<
   ir_version: 10,
   opset_import: ["" : 23],
   producer_name: "pytorch"
>
main_graph (int32[batch,77] text) => (float[batch,768] text_embedding) 
   <
      float[batch,77,768] add_1071
      float[batch,77,768] add_1092
      float[batch,77,768] add_119
      float[batch,77,768] add_1207
      float[batch,77,768] add_1228
      float[batch,77,768] add_1343
      float[batch,77,768] add_1364
      float[batch,77,768] add_140
      float[batch,77,768] add_1479
      float[batch,77,768] add_1500
      float[batch,1,768] add_1500_pooled
      float[batch,1,768] add_1615
      float[batch,1,768] add_1636
      float[batch,77,768] add_255
      float[batch,77,768] add_276
      float[batch,77,768] add_391
      float[batch,77,768] add_4
      float[batch,77,768] add_412
      float[batch,77,768] add_527
      float[batch,77,768] add_548
      float[batch,77,768] add_663
      float[batch,77,768] add_684
      float[batch,77,768] add_799
      float[batch,77,768] add_820
      float[batch,77,768] add_935
      float[batch,77,768] add_956
      int64[batch] argmax
      float[batch,1] clamp_min
      float[batch,77,768] embedding
      float[batch,77,3072] gelu
      float[batch,77,3072] gelu_1
      float[batch,77,3072] gelu_10
      float[batch,1,3072] gelu_11
      float[batch,77,3072] gelu_2
      float[batch,77,3072] gelu_3
      float[batch,77,3072] gelu_4
      float[batch,77,3072] gelu_5
      float[batch,77,3072] gelu_6
      float[batch,77,3072] gelu_7
      float[batch,77,3072] gelu_8
      float[batch,77,3072] gelu_9
      float[batch,77,768] layer_norm
      float[batch,77,768] layer_norm_1
      float[batch,77,768] layer_norm_10
      float[batch,77,768] layer_norm_11
      float[batch,77,768] layer_norm_12
      float[batch,77,768] layer_norm_13
      float[batch,77,768] layer_norm_14
      float[batch,77,768] layer_norm_15
      float[batch,77,768] layer_norm_16
      float[batch,77,768] layer_norm_17
      float[batch,77,768] layer_norm_18
      float[batch,77,768] layer_norm_19
      float[batch,77,768] layer_norm_2
      float[batch,77,768] layer_norm_20
      float[batch,77,768] layer_norm_21
      float[batch,77,768] layer_norm_22
      float[batch,1,768] layer_norm_23
      float[batch,77,768] layer_norm_3
      float[batch,77,768] layer_norm_4
      float[batch,77,768] layer_norm_5
      float[batch,77,768] layer_norm_6
      float[batch,77,768] layer_norm_7
      float[batch,77,768] layer_norm_8
      float[batch,77,768] layer_norm_9
      float[batch,1] linalg_vector_norm
      float[batch,77,3072] linear_10
      float[batch,77,768] linear_11
      float[batch,77,3072] linear_14
      float[batch,77,768] linear_15
      float[batch,77,3072] linear_18
      float[batch,77,768] linear_19
      float[batch,77,3072] linear_2
      float[batch,77,3072] linear_22
      float[batch,77,768] linear_23
      float[batch,77,3072] linear_26
      float[batch,77,768] linear_27
      float[batch,77,768] linear_3
      float[batch,77,3072] linear_30
      float[batch,77,768] linear_31
      float[batch,77,3072] linear_34
      float[batch,77,768] linear_35
      float[batch,77,3072] linear_38
      float[batch,77,768] linear_39
      float[batch,77,3072] linear_42
      float[batch,77,768] linear_43
      float[batch,1,3072] linear_46
      float[batch,1,768] linear_47
      float[batch,77,3072] linear_6
      float[batch,77,768] linear_7
      float[batch,768] matmul
      float[batch,77,768] node_scaled_dot_product_attention_10_k
      float[batch,77,768] node_scaled_dot_product_attention_10_out
      float[batch,77,768] node_scaled_dot_product_attention_10_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_10_q
      float[batch,77,2304] node_scaled_dot_product_attention_10_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_10_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_10_v
      float[batch,77,768] node_scaled_dot_product_attention_11_k
