Upload 2 files
Browse files
ncnn/RWKV-x070-World-0.1B-v2.8-20241210-ctx4096-ncnn.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c7dab7c856261b3f2fe33de419bfa87f957b5cd29ee305f01247158c5e7a50c9
|
3 |
+
size 382180960
|
ncnn/RWKV-x070-World-0.1B-v2.8-20241210-ctx4096-ncnn.param
ADDED
@@ -0,0 +1,1198 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
7767517
|
2 |
+
1196 1471
|
3 |
+
MemoryData model_info 0 1 model_info 0=5 21=1
|
4 |
+
Input input_0 0 1 token 0=1 1=1 2=1
|
5 |
+
Input input_1 0 1 state_0_in 0=768
|
6 |
+
Input input_2 0 1 state_1_in 0=64 1=64 2=12
|
7 |
+
Input input_3 0 1 state_2_in 0=768
|
8 |
+
Input input_4 0 1 state_3_in 0=768
|
9 |
+
Input input_5 0 1 state_4_in 0=64 1=64 2=12
|
10 |
+
Input input_6 0 1 state_5_in 0=768
|
11 |
+
Input input_7 0 1 state_6_in 0=768
|
12 |
+
Input input_8 0 1 state_7_in 0=64 1=64 2=12
|
13 |
+
Input input_9 0 1 state_8_in 0=768
|
14 |
+
Input input_10 0 1 state_9_in 0=768
|
15 |
+
Input input_11 0 1 state_10_in 0=64 1=64 2=12
|
16 |
+
Input input_12 0 1 state_11_in 0=768
|
17 |
+
Input input_13 0 1 state_12_in 0=768
|
18 |
+
Input input_14 0 1 state_13_in 0=64 1=64 2=12
|
19 |
+
Input input_15 0 1 state_14_in 0=768
|
20 |
+
Input input_16 0 1 state_15_in 0=768
|
21 |
+
Input input_17 0 1 state_16_in 0=64 1=64 2=12
|
22 |
+
Input input_18 0 1 state_17_in 0=768
|
23 |
+
Input input_19 0 1 state_18_in 0=768
|
24 |
+
Input input_20 0 1 state_19_in 0=64 1=64 2=12
|
25 |
+
Input input_21 0 1 state_20_in 0=768
|
26 |
+
Input input_22 0 1 state_21_in 0=768
|
27 |
+
Input input_23 0 1 state_22_in 0=64 1=64 2=12
|
28 |
+
Input input_24 0 1 state_23_in 0=768
|
29 |
+
Input input_25 0 1 state_24_in 0=768
|
30 |
+
Input input_26 0 1 state_25_in 0=64 1=64 2=12
|
31 |
+
Input input_27 0 1 state_26_in 0=768
|
32 |
+
Input input_28 0 1 state_27_in 0=768
|
33 |
+
Input input_29 0 1 state_28_in 0=64 1=64 2=12
|
34 |
+
Input input_30 0 1 state_29_in 0=768
|
35 |
+
Input input_31 0 1 state_30_in 0=768
|
36 |
+
Input input_32 0 1 state_31_in 0=64 1=64 2=12
|
37 |
+
Input input_33 0 1 state_32_in 0=768
|
38 |
+
Input input_34 0 1 state_33_in 0=768
|
39 |
+
Input input_35 0 1 state_34_in 0=64 1=64 2=12
|
40 |
+
Input input_36 0 1 state_35_in 0=768
|
41 |
+
Embed embedding 1 1 token emb 0=768 1=65536 3=50331648
|
42 |
+
Split split_emb 1 2 emb att_0_x_last att_0_x
|
43 |
+
LayerNorm layernorm_0 1 1 att_0_x att_0_xx 0=768 1=0.00001 2=1
|
44 |
+
Split split_att_0_xx 1 3 att_0_xx att_0_xx_0 att_0_xx_1 state_0_out
|
45 |
+
BinaryOp sub_0 2 1 state_0_in att_0_xx_0 att_0_sx 0=1
|
46 |
+
MemoryData data_att_0_maa 0 1 att_0_maa 0=768 1=6 21=0
|
47 |
+
BinaryOp mul_0 2 1 att_0_sx att_0_maa att_0_maa_x 0=2
|
48 |
+
BinaryOp add_0 2 1 att_0_xx_1 att_0_maa_x att_0_maa_rwkvag 0=0
|
49 |
+
Slice slice_att_0_maa_rwkvag 1 6 att_0_maa_rwkvag att_0_mr att_0_mw att_0_mk att_0_mv att_0_ma att_0_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
50 |
+
Split split_att_0_mv 1 2 att_0_mv att_0_mv_0 att_0_mv_1
|
51 |
+
Gemm gemm_att_0_key 1 1 att_0_mk att_0_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
52 |
+
Gemm gemm_att_0_value 1 1 att_0_mv_0 att_0_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
53 |
+
Gemm gemm_att_0_receptance 1 1 att_0_mr att_0_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
54 |
+
Gemm gemm_att_0_ma_lora_0 1 1 att_0_ma att_0_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
55 |
+
Gemm gemm_att_0_ma_lora_1 1 1 att_0_ma_lora_0 att_0_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
56 |
+
MemoryData data_att_0_a0 0 1 att_0_a0 0=768 21=0
|
57 |
+
BinaryOp add_1 2 1 att_0_ma_lora_1 att_0_a0 att_0_ma_lora_2 0=0
|
58 |
+
Sigmoid sigmoid_0 1 1 att_0_ma_lora_2 att_0_a
|
59 |
+
Split split_att_0_a 1 2 att_0_a att_0_a_0 att_0_a_1
|
60 |
+
Gemm gemm_att_0_mg_lora 1 1 att_0_mg att_0_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
61 |
+
Sigmoid sigmoid_1 1 1 att_0_mg_lora att_0_mg_lora_sigmoid
|
62 |
+
Gemm gemm_att_0_gate 1 1 att_0_mg_lora_sigmoid att_0_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
63 |
+
MemoryData data_att_0_k_k 0 1 att_0_k_k 0=768 21=0
|
64 |
+
MemoryData data_att_0_k_a 0 1 att_0_k_a 0=768 21=0
|
65 |
+
MemoryData data_att_0_r_k 0 1 att_0_r_k 0=768 21=0
|
66 |
+
Split split_att_0_key 1 2 att_0_key att_0_key_0 att_0_key_1
|
67 |
+
BinaryOp mul_1 2 1 att_0_key_0 att_0_k_k att_0_key_k 0=2
|
68 |
+
Reshape reshape_0 1 1 att_0_key_k att_0_key_k_reshape 0=64 1=1 2=12
|
69 |
+
Normalize l2norm_0 1 1 att_0_key_k_reshape att_0_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
70 |
+
Reshape reshape_1 1 1 att_0_key_k_norm att_0_key_k_norm_reshape 0=768
|
71 |
+
BinaryOp add_2 1 1 att_0_a_0 att_0_a_0_minus 0=0 1=1 2=-1.0
|
72 |
+
BinaryOp mul_2 2 1 att_0_a_0_minus att_0_k_a att_0_a_minus1_mulka 0=2
|
73 |
+
BinaryOp add_3 1 1 att_0_a_minus1_mulka att_0_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
74 |
+
BinaryOp mul_3 2 1 att_0_a_minus1_mulka_plus1 att_0_key_1 att_0_key_a 0=2
|
75 |
+
Split split_att_0_value 1 2 att_0_value att_0_value_final v_first_0
|
76 |
+
Gemm gemm_att_0_mw_lora 1 1 att_0_mw att_0_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
77 |
+
TanH tanh_att_0_mw_lora 1 1 att_0_mw_lora att_0_mw_lora_tanh
|
78 |
+
Gemm gemm_att_0_mw_lora_tanh_linear 1 1 att_0_mw_lora_tanh att_0_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
79 |
+
MemoryData data_att_0_td 0 1 att_0_td 0=768 21=0
|
80 |
+
BinaryOp add_4 2 1 att_0_mw_lora_tanh_linear att_0_td att_0_time_decay_pre 0=0
|
81 |
+
Sigmoid sigmoid_2 1 1 att_0_time_decay_pre att_0_time_decay_sigmoid
|
82 |
+
Exp exp_0 1 1 att_0_time_decay_sigmoid att_0_time_decay 1=-0.606531
|
83 |
+
Reshape reshape_2 1 1 att_0_time_decay att_0_time_decay_reshape 0=64 1=1 2=12
|
84 |
+
Split split_att_0_key_a 1 2 att_0_key_a att_0_key_a_0 att_0_key_a_1
|
85 |
+
Reshape reshape_3 1 1 att_0_key_a_0 att_0_key_a_reshape 0=64 1=1 2=12
|
86 |
+
Split split_att_0_value_final 1 3 att_0_value_final att_0_value_final_0 att_0_value_final_1 att_0_value_final_2
|
87 |
+
Reshape reshape_4 1 1 att_0_value_final_0 att_0_value_final_0_reshape 0=64 1=12
|
88 |
+
Reshape reshape_5 1 1 att_0_value_final_1 att_0_value_final_1_reshape 0=1 1=64 2=12
|
89 |
+
MatMul matmul_att_0_vk 2 1 att_0_value_final_1_reshape att_0_key_a_reshape att_0_vk 0=0
|
90 |
+
Split split_att_0_key_k_norm_reshape 1 2 att_0_key_k_norm_reshape att_0_kk_0 att_0_kk_1
|
91 |
+
BinaryOp mul_4 1 1 att_0_kk_0 att_0_kk_0_neg 0=2 1=1 2=-1.0
|
92 |
+
Reshape reshape_6 1 1 att_0_kk_0_neg att_0_kk_0_neg_reshape 0=1 1=64 2=12
|
93 |
+
BinaryOp mul_5 2 1 att_0_kk_1 att_0_a_1 att_0_kk_a 0=2
|
94 |
+
Reshape reshape_7 1 1 att_0_kk_a att_0_kk_a_reshape 0=64 1=1 2=12
|
95 |
+
MatMul matmul_att_0_ab 2 1 att_0_kk_0_neg_reshape att_0_kk_a_reshape att_0_ab 0=0
|
96 |
+
Split split_state_1_in 1 2 state_1_in att_0_state_prev_0 att_0_state_prev_1
|
97 |
+
BinaryOp mul_6 2 1 att_0_state_prev_0 att_0_time_decay_reshape att_0_state_td 0=2
|
98 |
+
MatMul matmul_att_0_sab 2 1 att_0_state_prev_1 att_0_ab att_0_sab 0=0
|
99 |
+
BinaryOp add_5 2 1 att_0_sab att_0_vk att_0_sab_vk 0=0
|
100 |
+
BinaryOp add_6 2 1 att_0_sab_vk att_0_state_td att_0_state_new 0=0
|
101 |
+
Split split_att_0_state_new 1 2 att_0_state_new att_0_state_new_0 state_1_out
|
102 |
+
Split split_att_0_receptance 1 2 att_0_receptance att_0_receptance_0 att_0_receptance_1
|
103 |
+
Reshape reshape_8 1 1 att_0_receptance_0 att_0_receptance_reshape 0=1 1=64 2=12
|
104 |
+
MatMul matmul_att_0_wkv_out 2 1 att_0_state_new_0 att_0_receptance_reshape att_0_wkv_out 0=0
|
105 |
+
Reshape reshape_9 1 1 att_0_wkv_out att_0_wkv_out_flatten 0=768
|
106 |
+
GroupNorm groupnorm_0 1 1 att_0_wkv_out_flatten att_0_x_gn 0=12 1=768 2=0.00064 3=1
|
107 |
+
BinaryOp mul_7 2 1 att_0_receptance_1 att_0_key_a_1 att_0_receptance_key_a 0=2
|
108 |
+
BinaryOp mul_8 2 1 att_0_receptance_key_a att_0_r_k att_0_rk 0=2
|
109 |
+
Reshape reshape_10 1 1 att_0_rk att_0_rk_reshape 0=64 1=12
|
110 |
+
Reduction sum_0 1 1 att_0_rk_reshape att_0_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
111 |
+
BinaryOp mul_9 2 1 att_0_value_final_0_reshape att_0_rk_sum att_0_value_final_0_rk 0=2
|
112 |
+
Reshape reshape_11 1 1 att_0_value_final_0_rk att_0_value_final_0_rk_reshape 0=768
|
113 |
+
BinaryOp add_7 2 1 att_0_value_final_0_rk_reshape att_0_x_gn att_0_x_gn_rkv 0=0
|
114 |
+
BinaryOp mul_10 2 1 att_0_x_gn_rkv att_0_gate att_0_x_gate 0=2
|
115 |
+
Gemm gemm_att_0_x_out 1 1 att_0_x_gate att_0_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
116 |
+
BinaryOp add_8 2 1 att_0_x_out att_0_x_last time_mixing_0_out 0=0
|
117 |
+
Split split_time_mixing_0_out 1 2 time_mixing_0_out ffn_0_x_last ffn_0_x
|
118 |
+
LayerNorm layernorm_1 1 1 ffn_0_x ffn_0_xx 0=768 1=0.00001 2=1
|
119 |
+
Split split_ffn_0_xx 1 3 ffn_0_xx ffn_0_xx_0 ffn_0_xx_1 state_2_out
|
120 |
+
BinaryOp sub_1 2 1 state_2_in ffn_0_xx_0 ffn_0_sx 0=1
|
121 |
+
Split split_ffn_0_sx 1 1 ffn_0_sx ffn_0_sx_0
|
122 |
+
MemoryData data_ffn_0_x_k 0 1 ffn_0_x_k 0=768 21=0
|
123 |
+
BinaryOp mul_11 2 1 ffn_0_sx_0 ffn_0_x_k ffn_0_xk 0=2
|
124 |
+
BinaryOp add_9 2 1 ffn_0_xk ffn_0_xx_1 ffn_0_xxk 0=0
|
125 |
+
Gemm gemm_ffn_0_key 1 1 ffn_0_xxk ffn_0_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
126 |
+
ReLU relu_0 1 1 ffn_0_key ffn_0_key_relu
|
127 |
+
UnaryOp square_ffn_0_key_relu 1 1 ffn_0_key_relu ffn_0_key_relu_square 0=4
|
128 |
+
Gemm gemm_ffn_0_value 1 1 ffn_0_key_relu_square ffn_0_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
129 |
+
BinaryOp add_10 2 1 ffn_0_value ffn_0_x_last channel_mixing_0_out 0=0
|
130 |
+
Split split_channel_mixing_0_out 1 2 channel_mixing_0_out att_1_x_last att_1_x
|
131 |
+
LayerNorm layernorm_2 1 1 att_1_x att_1_xx 0=768 1=0.00001 2=1
|
132 |
+
Split split_att_1_xx 1 3 att_1_xx att_1_xx_0 att_1_xx_1 state_3_out
|
133 |
+
BinaryOp sub_2 2 1 state_3_in att_1_xx_0 att_1_sx 0=1
|
134 |
+
MemoryData data_att_1_maa 0 1 att_1_maa 0=768 1=6 21=0
|
135 |
+
BinaryOp mul_12 2 1 att_1_sx att_1_maa att_1_maa_x 0=2
|
136 |
+
BinaryOp add_11 2 1 att_1_xx_1 att_1_maa_x att_1_maa_rwkvag 0=0
|
137 |
+
Slice slice_att_1_maa_rwkvag 1 6 att_1_maa_rwkvag att_1_mr att_1_mw att_1_mk att_1_mv att_1_ma att_1_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
138 |
+
Split split_att_1_mv 1 2 att_1_mv att_1_mv_0 att_1_mv_1
|
139 |
+
Gemm gemm_att_1_key 1 1 att_1_mk att_1_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
140 |
+
Gemm gemm_att_1_value 1 1 att_1_mv_0 att_1_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
141 |
+
Gemm gemm_att_1_receptance 1 1 att_1_mr att_1_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
142 |
+
Gemm gemm_att_1_ma_lora_0 1 1 att_1_ma att_1_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
143 |
+
Gemm gemm_att_1_ma_lora_1 1 1 att_1_ma_lora_0 att_1_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
144 |
+
MemoryData data_att_1_a0 0 1 att_1_a0 0=768 21=0
|
145 |
+
BinaryOp add_12 2 1 att_1_ma_lora_1 att_1_a0 att_1_ma_lora_2 0=0
|
146 |
+
Sigmoid sigmoid_3 1 1 att_1_ma_lora_2 att_1_a
|
147 |
+
Split split_att_1_a 1 2 att_1_a att_1_a_0 att_1_a_1
|
148 |
+
Gemm gemm_att_1_mg_lora 1 1 att_1_mg att_1_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
149 |
+
Sigmoid sigmoid_4 1 1 att_1_mg_lora att_1_mg_lora_sigmoid
|
150 |
+
Gemm gemm_att_1_gate 1 1 att_1_mg_lora_sigmoid att_1_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
151 |
+
MemoryData data_att_1_k_k 0 1 att_1_k_k 0=768 21=0
|
152 |
+
MemoryData data_att_1_k_a 0 1 att_1_k_a 0=768 21=0
|
153 |
+
MemoryData data_att_1_r_k 0 1 att_1_r_k 0=768 21=0
|
154 |
+
Split split_att_1_key 1 2 att_1_key att_1_key_0 att_1_key_1
|
155 |
+
BinaryOp mul_13 2 1 att_1_key_0 att_1_k_k att_1_key_k 0=2
|
156 |
+
Reshape reshape_12 1 1 att_1_key_k att_1_key_k_reshape 0=64 1=1 2=12
|
157 |
+
Normalize l2norm_1 1 1 att_1_key_k_reshape att_1_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
158 |
+
Reshape reshape_13 1 1 att_1_key_k_norm att_1_key_k_norm_reshape 0=768
|
159 |
+
BinaryOp add_13 1 1 att_1_a_0 att_1_a_0_minus 0=0 1=1 2=-1.0
|
160 |
+
BinaryOp mul_14 2 1 att_1_a_0_minus att_1_k_a att_1_a_minus1_mulka 0=2
|
161 |
+
BinaryOp add_14 1 1 att_1_a_minus1_mulka att_1_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
162 |
+
BinaryOp mul_15 2 1 att_1_a_minus1_mulka_plus1 att_1_key_1 att_1_key_a 0=2
