks.6.mlp.c_proj.bias")
   add_956 = Add (add_935, linear_27)
   layer_norm_14 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_956, "transformer.resblocks.7.ln_1.weight", "transformer.resblocks.7.ln_1.bias")
   [node_scaled_dot_product_attention_7_qkv_mm] node_scaled_dot_product_attention_7_qkv_mm_out = MatMul (layer_norm_14, val_23)
   [node_scaled_dot_product_attention_7_qkv_bias] node_scaled_dot_product_attention_7_qkv = Add (node_scaled_dot_product_attention_7_qkv_mm_out, "transformer.resblocks.7.attn.in_proj_bias")
   [node_scaled_dot_product_attention_7_qkv_split] node_scaled_dot_product_attention_7_q, node_scaled_dot_product_attention_7_k, node_scaled_dot_product_attention_7_v = Split <axis: int = -1> (node_scaled_dot_product_attention_7_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_7 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_7_q, node_scaled_dot_product_attention_7_k, node_scaled_dot_product_attention_7_v)
   [node_scaled_dot_product_attention_7_out_mm] node_scaled_dot_product_attention_7_out_mm_out = MatMul (scaled_dot_product_attention_7, node_scaled_dot_product_attention_7_wo_t)
   [node_scaled_dot_product_attention_7_out_bias] node_scaled_dot_product_attention_7_out = Add (node_scaled_dot_product_attention_7_out_mm_out, "transformer.resblocks.7.attn.out_proj.bias")
   add_1071 = Add (add_956, node_scaled_dot_product_attention_7_out)
   layer_norm_15 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1071, "transformer.resblocks.7.ln_2.weight", "transformer.resblocks.7.ln_2.bias")
   val_53 = MatMul (layer_norm_15, val_24)
   linear_30 = Add (val_53, "transformer.resblocks.7.mlp.c_fc.bias")
   gelu_7 = Gelu <approximate: string = "none"> (linear_30)
   val_54 = MatMul (gelu_7, val_25)
   linear_31 = Add (val_54, "transformer.resblocks.7.mlp.c_proj.bias")
   add_1092 = Add (add_1071, linear_31)
   layer_norm_16 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1092, "transformer.resblocks.8.ln_1.weight", "transformer.resblocks.8.ln_1.bias")
   [node_scaled_dot_product_attention_8_qkv_mm] node_scaled_dot_product_attention_8_qkv_mm_out = MatMul (layer_norm_16, val_26)
   [node_scaled_dot_product_attention_8_qkv_bias] node_scaled_dot_product_attention_8_qkv = Add (node_scaled_dot_product_attention_8_qkv_mm_out, "transformer.resblocks.8.attn.in_proj_bias")
   [node_scaled_dot_product_attention_8_qkv_split] node_scaled_dot_product_attention_8_q, node_scaled_dot_product_attention_8_k, node_scaled_dot_product_attention_8_v = Split <axis: int = -1> (node_scaled_dot_product_attention_8_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_8 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_8_q, node_scaled_dot_product_attention_8_k, node_scaled_dot_product_attention_8_v)
   [node_scaled_dot_product_attention_8_out_mm] node_scaled_dot_product_attention_8_out_mm_out = MatMul (scaled_dot_product_attention_8, node_scaled_dot_product_attention_8_wo_t)
   [node_scaled_dot_product_attention_8_out_bias] node_scaled_dot_product_attention_8_out = Add (node_scaled_dot_product_attention_8_out_mm_out, "transformer.resblocks.8.attn.out_proj.bias")