      float[batch,1,768] node_scaled_dot_product_attention_11_out
      float[batch,1,768] node_scaled_dot_product_attention_11_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_11_q
      float[batch,77,2304] node_scaled_dot_product_attention_11_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_11_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_11_v
      float[batch,77,768] node_scaled_dot_product_attention_1_k
      float[batch,77,768] node_scaled_dot_product_attention_1_out
      float[batch,77,768] node_scaled_dot_product_attention_1_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_1_q
      float[batch,77,2304] node_scaled_dot_product_attention_1_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_1_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_1_v
      float[batch,77,768] node_scaled_dot_product_attention_2_k
      float[batch,77,768] node_scaled_dot_product_attention_2_out
      float[batch,77,768] node_scaled_dot_product_attention_2_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_2_q
      float[batch,77,2304] node_scaled_dot_product_attention_2_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_2_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_2_v
      float[batch,77,768] node_scaled_dot_product_attention_3_k
      float[batch,77,768] node_scaled_dot_product_attention_3_out
      float[batch,77,768] node_scaled_dot_product_attention_3_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_3_q
      float[batch,77,2304] node_scaled_dot_product_attention_3_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_3_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_3_v
      float[batch,77,768] node_scaled_dot_product_attention_4_k
      float[batch,77,768] node_scaled_dot_product_attention_4_out
      float[batch,77,768] node_scaled_dot_product_attention_4_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_4_q
      float[batch,77,2304] node_scaled_dot_product_attention_4_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_4_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_4_v
      float[batch,77,768] node_scaled_dot_product_attention_5_k
      float[batch,77,768] node_scaled_dot_product_attention_5_out
      float[batch,77,768] node_scaled_dot_product_attention_5_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_5_q
      float[batch,77,2304] node_scaled_dot_product_attention_5_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_5_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_5_v
      float[batch,77,768] node_scaled_dot_product_attention_6_k
      float[batch,77,768] node_scaled_dot_product_attention_6_out
      float[batch,77,768] node_scaled_dot_product_attention_6_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_6_q
      float[batch,77,2304] node_scaled_dot_product_attention_6_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_6_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_6_v
      float[batch,77,768] node_scaled_dot_product_attention_7_k
      float[batch,77,768] node_scaled_dot_product_attention_7_out
      float[batch,77,768] node_scaled_dot_product_attention_7_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_7_q
      float[batch,77,2304] node_scaled_dot_product_attention_7_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_7_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_7_v
      float[batch,77,768] node_scaled_dot_product_attention_8_k
      float[batch,77,768] node_scaled_dot_product_attention_8_out
      float[batch,77,768] node_scaled_dot_product_attention_8_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_8_q
      float[batch,77,2304] node_scaled_dot_product_attention_8_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_8_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_8_v
      float[batch,77,768] node_scaled_dot_product_attention_9_k
      float[batch,77,768] node_scaled_dot_product_attention_9_out