|
163 |
+
Split split_att_1_value 1 2 att_1_value att_1_value_0 att_1_value_1
|
164 |
+
Split split_v_first_0 1 2 v_first_0 att_1_v_first v_first_1
|
165 |
+
BinaryOp minus_0 2 1 att_1_v_first att_1_value_1 att_1_vfirst_minus_value 0=1
|
166 |
+
Gemm gemm_att_1_mv_lora_0 1 1 att_1_mv_1 att_1_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
167 |
+
Gemm gemm_att_1_mv_lora_1 1 1 att_1_mv_lora_0 att_1_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
168 |
+
MemoryData data_att_1_v0 0 1 att_1_v0 0=768 21=0
|
169 |
+
BinaryOp add_15 2 1 att_1_mv_lora_1 att_1_v0 att_1_mv_lora_2 0=0
|
170 |
+
Sigmoid sigmoid_5 1 1 att_1_mv_lora_2 att_1_mv_lora_sigmoid
|
171 |
+
BinaryOp mul_16 2 1 att_1_mv_lora_sigmoid att_1_vfirst_minus_value att_1_vfirst_minus_value_mul 0=2
|
172 |
+
BinaryOp add_16 2 1 att_1_value_0 att_1_vfirst_minus_value_mul att_1_value_final 0=0
|
173 |
+
Gemm gemm_att_1_mw_lora 1 1 att_1_mw att_1_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
174 |
+
TanH tanh_att_1_mw_lora 1 1 att_1_mw_lora att_1_mw_lora_tanh
|
175 |
+
Gemm gemm_att_1_mw_lora_tanh_linear 1 1 att_1_mw_lora_tanh att_1_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
176 |
+
MemoryData data_att_1_td 0 1 att_1_td 0=768 21=0
|
177 |
+
BinaryOp add_17 2 1 att_1_mw_lora_tanh_linear att_1_td att_1_time_decay_pre 0=0
|
178 |
+
Sigmoid sigmoid_6 1 1 att_1_time_decay_pre att_1_time_decay_sigmoid
|
179 |
+
Exp exp_1 1 1 att_1_time_decay_sigmoid att_1_time_decay 1=-0.606531
|
180 |
+
Reshape reshape_14 1 1 att_1_time_decay att_1_time_decay_reshape 0=64 1=1 2=12
|
181 |
+
Split split_att_1_key_a 1 2 att_1_key_a att_1_key_a_0 att_1_key_a_1
|
182 |
+
Reshape reshape_15 1 1 att_1_key_a_0 att_1_key_a_reshape 0=64 1=1 2=12
|
183 |
+
Split split_att_1_value_final 1 3 att_1_value_final att_1_value_final_0 att_1_value_final_1 att_1_value_final_2
|
184 |
+
Reshape reshape_16 1 1 att_1_value_final_0 att_1_value_final_0_reshape 0=64 1=12
|
185 |
+
Reshape reshape_17 1 1 att_1_value_final_1 att_1_value_final_1_reshape 0=1 1=64 2=12
|
186 |
+
MatMul matmul_att_1_vk 2 1 att_1_value_final_1_reshape att_1_key_a_reshape att_1_vk 0=0
|
187 |
+
Split split_att_1_key_k_norm_reshape 1 2 att_1_key_k_norm_reshape att_1_kk_0 att_1_kk_1
|
188 |
+
BinaryOp mul_17 1 1 att_1_kk_0 att_1_kk_0_neg 0=2 1=1 2=-1.0
|
189 |
+
Reshape reshape_18 1 1 att_1_kk_0_neg att_1_kk_0_neg_reshape 0=1 1=64 2=12
|
190 |
+
BinaryOp mul_18 2 1 att_1_kk_1 att_1_a_1 att_1_kk_a 0=2
|
191 |
+
Reshape reshape_19 1 1 att_1_kk_a att_1_kk_a_reshape 0=64 1=1 2=12
|
192 |
+
MatMul matmul_att_1_ab 2 1 att_1_kk_0_neg_reshape att_1_kk_a_reshape att_1_ab 0=0
|
193 |
+
Split split_state_4_in 1 2 state_4_in att_1_state_prev_0 att_1_state_prev_1
|
194 |
+
BinaryOp mul_19 2 1 att_1_state_prev_0 att_1_time_decay_reshape att_1_state_td 0=2
|
195 |
+
MatMul matmul_att_1_sab 2 1 att_1_state_prev_1 att_1_ab att_1_sab 0=0
|
196 |
+
BinaryOp add_18 2 1 att_1_sab att_1_vk att_1_sab_vk 0=0
|
197 |
+
BinaryOp add_19 2 1 att_1_sab_vk att_1_state_td att_1_state_new 0=0
|
198 |
+
Split split_att_1_state_new 1 2 att_1_state_new att_1_state_new_0 state_4_out
|
199 |
+
Split split_att_1_receptance 1 2 att_1_receptance att_1_receptance_0 att_1_receptance_1
|
200 |
+
Reshape reshape_20 1 1 att_1_receptance_0 att_1_receptance_reshape 0=1 1=64 2=12
|
201 |
+
MatMul matmul_att_1_wkv_out 2 1 att_1_state_new_0 att_1_receptance_reshape att_1_wkv_out 0=0
|
202 |
+
Reshape reshape_21 1 1 att_1_wkv_out att_1_wkv_out_flatten 0=768
|
203 |
+
GroupNorm groupnorm_1 1 1 att_1_wkv_out_flatten att_1_x_gn 0=12 1=768 2=0.00064 3=1
|
204 |
+
BinaryOp mul_20 2 1 att_1_receptance_1 att_1_key_a_1 att_1_receptance_key_a 0=2
|
205 |
+
BinaryOp mul_21 2 1 att_1_receptance_key_a att_1_r_k att_1_rk 0=2
|
206 |
+
Reshape reshape_22 1 1 att_1_rk att_1_rk_reshape 0=64 1=12
|
207 |
+
Reduction sum_1 1 1 att_1_rk_reshape att_1_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
208 |
+
BinaryOp mul_22 2 1 att_1_value_final_0_reshape att_1_rk_sum att_1_value_final_0_rk 0=2
|
209 |
+
Reshape reshape_23 1 1 att_1_value_final_0_rk att_1_value_final_0_rk_reshape 0=768
|
210 |
+
BinaryOp add_20 2 1 att_1_value_final_0_rk_reshape att_1_x_gn att_1_x_gn_rkv 0=0
|
211 |
+
BinaryOp mul_23 2 1 att_1_x_gn_rkv att_1_gate att_1_x_gate 0=2
|
212 |
+
Gemm gemm_att_1_x_out 1 1 att_1_x_gate att_1_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
213 |
+
BinaryOp add_21 2 1 att_1_x_out att_1_x_last time_mixing_1_out 0=0
|
214 |
+
Split split_time_mixing_1_out 1 2 time_mixing_1_out ffn_1_x_last ffn_1_x
|
215 |
+
LayerNorm layernorm_3 1 1 ffn_1_x ffn_1_xx 0=768 1=0.00001 2=1
|
216 |
+
Split split_ffn_1_xx 1 3 ffn_1_xx ffn_1_xx_0 ffn_1_xx_1 state_5_out
|
217 |
+
BinaryOp sub_3 2 1 state_5_in ffn_1_xx_0 ffn_1_sx 0=1
|
218 |
+
Split split_ffn_1_sx 1 1 ffn_1_sx ffn_1_sx_0
|
219 |
+
MemoryData data_ffn_1_x_k 0 1 ffn_1_x_k 0=768 21=0
|
220 |
+
BinaryOp mul_24 2 1 ffn_1_sx_0 ffn_1_x_k ffn_1_xk 0=2
|
221 |
+
BinaryOp add_22 2 1 ffn_1_xk ffn_1_xx_1 ffn_1_xxk 0=0
|
222 |
+
Gemm gemm_ffn_1_key 1 1 ffn_1_xxk ffn_1_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
223 |
+
ReLU relu_1 1 1 ffn_1_key ffn_1_key_relu
|
224 |
+
UnaryOp square_ffn_1_key_relu 1 1 ffn_1_key_relu ffn_1_key_relu_square 0=4
|
225 |
+
Gemm gemm_ffn_1_value 1 1 ffn_1_key_relu_square ffn_1_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
226 |
+
BinaryOp add_23 2 1 ffn_1_value ffn_1_x_last channel_mixing_1_out 0=0
|
227 |
+
Split split_channel_mixing_1_out 1 2 channel_mixing_1_out att_2_x_last att_2_x
|
228 |
+
LayerNorm layernorm_4 1 1 att_2_x att_2_xx 0=768 1=0.00001 2=1
|
229 |
+
Split split_att_2_xx 1 3 att_2_xx att_2_xx_0 att_2_xx_1 state_6_out
|
230 |
+
BinaryOp sub_4 2 1 state_6_in att_2_xx_0 att_2_sx 0=1
|
231 |
+
MemoryData data_att_2_maa 0 1 att_2_maa 0=768 1=6 21=0
|
232 |
+
BinaryOp mul_25 2 1 att_2_sx att_2_maa att_2_maa_x 0=2
|
233 |
+
BinaryOp add_24 2 1 att_2_xx_1 att_2_maa_x att_2_maa_rwkvag 0=0
|
234 |
+
Slice slice_att_2_maa_rwkvag 1 6 att_2_maa_rwkvag att_2_mr att_2_mw att_2_mk att_2_mv att_2_ma att_2_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
235 |
+
Split split_att_2_mv 1 2 att_2_mv att_2_mv_0 att_2_mv_1
|
236 |
+
Gemm gemm_att_2_key 1 1 att_2_mk att_2_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
237 |
+
Gemm gemm_att_2_value 1 1 att_2_mv_0 att_2_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
238 |
+
Gemm gemm_att_2_receptance 1 1 att_2_mr att_2_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
239 |
+
Gemm gemm_att_2_ma_lora_0 1 1 att_2_ma att_2_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
240 |
+
Gemm gemm_att_2_ma_lora_1 1 1 att_2_ma_lora_0 att_2_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
241 |
+
MemoryData data_att_2_a0 0 1 att_2_a0 0=768 21=0
|
242 |
+
BinaryOp add_25 2 1 att_2_ma_lora_1 att_2_a0 att_2_ma_lora_2 0=0
|
243 |
+
Sigmoid sigmoid_7 1 1 att_2_ma_lora_2 att_2_a
|
244 |
+
Split split_att_2_a 1 2 att_2_a att_2_a_0 att_2_a_1
|
245 |
+
Gemm gemm_att_2_mg_lora 1 1 att_2_mg att_2_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
246 |
+
Sigmoid sigmoid_8 1 1 att_2_mg_lora att_2_mg_lora_sigmoid
|
247 |
+
Gemm gemm_att_2_gate 1 1 att_2_mg_lora_sigmoid att_2_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
248 |
+
MemoryData data_att_2_k_k 0 1 att_2_k_k 0=768 21=0
|
249 |
+
MemoryData data_att_2_k_a 0 1 att_2_k_a 0=768 21=0
|
250 |
+
MemoryData data_att_2_r_k 0 1 att_2_r_k 0=768 21=0
|
251 |
+
Split split_att_2_key 1 2 att_2_key att_2_key_0 att_2_key_1
|
252 |
+
BinaryOp mul_26 2 1 att_2_key_0 att_2_k_k att_2_key_k 0=2
|
253 |
+
Reshape reshape_24 1 1 att_2_key_k att_2_key_k_reshape 0=64 1=1 2=12
|
254 |
+
Normalize l2norm_2 1 1 att_2_key_k_reshape att_2_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
255 |
+
Reshape reshape_25 1 1 att_2_key_k_norm att_2_key_k_norm_reshape 0=768
|
256 |
+
BinaryOp add_26 1 1 att_2_a_0 att_2_a_0_minus 0=0 1=1 2=-1.0
|
257 |
+
BinaryOp mul_27 2 1 att_2_a_0_minus att_2_k_a att_2_a_minus1_mulka 0=2
|
258 |
+
BinaryOp add_27 1 1 att_2_a_minus1_mulka att_2_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
259 |
+
BinaryOp mul_28 2 1 att_2_a_minus1_mulka_plus1 att_2_key_1 att_2_key_a 0=2
|
260 |
+
Split split_att_2_value 1 2 att_2_value att_2_value_0 att_2_value_1
|
261 |
+
Split split_v_first_1 1 2 v_first_1 att_2_v_first v_first_2
|
262 |
+
BinaryOp minus_1 2 1 att_2_v_first att_2_value_1 att_2_vfirst_minus_value 0=1
|
263 |
+
Gemm gemm_att_2_mv_lora_0 1 1 att_2_mv_1 att_2_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
264 |
+
Gemm gemm_att_2_mv_lora_1 1 1 att_2_mv_lora_0 att_2_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
265 |
+
MemoryData data_att_2_v0 0 1 att_2_v0 0=768 21=0
|
266 |
+
BinaryOp add_28 2 1 att_2_mv_lora_1 att_2_v0 att_2_mv_lora_2 0=0
|
267 |
+
Sigmoid sigmoid_9 1 1 att_2_mv_lora_2 att_2_mv_lora_sigmoid
|
268 |
+
BinaryOp mul_29 2 1 att_2_mv_lora_sigmoid att_2_vfirst_minus_value att_2_vfirst_minus_value_mul 0=2
|
269 |
+
BinaryOp add_29 2 1 att_2_value_0 att_2_vfirst_minus_value_mul att_2_value_final 0=0
|
270 |
+
Gemm gemm_att_2_mw_lora 1 1 att_2_mw att_2_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
271 |
+
TanH tanh_att_2_mw_lora 1 1 att_2_mw_lora att_2_mw_lora_tanh
|
272 |
+
Gemm gemm_att_2_mw_lora_tanh_linear 1 1 att_2_mw_lora_tanh att_2_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
273 |
+
MemoryData data_att_2_td 0 1 att_2_td 0=768 21=0
|
274 |
+
BinaryOp add_30 2 1 att_2_mw_lora_tanh_linear att_2_td att_2_time_decay_pre 0=0
|
275 |
+
Sigmoid sigmoid_10 1 1 att_2_time_decay_pre att_2_time_decay_sigmoid
|
276 |
+
Exp exp_2 1 1 att_2_time_decay_sigmoid att_2_time_decay 1=-0.606531
|
277 |
+
Reshape reshape_26 1 1 att_2_time_decay att_2_time_decay_reshape 0=64 1=1 2=12
|
278 |
+
Split split_att_2_key_a 1 2 att_2_key_a att_2_key_a_0 att_2_key_a_1
|
279 |
+
Reshape reshape_27 1 1 att_2_key_a_0 att_2_key_a_reshape 0=64 1=1 2=12
|
280 |
+
Split split_att_2_value_final 1 3 att_2_value_final att_2_value_final_0 att_2_value_final_1 att_2_value_final_2
|
281 |
+
Reshape reshape_28 1 1 att_2_value_final_0 att_2_value_final_0_reshape 0=64 1=12
|
282 |
+
Reshape reshape_29 1 1 att_2_value_final_1 att_2_value_final_1_reshape 0=1 1=64 2=12
|
283 |
+
MatMul matmul_att_2_vk 2 1 att_2_value_final_1_reshape att_2_key_a_reshape att_2_vk 0=0
|
284 |
+
Split split_att_2_key_k_norm_reshape 1 2 att_2_key_k_norm_reshape att_2_kk_0 att_2_kk_1
|
285 |
+
BinaryOp mul_30 1 1 att_2_kk_0 att_2_kk_0_neg 0=2 1=1 2=-1.0
|
286 |
+
Reshape reshape_30 1 1 att_2_kk_0_neg att_2_kk_0_neg_reshape 0=1 1=64 2=12
|
287 |
+
BinaryOp mul_31 2 1 att_2_kk_1 att_2_a_1 att_2_kk_a 0=2
|
288 |
+
Reshape reshape_31 1 1 att_2_kk_a att_2_kk_a_reshape 0=64 1=1 2=12
|
289 |
+
MatMul matmul_att_2_ab 2 1 att_2_kk_0_neg_reshape att_2_kk_a_reshape att_2_ab 0=0
|
290 |
+
Split split_state_7_in 1 2 state_7_in att_2_state_prev_0 att_2_state_prev_1
|
291 |
+
BinaryOp mul_32 2 1 att_2_state_prev_0 att_2_time_decay_reshape att_2_state_td 0=2
|
292 |
+
MatMul matmul_att_2_sab 2 1 att_2_state_prev_1 att_2_ab att_2_sab 0=0
|
293 |
+
BinaryOp add_31 2 1 att_2_sab att_2_vk att_2_sab_vk 0=0
|
294 |
+
BinaryOp add_32 2 1 att_2_sab_vk att_2_state_td att_2_state_new 0=0
|
295 |
+
Split split_att_2_state_new 1 2 att_2_state_new att_2_state_new_0 state_7_out
|
296 |
+
Split split_att_2_receptance 1 2 att_2_receptance att_2_receptance_0 att_2_receptance_1
|
297 |
+
Reshape reshape_32 1 1 att_2_receptance_0 att_2_receptance_reshape 0=1 1=64 2=12
|
298 |
+
MatMul matmul_att_2_wkv_out 2 1 att_2_state_new_0 att_2_receptance_reshape att_2_wkv_out 0=0
|
299 |
+
Reshape reshape_33 1 1 att_2_wkv_out att_2_wkv_out_flatten 0=768
|
300 |
+
GroupNorm groupnorm_2 1 1 att_2_wkv_out_flatten att_2_x_gn 0=12 1=768 2=0.00064 3=1
|
301 |
+
BinaryOp mul_33 2 1 att_2_receptance_1 att_2_key_a_1 att_2_receptance_key_a 0=2
|
302 |
+
BinaryOp mul_34 2 1 att_2_receptance_key_a att_2_r_k att_2_rk 0=2
|
303 |
+
Reshape reshape_34 1 1 att_2_rk att_2_rk_reshape 0=64 1=12
|
304 |
+
Reduction sum_2 1 1 att_2_rk_reshape att_2_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
305 |
+
BinaryOp mul_35 2 1 att_2_value_final_0_reshape att_2_rk_sum att_2_value_final_0_rk 0=2
|
306 |
+
Reshape reshape_35 1 1 att_2_value_final_0_rk att_2_value_final_0_rk_reshape 0=768
|
307 |
+