   add_1207 = Add (add_1092, node_scaled_dot_product_attention_8_out)
   layer_norm_17 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1207, "transformer.resblocks.8.ln_2.weight", "transformer.resblocks.8.ln_2.bias")
   val_55 = MatMul (layer_norm_17, val_27)
   linear_34 = Add (val_55, "transformer.resblocks.8.mlp.c_fc.bias")
   gelu_8 = Gelu <approximate: string = "none"> (linear_34)
   val_56 = MatMul (gelu_8, val_28)
   linear_35 = Add (val_56, "transformer.resblocks.8.mlp.c_proj.bias")
   add_1228 = Add (add_1207, linear_35)
   layer_norm_18 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1228, "transformer.resblocks.9.ln_1.weight", "transformer.resblocks.9.ln_1.bias")
   [node_scaled_dot_product_attention_9_qkv_mm] node_scaled_dot_product_attention_9_qkv_mm_out = MatMul (layer_norm_18, val_29)
   [node_scaled_dot_product_attention_9_qkv_bias] node_scaled_dot_product_attention_9_qkv = Add (node_scaled_dot_product_attention_9_qkv_mm_out, "transformer.resblocks.9.attn.in_proj_bias")
   [node_scaled_dot_product_attention_9_qkv_split] node_scaled_dot_product_attention_9_q, node_scaled_dot_product_attention_9_k, node_scaled_dot_product_attention_9_v = Split <axis: int = -1> (node_scaled_dot_product_attention_9_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_9 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_9_q, node_scaled_dot_product_attention_9_k, node_scaled_dot_product_attention_9_v)
   [node_scaled_dot_product_attention_9_out_mm] node_scaled_dot_product_attention_9_out_mm_out = MatMul (scaled_dot_product_attention_9, node_scaled_dot_product_attention_9_wo_t)
   [node_scaled_dot_product_attention_9_out_bias] node_scaled_dot_product_attention_9_out = Add (node_scaled_dot_product_attention_9_out_mm_out, "transformer.resblocks.9.attn.out_proj.bias")
   add_1343 = Add (add_1228, node_scaled_dot_product_attention_9_out)
   layer_norm_19 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1343, "transformer.resblocks.9.ln_2.weight", "transformer.resblocks.9.ln_2.bias")
   val_57 = MatMul (layer_norm_19, val_30)
   linear_38 = Add (val_57, "transformer.resblocks.9.mlp.c_fc.bias")
   gelu_9 = Gelu <approximate: string = "none"> (linear_38)
   val_58 = MatMul (gelu_9, val_31)
   linear_39 = Add (val_58, "transformer.resblocks.9.mlp.c_proj.bias")
   add_1364 = Add (add_1343, linear_39)
   layer_norm_20 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1364, "transformer.resblocks.10.ln_1.weight", "transformer.resblocks.10.ln_1.bias")
   [node_scaled_dot_product_attention_10_qkv_mm] node_scaled_dot_product_attention_10_qkv_mm_out = MatMul (layer_norm_20, val_32)
   [node_scaled_dot_product_attention_10_qkv_bias] node_scaled_dot_product_attention_10_qkv = Add (node_scaled_dot_product_attention_10_qkv_mm_out, "transformer.resblocks.10.attn.in_proj_bias")
   [node_scaled_dot_product_attention_10_qkv_split] node_scaled_dot_product_attention_10_q, node_scaled_dot_product_attention_10_k, node_scaled_dot_product_attention_10_v = Split <axis: int = -1> (node_scaled_dot_product_attention_10_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_10 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_10_q, node_scaled_dot_product_attention_10_k, node_scaled_dot_product_attention_10_v)