      float[batch,77,768] node_scaled_dot_product_attention_9_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_9_q
      float[batch,77,2304] node_scaled_dot_product_attention_9_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_9_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_9_v
      float[batch,77,768] node_scaled_dot_product_attention_k
      float[batch,77,768] node_scaled_dot_product_attention_out
      float[batch,77,768] node_scaled_dot_product_attention_out_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_q
      float[batch,77,2304] node_scaled_dot_product_attention_qkv
      float[batch,77,2304] node_scaled_dot_product_attention_qkv_mm_out
      float[batch,77,768] node_scaled_dot_product_attention_v
      float[batch,77,768] scaled_dot_product_attention
      float[batch,77,768] scaled_dot_product_attention_1
      float[batch,77,768] scaled_dot_product_attention_10
      float[batch,77,768] scaled_dot_product_attention_11
      float[batch,1,768] scaled_dot_product_attention_11_pooled
      float[batch,77,768] scaled_dot_product_attention_2
      float[batch,77,768] scaled_dot_product_attention_3
      float[batch,77,768] scaled_dot_product_attention_4
      float[batch,77,768] scaled_dot_product_attention_5
      float[batch,77,768] scaled_dot_product_attention_6
      float[batch,77,768] scaled_dot_product_attention_7
      float[batch,77,768] scaled_dot_product_attention_8
      float[batch,77,768] scaled_dot_product_attention_9
      float[batch,77,3072] val_39
      float[batch,77,768] val_40
      float[batch,77,3072] val_41
      float[batch,77,768] val_42
      float[batch,77,3072] val_43
      float[batch,77,768] val_44
      float[batch,77,3072] val_45
      float[batch,77,768] val_46
      float[batch,77,3072] val_47
      float[batch,77,768] val_48
      float[batch,77,3072] val_49
      float[batch,77,768] val_50
      float[batch,77,3072] val_51
      float[batch,77,768] val_52
      float[batch,77,3072] val_53
      float[batch,77,768] val_54
      float[batch,77,3072] val_55
      float[batch,77,768] val_56
      float[batch,77,3072] val_57
      float[batch,77,768] val_58
      float[batch,77,3072] val_59
      float[batch,77,768] val_60
      float[batch,77] val_61
      float[batch,1,77] val_62
      float[batch,1,3072] val_63
      float[batch,1,768] val_64
      float[batch,1,768] val_65
      float[batch,768] 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_3x768)
   scaled_dot_product_attention = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_1 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_2 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_3 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_4 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_5 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_6 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_7 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_8 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_9 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_10 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768)
   scaled_dot_product_attention_11 = Attention <is_causal: int = 1, kv_num_heads: int = 12, q_num_heads: int = 12, 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_3x768 INT64[3] eca50b2daf34
ln_final.bias FLOAT[768] dba19084d90b
ln_final.weight FLOAT[768] b558c4ac6e38
node_scaled_dot_product_attention_10_wo_t FLOAT[768,768] 9c29135c341b
node_scaled_dot_product_attention_11_wo_t FLOAT[768,768] 5482e0adfd4b
node_scaled_dot_product_attention_1_wo_t FLOAT[768,768] 510a6c1a7f62
node_scaled_dot_product_attention_2_wo_t FLOAT[768,768] 15b2edc712cd
node_scaled_dot_product_attention_3_wo_t FLOAT[768,768] 121810cb2bdc
node_scaled_dot_product_attention_4_wo_t FLOAT[768,768] 36423c7c7dc8
node_scaled_dot_product_attention_5_wo_t FLOAT[768,768] f5f46c8dec9e
node_scaled_dot_product_attention_6_wo_t FLOAT[768,768] 7b90e7139301