BinaryOp add_33 2 1 att_2_value_final_0_rk_reshape att_2_x_gn att_2_x_gn_rkv 0=0
|
308 |
+
BinaryOp mul_36 2 1 att_2_x_gn_rkv att_2_gate att_2_x_gate 0=2
|
309 |
+
Gemm gemm_att_2_x_out 1 1 att_2_x_gate att_2_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
310 |
+
BinaryOp add_34 2 1 att_2_x_out att_2_x_last time_mixing_2_out 0=0
|
311 |
+
Split split_time_mixing_2_out 1 2 time_mixing_2_out ffn_2_x_last ffn_2_x
|
312 |
+
LayerNorm layernorm_5 1 1 ffn_2_x ffn_2_xx 0=768 1=0.00001 2=1
|
313 |
+
Split split_ffn_2_xx 1 3 ffn_2_xx ffn_2_xx_0 ffn_2_xx_1 state_8_out
|
314 |
+
BinaryOp sub_5 2 1 state_8_in ffn_2_xx_0 ffn_2_sx 0=1
|
315 |
+
Split split_ffn_2_sx 1 1 ffn_2_sx ffn_2_sx_0
|
316 |
+
MemoryData data_ffn_2_x_k 0 1 ffn_2_x_k 0=768 21=0
|
317 |
+
BinaryOp mul_37 2 1 ffn_2_sx_0 ffn_2_x_k ffn_2_xk 0=2
|
318 |
+
BinaryOp add_35 2 1 ffn_2_xk ffn_2_xx_1 ffn_2_xxk 0=0
|
319 |
+
Gemm gemm_ffn_2_key 1 1 ffn_2_xxk ffn_2_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
320 |
+
ReLU relu_2 1 1 ffn_2_key ffn_2_key_relu
|
321 |
+
UnaryOp square_ffn_2_key_relu 1 1 ffn_2_key_relu ffn_2_key_relu_square 0=4
|
322 |
+
Gemm gemm_ffn_2_value 1 1 ffn_2_key_relu_square ffn_2_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
323 |
+
BinaryOp add_36 2 1 ffn_2_value ffn_2_x_last channel_mixing_2_out 0=0
|
324 |
+
Split split_channel_mixing_2_out 1 2 channel_mixing_2_out att_3_x_last att_3_x
|
325 |
+
LayerNorm layernorm_6 1 1 att_3_x att_3_xx 0=768 1=0.00001 2=1
|
326 |
+
Split split_att_3_xx 1 3 att_3_xx att_3_xx_0 att_3_xx_1 state_9_out
|
327 |
+
BinaryOp sub_6 2 1 state_9_in att_3_xx_0 att_3_sx 0=1
|
328 |
+
MemoryData data_att_3_maa 0 1 att_3_maa 0=768 1=6 21=0
|
329 |
+
BinaryOp mul_38 2 1 att_3_sx att_3_maa att_3_maa_x 0=2
|
330 |
+
BinaryOp add_37 2 1 att_3_xx_1 att_3_maa_x att_3_maa_rwkvag 0=0
|
331 |
+
Slice slice_att_3_maa_rwkvag 1 6 att_3_maa_rwkvag att_3_mr att_3_mw att_3_mk att_3_mv att_3_ma att_3_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
332 |
+
Split split_att_3_mv 1 2 att_3_mv att_3_mv_0 att_3_mv_1
|
333 |
+
Gemm gemm_att_3_key 1 1 att_3_mk att_3_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
334 |
+
Gemm gemm_att_3_value 1 1 att_3_mv_0 att_3_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
335 |
+
Gemm gemm_att_3_receptance 1 1 att_3_mr att_3_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
336 |
+
Gemm gemm_att_3_ma_lora_0 1 1 att_3_ma att_3_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
337 |
+
Gemm gemm_att_3_ma_lora_1 1 1 att_3_ma_lora_0 att_3_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
338 |
+
MemoryData data_att_3_a0 0 1 att_3_a0 0=768 21=0
|
339 |
+
BinaryOp add_38 2 1 att_3_ma_lora_1 att_3_a0 att_3_ma_lora_2 0=0
|
340 |
+
Sigmoid sigmoid_11 1 1 att_3_ma_lora_2 att_3_a
|
341 |
+
Split split_att_3_a 1 2 att_3_a att_3_a_0 att_3_a_1
|
342 |
+
Gemm gemm_att_3_mg_lora 1 1 att_3_mg att_3_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
343 |
+
Sigmoid sigmoid_12 1 1 att_3_mg_lora att_3_mg_lora_sigmoid
|
344 |
+
Gemm gemm_att_3_gate 1 1 att_3_mg_lora_sigmoid att_3_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
345 |
+
MemoryData data_att_3_k_k 0 1 att_3_k_k 0=768 21=0
|
346 |
+
MemoryData data_att_3_k_a 0 1 att_3_k_a 0=768 21=0
|
347 |
+
MemoryData data_att_3_r_k 0 1 att_3_r_k 0=768 21=0
|
348 |
+
Split split_att_3_key 1 2 att_3_key att_3_key_0 att_3_key_1
|
349 |
+
BinaryOp mul_39 2 1 att_3_key_0 att_3_k_k att_3_key_k 0=2
|
350 |
+
Reshape reshape_36 1 1 att_3_key_k att_3_key_k_reshape 0=64 1=1 2=12
|
351 |
+
Normalize l2norm_3 1 1 att_3_key_k_reshape att_3_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
352 |
+
Reshape reshape_37 1 1 att_3_key_k_norm att_3_key_k_norm_reshape 0=768
|
353 |
+
BinaryOp add_39 1 1 att_3_a_0 att_3_a_0_minus 0=0 1=1 2=-1.0
|
354 |
+
BinaryOp mul_40 2 1 att_3_a_0_minus att_3_k_a att_3_a_minus1_mulka 0=2
|
355 |
+
BinaryOp add_40 1 1 att_3_a_minus1_mulka att_3_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
356 |
+
BinaryOp mul_41 2 1 att_3_a_minus1_mulka_plus1 att_3_key_1 att_3_key_a 0=2
|
357 |
+
Split split_att_3_value 1 2 att_3_value att_3_value_0 att_3_value_1
|
358 |
+
Split split_v_first_2 1 2 v_first_2 att_3_v_first v_first_3
|
359 |
+
BinaryOp minus_2 2 1 att_3_v_first att_3_value_1 att_3_vfirst_minus_value 0=1
|
360 |
+
Gemm gemm_att_3_mv_lora_0 1 1 att_3_mv_1 att_3_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
361 |
+
Gemm gemm_att_3_mv_lora_1 1 1 att_3_mv_lora_0 att_3_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
362 |
+
MemoryData data_att_3_v0 0 1 att_3_v0 0=768 21=0
|
363 |
+
BinaryOp add_41 2 1 att_3_mv_lora_1 att_3_v0 att_3_mv_lora_2 0=0
|
364 |
+
Sigmoid sigmoid_13 1 1 att_3_mv_lora_2 att_3_mv_lora_sigmoid
|
365 |
+
BinaryOp mul_42 2 1 att_3_mv_lora_sigmoid att_3_vfirst_minus_value att_3_vfirst_minus_value_mul 0=2
|
366 |
+
BinaryOp add_42 2 1 att_3_value_0 att_3_vfirst_minus_value_mul att_3_value_final 0=0
|
367 |
+
Gemm gemm_att_3_mw_lora 1 1 att_3_mw att_3_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
368 |
+
TanH tanh_att_3_mw_lora 1 1 att_3_mw_lora att_3_mw_lora_tanh
|
369 |
+
Gemm gemm_att_3_mw_lora_tanh_linear 1 1 att_3_mw_lora_tanh att_3_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
370 |
+
MemoryData data_att_3_td 0 1 att_3_td 0=768 21=0
|
371 |
+
BinaryOp add_43 2 1 att_3_mw_lora_tanh_linear att_3_td att_3_time_decay_pre 0=0
|
372 |
+
Sigmoid sigmoid_14 1 1 att_3_time_decay_pre att_3_time_decay_sigmoid
|
373 |
+
Exp exp_3 1 1 att_3_time_decay_sigmoid att_3_time_decay 1=-0.606531
|
374 |
+
Reshape reshape_38 1 1 att_3_time_decay att_3_time_decay_reshape 0=64 1=1 2=12
|
375 |
+
Split split_att_3_key_a 1 2 att_3_key_a att_3_key_a_0 att_3_key_a_1
|
376 |
+
Reshape reshape_39 1 1 att_3_key_a_0 att_3_key_a_reshape 0=64 1=1 2=12
|
377 |
+
Split split_att_3_value_final 1 3 att_3_value_final att_3_value_final_0 att_3_value_final_1 att_3_value_final_2
|
378 |
+
Reshape reshape_40 1 1 att_3_value_final_0 att_3_value_final_0_reshape 0=64 1=12
|
379 |
+
Reshape reshape_41 1 1 att_3_value_final_1 att_3_value_final_1_reshape 0=1 1=64 2=12
|
380 |
+
MatMul matmul_att_3_vk 2 1 att_3_value_final_1_reshape att_3_key_a_reshape att_3_vk 0=0
|
381 |
+
Split split_att_3_key_k_norm_reshape 1 2 att_3_key_k_norm_reshape att_3_kk_0 att_3_kk_1
|
382 |
+
BinaryOp mul_43 1 1 att_3_kk_0 att_3_kk_0_neg 0=2 1=1 2=-1.0
|
383 |
+
Reshape reshape_42 1 1 att_3_kk_0_neg att_3_kk_0_neg_reshape 0=1 1=64 2=12
|
384 |
+
BinaryOp mul_44 2 1 att_3_kk_1 att_3_a_1 att_3_kk_a 0=2
|
385 |
+
Reshape reshape_43 1 1 att_3_kk_a att_3_kk_a_reshape 0=64 1=1 2=12
|
386 |
+
MatMul matmul_att_3_ab 2 1 att_3_kk_0_neg_reshape att_3_kk_a_reshape att_3_ab 0=0
|
387 |
+
Split split_state_10_in 1 2 state_10_in att_3_state_prev_0 att_3_state_prev_1
|
388 |
+
BinaryOp mul_45 2 1 att_3_state_prev_0 att_3_time_decay_reshape att_3_state_td 0=2
|
389 |
+
MatMul matmul_att_3_sab 2 1 att_3_state_prev_1 att_3_ab att_3_sab 0=0
|
390 |
+
BinaryOp add_44 2 1 att_3_sab att_3_vk att_3_sab_vk 0=0
|
391 |
+
BinaryOp add_45 2 1 att_3_sab_vk att_3_state_td att_3_state_new 0=0
|
392 |
+
Split split_att_3_state_new 1 2 att_3_state_new att_3_state_new_0 state_10_out
|
393 |
+
Split split_att_3_receptance 1 2 att_3_receptance att_3_receptance_0 att_3_receptance_1
|
394 |
+
Reshape reshape_44 1 1 att_3_receptance_0 att_3_receptance_reshape 0=1 1=64 2=12
|
395 |
+
MatMul matmul_att_3_wkv_out 2 1 att_3_state_new_0 att_3_receptance_reshape att_3_wkv_out 0=0
|
396 |
+
Reshape reshape_45 1 1 att_3_wkv_out att_3_wkv_out_flatten 0=768
|
397 |
+
GroupNorm groupnorm_3 1 1 att_3_wkv_out_flatten att_3_x_gn 0=12 1=768 2=0.00064 3=1
|
398 |
+
BinaryOp mul_46 2 1 att_3_receptance_1 att_3_key_a_1 att_3_receptance_key_a 0=2
|
399 |
+
BinaryOp mul_47 2 1 att_3_receptance_key_a att_3_r_k att_3_rk 0=2
|
400 |
+
Reshape reshape_46 1 1 att_3_rk att_3_rk_reshape 0=64 1=12
|
401 |
+
Reduction sum_3 1 1 att_3_rk_reshape att_3_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
402 |
+
BinaryOp mul_48 2 1 att_3_value_final_0_reshape att_3_rk_sum att_3_value_final_0_rk 0=2
|
403 |
+
Reshape reshape_47 1 1 att_3_value_final_0_rk att_3_value_final_0_rk_reshape 0=768
|
404 |
+
BinaryOp add_46 2 1 att_3_value_final_0_rk_reshape att_3_x_gn att_3_x_gn_rkv 0=0
|
405 |
+
BinaryOp mul_49 2 1 att_3_x_gn_rkv att_3_gate att_3_x_gate 0=2
|
406 |
+
Gemm gemm_att_3_x_out 1 1 att_3_x_gate att_3_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
407 |
+
BinaryOp add_47 2 1 att_3_x_out att_3_x_last time_mixing_3_out 0=0
|
408 |
+
Split split_time_mixing_3_out 1 2 time_mixing_3_out ffn_3_x_last ffn_3_x
|
409 |
+
LayerNorm layernorm_7 1 1 ffn_3_x ffn_3_xx 0=768 1=0.00001 2=1
|
410 |
+
Split split_ffn_3_xx 1 3 ffn_3_xx ffn_3_xx_0 ffn_3_xx_1 state_11_out
|
411 |
+
BinaryOp sub_7 2 1 state_11_in ffn_3_xx_0 ffn_3_sx 0=1
|
412 |
+
Split split_ffn_3_sx 1 1 ffn_3_sx ffn_3_sx_0
|
413 |
+
MemoryData data_ffn_3_x_k 0 1 ffn_3_x_k 0=768 21=0
|
414 |
+
BinaryOp mul_50 2 1 ffn_3_sx_0 ffn_3_x_k ffn_3_xk 0=2
|
415 |
+
BinaryOp add_48 2 1 ffn_3_xk ffn_3_xx_1 ffn_3_xxk 0=0
|
416 |
+
Gemm gemm_ffn_3_key 1 1 ffn_3_xxk ffn_3_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
417 |
+
ReLU relu_3 1 1 ffn_3_key ffn_3_key_relu
|
418 |
+
UnaryOp square_ffn_3_key_relu 1 1 ffn_3_key_relu ffn_3_key_relu_square 0=4
|
419 |
+
Gemm gemm_ffn_3_value 1 1 ffn_3_key_relu_square ffn_3_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
420 |
+
BinaryOp add_49 2 1 ffn_3_value ffn_3_x_last channel_mixing_3_out 0=0
|
421 |
+
Split split_channel_mixing_3_out 1 2 channel_mixing_3_out att_4_x_last att_4_x
|
422 |
+
LayerNorm layernorm_8 1 1 att_4_x att_4_xx 0=768 1=0.00001 2=1
|
423 |
+
Split split_att_4_xx 1 3 att_4_xx att_4_xx_0 att_4_xx_1 state_12_out
|
424 |
+
BinaryOp sub_8 2 1 state_12_in att_4_xx_0 att_4_sx 0=1
|
425 |
+
MemoryData data_att_4_maa 0 1 att_4_maa 0=768 1=6 21=0
|
426 |
+
BinaryOp mul_51 2 1 att_4_sx att_4_maa att_4_maa_x 0=2
|
427 |
+
BinaryOp add_50 2 1 att_4_xx_1 att_4_maa_x att_4_maa_rwkvag 0=0
|
428 |
+
Slice slice_att_4_maa_rwkvag 1 6 att_4_maa_rwkvag att_4_mr att_4_mw att_4_mk att_4_mv att_4_ma att_4_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
429 |
+
Split split_att_4_mv 1 2 att_4_mv att_4_mv_0 att_4_mv_1
|
430 |
+
Gemm gemm_att_4_key 1 1 att_4_mk att_4_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
431 |
+
Gemm gemm_att_4_value 1 1 att_4_mv_0 att_4_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
432 |
+
Gemm gemm_att_4_receptance 1 1 att_4_mr att_4_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
433 |
+
Gemm gemm_att_4_ma_lora_0 1 1 att_4_ma att_4_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
434 |
+
Gemm gemm_att_4_ma_lora_1 1 1 att_4_ma_lora_0 att_4_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
435 |
+
MemoryData data_att_4_a0 0 1 att_4_a0 0=768 21=0
|
436 |
+
BinaryOp add_51 2 1 att_4_ma_lora_1 att_4_a0 att_4_ma_lora_2 0=0
|
437 |
+
Sigmoid sigmoid_15 1 1 att_4_ma_lora_2 att_4_a
|
438 |
+
Split split_att_4_a 1 2 att_4_a att_4_a_0 att_4_a_1
|
439 |
+
Gemm gemm_att_4_mg_lora 1 1 att_4_mg att_4_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
440 |
+
Sigmoid sigmoid_16 1 1 att_4_mg_lora att_4_mg_lora_sigmoid
|
441 |
+
Gemm gemm_att_4_gate 1 1 att_4_mg_lora_sigmoid att_4_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
442 |
+
MemoryData data_att_4_k_k 0 1 att_4_k_k 0=768 21=0
|
443 |
+
MemoryData data_att_4_k_a 0 1 att_4_k_a 0=768 21=0
|
444 |
+
MemoryData data_att_4_r_k 0 1 att_4_r_k 0=768 21=0
|
445 |
+
Split split_att_4_key 1 2 att_4_key att_4_key_0 att_4_key_1
|
446 |
+
BinaryOp mul_52 2 1 att_4_key_0 att_4_k_k att_4_key_k 0=2
|
447 |
+
Reshape reshape_48 1 1 att_4_key_k att_4_key_k_reshape 0=64 1=1 2=12
|
448 |
+
Normalize l2norm_4 1 1 att_4_key_k_reshape att_4_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
449 |
+
Reshape reshape_49 1 1 att_4_key_k_norm att_4_key_k_norm_reshape 0=768
|
450 |
+
BinaryOp add_52 1 1 att_4_a_0 att_4_a_0_minus 0=0 1=1 2=-1.0
|
451 |
+
BinaryOp mul_53 2 1 att_4_a_0_minus att_4_k_a att_4_a_minus1_mulka 0=2
|
452 |
+
BinaryOp add_53 1 1 att_4_a_minus1_mulka att_4_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
453 |
+
BinaryOp mul_54 2 1 att_4_a_minus1_mulka_plus1 att_4_key_1 att_4_key_a 0=2
|
454 |
+
Split split_att_4_value 1 2 att_4_value att_4_value_0 att_4_value_1