   [node_scaled_dot_product_attention_10_out_mm] node_scaled_dot_product_attention_10_out_mm_out = MatMul (scaled_dot_product_attention_10, node_scaled_dot_product_attention_10_wo_t)
   [node_scaled_dot_product_attention_10_out_bias] node_scaled_dot_product_attention_10_out = Add (node_scaled_dot_product_attention_10_out_mm_out, "transformer.resblocks.10.attn.out_proj.bias")
   add_1479 = Add (add_1364, node_scaled_dot_product_attention_10_out)
   layer_norm_21 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1479, "transformer.resblocks.10.ln_2.weight", "transformer.resblocks.10.ln_2.bias")
   val_59 = MatMul (layer_norm_21, val_33)
   linear_42 = Add (val_59, "transformer.resblocks.10.mlp.c_fc.bias")
   gelu_10 = Gelu <approximate: string = "none"> (linear_42)
   val_60 = MatMul (gelu_10, val_34)
   linear_43 = Add (val_60, "transformer.resblocks.10.mlp.c_proj.bias")
   add_1500 = Add (add_1479, linear_43)
   layer_norm_22 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1500, "transformer.resblocks.11.ln_1.weight", "transformer.resblocks.11.ln_1.bias")
   [node_scaled_dot_product_attention_11_qkv_mm] node_scaled_dot_product_attention_11_qkv_mm_out = MatMul (layer_norm_22, val_35)
   [node_scaled_dot_product_attention_11_qkv_bias] node_scaled_dot_product_attention_11_qkv = Add (node_scaled_dot_product_attention_11_qkv_mm_out, "transformer.resblocks.11.attn.in_proj_bias")
   [node_scaled_dot_product_attention_11_qkv_split] node_scaled_dot_product_attention_11_q, node_scaled_dot_product_attention_11_k, node_scaled_dot_product_attention_11_v = Split <axis: int = -1> (node_scaled_dot_product_attention_11_qkv, attn3d_split_3x640)
   scaled_dot_product_attention_11 = Attention <is_causal: int = 1, kv_num_heads: int = 10, q_num_heads: int = 10, qk_matmul_output_mode: int = 0, softcap: float = 0> (node_scaled_dot_product_attention_11_q, node_scaled_dot_product_attention_11_k, node_scaled_dot_product_attention_11_v)
   [node_argmax] argmax = ArgMax <axis: int = -1, keepdims: int = 0, select_last_index: int = 0> (text)
   val_61 = Gather <axis: int = 0> (val_755_eye, argmax)
   val_62 = Unsqueeze (val_61, val_755_axes1)
   [pool_hoist_scaled_dot_product_attention_11] scaled_dot_product_attention_11_pooled = MatMul (val_62, scaled_dot_product_attention_11)
   [node_scaled_dot_product_attention_11_out_mm] node_scaled_dot_product_attention_11_out_mm_out = MatMul (scaled_dot_product_attention_11_pooled, node_scaled_dot_product_attention_11_wo_t)
   [node_scaled_dot_product_attention_11_out_bias] node_scaled_dot_product_attention_11_out = Add (node_scaled_dot_product_attention_11_out_mm_out, "transformer.resblocks.11.attn.out_proj.bias")
   [pool_hoist_add_1500] add_1500_pooled = MatMul (val_62, add_1500)
   add_1615 = Add (add_1500_pooled, node_scaled_dot_product_attention_11_out)
   layer_norm_23 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1615, "transformer.resblocks.11.ln_2.weight", "transformer.resblocks.11.ln_2.bias")