node_scaled_dot_product_attention_7_wo_t FLOAT[768,768] 1366123ce262
node_scaled_dot_product_attention_8_wo_t FLOAT[768,768] 1b5ed0d56361
node_scaled_dot_product_attention_9_wo_t FLOAT[768,768] ee2d9d8da4b1
node_scaled_dot_product_attention_wo_t FLOAT[768,768] c8141613ecf2
positional_embedding FLOAT[77,768] 5f16dd1fc44e
text_projection FLOAT[768,768] c7676f9eb616
token_embedding.weight_fp16 FLOAT16[49408,768] 35540399de50
transformer.resblocks.0.attn.in_proj_bias FLOAT[2304] 26eb8c55f6ee
transformer.resblocks.0.attn.out_proj.bias FLOAT[768] f22968b0dc0c
transformer.resblocks.0.ln_1.bias FLOAT[768] a4c53ca350e4
transformer.resblocks.0.ln_1.weight FLOAT[768] 0851e4f4f882
transformer.resblocks.0.ln_2.bias FLOAT[768] f4b22a0217a9
transformer.resblocks.0.ln_2.weight FLOAT[768] 4cda2f17081f
transformer.resblocks.0.mlp.c_fc.bias FLOAT[3072] 847b227c6e17
transformer.resblocks.0.mlp.c_proj.bias FLOAT[768] 12679d2237e4
transformer.resblocks.1.attn.in_proj_bias FLOAT[2304] decfe6e317bd
transformer.resblocks.1.attn.out_proj.bias FLOAT[768] 01ef02edb4a2
transformer.resblocks.1.ln_1.bias FLOAT[768] a39ca069958d
transformer.resblocks.1.ln_1.weight FLOAT[768] 294ea5a480f1
transformer.resblocks.1.ln_2.bias FLOAT[768] 53647a13720d
transformer.resblocks.1.ln_2.weight FLOAT[768] 5f92a36ca43c
transformer.resblocks.1.mlp.c_fc.bias FLOAT[3072] b017930dd91d
transformer.resblocks.1.mlp.c_proj.bias FLOAT[768] 93043e95bb7d
transformer.resblocks.10.attn.in_proj_bias FLOAT[2304] b5a89ab0b04b
transformer.resblocks.10.attn.out_proj.bias FLOAT[768] c0bb6f3319e6
transformer.resblocks.10.ln_1.bias FLOAT[768] 0e2fc54879b1
transformer.resblocks.10.ln_1.weight FLOAT[768] 32c8d658d754
transformer.resblocks.10.ln_2.bias FLOAT[768] 19a0d3e5855e
transformer.resblocks.10.ln_2.weight FLOAT[768] ad658175f96c
transformer.resblocks.10.mlp.c_fc.bias FLOAT[3072] a73c20f5540c
transformer.resblocks.10.mlp.c_proj.bias FLOAT[768] 0f886f533529
transformer.resblocks.11.attn.in_proj_bias FLOAT[2304] 8e461d08e6f8
transformer.resblocks.11.attn.out_proj.bias FLOAT[768] bf37c9466e17
transformer.resblocks.11.ln_1.bias FLOAT[768] 6e0d2c68c6cc
transformer.resblocks.11.ln_1.weight FLOAT[768] 7efbee7c4565
transformer.resblocks.11.ln_2.bias FLOAT[768] 8e0c2e643ed7
transformer.resblocks.11.ln_2.weight FLOAT[768] dc9e8ae0aede
transformer.resblocks.11.mlp.c_fc.bias FLOAT[3072] cd0d0466a3fd
transformer.resblocks.11.mlp.c_proj.bias FLOAT[768] edd2dbea2ffe
transformer.resblocks.2.attn.in_proj_bias FLOAT[2304] 9666ebb96405
transformer.resblocks.2.attn.out_proj.bias FLOAT[768] 1463435a07fb
transformer.resblocks.2.ln_1.bias FLOAT[768] 78823461abcd
transformer.resblocks.2.ln_1.weight FLOAT[768] f3ad62968853
transformer.resblocks.2.ln_2.bias FLOAT[768] 3b3d1d096c91
transformer.resblocks.2.ln_2.weight FLOAT[768] 5df5c5eb6fd3
transformer.resblocks.2.mlp.c_fc.bias FLOAT[3072] 1133494758ec
transformer.resblocks.2.mlp.c_proj.bias FLOAT[768] eab4287b962b
transformer.resblocks.3.attn.in_proj_bias FLOAT[2304] 2b52a1baaef9
transformer.resblocks.3.attn.out_proj.bias FLOAT[768] 59ddfc83ef8b
transformer.resblocks.3.ln_1.bias FLOAT[768] c8be2da71bd0
transformer.resblocks.3.ln_1.weight FLOAT[768] 5d7a02bbf439
transformer.resblocks.3.ln_2.bias FLOAT[768] 07d680b1f04c
transformer.resblocks.3.ln_2.weight FLOAT[768] 349131c8a2d2
transformer.resblocks.3.mlp.c_fc.bias FLOAT[3072] e44c994d07fc
transformer.resblocks.3.mlp.c_proj.bias FLOAT[768] fcaa21fc72a4
transformer.resblocks.4.attn.in_proj_bias FLOAT[2304] eb84d55f0c3e
transformer.resblocks.4.attn.out_proj.bias FLOAT[768] 53e0f650f47e
transformer.resblocks.4.ln_1.bias FLOAT[768] 16a61e013ddd
transformer.resblocks.4.ln_1.weight FLOAT[768] e9ea7fd1868c
transformer.resblocks.4.ln_2.bias FLOAT[768] b1ac526e4301
transformer.resblocks.4.ln_2.weight FLOAT[768] fde28576ba34
transformer.resblocks.4.mlp.c_fc.bias FLOAT[3072] 8223b5d1fb09
transformer.resblocks.4.mlp.c_proj.bias FLOAT[768] 379ee8747604