|
455 |
+
Split split_v_first_3 1 2 v_first_3 att_4_v_first v_first_4
|
456 |
+
BinaryOp minus_3 2 1 att_4_v_first att_4_value_1 att_4_vfirst_minus_value 0=1
|
457 |
+
Gemm gemm_att_4_mv_lora_0 1 1 att_4_mv_1 att_4_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
458 |
+
Gemm gemm_att_4_mv_lora_1 1 1 att_4_mv_lora_0 att_4_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
459 |
+
MemoryData data_att_4_v0 0 1 att_4_v0 0=768 21=0
|
460 |
+
BinaryOp add_54 2 1 att_4_mv_lora_1 att_4_v0 att_4_mv_lora_2 0=0
|
461 |
+
Sigmoid sigmoid_17 1 1 att_4_mv_lora_2 att_4_mv_lora_sigmoid
|
462 |
+
BinaryOp mul_55 2 1 att_4_mv_lora_sigmoid att_4_vfirst_minus_value att_4_vfirst_minus_value_mul 0=2
|
463 |
+
BinaryOp add_55 2 1 att_4_value_0 att_4_vfirst_minus_value_mul att_4_value_final 0=0
|
464 |
+
Gemm gemm_att_4_mw_lora 1 1 att_4_mw att_4_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
465 |
+
TanH tanh_att_4_mw_lora 1 1 att_4_mw_lora att_4_mw_lora_tanh
|
466 |
+
Gemm gemm_att_4_mw_lora_tanh_linear 1 1 att_4_mw_lora_tanh att_4_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
467 |
+
MemoryData data_att_4_td 0 1 att_4_td 0=768 21=0
|
468 |
+
BinaryOp add_56 2 1 att_4_mw_lora_tanh_linear att_4_td att_4_time_decay_pre 0=0
|
469 |
+
Sigmoid sigmoid_18 1 1 att_4_time_decay_pre att_4_time_decay_sigmoid
|
470 |
+
Exp exp_4 1 1 att_4_time_decay_sigmoid att_4_time_decay 1=-0.606531
|
471 |
+
Reshape reshape_50 1 1 att_4_time_decay att_4_time_decay_reshape 0=64 1=1 2=12
|
472 |
+
Split split_att_4_key_a 1 2 att_4_key_a att_4_key_a_0 att_4_key_a_1
|
473 |
+
Reshape reshape_51 1 1 att_4_key_a_0 att_4_key_a_reshape 0=64 1=1 2=12
|
474 |
+
Split split_att_4_value_final 1 3 att_4_value_final att_4_value_final_0 att_4_value_final_1 att_4_value_final_2
|
475 |
+
Reshape reshape_52 1 1 att_4_value_final_0 att_4_value_final_0_reshape 0=64 1=12
|
476 |
+
Reshape reshape_53 1 1 att_4_value_final_1 att_4_value_final_1_reshape 0=1 1=64 2=12
|
477 |
+
MatMul matmul_att_4_vk 2 1 att_4_value_final_1_reshape att_4_key_a_reshape att_4_vk 0=0
|
478 |
+
Split split_att_4_key_k_norm_reshape 1 2 att_4_key_k_norm_reshape att_4_kk_0 att_4_kk_1
|
479 |
+
BinaryOp mul_56 1 1 att_4_kk_0 att_4_kk_0_neg 0=2 1=1 2=-1.0
|
480 |
+
Reshape reshape_54 1 1 att_4_kk_0_neg att_4_kk_0_neg_reshape 0=1 1=64 2=12
|
481 |
+
BinaryOp mul_57 2 1 att_4_kk_1 att_4_a_1 att_4_kk_a 0=2
|
482 |
+
Reshape reshape_55 1 1 att_4_kk_a att_4_kk_a_reshape 0=64 1=1 2=12
|
483 |
+
MatMul matmul_att_4_ab 2 1 att_4_kk_0_neg_reshape att_4_kk_a_reshape att_4_ab 0=0
|
484 |
+
Split split_state_13_in 1 2 state_13_in att_4_state_prev_0 att_4_state_prev_1
|
485 |
+
BinaryOp mul_58 2 1 att_4_state_prev_0 att_4_time_decay_reshape att_4_state_td 0=2
|
486 |
+
MatMul matmul_att_4_sab 2 1 att_4_state_prev_1 att_4_ab att_4_sab 0=0
|
487 |
+
BinaryOp add_57 2 1 att_4_sab att_4_vk att_4_sab_vk 0=0
|
488 |
+
BinaryOp add_58 2 1 att_4_sab_vk att_4_state_td att_4_state_new 0=0
|
489 |
+
Split split_att_4_state_new 1 2 att_4_state_new att_4_state_new_0 state_13_out
|
490 |
+
Split split_att_4_receptance 1 2 att_4_receptance att_4_receptance_0 att_4_receptance_1
|
491 |
+
Reshape reshape_56 1 1 att_4_receptance_0 att_4_receptance_reshape 0=1 1=64 2=12
|
492 |
+
MatMul matmul_att_4_wkv_out 2 1 att_4_state_new_0 att_4_receptance_reshape att_4_wkv_out 0=0
|
493 |
+
Reshape reshape_57 1 1 att_4_wkv_out att_4_wkv_out_flatten 0=768
|
494 |
+
GroupNorm groupnorm_4 1 1 att_4_wkv_out_flatten att_4_x_gn 0=12 1=768 2=0.00064 3=1
|
495 |
+
BinaryOp mul_59 2 1 att_4_receptance_1 att_4_key_a_1 att_4_receptance_key_a 0=2
|
496 |
+
BinaryOp mul_60 2 1 att_4_receptance_key_a att_4_r_k att_4_rk 0=2
|
497 |
+
Reshape reshape_58 1 1 att_4_rk att_4_rk_reshape 0=64 1=12
|
498 |
+
Reduction sum_4 1 1 att_4_rk_reshape att_4_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
499 |
+
BinaryOp mul_61 2 1 att_4_value_final_0_reshape att_4_rk_sum att_4_value_final_0_rk 0=2
|
500 |
+
Reshape reshape_59 1 1 att_4_value_final_0_rk att_4_value_final_0_rk_reshape 0=768
|
501 |
+
BinaryOp add_59 2 1 att_4_value_final_0_rk_reshape att_4_x_gn att_4_x_gn_rkv 0=0
|
502 |
+
BinaryOp mul_62 2 1 att_4_x_gn_rkv att_4_gate att_4_x_gate 0=2
|
503 |
+
Gemm gemm_att_4_x_out 1 1 att_4_x_gate att_4_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
504 |
+
BinaryOp add_60 2 1 att_4_x_out att_4_x_last time_mixing_4_out 0=0
|
505 |
+
Split split_time_mixing_4_out 1 2 time_mixing_4_out ffn_4_x_last ffn_4_x
|
506 |
+
LayerNorm layernorm_9 1 1 ffn_4_x ffn_4_xx 0=768 1=0.00001 2=1
|
507 |
+
Split split_ffn_4_xx 1 3 ffn_4_xx ffn_4_xx_0 ffn_4_xx_1 state_14_out
|
508 |
+
BinaryOp sub_9 2 1 state_14_in ffn_4_xx_0 ffn_4_sx 0=1
|
509 |
+
Split split_ffn_4_sx 1 1 ffn_4_sx ffn_4_sx_0
|
510 |
+
MemoryData data_ffn_4_x_k 0 1 ffn_4_x_k 0=768 21=0
|
511 |
+
BinaryOp mul_63 2 1 ffn_4_sx_0 ffn_4_x_k ffn_4_xk 0=2
|
512 |
+
BinaryOp add_61 2 1 ffn_4_xk ffn_4_xx_1 ffn_4_xxk 0=0
|
513 |
+
Gemm gemm_ffn_4_key 1 1 ffn_4_xxk ffn_4_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
514 |
+
ReLU relu_4 1 1 ffn_4_key ffn_4_key_relu
|
515 |
+
UnaryOp square_ffn_4_key_relu 1 1 ffn_4_key_relu ffn_4_key_relu_square 0=4
|
516 |
+
Gemm gemm_ffn_4_value 1 1 ffn_4_key_relu_square ffn_4_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
517 |
+
BinaryOp add_62 2 1 ffn_4_value ffn_4_x_last channel_mixing_4_out 0=0
|
518 |
+
Split split_channel_mixing_4_out 1 2 channel_mixing_4_out att_5_x_last att_5_x
|
519 |
+
LayerNorm layernorm_10 1 1 att_5_x att_5_xx 0=768 1=0.00001 2=1
|
520 |
+
Split split_att_5_xx 1 3 att_5_xx att_5_xx_0 att_5_xx_1 state_15_out
|
521 |
+
BinaryOp sub_10 2 1 state_15_in att_5_xx_0 att_5_sx 0=1
|
522 |
+
MemoryData data_att_5_maa 0 1 att_5_maa 0=768 1=6 21=0
|
523 |
+
BinaryOp mul_64 2 1 att_5_sx att_5_maa att_5_maa_x 0=2
|
524 |
+
BinaryOp add_63 2 1 att_5_xx_1 att_5_maa_x att_5_maa_rwkvag 0=0
|
525 |
+
Slice slice_att_5_maa_rwkvag 1 6 att_5_maa_rwkvag att_5_mr att_5_mw att_5_mk att_5_mv att_5_ma att_5_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
526 |
+
Split split_att_5_mv 1 2 att_5_mv att_5_mv_0 att_5_mv_1
|
527 |
+
Gemm gemm_att_5_key 1 1 att_5_mk att_5_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
528 |
+
Gemm gemm_att_5_value 1 1 att_5_mv_0 att_5_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
529 |
+
Gemm gemm_att_5_receptance 1 1 att_5_mr att_5_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
530 |
+
Gemm gemm_att_5_ma_lora_0 1 1 att_5_ma att_5_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
531 |
+
Gemm gemm_att_5_ma_lora_1 1 1 att_5_ma_lora_0 att_5_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
532 |
+
MemoryData data_att_5_a0 0 1 att_5_a0 0=768 21=0
|
533 |
+
BinaryOp add_64 2 1 att_5_ma_lora_1 att_5_a0 att_5_ma_lora_2 0=0
|
534 |
+
Sigmoid sigmoid_19 1 1 att_5_ma_lora_2 att_5_a
|
535 |
+
Split split_att_5_a 1 2 att_5_a att_5_a_0 att_5_a_1
|
536 |
+
Gemm gemm_att_5_mg_lora 1 1 att_5_mg att_5_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
537 |
+
Sigmoid sigmoid_20 1 1 att_5_mg_lora att_5_mg_lora_sigmoid
|
538 |
+
Gemm gemm_att_5_gate 1 1 att_5_mg_lora_sigmoid att_5_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
539 |
+
MemoryData data_att_5_k_k 0 1 att_5_k_k 0=768 21=0
|
540 |
+
MemoryData data_att_5_k_a 0 1 att_5_k_a 0=768 21=0
|
541 |
+
MemoryData data_att_5_r_k 0 1 att_5_r_k 0=768 21=0
|
542 |
+
Split split_att_5_key 1 2 att_5_key att_5_key_0 att_5_key_1
|
543 |
+
BinaryOp mul_65 2 1 att_5_key_0 att_5_k_k att_5_key_k 0=2
|
544 |
+
Reshape reshape_60 1 1 att_5_key_k att_5_key_k_reshape 0=64 1=1 2=12
|
545 |
+
Normalize l2norm_5 1 1 att_5_key_k_reshape att_5_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
546 |
+
Reshape reshape_61 1 1 att_5_key_k_norm att_5_key_k_norm_reshape 0=768
|
547 |
+
BinaryOp add_65 1 1 att_5_a_0 att_5_a_0_minus 0=0 1=1 2=-1.0
|
548 |
+
BinaryOp mul_66 2 1 att_5_a_0_minus att_5_k_a att_5_a_minus1_mulka 0=2
|
549 |
+
BinaryOp add_66 1 1 att_5_a_minus1_mulka att_5_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
550 |
+
BinaryOp mul_67 2 1 att_5_a_minus1_mulka_plus1 att_5_key_1 att_5_key_a 0=2
|
551 |
+
Split split_att_5_value 1 2 att_5_value att_5_value_0 att_5_value_1
|
552 |
+
Split split_v_first_4 1 2 v_first_4 att_5_v_first v_first_5
|
553 |
+
BinaryOp minus_4 2 1 att_5_v_first att_5_value_1 att_5_vfirst_minus_value 0=1
|
554 |
+
Gemm gemm_att_5_mv_lora_0 1 1 att_5_mv_1 att_5_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
555 |
+
Gemm gemm_att_5_mv_lora_1 1 1 att_5_mv_lora_0 att_5_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
556 |
+
MemoryData data_att_5_v0 0 1 att_5_v0 0=768 21=0
|
557 |
+
BinaryOp add_67 2 1 att_5_mv_lora_1 att_5_v0 att_5_mv_lora_2 0=0
|
558 |
+
Sigmoid sigmoid_21 1 1 att_5_mv_lora_2 att_5_mv_lora_sigmoid
|
559 |
+
BinaryOp mul_68 2 1 att_5_mv_lora_sigmoid att_5_vfirst_minus_value att_5_vfirst_minus_value_mul 0=2
|
560 |
+
BinaryOp add_68 2 1 att_5_value_0 att_5_vfirst_minus_value_mul att_5_value_final 0=0
|
561 |
+
Gemm gemm_att_5_mw_lora 1 1 att_5_mw att_5_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
562 |
+
TanH tanh_att_5_mw_lora 1 1 att_5_mw_lora att_5_mw_lora_tanh
|
563 |
+
Gemm gemm_att_5_mw_lora_tanh_linear 1 1 att_5_mw_lora_tanh att_5_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
564 |
+
MemoryData data_att_5_td 0 1 att_5_td 0=768 21=0
|
565 |
+
BinaryOp add_69 2 1 att_5_mw_lora_tanh_linear att_5_td att_5_time_decay_pre 0=0
|
566 |
+
Sigmoid sigmoid_22 1 1 att_5_time_decay_pre att_5_time_decay_sigmoid
|
567 |
+
Exp exp_5 1 1 att_5_time_decay_sigmoid att_5_time_decay 1=-0.606531
|
568 |
+
Reshape reshape_62 1 1 att_5_time_decay att_5_time_decay_reshape 0=64 1=1 2=12
|
569 |
+
Split split_att_5_key_a 1 2 att_5_key_a att_5_key_a_0 att_5_key_a_1
|
570 |
+
Reshape reshape_63 1 1 att_5_key_a_0 att_5_key_a_reshape 0=64 1=1 2=12
|
571 |
+
Split split_att_5_value_final 1 3 att_5_value_final att_5_value_final_0 att_5_value_final_1 att_5_value_final_2
|
572 |
+
Reshape reshape_64 1 1 att_5_value_final_0 att_5_value_final_0_reshape 0=64 1=12
|
573 |
+
Reshape reshape_65 1 1 att_5_value_final_1 att_5_value_final_1_reshape 0=1 1=64 2=12
|
574 |
+
MatMul matmul_att_5_vk 2 1 att_5_value_final_1_reshape att_5_key_a_reshape att_5_vk 0=0
|
575 |
+
Split split_att_5_key_k_norm_reshape 1 2 att_5_key_k_norm_reshape att_5_kk_0 att_5_kk_1
|
576 |
+
BinaryOp mul_69 1 1 att_5_kk_0 att_5_kk_0_neg 0=2 1=1 2=-1.0
|
577 |
+
Reshape reshape_66 1 1 att_5_kk_0_neg att_5_kk_0_neg_reshape 0=1 1=64 2=12
|
578 |
+
BinaryOp mul_70 2 1 att_5_kk_1 att_5_a_1 att_5_kk_a 0=2
|
579 |
+
Reshape reshape_67 1 1 att_5_kk_a att_5_kk_a_reshape 0=64 1=1 2=12
|
580 |
+
MatMul matmul_att_5_ab 2 1 att_5_kk_0_neg_reshape att_5_kk_a_reshape att_5_ab 0=0
|
581 |
+
Split split_state_16_in 1 2 state_16_in att_5_state_prev_0 att_5_state_prev_1
|
582 |
+
BinaryOp mul_71 2 1 att_5_state_prev_0 att_5_time_decay_reshape att_5_state_td 0=2
|
583 |
+
MatMul matmul_att_5_sab 2 1 att_5_state_prev_1 att_5_ab att_5_sab 0=0
|
584 |
+
BinaryOp add_70 2 1 att_5_sab att_5_vk att_5_sab_vk 0=0
|
585 |
+
BinaryOp add_71 2 1 att_5_sab_vk att_5_state_td att_5_state_new 0=0
|
586 |
+
Split split_att_5_state_new 1 2 att_5_state_new att_5_state_new_0 state_16_out
|
587 |
+
Split split_att_5_receptance 1 2 att_5_receptance att_5_receptance_0 att_5_receptance_1
|
588 |
+
Reshape reshape_68 1 1 att_5_receptance_0 att_5_receptance_reshape 0=1 1=64 2=12
|
589 |
+
MatMul matmul_att_5_wkv_out 2 1 att_5_state_new_0 att_5_receptance_reshape att_5_wkv_out 0=0
|
590 |
+
Reshape reshape_69 1 1 att_5_wkv_out att_5_wkv_out_flatten 0=768
|
591 |
+
GroupNorm groupnorm_5 1 1 att_5_wkv_out_flatten att_5_x_gn 0=12 1=768 2=0.00064 3=1
|
592 |
+
BinaryOp mul_72 2 1 att_5_receptance_1 att_5_key_a_1 att_5_receptance_key_a 0=2
|
593 |
+
BinaryOp mul_73 2 1 att_5_receptance_key_a att_5_r_k att_5_rk 0=2
|
594 |
+
Reshape reshape_70 1 1 att_5_rk att_5_rk_reshape 0=64 1=12
|
595 |
+
Reduction sum_5 1 1 att_5_rk_reshape att_5_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
596 |
+
BinaryOp mul_74 2 1 att_5_value_final_0_reshape att_5_rk_sum att_5_value_final_0_rk 0=2
|
597 |
+
Reshape reshape_71 1 1 att_5_value_final_0_rk att_5_value_final_0_rk_reshape 0=768
|
598 |
+
BinaryOp add_72 2 1 att_5_value_final_0_rk_reshape att_5_x_gn att_5_x_gn_rkv 0=0