   val_63 = MatMul (layer_norm_23, val_36)
   linear_46 = Add (val_63, "transformer.resblocks.11.mlp.c_fc.bias")
   gelu_11 = Gelu <approximate: string = "none"> (linear_46)
   val_64 = MatMul (gelu_11, val_37)
   linear_47 = Add (val_64, "transformer.resblocks.11.mlp.c_proj.bias")
   add_1636 = Add (add_1615, linear_47)
   val_65 = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> (add_1636, "ln_final.weight", "ln_final.bias")
   val_66 = Squeeze (val_65, val_755_axes1)
   [node_matmul] matmul = MatMul (val_66, text_projection)
   [node_linalg_vector_norm] linalg_vector_norm = ReduceL2 <keepdims: int = 1, noop_with_empty_axes: int = 0> (matmul, val_0)
   [node_clamp_min] clamp_min = Clip (linalg_vector_norm, val_1)
   [node_div] text_embedding = Div (matmul, clamp_min)
}

weights:
attn3d_split_3x640 INT64[3] 96f437355f6e
ln_final.bias FLOAT[640] 2610f3f13c04
ln_final.weight FLOAT[640] bbc6361978ef
node_scaled_dot_product_attention_10_wo_t FLOAT[640,640] 72957367bc0b
node_scaled_dot_product_attention_11_wo_t FLOAT[640,640] 0ae34099890c
node_scaled_dot_product_attention_1_wo_t FLOAT[640,640] d00ab3994066
node_scaled_dot_product_attention_2_wo_t FLOAT[640,640] 8b1ad4a13581
node_scaled_dot_product_attention_3_wo_t FLOAT[640,640] 4fea02ff4795
node_scaled_dot_product_attention_4_wo_t FLOAT[640,640] a7f301f1d2ea
node_scaled_dot_product_attention_5_wo_t FLOAT[640,640] 45fdb4d4866d
node_scaled_dot_product_attention_6_wo_t FLOAT[640,640] 06a132f50b64
node_scaled_dot_product_attention_7_wo_t FLOAT[640,640] 194a7d77a130
node_scaled_dot_product_attention_8_wo_t FLOAT[640,640] 7ca454100438
node_scaled_dot_product_attention_9_wo_t FLOAT[640,640] 7b400e6f4628
node_scaled_dot_product_attention_wo_t FLOAT[640,640] fdac9343e635
positional_embedding FLOAT[77,640] fa836ed05c50
text_projection FLOAT[640,640] b41e9a00db5e
token_embedding.weight_fp16 FLOAT16[49408,640] e8f205ab41a4
transformer.resblocks.0.attn.in_proj_bias FLOAT[1920] f9f14729a87d
transformer.resblocks.0.attn.out_proj.bias FLOAT[640] b589e7a6c36f
transformer.resblocks.0.ln_1.bias FLOAT[640] 36ad86e9af58
transformer.resblocks.0.ln_1.weight FLOAT[640] 35ee8a6e14f1
transformer.resblocks.0.ln_2.bias FLOAT[640] 6df8e9a66e1c
transformer.resblocks.0.ln_2.weight FLOAT[640] 74d9e8b9fa1f
transformer.resblocks.0.mlp.c_fc.bias FLOAT[2560] 0a521d40b6c7
transformer.resblocks.0.mlp.c_proj.bias FLOAT[640] 18da244a1a9b
transformer.resblocks.1.attn.in_proj_bias FLOAT[1920] 13df81528c3f
transformer.resblocks.1.attn.out_proj.bias FLOAT[640] 002d47959b81
transformer.resblocks.1.ln_1.bias FLOAT[640] 96490d1095d1
transformer.resblocks.1.ln_1.weight FLOAT[640] 8e34c486a7a3
transformer.resblocks.1.ln_2.bias FLOAT[640] c97f36f843f6
transformer.resblocks.1.ln_2.weight FLOAT[640] a44e3fddf7ed
transformer.resblocks.1.mlp.c_fc.bias FLOAT[2560] c77111e1abb7
transformer.resblocks.1.mlp.c_proj.bias FLOAT[640] 2933029d5795
transformer.resblocks.10.attn.in_proj_bias FLOAT[1920] 8babb520d39c
transformer.resblocks.10.attn.out_proj.bias FLOAT[640] aa4e0deff979
transformer.resblocks.10.ln_1.bias FLOAT[640] f27c80fb4fd5