transformer.resblocks.5.attn.in_proj_bias FLOAT[2304] ff94f77514df
transformer.resblocks.5.attn.out_proj.bias FLOAT[768] 1ca7cbd1471c
transformer.resblocks.5.ln_1.bias FLOAT[768] 2f9ea66e0577
transformer.resblocks.5.ln_1.weight FLOAT[768] 36f0b6a808c1
transformer.resblocks.5.ln_2.bias FLOAT[768] ec08e034c6ff
transformer.resblocks.5.ln_2.weight FLOAT[768] ac111777f520
transformer.resblocks.5.mlp.c_fc.bias FLOAT[3072] 5c278c0c3274
transformer.resblocks.5.mlp.c_proj.bias FLOAT[768] d9910081cc44
transformer.resblocks.6.attn.in_proj_bias FLOAT[2304] d62604830b92
transformer.resblocks.6.attn.out_proj.bias FLOAT[768] b705ca052d30
transformer.resblocks.6.ln_1.bias FLOAT[768] 07d82afbab22
transformer.resblocks.6.ln_1.weight FLOAT[768] 8256675e19d0
transformer.resblocks.6.ln_2.bias FLOAT[768] 953404a3d37b
transformer.resblocks.6.ln_2.weight FLOAT[768] b17ab6076aab
transformer.resblocks.6.mlp.c_fc.bias FLOAT[3072] 37545d0d9128
transformer.resblocks.6.mlp.c_proj.bias FLOAT[768] 79a685ff0011
transformer.resblocks.7.attn.in_proj_bias FLOAT[2304] d886811316d2
transformer.resblocks.7.attn.out_proj.bias FLOAT[768] 835bd9cdae2d
transformer.resblocks.7.ln_1.bias FLOAT[768] b88f06a36824
transformer.resblocks.7.ln_1.weight FLOAT[768] a4be6aa18e1c
transformer.resblocks.7.ln_2.bias FLOAT[768] 2cdd15269d84
transformer.resblocks.7.ln_2.weight FLOAT[768] b3239799bdbf
transformer.resblocks.7.mlp.c_fc.bias FLOAT[3072] ef541d16508d
transformer.resblocks.7.mlp.c_proj.bias FLOAT[768] 8de30eac19d2
transformer.resblocks.8.attn.in_proj_bias FLOAT[2304] bfe69edc540d
transformer.resblocks.8.attn.out_proj.bias FLOAT[768] 09894ad69af0
transformer.resblocks.8.ln_1.bias FLOAT[768] 0abe4ed09bc3
transformer.resblocks.8.ln_1.weight FLOAT[768] 9f1d141bf733
transformer.resblocks.8.ln_2.bias FLOAT[768] 932fb764600a
transformer.resblocks.8.ln_2.weight FLOAT[768] 619fca305b9c
transformer.resblocks.8.mlp.c_fc.bias FLOAT[3072] 766096a4e090
transformer.resblocks.8.mlp.c_proj.bias FLOAT[768] 9d3f26ff64e3
transformer.resblocks.9.attn.in_proj_bias FLOAT[2304] acdc8a659353
transformer.resblocks.9.attn.out_proj.bias FLOAT[768] 3f216f97b2f6
transformer.resblocks.9.ln_1.bias FLOAT[768] 00bc262927d1
transformer.resblocks.9.ln_1.weight FLOAT[768] 7836027ca7db
transformer.resblocks.9.ln_2.bias FLOAT[768] bfdc38862ddf
transformer.resblocks.9.ln_2.weight FLOAT[768] 590573170b29
transformer.resblocks.9.mlp.c_fc.bias FLOAT[3072] 60efb9311587
transformer.resblocks.9.mlp.c_proj.bias FLOAT[768] 26002a9af872
val_0 INT64[1] 12a3ae445661
val_1 FLOAT[] 6708d9be4956
val_10 FLOAT[3072,768] 24ac16dcc61d
val_11 FLOAT[768,2304] 8b3727b50c7a
val_12 FLOAT[768,3072] a1bf18910032
val_13 FLOAT[3072,768] 6b0461d463df
val_14 FLOAT[768,2304] cc051ebc1112
val_15 FLOAT[768,3072] 7432131a70c7
val_16 FLOAT[3072,768] b395ec090fb9
val_17 FLOAT[768,2304] dd20f30e4ed6
val_18 FLOAT[768,3072] 5a5c97b57e5c
val_19 FLOAT[3072,768] c0a3f7969ef4
val_2 FLOAT[768,2304] 98ef4ca875d5
val_20 FLOAT[768,2304] 4847af810307
val_21 FLOAT[768,3072] a51cbbfb819e
val_22 FLOAT[3072,768] ab6ccb1d9924
val_23 FLOAT[768,2304] e25c52182f6d
val_24 FLOAT[768,3072] 02da0af43baf
val_25 FLOAT[3072,768] caef7c45dead
val_26 FLOAT[768,2304] 70ca90b566ad
val_27 FLOAT[768,3072] 81338c391bc3
val_28 FLOAT[3072,768] 12a481029464
val_29 FLOAT[768,2304] 19e05964b861
val_3 FLOAT[768,3072] b0a11fe04530
val_30 FLOAT[768,3072] b657bfff47b2
val_31 FLOAT[3072,768] b6c4743c42cb
val_32 FLOAT[768,2304] 9a267e68cc16
val_33 FLOAT[768,3072] 68a924995109
val_34 FLOAT[3072,768] ee6f12848617
val_35 FLOAT[768,2304] 6d5336947763
val_36 FLOAT[768,3072] c6fafd2605fc
val_37 FLOAT[3072,768] f364208d3cde
val_4 FLOAT[3072,768] 5dc64a2d6992
val_5 FLOAT[768,2304] 02dbdeb4a500
val_6 FLOAT[768,3072] c30e44280a1c
val_7 FLOAT[3072,768] 0bd58b00429c
val_755_axes1 INT64[1] 7c9fa136d441
val_755_eye FLOAT[77,77] 39ccea14e08c
val_8 FLOAT[768,2304] ab766f5157d5
val_9 FLOAT[768,3072] af61028b2540