|
599 |
+
BinaryOp mul_75 2 1 att_5_x_gn_rkv att_5_gate att_5_x_gate 0=2
|
600 |
+
Gemm gemm_att_5_x_out 1 1 att_5_x_gate att_5_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
601 |
+
BinaryOp add_73 2 1 att_5_x_out att_5_x_last time_mixing_5_out 0=0
|
602 |
+
Split split_time_mixing_5_out 1 2 time_mixing_5_out ffn_5_x_last ffn_5_x
|
603 |
+
LayerNorm layernorm_11 1 1 ffn_5_x ffn_5_xx 0=768 1=0.00001 2=1
|
604 |
+
Split split_ffn_5_xx 1 3 ffn_5_xx ffn_5_xx_0 ffn_5_xx_1 state_17_out
|
605 |
+
BinaryOp sub_11 2 1 state_17_in ffn_5_xx_0 ffn_5_sx 0=1
|
606 |
+
Split split_ffn_5_sx 1 1 ffn_5_sx ffn_5_sx_0
|
607 |
+
MemoryData data_ffn_5_x_k 0 1 ffn_5_x_k 0=768 21=0
|
608 |
+
BinaryOp mul_76 2 1 ffn_5_sx_0 ffn_5_x_k ffn_5_xk 0=2
|
609 |
+
BinaryOp add_74 2 1 ffn_5_xk ffn_5_xx_1 ffn_5_xxk 0=0
|
610 |
+
Gemm gemm_ffn_5_key 1 1 ffn_5_xxk ffn_5_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
611 |
+
ReLU relu_5 1 1 ffn_5_key ffn_5_key_relu
|
612 |
+
UnaryOp square_ffn_5_key_relu 1 1 ffn_5_key_relu ffn_5_key_relu_square 0=4
|
613 |
+
Gemm gemm_ffn_5_value 1 1 ffn_5_key_relu_square ffn_5_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
614 |
+
BinaryOp add_75 2 1 ffn_5_value ffn_5_x_last channel_mixing_5_out 0=0
|
615 |
+
Split split_channel_mixing_5_out 1 2 channel_mixing_5_out att_6_x_last att_6_x
|
616 |
+
LayerNorm layernorm_12 1 1 att_6_x att_6_xx 0=768 1=0.00001 2=1
|
617 |
+
Split split_att_6_xx 1 3 att_6_xx att_6_xx_0 att_6_xx_1 state_18_out
|
618 |
+
BinaryOp sub_12 2 1 state_18_in att_6_xx_0 att_6_sx 0=1
|
619 |
+
MemoryData data_att_6_maa 0 1 att_6_maa 0=768 1=6 21=0
|
620 |
+
BinaryOp mul_77 2 1 att_6_sx att_6_maa att_6_maa_x 0=2
|
621 |
+
BinaryOp add_76 2 1 att_6_xx_1 att_6_maa_x att_6_maa_rwkvag 0=0
|
622 |
+
Slice slice_att_6_maa_rwkvag 1 6 att_6_maa_rwkvag att_6_mr att_6_mw att_6_mk att_6_mv att_6_ma att_6_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
623 |
+
Split split_att_6_mv 1 2 att_6_mv att_6_mv_0 att_6_mv_1
|
624 |
+
Gemm gemm_att_6_key 1 1 att_6_mk att_6_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
625 |
+
Gemm gemm_att_6_value 1 1 att_6_mv_0 att_6_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
626 |
+
Gemm gemm_att_6_receptance 1 1 att_6_mr att_6_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
627 |
+
Gemm gemm_att_6_ma_lora_0 1 1 att_6_ma att_6_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
628 |
+
Gemm gemm_att_6_ma_lora_1 1 1 att_6_ma_lora_0 att_6_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
629 |
+
MemoryData data_att_6_a0 0 1 att_6_a0 0=768 21=0
|
630 |
+
BinaryOp add_77 2 1 att_6_ma_lora_1 att_6_a0 att_6_ma_lora_2 0=0
|
631 |
+
Sigmoid sigmoid_23 1 1 att_6_ma_lora_2 att_6_a
|
632 |
+
Split split_att_6_a 1 2 att_6_a att_6_a_0 att_6_a_1
|
633 |
+
Gemm gemm_att_6_mg_lora 1 1 att_6_mg att_6_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
634 |
+
Sigmoid sigmoid_24 1 1 att_6_mg_lora att_6_mg_lora_sigmoid
|
635 |
+
Gemm gemm_att_6_gate 1 1 att_6_mg_lora_sigmoid att_6_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
636 |
+
MemoryData data_att_6_k_k 0 1 att_6_k_k 0=768 21=0
|
637 |
+
MemoryData data_att_6_k_a 0 1 att_6_k_a 0=768 21=0
|
638 |
+
MemoryData data_att_6_r_k 0 1 att_6_r_k 0=768 21=0
|
639 |
+
Split split_att_6_key 1 2 att_6_key att_6_key_0 att_6_key_1
|
640 |
+
BinaryOp mul_78 2 1 att_6_key_0 att_6_k_k att_6_key_k 0=2
|
641 |
+
Reshape reshape_72 1 1 att_6_key_k att_6_key_k_reshape 0=64 1=1 2=12
|
642 |
+
Normalize l2norm_6 1 1 att_6_key_k_reshape att_6_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
643 |
+
Reshape reshape_73 1 1 att_6_key_k_norm att_6_key_k_norm_reshape 0=768
|
644 |
+
BinaryOp add_78 1 1 att_6_a_0 att_6_a_0_minus 0=0 1=1 2=-1.0
|
645 |
+
BinaryOp mul_79 2 1 att_6_a_0_minus att_6_k_a att_6_a_minus1_mulka 0=2
|
646 |
+
BinaryOp add_79 1 1 att_6_a_minus1_mulka att_6_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
647 |
+
BinaryOp mul_80 2 1 att_6_a_minus1_mulka_plus1 att_6_key_1 att_6_key_a 0=2
|
648 |
+
Split split_att_6_value 1 2 att_6_value att_6_value_0 att_6_value_1
|
649 |
+
Split split_v_first_5 1 2 v_first_5 att_6_v_first v_first_6
|
650 |
+
BinaryOp minus_5 2 1 att_6_v_first att_6_value_1 att_6_vfirst_minus_value 0=1
|
651 |
+
Gemm gemm_att_6_mv_lora_0 1 1 att_6_mv_1 att_6_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
652 |
+
Gemm gemm_att_6_mv_lora_1 1 1 att_6_mv_lora_0 att_6_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
653 |
+
MemoryData data_att_6_v0 0 1 att_6_v0 0=768 21=0
|
654 |
+
BinaryOp add_80 2 1 att_6_mv_lora_1 att_6_v0 att_6_mv_lora_2 0=0
|
655 |
+
Sigmoid sigmoid_25 1 1 att_6_mv_lora_2 att_6_mv_lora_sigmoid
|
656 |
+
BinaryOp mul_81 2 1 att_6_mv_lora_sigmoid att_6_vfirst_minus_value att_6_vfirst_minus_value_mul 0=2
|
657 |
+
BinaryOp add_81 2 1 att_6_value_0 att_6_vfirst_minus_value_mul att_6_value_final 0=0
|
658 |
+
Gemm gemm_att_6_mw_lora 1 1 att_6_mw att_6_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
659 |
+
TanH tanh_att_6_mw_lora 1 1 att_6_mw_lora att_6_mw_lora_tanh
|
660 |
+
Gemm gemm_att_6_mw_lora_tanh_linear 1 1 att_6_mw_lora_tanh att_6_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
661 |
+
MemoryData data_att_6_td 0 1 att_6_td 0=768 21=0
|
662 |
+
BinaryOp add_82 2 1 att_6_mw_lora_tanh_linear att_6_td att_6_time_decay_pre 0=0
|
663 |
+
Sigmoid sigmoid_26 1 1 att_6_time_decay_pre att_6_time_decay_sigmoid
|
664 |
+
Exp exp_6 1 1 att_6_time_decay_sigmoid att_6_time_decay 1=-0.606531
|
665 |
+
Reshape reshape_74 1 1 att_6_time_decay att_6_time_decay_reshape 0=64 1=1 2=12
|
666 |
+
Split split_att_6_key_a 1 2 att_6_key_a att_6_key_a_0 att_6_key_a_1
|
667 |
+
Reshape reshape_75 1 1 att_6_key_a_0 att_6_key_a_reshape 0=64 1=1 2=12
|
668 |
+
Split split_att_6_value_final 1 3 att_6_value_final att_6_value_final_0 att_6_value_final_1 att_6_value_final_2
|
669 |
+
Reshape reshape_76 1 1 att_6_value_final_0 att_6_value_final_0_reshape 0=64 1=12
|
670 |
+
Reshape reshape_77 1 1 att_6_value_final_1 att_6_value_final_1_reshape 0=1 1=64 2=12
|
671 |
+
MatMul matmul_att_6_vk 2 1 att_6_value_final_1_reshape att_6_key_a_reshape att_6_vk 0=0
|
672 |
+
Split split_att_6_key_k_norm_reshape 1 2 att_6_key_k_norm_reshape att_6_kk_0 att_6_kk_1
|
673 |
+
BinaryOp mul_82 1 1 att_6_kk_0 att_6_kk_0_neg 0=2 1=1 2=-1.0
|
674 |
+
Reshape reshape_78 1 1 att_6_kk_0_neg att_6_kk_0_neg_reshape 0=1 1=64 2=12
|
675 |
+
BinaryOp mul_83 2 1 att_6_kk_1 att_6_a_1 att_6_kk_a 0=2
|
676 |
+
Reshape reshape_79 1 1 att_6_kk_a att_6_kk_a_reshape 0=64 1=1 2=12
|
677 |
+
MatMul matmul_att_6_ab 2 1 att_6_kk_0_neg_reshape att_6_kk_a_reshape att_6_ab 0=0
|
678 |
+
Split split_state_19_in 1 2 state_19_in att_6_state_prev_0 att_6_state_prev_1
|
679 |
+
BinaryOp mul_84 2 1 att_6_state_prev_0 att_6_time_decay_reshape att_6_state_td 0=2
|
680 |
+
MatMul matmul_att_6_sab 2 1 att_6_state_prev_1 att_6_ab att_6_sab 0=0
|
681 |
+
BinaryOp add_83 2 1 att_6_sab att_6_vk att_6_sab_vk 0=0
|
682 |
+
BinaryOp add_84 2 1 att_6_sab_vk att_6_state_td att_6_state_new 0=0
|
683 |
+
Split split_att_6_state_new 1 2 att_6_state_new att_6_state_new_0 state_19_out
|
684 |
+
Split split_att_6_receptance 1 2 att_6_receptance att_6_receptance_0 att_6_receptance_1
|
685 |
+
Reshape reshape_80 1 1 att_6_receptance_0 att_6_receptance_reshape 0=1 1=64 2=12
|
686 |
+
MatMul matmul_att_6_wkv_out 2 1 att_6_state_new_0 att_6_receptance_reshape att_6_wkv_out 0=0
|
687 |
+
Reshape reshape_81 1 1 att_6_wkv_out att_6_wkv_out_flatten 0=768
|
688 |
+
GroupNorm groupnorm_6 1 1 att_6_wkv_out_flatten att_6_x_gn 0=12 1=768 2=0.00064 3=1
|
689 |
+
BinaryOp mul_85 2 1 att_6_receptance_1 att_6_key_a_1 att_6_receptance_key_a 0=2
|
690 |
+
BinaryOp mul_86 2 1 att_6_receptance_key_a att_6_r_k att_6_rk 0=2
|
691 |
+
Reshape reshape_82 1 1 att_6_rk att_6_rk_reshape 0=64 1=12
|
692 |
+
Reduction sum_6 1 1 att_6_rk_reshape att_6_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
693 |
+
BinaryOp mul_87 2 1 att_6_value_final_0_reshape att_6_rk_sum att_6_value_final_0_rk 0=2
|
694 |
+
Reshape reshape_83 1 1 att_6_value_final_0_rk att_6_value_final_0_rk_reshape 0=768
|
695 |
+
BinaryOp add_85 2 1 att_6_value_final_0_rk_reshape att_6_x_gn att_6_x_gn_rkv 0=0
|
696 |
+
BinaryOp mul_88 2 1 att_6_x_gn_rkv att_6_gate att_6_x_gate 0=2
|
697 |
+
Gemm gemm_att_6_x_out 1 1 att_6_x_gate att_6_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
698 |
+
BinaryOp add_86 2 1 att_6_x_out att_6_x_last time_mixing_6_out 0=0
|
699 |
+
Split split_time_mixing_6_out 1 2 time_mixing_6_out ffn_6_x_last ffn_6_x
|
700 |
+
LayerNorm layernorm_13 1 1 ffn_6_x ffn_6_xx 0=768 1=0.00001 2=1
|
701 |
+
Split split_ffn_6_xx 1 3 ffn_6_xx ffn_6_xx_0 ffn_6_xx_1 state_20_out
|
702 |
+
BinaryOp sub_13 2 1 state_20_in ffn_6_xx_0 ffn_6_sx 0=1
|
703 |
+
Split split_ffn_6_sx 1 1 ffn_6_sx ffn_6_sx_0
|
704 |
+
MemoryData data_ffn_6_x_k 0 1 ffn_6_x_k 0=768 21=0
|
705 |
+
BinaryOp mul_89 2 1 ffn_6_sx_0 ffn_6_x_k ffn_6_xk 0=2
|
706 |
+
BinaryOp add_87 2 1 ffn_6_xk ffn_6_xx_1 ffn_6_xxk 0=0
|
707 |
+
Gemm gemm_ffn_6_key 1 1 ffn_6_xxk ffn_6_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
708 |
+
ReLU relu_6 1 1 ffn_6_key ffn_6_key_relu
|
709 |
+
UnaryOp square_ffn_6_key_relu 1 1 ffn_6_key_relu ffn_6_key_relu_square 0=4
|
710 |
+
Gemm gemm_ffn_6_value 1 1 ffn_6_key_relu_square ffn_6_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
711 |
+
BinaryOp add_88 2 1 ffn_6_value ffn_6_x_last channel_mixing_6_out 0=0
|
712 |
+
Split split_channel_mixing_6_out 1 2 channel_mixing_6_out att_7_x_last att_7_x
|
713 |
+
LayerNorm layernorm_14 1 1 att_7_x att_7_xx 0=768 1=0.00001 2=1
|
714 |
+
Split split_att_7_xx 1 3 att_7_xx att_7_xx_0 att_7_xx_1 state_21_out
|
715 |
+
BinaryOp sub_14 2 1 state_21_in att_7_xx_0 att_7_sx 0=1
|
716 |
+
MemoryData data_att_7_maa 0 1 att_7_maa 0=768 1=6 21=0
|
717 |
+
BinaryOp mul_90 2 1 att_7_sx att_7_maa att_7_maa_x 0=2
|
718 |
+
BinaryOp add_89 2 1 att_7_xx_1 att_7_maa_x att_7_maa_rwkvag 0=0
|
719 |
+
Slice slice_att_7_maa_rwkvag 1 6 att_7_maa_rwkvag att_7_mr att_7_mw att_7_mk att_7_mv att_7_ma att_7_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
720 |
+
Split split_att_7_mv 1 2 att_7_mv att_7_mv_0 att_7_mv_1
|
721 |
+
Gemm gemm_att_7_key 1 1 att_7_mk att_7_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
722 |
+
Gemm gemm_att_7_value 1 1 att_7_mv_0 att_7_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
723 |
+
Gemm gemm_att_7_receptance 1 1 att_7_mr att_7_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
724 |
+
Gemm gemm_att_7_ma_lora_0 1 1 att_7_ma att_7_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
725 |
+
Gemm gemm_att_7_ma_lora_1 1 1 att_7_ma_lora_0 att_7_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
726 |
+
MemoryData data_att_7_a0 0 1 att_7_a0 0=768 21=0
|
727 |
+
BinaryOp add_90 2 1 att_7_ma_lora_1 att_7_a0 att_7_ma_lora_2 0=0
|
728 |
+
Sigmoid sigmoid_27 1 1 att_7_ma_lora_2 att_7_a
|
729 |
+
Split split_att_7_a 1 2 att_7_a att_7_a_0 att_7_a_1
|
730 |
+
Gemm gemm_att_7_mg_lora 1 1 att_7_mg att_7_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
731 |
+
Sigmoid sigmoid_28 1 1 att_7_mg_lora att_7_mg_lora_sigmoid
|
732 |
+
Gemm gemm_att_7_gate 1 1 att_7_mg_lora_sigmoid att_7_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
733 |
+
MemoryData data_att_7_k_k 0 1 att_7_k_k 0=768 21=0
|
734 |
+
MemoryData data_att_7_k_a 0 1 att_7_k_a 0=768 21=0
|
735 |
+
MemoryData data_att_7_r_k 0 1 att_7_r_k 0=768 21=0
|
736 |
+
Split split_att_7_key 1 2 att_7_key att_7_key_0 att_7_key_1
|
737 |
+
BinaryOp mul_91 2 1 att_7_key_0 att_7_k_k att_7_key_k 0=2
|
738 |
+
Reshape reshape_84 1 1 att_7_key_k att_7_key_k_reshape 0=64 1=1 2=12
|
739 |
+
Normalize l2norm_7 1 1 att_7_key_k_reshape att_7_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
740 |
+
Reshape reshape_85 1 1 att_7_key_k_norm att_7_key_k_norm_reshape 0=768
|
741 |
+
BinaryOp add_91 1 1 att_7_a_0 att_7_a_0_minus 0=0 1=1 2=-1.0
|
742 |
+
BinaryOp mul_92 2 1 att_7_a_0_minus att_7_k_a att_7_a_minus1_mulka 0=2
|
743 |
+
BinaryOp add_92 1 1 att_7_a_minus1_mulka att_7_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
744 |
+
BinaryOp mul_93 2 1 att_7_a_minus1_mulka_plus1 att_7_key_1 att_7_key_a 0=2
|
745 |
+
Split split_att_7_value 1 2 att_7_value att_7_value_0 att_7_value_1