transformer.resblocks.10.ln_1.weight FLOAT[640] 0eb0b4dfe903
transformer.resblocks.10.ln_2.bias FLOAT[640] a08c573253a1
transformer.resblocks.10.ln_2.weight FLOAT[640] 2f16f3af0174
transformer.resblocks.10.mlp.c_fc.bias FLOAT[2560] a7c96eb07f88
transformer.resblocks.10.mlp.c_proj.bias FLOAT[640] 2e6f4f347815
transformer.resblocks.11.attn.in_proj_bias FLOAT[1920] d029d535140d
transformer.resblocks.11.attn.out_proj.bias FLOAT[640] a94f93f3cb88
transformer.resblocks.11.ln_1.bias FLOAT[640] ca93d61bbb69
transformer.resblocks.11.ln_1.weight FLOAT[640] 6ccf48d581af
transformer.resblocks.11.ln_2.bias FLOAT[640] c61c764395c7
transformer.resblocks.11.ln_2.weight FLOAT[640] 8b1d7cb95479
transformer.resblocks.11.mlp.c_fc.bias FLOAT[2560] 4b0efdb97954
transformer.resblocks.11.mlp.c_proj.bias FLOAT[640] a3736b2abc6a
transformer.resblocks.2.attn.in_proj_bias FLOAT[1920] 57bb6c5b8323
transformer.resblocks.2.attn.out_proj.bias FLOAT[640] 92e2a6a3e853
transformer.resblocks.2.ln_1.bias FLOAT[640] 74bfac4ba4dc
transformer.resblocks.2.ln_1.weight FLOAT[640] e6551460898f
transformer.resblocks.2.ln_2.bias FLOAT[640] 08b09b246fb5
transformer.resblocks.2.ln_2.weight FLOAT[640] 3bbdc400a9a0
transformer.resblocks.2.mlp.c_fc.bias FLOAT[2560] 72e6d5bf1495
transformer.resblocks.2.mlp.c_proj.bias FLOAT[640] b669b1b37e48
transformer.resblocks.3.attn.in_proj_bias FLOAT[1920] 533c1c7cab5a
transformer.resblocks.3.attn.out_proj.bias FLOAT[640] b93991f2507b
transformer.resblocks.3.ln_1.bias FLOAT[640] b488b856e7aa
transformer.resblocks.3.ln_1.weight FLOAT[640] acafde77e6e7
transformer.resblocks.3.ln_2.bias FLOAT[640] c91ddbdf332b
transformer.resblocks.3.ln_2.weight FLOAT[640] aa5955413c23
transformer.resblocks.3.mlp.c_fc.bias FLOAT[2560] 1da2cf75902a
transformer.resblocks.3.mlp.c_proj.bias FLOAT[640] 8ea65818597a
transformer.resblocks.4.attn.in_proj_bias FLOAT[1920] 8a39ba9e7924
transformer.resblocks.4.attn.out_proj.bias FLOAT[640] 24d2ae37fb75
transformer.resblocks.4.ln_1.bias FLOAT[640] 87ef355af11d
transformer.resblocks.4.ln_1.weight FLOAT[640] b0322d6d9841
transformer.resblocks.4.ln_2.bias FLOAT[640] e5d5a298615f
transformer.resblocks.4.ln_2.weight FLOAT[640] d809bee4b498
transformer.resblocks.4.mlp.c_fc.bias FLOAT[2560] a44bfae94d69
transformer.resblocks.4.mlp.c_proj.bias FLOAT[640] 7daa0eee2be0
transformer.resblocks.5.attn.in_proj_bias FLOAT[1920] 9a1ad7b29f49
transformer.resblocks.5.attn.out_proj.bias FLOAT[640] a2ee3348d44e
transformer.resblocks.5.ln_1.bias FLOAT[640] 542ab252169f
transformer.resblocks.5.ln_1.weight FLOAT[640] d694a4e8656c
transformer.resblocks.5.ln_2.bias FLOAT[640] b956b6a021d4
transformer.resblocks.5.ln_2.weight FLOAT[640] 80feccecb1aa
transformer.resblocks.5.mlp.c_fc.bias FLOAT[2560] 5146d34ab3a7
transformer.resblocks.5.mlp.c_proj.bias FLOAT[640] 67a75387d2e2
transformer.resblocks.6.attn.in_proj_bias FLOAT[1920] 50d5c41deef0
transformer.resblocks.6.attn.out_proj.bias FLOAT[640] 6b02c111b294
transformer.resblocks.6.ln_1.bias FLOAT[640] ef6e5c152ebd
transformer.resblocks.6.ln_1.weight FLOAT[640] 0413e6a38248
transformer.resblocks.6.ln_2.bias FLOAT[640] 2f15931f4e3b