|
746 |
+
Split split_v_first_6 1 2 v_first_6 att_7_v_first v_first_7
|
747 |
+
BinaryOp minus_6 2 1 att_7_v_first att_7_value_1 att_7_vfirst_minus_value 0=1
|
748 |
+
Gemm gemm_att_7_mv_lora_0 1 1 att_7_mv_1 att_7_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
749 |
+
Gemm gemm_att_7_mv_lora_1 1 1 att_7_mv_lora_0 att_7_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
750 |
+
MemoryData data_att_7_v0 0 1 att_7_v0 0=768 21=0
|
751 |
+
BinaryOp add_93 2 1 att_7_mv_lora_1 att_7_v0 att_7_mv_lora_2 0=0
|
752 |
+
Sigmoid sigmoid_29 1 1 att_7_mv_lora_2 att_7_mv_lora_sigmoid
|
753 |
+
BinaryOp mul_94 2 1 att_7_mv_lora_sigmoid att_7_vfirst_minus_value att_7_vfirst_minus_value_mul 0=2
|
754 |
+
BinaryOp add_94 2 1 att_7_value_0 att_7_vfirst_minus_value_mul att_7_value_final 0=0
|
755 |
+
Gemm gemm_att_7_mw_lora 1 1 att_7_mw att_7_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
756 |
+
TanH tanh_att_7_mw_lora 1 1 att_7_mw_lora att_7_mw_lora_tanh
|
757 |
+
Gemm gemm_att_7_mw_lora_tanh_linear 1 1 att_7_mw_lora_tanh att_7_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
758 |
+
MemoryData data_att_7_td 0 1 att_7_td 0=768 21=0
|
759 |
+
BinaryOp add_95 2 1 att_7_mw_lora_tanh_linear att_7_td att_7_time_decay_pre 0=0
|
760 |
+
Sigmoid sigmoid_30 1 1 att_7_time_decay_pre att_7_time_decay_sigmoid
|
761 |
+
Exp exp_7 1 1 att_7_time_decay_sigmoid att_7_time_decay 1=-0.606531
|
762 |
+
Reshape reshape_86 1 1 att_7_time_decay att_7_time_decay_reshape 0=64 1=1 2=12
|
763 |
+
Split split_att_7_key_a 1 2 att_7_key_a att_7_key_a_0 att_7_key_a_1
|
764 |
+
Reshape reshape_87 1 1 att_7_key_a_0 att_7_key_a_reshape 0=64 1=1 2=12
|
765 |
+
Split split_att_7_value_final 1 3 att_7_value_final att_7_value_final_0 att_7_value_final_1 att_7_value_final_2
|
766 |
+
Reshape reshape_88 1 1 att_7_value_final_0 att_7_value_final_0_reshape 0=64 1=12
|
767 |
+
Reshape reshape_89 1 1 att_7_value_final_1 att_7_value_final_1_reshape 0=1 1=64 2=12
|
768 |
+
MatMul matmul_att_7_vk 2 1 att_7_value_final_1_reshape att_7_key_a_reshape att_7_vk 0=0
|
769 |
+
Split split_att_7_key_k_norm_reshape 1 2 att_7_key_k_norm_reshape att_7_kk_0 att_7_kk_1
|
770 |
+
BinaryOp mul_95 1 1 att_7_kk_0 att_7_kk_0_neg 0=2 1=1 2=-1.0
|
771 |
+
Reshape reshape_90 1 1 att_7_kk_0_neg att_7_kk_0_neg_reshape 0=1 1=64 2=12
|
772 |
+
BinaryOp mul_96 2 1 att_7_kk_1 att_7_a_1 att_7_kk_a 0=2
|
773 |
+
Reshape reshape_91 1 1 att_7_kk_a att_7_kk_a_reshape 0=64 1=1 2=12
|
774 |
+
MatMul matmul_att_7_ab 2 1 att_7_kk_0_neg_reshape att_7_kk_a_reshape att_7_ab 0=0
|
775 |
+
Split split_state_22_in 1 2 state_22_in att_7_state_prev_0 att_7_state_prev_1
|
776 |
+
BinaryOp mul_97 2 1 att_7_state_prev_0 att_7_time_decay_reshape att_7_state_td 0=2
|
777 |
+
MatMul matmul_att_7_sab 2 1 att_7_state_prev_1 att_7_ab att_7_sab 0=0
|
778 |
+
BinaryOp add_96 2 1 att_7_sab att_7_vk att_7_sab_vk 0=0
|
779 |
+
BinaryOp add_97 2 1 att_7_sab_vk att_7_state_td att_7_state_new 0=0
|
780 |
+
Split split_att_7_state_new 1 2 att_7_state_new att_7_state_new_0 state_22_out
|
781 |
+
Split split_att_7_receptance 1 2 att_7_receptance att_7_receptance_0 att_7_receptance_1
|
782 |
+
Reshape reshape_92 1 1 att_7_receptance_0 att_7_receptance_reshape 0=1 1=64 2=12
|
783 |
+
MatMul matmul_att_7_wkv_out 2 1 att_7_state_new_0 att_7_receptance_reshape att_7_wkv_out 0=0
|
784 |
+
Reshape reshape_93 1 1 att_7_wkv_out att_7_wkv_out_flatten 0=768
|
785 |
+
GroupNorm groupnorm_7 1 1 att_7_wkv_out_flatten att_7_x_gn 0=12 1=768 2=0.00064 3=1
|
786 |
+
BinaryOp mul_98 2 1 att_7_receptance_1 att_7_key_a_1 att_7_receptance_key_a 0=2
|
787 |
+
BinaryOp mul_99 2 1 att_7_receptance_key_a att_7_r_k att_7_rk 0=2
|
788 |
+
Reshape reshape_94 1 1 att_7_rk att_7_rk_reshape 0=64 1=12
|
789 |
+
Reduction sum_7 1 1 att_7_rk_reshape att_7_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
790 |
+
BinaryOp mul_100 2 1 att_7_value_final_0_reshape att_7_rk_sum att_7_value_final_0_rk 0=2
|
791 |
+
Reshape reshape_95 1 1 att_7_value_final_0_rk att_7_value_final_0_rk_reshape 0=768
|
792 |
+
BinaryOp add_98 2 1 att_7_value_final_0_rk_reshape att_7_x_gn att_7_x_gn_rkv 0=0
|
793 |
+
BinaryOp mul_101 2 1 att_7_x_gn_rkv att_7_gate att_7_x_gate 0=2
|
794 |
+
Gemm gemm_att_7_x_out 1 1 att_7_x_gate att_7_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
795 |
+
BinaryOp add_99 2 1 att_7_x_out att_7_x_last time_mixing_7_out 0=0
|
796 |
+
Split split_time_mixing_7_out 1 2 time_mixing_7_out ffn_7_x_last ffn_7_x
|
797 |
+
LayerNorm layernorm_15 1 1 ffn_7_x ffn_7_xx 0=768 1=0.00001 2=1
|
798 |
+
Split split_ffn_7_xx 1 3 ffn_7_xx ffn_7_xx_0 ffn_7_xx_1 state_23_out
|
799 |
+
BinaryOp sub_15 2 1 state_23_in ffn_7_xx_0 ffn_7_sx 0=1
|
800 |
+
Split split_ffn_7_sx 1 1 ffn_7_sx ffn_7_sx_0
|
801 |
+
MemoryData data_ffn_7_x_k 0 1 ffn_7_x_k 0=768 21=0
|
802 |
+
BinaryOp mul_102 2 1 ffn_7_sx_0 ffn_7_x_k ffn_7_xk 0=2
|
803 |
+
BinaryOp add_100 2 1 ffn_7_xk ffn_7_xx_1 ffn_7_xxk 0=0
|
804 |
+
Gemm gemm_ffn_7_key 1 1 ffn_7_xxk ffn_7_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
805 |
+
ReLU relu_7 1 1 ffn_7_key ffn_7_key_relu
|
806 |
+
UnaryOp square_ffn_7_key_relu 1 1 ffn_7_key_relu ffn_7_key_relu_square 0=4
|
807 |
+
Gemm gemm_ffn_7_value 1 1 ffn_7_key_relu_square ffn_7_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
808 |
+
BinaryOp add_101 2 1 ffn_7_value ffn_7_x_last channel_mixing_7_out 0=0
|
809 |
+
Split split_channel_mixing_7_out 1 2 channel_mixing_7_out att_8_x_last att_8_x
|
810 |
+
LayerNorm layernorm_16 1 1 att_8_x att_8_xx 0=768 1=0.00001 2=1
|
811 |
+
Split split_att_8_xx 1 3 att_8_xx att_8_xx_0 att_8_xx_1 state_24_out
|
812 |
+
BinaryOp sub_16 2 1 state_24_in att_8_xx_0 att_8_sx 0=1
|
813 |
+
MemoryData data_att_8_maa 0 1 att_8_maa 0=768 1=6 21=0
|
814 |
+
BinaryOp mul_103 2 1 att_8_sx att_8_maa att_8_maa_x 0=2
|
815 |
+
BinaryOp add_102 2 1 att_8_xx_1 att_8_maa_x att_8_maa_rwkvag 0=0
|
816 |
+
Slice slice_att_8_maa_rwkvag 1 6 att_8_maa_rwkvag att_8_mr att_8_mw att_8_mk att_8_mv att_8_ma att_8_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
817 |
+
Split split_att_8_mv 1 2 att_8_mv att_8_mv_0 att_8_mv_1
|
818 |
+
Gemm gemm_att_8_key 1 1 att_8_mk att_8_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
819 |
+
Gemm gemm_att_8_value 1 1 att_8_mv_0 att_8_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
820 |
+
Gemm gemm_att_8_receptance 1 1 att_8_mr att_8_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
821 |
+
Gemm gemm_att_8_ma_lora_0 1 1 att_8_ma att_8_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
822 |
+
Gemm gemm_att_8_ma_lora_1 1 1 att_8_ma_lora_0 att_8_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
823 |
+
MemoryData data_att_8_a0 0 1 att_8_a0 0=768 21=0
|
824 |
+
BinaryOp add_103 2 1 att_8_ma_lora_1 att_8_a0 att_8_ma_lora_2 0=0
|
825 |
+
Sigmoid sigmoid_31 1 1 att_8_ma_lora_2 att_8_a
|
826 |
+
Split split_att_8_a 1 2 att_8_a att_8_a_0 att_8_a_1
|
827 |
+
Gemm gemm_att_8_mg_lora 1 1 att_8_mg att_8_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
828 |
+
Sigmoid sigmoid_32 1 1 att_8_mg_lora att_8_mg_lora_sigmoid
|
829 |
+
Gemm gemm_att_8_gate 1 1 att_8_mg_lora_sigmoid att_8_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
830 |
+
MemoryData data_att_8_k_k 0 1 att_8_k_k 0=768 21=0
|
831 |
+
MemoryData data_att_8_k_a 0 1 att_8_k_a 0=768 21=0
|
832 |
+
MemoryData data_att_8_r_k 0 1 att_8_r_k 0=768 21=0
|
833 |
+
Split split_att_8_key 1 2 att_8_key att_8_key_0 att_8_key_1
|
834 |
+
BinaryOp mul_104 2 1 att_8_key_0 att_8_k_k att_8_key_k 0=2
|
835 |
+
Reshape reshape_96 1 1 att_8_key_k att_8_key_k_reshape 0=64 1=1 2=12
|
836 |
+
Normalize l2norm_8 1 1 att_8_key_k_reshape att_8_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
837 |
+
Reshape reshape_97 1 1 att_8_key_k_norm att_8_key_k_norm_reshape 0=768
|
838 |
+
BinaryOp add_104 1 1 att_8_a_0 att_8_a_0_minus 0=0 1=1 2=-1.0
|
839 |
+
BinaryOp mul_105 2 1 att_8_a_0_minus att_8_k_a att_8_a_minus1_mulka 0=2
|
840 |
+
BinaryOp add_105 1 1 att_8_a_minus1_mulka att_8_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
841 |
+
BinaryOp mul_106 2 1 att_8_a_minus1_mulka_plus1 att_8_key_1 att_8_key_a 0=2
|
842 |
+
Split split_att_8_value 1 2 att_8_value att_8_value_0 att_8_value_1
|
843 |
+
Split split_v_first_7 1 2 v_first_7 att_8_v_first v_first_8
|
844 |
+
BinaryOp minus_7 2 1 att_8_v_first att_8_value_1 att_8_vfirst_minus_value 0=1
|
845 |
+
Gemm gemm_att_8_mv_lora_0 1 1 att_8_mv_1 att_8_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
846 |
+
Gemm gemm_att_8_mv_lora_1 1 1 att_8_mv_lora_0 att_8_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
847 |
+
MemoryData data_att_8_v0 0 1 att_8_v0 0=768 21=0
|
848 |
+
BinaryOp add_106 2 1 att_8_mv_lora_1 att_8_v0 att_8_mv_lora_2 0=0
|
849 |
+
Sigmoid sigmoid_33 1 1 att_8_mv_lora_2 att_8_mv_lora_sigmoid
|
850 |
+
BinaryOp mul_107 2 1 att_8_mv_lora_sigmoid att_8_vfirst_minus_value att_8_vfirst_minus_value_mul 0=2
|
851 |
+
BinaryOp add_107 2 1 att_8_value_0 att_8_vfirst_minus_value_mul att_8_value_final 0=0
|
852 |
+
Gemm gemm_att_8_mw_lora 1 1 att_8_mw att_8_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
853 |
+
TanH tanh_att_8_mw_lora 1 1 att_8_mw_lora att_8_mw_lora_tanh
|
854 |
+
Gemm gemm_att_8_mw_lora_tanh_linear 1 1 att_8_mw_lora_tanh att_8_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
855 |
+
MemoryData data_att_8_td 0 1 att_8_td 0=768 21=0
|
856 |
+
BinaryOp add_108 2 1 att_8_mw_lora_tanh_linear att_8_td att_8_time_decay_pre 0=0
|
857 |
+
Sigmoid sigmoid_34 1 1 att_8_time_decay_pre att_8_time_decay_sigmoid
|
858 |
+
Exp exp_8 1 1 att_8_time_decay_sigmoid att_8_time_decay 1=-0.606531
|
859 |
+
Reshape reshape_98 1 1 att_8_time_decay att_8_time_decay_reshape 0=64 1=1 2=12
|
860 |
+
Split split_att_8_key_a 1 2 att_8_key_a att_8_key_a_0 att_8_key_a_1
|
861 |
+
Reshape reshape_99 1 1 att_8_key_a_0 att_8_key_a_reshape 0=64 1=1 2=12
|
862 |
+
Split split_att_8_value_final 1 3 att_8_value_final att_8_value_final_0 att_8_value_final_1 att_8_value_final_2
|
863 |
+
Reshape reshape_100 1 1 att_8_value_final_0 att_8_value_final_0_reshape 0=64 1=12
|
864 |
+
Reshape reshape_101 1 1 att_8_value_final_1 att_8_value_final_1_reshape 0=1 1=64 2=12
|
865 |
+
MatMul matmul_att_8_vk 2 1 att_8_value_final_1_reshape att_8_key_a_reshape att_8_vk 0=0
|
866 |
+
Split split_att_8_key_k_norm_reshape 1 2 att_8_key_k_norm_reshape att_8_kk_0 att_8_kk_1
|
867 |
+
BinaryOp mul_108 1 1 att_8_kk_0 att_8_kk_0_neg 0=2 1=1 2=-1.0
|
868 |
+
Reshape reshape_102 1 1 att_8_kk_0_neg att_8_kk_0_neg_reshape 0=1 1=64 2=12
|
869 |
+
BinaryOp mul_109 2 1 att_8_kk_1 att_8_a_1 att_8_kk_a 0=2
|
870 |
+
Reshape reshape_103 1 1 att_8_kk_a att_8_kk_a_reshape 0=64 1=1 2=12
|
871 |
+
MatMul matmul_att_8_ab 2 1 att_8_kk_0_neg_reshape att_8_kk_a_reshape att_8_ab 0=0
|
872 |
+
Split split_state_25_in 1 2 state_25_in att_8_state_prev_0 att_8_state_prev_1
|
873 |
+
BinaryOp mul_110 2 1 att_8_state_prev_0 att_8_time_decay_reshape att_8_state_td 0=2
|
874 |
+
MatMul matmul_att_8_sab 2 1 att_8_state_prev_1 att_8_ab att_8_sab 0=0
|
875 |
+
BinaryOp add_109 2 1 att_8_sab att_8_vk att_8_sab_vk 0=0
|
876 |
+
BinaryOp add_110 2 1 att_8_sab_vk att_8_state_td att_8_state_new 0=0
|
877 |
+
Split split_att_8_state_new 1 2 att_8_state_new att_8_state_new_0 state_25_out
|
878 |
+
Split split_att_8_receptance 1 2 att_8_receptance att_8_receptance_0 att_8_receptance_1
|
879 |
+
Reshape reshape_104 1 1 att_8_receptance_0 att_8_receptance_reshape 0=1 1=64 2=12
|
880 |
+
MatMul matmul_att_8_wkv_out 2 1 att_8_state_new_0 att_8_receptance_reshape att_8_wkv_out 0=0
|
881 |
+
Reshape reshape_105 1 1 att_8_wkv_out att_8_wkv_out_flatten 0=768
|
882 |
+
GroupNorm groupnorm_8 1 1 att_8_wkv_out_flatten att_8_x_gn 0=12 1=768 2=0.00064 3=1
|
883 |
+
BinaryOp mul_111 2 1 att_8_receptance_1 att_8_key_a_1 att_8_receptance_key_a 0=2
|
884 |
+
BinaryOp mul_112 2 1 att_8_receptance_key_a att_8_r_k att_8_rk 0=2
|
885 |
+
Reshape reshape_106 1 1 att_8_rk att_8_rk_reshape 0=64 1=12
|
886 |
+
Reduction sum_8 1 1 att_8_rk_reshape att_8_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
887 |
+
BinaryOp mul_113 2 1 att_8_value_final_0_reshape att_8_rk_sum att_8_value_final_0_rk 0=2
|
888 |
+
Reshape reshape_107 1 1 att_8_value_final_0_rk att_8_value_final_0_rk_reshape 0=768
|
889 |
+