transformer.resblocks.6.ln_2.weight FLOAT[640] 46513a95b743
transformer.resblocks.6.mlp.c_fc.bias FLOAT[2560] 91f022a478e1
transformer.resblocks.6.mlp.c_proj.bias FLOAT[640] c9f20ca02b8b
transformer.resblocks.7.attn.in_proj_bias FLOAT[1920] d712c8b18a2b
transformer.resblocks.7.attn.out_proj.bias FLOAT[640] f22d9c09352b
transformer.resblocks.7.ln_1.bias FLOAT[640] 04ff81d9b9de
transformer.resblocks.7.ln_1.weight FLOAT[640] 60888aeab76d
transformer.resblocks.7.ln_2.bias FLOAT[640] a13bf3d22da3
transformer.resblocks.7.ln_2.weight FLOAT[640] 65d48358d8f1
transformer.resblocks.7.mlp.c_fc.bias FLOAT[2560] 8e2b6500003b
transformer.resblocks.7.mlp.c_proj.bias FLOAT[640] b1db5d09c688
transformer.resblocks.8.attn.in_proj_bias FLOAT[1920] fe933f7246cb
transformer.resblocks.8.attn.out_proj.bias FLOAT[640] 9f1f34e16825
transformer.resblocks.8.ln_1.bias FLOAT[640] 36a1b4cc4217
transformer.resblocks.8.ln_1.weight FLOAT[640] 3874c0dc0eca
transformer.resblocks.8.ln_2.bias FLOAT[640] ae8ce1c4e525
transformer.resblocks.8.ln_2.weight FLOAT[640] 05ab1c4e51ad
transformer.resblocks.8.mlp.c_fc.bias FLOAT[2560] edd54e1f4a38
transformer.resblocks.8.mlp.c_proj.bias FLOAT[640] 90a2d973abdd
transformer.resblocks.9.attn.in_proj_bias FLOAT[1920] f03083d4d58d
transformer.resblocks.9.attn.out_proj.bias FLOAT[640] 456d6c10ef26
transformer.resblocks.9.ln_1.bias FLOAT[640] be87f78d2b16
transformer.resblocks.9.ln_1.weight FLOAT[640] e51b0cc96b41
transformer.resblocks.9.ln_2.bias FLOAT[640] 68f458c34455
transformer.resblocks.9.ln_2.weight FLOAT[640] 624dfb1b52a2
transformer.resblocks.9.mlp.c_fc.bias FLOAT[2560] e4437cbb550b
transformer.resblocks.9.mlp.c_proj.bias FLOAT[640] 41f4caf6c80b
val_0 INT64[1] 12a3ae445661
val_1 FLOAT[] 6708d9be4956
val_10 FLOAT[2560,640] 36a5936ce6ba
val_11 FLOAT[640,1920] 455e4617f9e3
val_12 FLOAT[640,2560] cbb7c629d8b2
val_13 FLOAT[2560,640] c3e0708e2496
val_14 FLOAT[640,1920] 7639d3683880
val_15 FLOAT[640,2560] f73cab321ca9
val_16 FLOAT[2560,640] f45c14cd32aa
val_17 FLOAT[640,1920] 3ba754ebd2b4
val_18 FLOAT[640,2560] fcbdc30bf353
val_19 FLOAT[2560,640] 051265270a87
val_2 FLOAT[640,1920] 4d884409852d
val_20 FLOAT[640,1920] f0ea42264bd6
val_21 FLOAT[640,2560] 62e98fd54365
val_22 FLOAT[2560,640] 717a394ed3de
val_23 FLOAT[640,1920] 1b19293358be
val_24 FLOAT[640,2560] 9a50acc3a4d0
val_25 FLOAT[2560,640] 6a5b952bbe19
val_26 FLOAT[640,1920] 9aa5bdad9721
val_27 FLOAT[640,2560] bc3ebab62871
val_28 FLOAT[2560,640] 80270659c045
val_29 FLOAT[640,1920] 6c4d62dcd425
val_3 FLOAT[640,2560] 79ba0bf52047
val_30 FLOAT[640,2560] ba4b0a5fd527
val_31 FLOAT[2560,640] f755bbe841e3
val_32 FLOAT[640,1920] d97256bfa0c0
val_33 FLOAT[640,2560] 48b5674790ff
val_34 FLOAT[2560,640] c85544302be5
val_35 FLOAT[640,1920] 8cb80817037a
val_36 FLOAT[640,2560] 887a5e56f897
val_37 FLOAT[2560,640] d3ff8f09820a
val_4 FLOAT[2560,640] c60111dc058e
val_5 FLOAT[640,1920] bd42d4e0e63e
val_6 FLOAT[640,2560] ed19ce4d1bce
val_7 FLOAT[2560,640] d46a866ca596
val_755_axes1 INT64[1] 7c9fa136d441
val_755_eye FLOAT[77,77] 39ccea14e08c
val_8 FLOAT[640,1920] 51c38a1f2134
val_9 FLOAT[640,2560] 3b5a9c71979c