BinaryOp add_111 2 1 att_8_value_final_0_rk_reshape att_8_x_gn att_8_x_gn_rkv 0=0
|
890 |
+
BinaryOp mul_114 2 1 att_8_x_gn_rkv att_8_gate att_8_x_gate 0=2
|
891 |
+
Gemm gemm_att_8_x_out 1 1 att_8_x_gate att_8_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
892 |
+
BinaryOp add_112 2 1 att_8_x_out att_8_x_last time_mixing_8_out 0=0
|
893 |
+
Split split_time_mixing_8_out 1 2 time_mixing_8_out ffn_8_x_last ffn_8_x
|
894 |
+
LayerNorm layernorm_17 1 1 ffn_8_x ffn_8_xx 0=768 1=0.00001 2=1
|
895 |
+
Split split_ffn_8_xx 1 3 ffn_8_xx ffn_8_xx_0 ffn_8_xx_1 state_26_out
|
896 |
+
BinaryOp sub_17 2 1 state_26_in ffn_8_xx_0 ffn_8_sx 0=1
|
897 |
+
Split split_ffn_8_sx 1 1 ffn_8_sx ffn_8_sx_0
|
898 |
+
MemoryData data_ffn_8_x_k 0 1 ffn_8_x_k 0=768 21=0
|
899 |
+
BinaryOp mul_115 2 1 ffn_8_sx_0 ffn_8_x_k ffn_8_xk 0=2
|
900 |
+
BinaryOp add_113 2 1 ffn_8_xk ffn_8_xx_1 ffn_8_xxk 0=0
|
901 |
+
Gemm gemm_ffn_8_key 1 1 ffn_8_xxk ffn_8_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
902 |
+
ReLU relu_8 1 1 ffn_8_key ffn_8_key_relu
|
903 |
+
UnaryOp square_ffn_8_key_relu 1 1 ffn_8_key_relu ffn_8_key_relu_square 0=4
|
904 |
+
Gemm gemm_ffn_8_value 1 1 ffn_8_key_relu_square ffn_8_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
905 |
+
BinaryOp add_114 2 1 ffn_8_value ffn_8_x_last channel_mixing_8_out 0=0
|
906 |
+
Split split_channel_mixing_8_out 1 2 channel_mixing_8_out att_9_x_last att_9_x
|
907 |
+
LayerNorm layernorm_18 1 1 att_9_x att_9_xx 0=768 1=0.00001 2=1
|
908 |
+
Split split_att_9_xx 1 3 att_9_xx att_9_xx_0 att_9_xx_1 state_27_out
|
909 |
+
BinaryOp sub_18 2 1 state_27_in att_9_xx_0 att_9_sx 0=1
|
910 |
+
MemoryData data_att_9_maa 0 1 att_9_maa 0=768 1=6 21=0
|
911 |
+
BinaryOp mul_116 2 1 att_9_sx att_9_maa att_9_maa_x 0=2
|
912 |
+
BinaryOp add_115 2 1 att_9_xx_1 att_9_maa_x att_9_maa_rwkvag 0=0
|
913 |
+
Slice slice_att_9_maa_rwkvag 1 6 att_9_maa_rwkvag att_9_mr att_9_mw att_9_mk att_9_mv att_9_ma att_9_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
914 |
+
Split split_att_9_mv 1 2 att_9_mv att_9_mv_0 att_9_mv_1
|
915 |
+
Gemm gemm_att_9_key 1 1 att_9_mk att_9_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
916 |
+
Gemm gemm_att_9_value 1 1 att_9_mv_0 att_9_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
917 |
+
Gemm gemm_att_9_receptance 1 1 att_9_mr att_9_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
918 |
+
Gemm gemm_att_9_ma_lora_0 1 1 att_9_ma att_9_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
919 |
+
Gemm gemm_att_9_ma_lora_1 1 1 att_9_ma_lora_0 att_9_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
920 |
+
MemoryData data_att_9_a0 0 1 att_9_a0 0=768 21=0
|
921 |
+
BinaryOp add_116 2 1 att_9_ma_lora_1 att_9_a0 att_9_ma_lora_2 0=0
|
922 |
+
Sigmoid sigmoid_35 1 1 att_9_ma_lora_2 att_9_a
|
923 |
+
Split split_att_9_a 1 2 att_9_a att_9_a_0 att_9_a_1
|
924 |
+
Gemm gemm_att_9_mg_lora 1 1 att_9_mg att_9_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
925 |
+
Sigmoid sigmoid_36 1 1 att_9_mg_lora att_9_mg_lora_sigmoid
|
926 |
+
Gemm gemm_att_9_gate 1 1 att_9_mg_lora_sigmoid att_9_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
927 |
+
MemoryData data_att_9_k_k 0 1 att_9_k_k 0=768 21=0
|
928 |
+
MemoryData data_att_9_k_a 0 1 att_9_k_a 0=768 21=0
|
929 |
+
MemoryData data_att_9_r_k 0 1 att_9_r_k 0=768 21=0
|
930 |
+
Split split_att_9_key 1 2 att_9_key att_9_key_0 att_9_key_1
|
931 |
+
BinaryOp mul_117 2 1 att_9_key_0 att_9_k_k att_9_key_k 0=2
|
932 |
+
Reshape reshape_108 1 1 att_9_key_k att_9_key_k_reshape 0=64 1=1 2=12
|
933 |
+
Normalize l2norm_9 1 1 att_9_key_k_reshape att_9_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
934 |
+
Reshape reshape_109 1 1 att_9_key_k_norm att_9_key_k_norm_reshape 0=768
|
935 |
+
BinaryOp add_117 1 1 att_9_a_0 att_9_a_0_minus 0=0 1=1 2=-1.0
|
936 |
+
BinaryOp mul_118 2 1 att_9_a_0_minus att_9_k_a att_9_a_minus1_mulka 0=2
|
937 |
+
BinaryOp add_118 1 1 att_9_a_minus1_mulka att_9_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
938 |
+
BinaryOp mul_119 2 1 att_9_a_minus1_mulka_plus1 att_9_key_1 att_9_key_a 0=2
|
939 |
+
Split split_att_9_value 1 2 att_9_value att_9_value_0 att_9_value_1
|
940 |
+
Split split_v_first_8 1 2 v_first_8 att_9_v_first v_first_9
|
941 |
+
BinaryOp minus_8 2 1 att_9_v_first att_9_value_1 att_9_vfirst_minus_value 0=1
|
942 |
+
Gemm gemm_att_9_mv_lora_0 1 1 att_9_mv_1 att_9_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
943 |
+
Gemm gemm_att_9_mv_lora_1 1 1 att_9_mv_lora_0 att_9_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
944 |
+
MemoryData data_att_9_v0 0 1 att_9_v0 0=768 21=0
|
945 |
+
BinaryOp add_119 2 1 att_9_mv_lora_1 att_9_v0 att_9_mv_lora_2 0=0
|
946 |
+
Sigmoid sigmoid_37 1 1 att_9_mv_lora_2 att_9_mv_lora_sigmoid
|
947 |
+
BinaryOp mul_120 2 1 att_9_mv_lora_sigmoid att_9_vfirst_minus_value att_9_vfirst_minus_value_mul 0=2
|
948 |
+
BinaryOp add_120 2 1 att_9_value_0 att_9_vfirst_minus_value_mul att_9_value_final 0=0
|
949 |
+
Gemm gemm_att_9_mw_lora 1 1 att_9_mw att_9_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
950 |
+
TanH tanh_att_9_mw_lora 1 1 att_9_mw_lora att_9_mw_lora_tanh
|
951 |
+
Gemm gemm_att_9_mw_lora_tanh_linear 1 1 att_9_mw_lora_tanh att_9_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
952 |
+
MemoryData data_att_9_td 0 1 att_9_td 0=768 21=0
|
953 |
+
BinaryOp add_121 2 1 att_9_mw_lora_tanh_linear att_9_td att_9_time_decay_pre 0=0
|
954 |
+
Sigmoid sigmoid_38 1 1 att_9_time_decay_pre att_9_time_decay_sigmoid
|
955 |
+
Exp exp_9 1 1 att_9_time_decay_sigmoid att_9_time_decay 1=-0.606531
|
956 |
+
Reshape reshape_110 1 1 att_9_time_decay att_9_time_decay_reshape 0=64 1=1 2=12
|
957 |
+
Split split_att_9_key_a 1 2 att_9_key_a att_9_key_a_0 att_9_key_a_1
|
958 |
+
Reshape reshape_111 1 1 att_9_key_a_0 att_9_key_a_reshape 0=64 1=1 2=12
|
959 |
+
Split split_att_9_value_final 1 3 att_9_value_final att_9_value_final_0 att_9_value_final_1 att_9_value_final_2
|
960 |
+
Reshape reshape_112 1 1 att_9_value_final_0 att_9_value_final_0_reshape 0=64 1=12
|
961 |
+
Reshape reshape_113 1 1 att_9_value_final_1 att_9_value_final_1_reshape 0=1 1=64 2=12
|
962 |
+
MatMul matmul_att_9_vk 2 1 att_9_value_final_1_reshape att_9_key_a_reshape att_9_vk 0=0
|
963 |
+
Split split_att_9_key_k_norm_reshape 1 2 att_9_key_k_norm_reshape att_9_kk_0 att_9_kk_1
|
964 |
+
BinaryOp mul_121 1 1 att_9_kk_0 att_9_kk_0_neg 0=2 1=1 2=-1.0
|
965 |
+
Reshape reshape_114 1 1 att_9_kk_0_neg att_9_kk_0_neg_reshape 0=1 1=64 2=12
|
966 |
+
BinaryOp mul_122 2 1 att_9_kk_1 att_9_a_1 att_9_kk_a 0=2
|
967 |
+
Reshape reshape_115 1 1 att_9_kk_a att_9_kk_a_reshape 0=64 1=1 2=12
|
968 |
+
MatMul matmul_att_9_ab 2 1 att_9_kk_0_neg_reshape att_9_kk_a_reshape att_9_ab 0=0
|
969 |
+
Split split_state_28_in 1 2 state_28_in att_9_state_prev_0 att_9_state_prev_1
|
970 |
+
BinaryOp mul_123 2 1 att_9_state_prev_0 att_9_time_decay_reshape att_9_state_td 0=2
|
971 |
+
MatMul matmul_att_9_sab 2 1 att_9_state_prev_1 att_9_ab att_9_sab 0=0
|
972 |
+
BinaryOp add_122 2 1 att_9_sab att_9_vk att_9_sab_vk 0=0
|
973 |
+
BinaryOp add_123 2 1 att_9_sab_vk att_9_state_td att_9_state_new 0=0
|
974 |
+
Split split_att_9_state_new 1 2 att_9_state_new att_9_state_new_0 state_28_out
|
975 |
+
Split split_att_9_receptance 1 2 att_9_receptance att_9_receptance_0 att_9_receptance_1
|
976 |
+
Reshape reshape_116 1 1 att_9_receptance_0 att_9_receptance_reshape 0=1 1=64 2=12
|
977 |
+
MatMul matmul_att_9_wkv_out 2 1 att_9_state_new_0 att_9_receptance_reshape att_9_wkv_out 0=0
|
978 |
+
Reshape reshape_117 1 1 att_9_wkv_out att_9_wkv_out_flatten 0=768
|
979 |
+
GroupNorm groupnorm_9 1 1 att_9_wkv_out_flatten att_9_x_gn 0=12 1=768 2=0.00064 3=1
|
980 |
+
BinaryOp mul_124 2 1 att_9_receptance_1 att_9_key_a_1 att_9_receptance_key_a 0=2
|
981 |
+
BinaryOp mul_125 2 1 att_9_receptance_key_a att_9_r_k att_9_rk 0=2
|
982 |
+
Reshape reshape_118 1 1 att_9_rk att_9_rk_reshape 0=64 1=12
|
983 |
+
Reduction sum_9 1 1 att_9_rk_reshape att_9_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
984 |
+
BinaryOp mul_126 2 1 att_9_value_final_0_reshape att_9_rk_sum att_9_value_final_0_rk 0=2
|
985 |
+
Reshape reshape_119 1 1 att_9_value_final_0_rk att_9_value_final_0_rk_reshape 0=768
|
986 |
+
BinaryOp add_124 2 1 att_9_value_final_0_rk_reshape att_9_x_gn att_9_x_gn_rkv 0=0
|
987 |
+
BinaryOp mul_127 2 1 att_9_x_gn_rkv att_9_gate att_9_x_gate 0=2
|
988 |
+
Gemm gemm_att_9_x_out 1 1 att_9_x_gate att_9_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
989 |
+
BinaryOp add_125 2 1 att_9_x_out att_9_x_last time_mixing_9_out 0=0
|
990 |
+
Split split_time_mixing_9_out 1 2 time_mixing_9_out ffn_9_x_last ffn_9_x
|
991 |
+
LayerNorm layernorm_19 1 1 ffn_9_x ffn_9_xx 0=768 1=0.00001 2=1
|
992 |
+
Split split_ffn_9_xx 1 3 ffn_9_xx ffn_9_xx_0 ffn_9_xx_1 state_29_out
|
993 |
+
BinaryOp sub_19 2 1 state_29_in ffn_9_xx_0 ffn_9_sx 0=1
|
994 |
+
Split split_ffn_9_sx 1 1 ffn_9_sx ffn_9_sx_0
|
995 |
+
MemoryData data_ffn_9_x_k 0 1 ffn_9_x_k 0=768 21=0
|
996 |
+
BinaryOp mul_128 2 1 ffn_9_sx_0 ffn_9_x_k ffn_9_xk 0=2
|
997 |
+
BinaryOp add_126 2 1 ffn_9_xk ffn_9_xx_1 ffn_9_xxk 0=0
|
998 |
+
Gemm gemm_ffn_9_key 1 1 ffn_9_xxk ffn_9_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
999 |
+
ReLU relu_9 1 1 ffn_9_key ffn_9_key_relu
|
1000 |
+
UnaryOp square_ffn_9_key_relu 1 1 ffn_9_key_relu ffn_9_key_relu_square 0=4
|
1001 |
+
Gemm gemm_ffn_9_value 1 1 ffn_9_key_relu_square ffn_9_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
1002 |
+
BinaryOp add_127 2 1 ffn_9_value ffn_9_x_last channel_mixing_9_out 0=0
|
1003 |
+
Split split_channel_mixing_9_out 1 2 channel_mixing_9_out att_10_x_last att_10_x
|
1004 |
+
LayerNorm layernorm_20 1 1 att_10_x att_10_xx 0=768 1=0.00001 2=1
|
1005 |
+
Split split_att_10_xx 1 3 att_10_xx att_10_xx_0 att_10_xx_1 state_30_out
|
1006 |
+
BinaryOp sub_20 2 1 state_30_in att_10_xx_0 att_10_sx 0=1
|
1007 |
+
MemoryData data_att_10_maa 0 1 att_10_maa 0=768 1=6 21=0
|
1008 |
+
BinaryOp mul_129 2 1 att_10_sx att_10_maa att_10_maa_x 0=2
|
1009 |
+
BinaryOp add_128 2 1 att_10_xx_1 att_10_maa_x att_10_maa_rwkvag 0=0
|
1010 |
+
Slice slice_att_10_maa_rwkvag 1 6 att_10_maa_rwkvag att_10_mr att_10_mw att_10_mk att_10_mv att_10_ma att_10_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
1011 |
+
Split split_att_10_mv 1 2 att_10_mv att_10_mv_0 att_10_mv_1
|
1012 |
+
Gemm gemm_att_10_key 1 1 att_10_mk att_10_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1013 |
+
Gemm gemm_att_10_value 1 1 att_10_mv_0 att_10_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1014 |
+
Gemm gemm_att_10_receptance 1 1 att_10_mr att_10_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1015 |
+
Gemm gemm_att_10_ma_lora_0 1 1 att_10_ma att_10_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
1016 |
+
Gemm gemm_att_10_ma_lora_1 1 1 att_10_ma_lora_0 att_10_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
1017 |
+
MemoryData data_att_10_a0 0 1 att_10_a0 0=768 21=0
|
1018 |
+
BinaryOp add_129 2 1 att_10_ma_lora_1 att_10_a0 att_10_ma_lora_2 0=0
|
1019 |
+
Sigmoid sigmoid_39 1 1 att_10_ma_lora_2 att_10_a
|
1020 |
+
Split split_att_10_a 1 2 att_10_a att_10_a_0 att_10_a_1
|
1021 |
+
Gemm gemm_att_10_mg_lora 1 1 att_10_mg att_10_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
1022 |
+
Sigmoid sigmoid_40 1 1 att_10_mg_lora att_10_mg_lora_sigmoid
|
1023 |
+
Gemm gemm_att_10_gate 1 1 att_10_mg_lora_sigmoid att_10_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
1024 |
+
MemoryData data_att_10_k_k 0 1 att_10_k_k 0=768 21=0
|
1025 |
+
MemoryData data_att_10_k_a 0 1 att_10_k_a 0=768 21=0
|
1026 |
+
MemoryData data_att_10_r_k 0 1 att_10_r_k 0=768 21=0
|
1027 |
+
Split split_att_10_key 1 2 att_10_key att_10_key_0 att_10_key_1
|
1028 |
+
BinaryOp mul_130 2 1 att_10_key_0 att_10_k_k att_10_key_k 0=2
|
1029 |
+
Reshape reshape_120 1 1 att_10_key_k att_10_key_k_reshape 0=64 1=1 2=12
|
1030 |
+
Normalize l2norm_10 1 1 att_10_key_k_reshape att_10_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
1031 |
+
Reshape reshape_121 1 1 att_10_key_k_norm att_10_key_k_norm_reshape 0=768
|
1032 |
+
BinaryOp add_130 1 1 att_10_a_0 att_10_a_0_minus 0=0 1=1 2=-1.0
|
1033 |
+
BinaryOp mul_131 2 1 att_10_a_0_minus att_10_k_a att_10_a_minus1_mulka 0=2
|
1034 |
+
BinaryOp add_131 1 1 att_10_a_minus1_mulka att_10_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
1035 |
+
BinaryOp mul_132 2 1 att_10_a_minus1_mulka_plus1 att_10_key_1 att_10_key_a 0=2
|
1036 |
+
Split split_att_10_value 1 2 att_10_value att_10_value_0 att_10_value_1
|
1037 |
+
Split split_v_first_9 1 2 v_first_9 att_10_v_first v_first_10
|
1038 |
+
BinaryOp minus_9 2 1 att_10_v_first att_10_value_1 att_10_vfirst_minus_value 0=1
|
1039 |
+
Gemm gemm_att_10_mv_lora_0 1 1 att_10_mv_1 att_10_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
1040 |
+
Gemm gemm_att_10_mv_lora_1 1 1 att_10_mv_lora_0 att_10_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
1041 |
+
MemoryData data_att_10_v0 0 1 att_10_v0 0=768 21=0
|
1042 |
+
BinaryOp add_132 2 1 att_10_mv_lora_1 att_10_v0 att_10_mv_lora_2 0=0
|
1043 |
+
Sigmoid sigmoid_41 1 1 att_10_mv_lora_2 att_10_mv_lora_sigmoid
|
1044 |
+
BinaryOp mul_133 2 1 att_10_mv_lora_sigmoid att_10_vfirst_minus_value att_10_vfirst_minus_value_mul 0=2
|
1045 |
+
BinaryOp add_133 2 1 att_10_value_0 att_10_vfirst_minus_value_mul att_10_value_final 0=0
|
1046 |
+
Gemm gemm_att_10_mw_lora 1 1 att_10_mw att_10_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
1047 |
+
TanH tanh_att_10_mw_lora 1 1 att_10_mw_lora att_10_mw_lora_tanh
|
1048 |
+
Gemm gemm_att_10_mw_lora_tanh_linear 1 1 att_10_mw_lora_tanh att_10_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
1049 |
+
MemoryData data_att_10_td 0 1 att_10_td 0=768 21=0
|
1050 |
+
BinaryOp add_134 2 1 att_10_mw_lora_tanh_linear att_10_td att_10_time_decay_pre 0=0
|
1051 |
+
Sigmoid sigmoid_42 1 1 att_10_time_decay_pre att_10_time_decay_sigmoid
|
1052 |
+
Exp exp_10 1 1 att_10_time_decay_sigmoid att_10_time_decay 1=-0.606531
|
1053 |
+
Reshape reshape_122 1 1 att_10_time_decay att_10_time_decay_reshape 0=64 1=1 2=12
|
1054 |
+
Split split_att_10_key_a 1 2 att_10_key_a att_10_key_a_0 att_10_key_a_1
|
1055 |
+
Reshape reshape_123 1 1 att_10_key_a_0 att_10_key_a_reshape 0=64 1=1 2=12
|
1056 |
+
Split split_att_10_value_final 1 3 att_10_value_final att_10_value_final_0 att_10_value_final_1 att_10_value_final_2
|
1057 |
+
Reshape reshape_124 1 1 att_10_value_final_0 att_10_value_final_0_reshape 0=64 1=12
|
1058 |
+
Reshape reshape_125 1 1 att_10_value_final_1 att_10_value_final_1_reshape 0=1 1=64 2=12
|
1059 |
+
MatMul matmul_att_10_vk 2 1 att_10_value_final_1_reshape att_10_key_a_reshape att_10_vk 0=0
|
1060 |
+
Split split_att_10_key_k_norm_reshape 1 2 att_10_key_k_norm_reshape att_10_kk_0 att_10_kk_1
|
1061 |
+
BinaryOp mul_134 1 1 att_10_kk_0 att_10_kk_0_neg 0=2 1=1 2=-1.0
|
1062 |
+
Reshape reshape_126 1 1 att_10_kk_0_neg att_10_kk_0_neg_reshape 0=1 1=64 2=12
|
1063 |
+
BinaryOp mul_135 2 1 att_10_kk_1 att_10_a_1 att_10_kk_a 0=2
|
1064 |
+
Reshape reshape_127 1 1 att_10_kk_a att_10_kk_a_reshape 0=64 1=1 2=12
|
1065 |
+
MatMul matmul_att_10_ab 2 1 att_10_kk_0_neg_reshape att_10_kk_a_reshape att_10_ab 0=0
|
1066 |
+
Split split_state_31_in 1 2 state_31_in att_10_state_prev_0 att_10_state_prev_1
|
1067 |
+
BinaryOp mul_136 2 1 att_10_state_prev_0 att_10_time_decay_reshape att_10_state_td 0=2
|
1068 |
+
MatMul matmul_att_10_sab 2 1 att_10_state_prev_1 att_10_ab att_10_sab 0=0
|
1069 |
+
BinaryOp add_135 2 1 att_10_sab att_10_vk att_10_sab_vk 0=0
|
1070 |
+
BinaryOp add_136 2 1 att_10_sab_vk att_10_state_td att_10_state_new 0=0
|
1071 |
+
Split split_att_10_state_new 1 2 att_10_state_new att_10_state_new_0 state_31_out
|
1072 |
+
Split split_att_10_receptance 1 2 att_10_receptance att_10_receptance_0 att_10_receptance_1
|
1073 |
+
Reshape reshape_128 1 1 att_10_receptance_0 att_10_receptance_reshape 0=1 1=64 2=12
|
1074 |
+
MatMul matmul_att_10_wkv_out 2 1 att_10_state_new_0 att_10_receptance_reshape att_10_wkv_out 0=0
|
1075 |
+
Reshape reshape_129 1 1 att_10_wkv_out att_10_wkv_out_flatten 0=768
|
1076 |
+
GroupNorm groupnorm_10 1 1 att_10_wkv_out_flatten att_10_x_gn 0=12 1=768 2=0.00064 3=1
|
1077 |
+
BinaryOp mul_137 2 1 att_10_receptance_1 att_10_key_a_1 att_10_receptance_key_a 0=2
|
1078 |
+
BinaryOp mul_138 2 1 att_10_receptance_key_a att_10_r_k att_10_rk 0=2
|
1079 |
+
Reshape reshape_130 1 1 att_10_rk att_10_rk_reshape 0=64 1=12
|
1080 |
+
Reduction sum_10 1 1 att_10_rk_reshape att_10_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
1081 |
+
BinaryOp mul_139 2 1 att_10_value_final_0_reshape att_10_rk_sum att_10_value_final_0_rk 0=2
|
1082 |
+
Reshape reshape_131 1 1 att_10_value_final_0_rk att_10_value_final_0_rk_reshape 0=768
|
1083 |
+
BinaryOp add_137 2 1 att_10_value_final_0_rk_reshape att_10_x_gn att_10_x_gn_rkv 0=0
|
1084 |
+
BinaryOp mul_140 2 1 att_10_x_gn_rkv att_10_gate att_10_x_gate 0=2
|
1085 |
+
Gemm gemm_att_10_x_out 1 1 att_10_x_gate att_10_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1086 |
+
BinaryOp add_138 2 1 att_10_x_out att_10_x_last time_mixing_10_out 0=0
|
1087 |
+
Split split_time_mixing_10_out 1 2 time_mixing_10_out ffn_10_x_last ffn_10_x
|
1088 |
+
LayerNorm layernorm_21 1 1 ffn_10_x ffn_10_xx 0=768 1=0.00001 2=1
|
1089 |
+
Split split_ffn_10_xx 1 3 ffn_10_xx ffn_10_xx_0 ffn_10_xx_1 state_32_out
|
1090 |
+
BinaryOp sub_21 2 1 state_32_in ffn_10_xx_0 ffn_10_sx 0=1
|
1091 |
+
Split split_ffn_10_sx 1 1 ffn_10_sx ffn_10_sx_0
|
1092 |
+
MemoryData data_ffn_10_x_k 0 1 ffn_10_x_k 0=768 21=0
|
1093 |
+
BinaryOp mul_141 2 1 ffn_10_sx_0 ffn_10_x_k ffn_10_xk 0=2
|
1094 |
+
BinaryOp add_139 2 1 ffn_10_xk ffn_10_xx_1 ffn_10_xxk 0=0
|
1095 |
+
Gemm gemm_ffn_10_key 1 1 ffn_10_xxk ffn_10_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
1096 |
+
ReLU relu_10 1 1 ffn_10_key ffn_10_key_relu
|
1097 |
+
UnaryOp square_ffn_10_key_relu 1 1 ffn_10_key_relu ffn_10_key_relu_square 0=4
|
1098 |
+
Gemm gemm_ffn_10_value 1 1 ffn_10_key_relu_square ffn_10_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
1099 |
+
BinaryOp add_140 2 1 ffn_10_value ffn_10_x_last channel_mixing_10_out 0=0
|
1100 |
+
Split split_channel_mixing_10_out 1 2 channel_mixing_10_out att_11_x_last att_11_x
|
1101 |
+
LayerNorm layernorm_22 1 1 att_11_x att_11_xx 0=768 1=0.00001 2=1
|
1102 |
+
Split split_att_11_xx 1 3 att_11_xx att_11_xx_0 att_11_xx_1 state_33_out
|
1103 |
+
BinaryOp sub_22 2 1 state_33_in att_11_xx_0 att_11_sx 0=1
|
1104 |
+
MemoryData data_att_11_maa 0 1 att_11_maa 0=768 1=6 21=0
|
1105 |
+
BinaryOp mul_142 2 1 att_11_sx att_11_maa att_11_maa_x 0=2
|
1106 |
+
BinaryOp add_141 2 1 att_11_xx_1 att_11_maa_x att_11_maa_rwkvag 0=0
|
1107 |
+
Slice slice_att_11_maa_rwkvag 1 6 att_11_maa_rwkvag att_11_mr att_11_mw att_11_mk att_11_mv att_11_ma att_11_mg -23300=6,-233,-233,-233,-233,-233,-233 1=0
|
1108 |
+
Split split_att_11_mv 1 2 att_11_mv att_11_mv_0 att_11_mv_1
|
1109 |
+
Gemm gemm_att_11_key 1 1 att_11_mk att_11_key 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1110 |
+
Gemm gemm_att_11_value 1 1 att_11_mv_0 att_11_value 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1111 |
+
Gemm gemm_att_11_receptance 1 1 att_11_mr att_11_receptance 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1112 |
+
Gemm gemm_att_11_ma_lora_0 1 1 att_11_ma att_11_ma_lora_0 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
1113 |
+
Gemm gemm_att_11_ma_lora_1 1 1 att_11_ma_lora_0 att_11_ma_lora_1 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
1114 |
+
MemoryData data_att_11_a0 0 1 att_11_a0 0=768 21=0
|
1115 |
+
BinaryOp add_142 2 1 att_11_ma_lora_1 att_11_a0 att_11_ma_lora_2 0=0
|
1116 |
+
Sigmoid sigmoid_43 1 1 att_11_ma_lora_2 att_11_a
|
1117 |
+
Split split_att_11_a 1 2 att_11_a att_11_a_0 att_11_a_1
|
1118 |
+
Gemm gemm_att_11_mg_lora 1 1 att_11_mg att_11_mg_lora 4=0 5=1 6=0 7=0 8=128 9=768 10=-1
|
1119 |
+
Sigmoid sigmoid_44 1 1 att_11_mg_lora att_11_mg_lora_sigmoid
|
1120 |
+
Gemm gemm_att_11_gate 1 1 att_11_mg_lora_sigmoid att_11_gate 4=0 5=1 6=0 7=0 8=768 9=128 10=-1
|
1121 |
+
MemoryData data_att_11_k_k 0 1 att_11_k_k 0=768 21=0
|
1122 |
+
MemoryData data_att_11_k_a 0 1 att_11_k_a 0=768 21=0
|
1123 |
+
MemoryData data_att_11_r_k 0 1 att_11_r_k 0=768 21=0
|
1124 |
+
Split split_att_11_key 1 2 att_11_key att_11_key_0 att_11_key_1
|
1125 |
+
BinaryOp mul_143 2 1 att_11_key_0 att_11_k_k att_11_key_k 0=2
|
1126 |
+
Reshape reshape_132 1 1 att_11_key_k att_11_key_k_reshape 0=64 1=1 2=12
|
1127 |
+
Normalize l2norm_11 1 1 att_11_key_k_reshape att_11_key_k_norm 0=1 1=1 2=0.0000001 3=1 4=0 9=1
|
1128 |
+
Reshape reshape_133 1 1 att_11_key_k_norm att_11_key_k_norm_reshape 0=768
|
1129 |
+
BinaryOp add_143 1 1 att_11_a_0 att_11_a_0_minus 0=0 1=1 2=-1.0
|
1130 |
+
BinaryOp mul_144 2 1 att_11_a_0_minus att_11_k_a att_11_a_minus1_mulka 0=2
|
1131 |
+
BinaryOp add_144 1 1 att_11_a_minus1_mulka att_11_a_minus1_mulka_plus1 0=0 1=1 2=1.0
|
1132 |
+
BinaryOp mul_145 2 1 att_11_a_minus1_mulka_plus1 att_11_key_1 att_11_key_a 0=2
|
1133 |
+
Split split_att_11_value 1 2 att_11_value att_11_value_0 att_11_value_1
|
1134 |
+
Split split_v_first_10 1 2 v_first_10 att_11_v_first v_first_11
|
1135 |
+
BinaryOp minus_10 2 1 att_11_v_first att_11_value_1 att_11_vfirst_minus_value 0=1
|
1136 |
+
Gemm gemm_att_11_mv_lora_0 1 1 att_11_mv_1 att_11_mv_lora_0 4=0 5=1 6=0 7=0 8=32 9=768 10=-1
|
1137 |
+
Gemm gemm_att_11_mv_lora_1 1 1 att_11_mv_lora_0 att_11_mv_lora_1 4=0 5=1 6=0 7=0 8=768 9=32 10=-1
|
1138 |
+
MemoryData data_att_11_v0 0 1 att_11_v0 0=768 21=0
|
1139 |
+
BinaryOp add_145 2 1 att_11_mv_lora_1 att_11_v0 att_11_mv_lora_2 0=0
|
1140 |
+
Sigmoid sigmoid_45 1 1 att_11_mv_lora_2 att_11_mv_lora_sigmoid
|
1141 |
+
BinaryOp mul_146 2 1 att_11_mv_lora_sigmoid att_11_vfirst_minus_value att_11_vfirst_minus_value_mul 0=2
|
1142 |
+
BinaryOp add_146 2 1 att_11_value_0 att_11_vfirst_minus_value_mul att_11_value_final 0=0
|
1143 |
+
Gemm gemm_att_11_mw_lora 1 1 att_11_mw att_11_mw_lora 4=0 5=1 6=0 7=0 8=64 9=768 10=-1
|
1144 |
+
TanH tanh_att_11_mw_lora 1 1 att_11_mw_lora att_11_mw_lora_tanh
|
1145 |
+
Gemm gemm_att_11_mw_lora_tanh_linear 1 1 att_11_mw_lora_tanh att_11_mw_lora_tanh_linear 4=0 5=1 6=0 7=0 8=768 9=64 10=-1
|
1146 |
+
MemoryData data_att_11_td 0 1 att_11_td 0=768 21=0
|
1147 |
+
BinaryOp add_147 2 1 att_11_mw_lora_tanh_linear att_11_td att_11_time_decay_pre 0=0
|
1148 |
+
Sigmoid sigmoid_46 1 1 att_11_time_decay_pre att_11_time_decay_sigmoid
|
1149 |
+
Exp exp_11 1 1 att_11_time_decay_sigmoid att_11_time_decay 1=-0.606531
|
1150 |
+
Reshape reshape_134 1 1 att_11_time_decay att_11_time_decay_reshape 0=64 1=1 2=12
|
1151 |
+
Split split_att_11_key_a 1 2 att_11_key_a att_11_key_a_0 att_11_key_a_1
|
1152 |
+
Reshape reshape_135 1 1 att_11_key_a_0 att_11_key_a_reshape 0=64 1=1 2=12
|
1153 |
+
Split split_att_11_value_final 1 3 att_11_value_final att_11_value_final_0 att_11_value_final_1 att_11_value_final_2
|
1154 |
+
Reshape reshape_136 1 1 att_11_value_final_0 att_11_value_final_0_reshape 0=64 1=12
|
1155 |
+
Reshape reshape_137 1 1 att_11_value_final_1 att_11_value_final_1_reshape 0=1 1=64 2=12
|
1156 |
+
MatMul matmul_att_11_vk 2 1 att_11_value_final_1_reshape att_11_key_a_reshape att_11_vk 0=0
|
1157 |
+
Split split_att_11_key_k_norm_reshape 1 2 att_11_key_k_norm_reshape att_11_kk_0 att_11_kk_1
|
1158 |
+
BinaryOp mul_147 1 1 att_11_kk_0 att_11_kk_0_neg 0=2 1=1 2=-1.0
|
1159 |
+
Reshape reshape_138 1 1 att_11_kk_0_neg att_11_kk_0_neg_reshape 0=1 1=64 2=12
|
1160 |
+
BinaryOp mul_148 2 1 att_11_kk_1 att_11_a_1 att_11_kk_a 0=2
|
1161 |
+
Reshape reshape_139 1 1 att_11_kk_a att_11_kk_a_reshape 0=64 1=1 2=12
|
1162 |
+
MatMul matmul_att_11_ab 2 1 att_11_kk_0_neg_reshape att_11_kk_a_reshape att_11_ab 0=0
|
1163 |
+
Split split_state_34_in 1 2 state_34_in att_11_state_prev_0 att_11_state_prev_1
|
1164 |
+
BinaryOp mul_149 2 1 att_11_state_prev_0 att_11_time_decay_reshape att_11_state_td 0=2
|
1165 |
+
MatMul matmul_att_11_sab 2 1 att_11_state_prev_1 att_11_ab att_11_sab 0=0
|
1166 |
+
BinaryOp add_148 2 1 att_11_sab att_11_vk att_11_sab_vk 0=0
|
1167 |
+
BinaryOp add_149 2 1 att_11_sab_vk att_11_state_td att_11_state_new 0=0
|
1168 |
+
Split split_att_11_state_new 1 2 att_11_state_new att_11_state_new_0 state_34_out
|
1169 |
+
Split split_att_11_receptance 1 2 att_11_receptance att_11_receptance_0 att_11_receptance_1
|
1170 |
+
Reshape reshape_140 1 1 att_11_receptance_0 att_11_receptance_reshape 0=1 1=64 2=12
|
1171 |
+
MatMul matmul_att_11_wkv_out 2 1 att_11_state_new_0 att_11_receptance_reshape att_11_wkv_out 0=0
|
1172 |
+
Reshape reshape_141 1 1 att_11_wkv_out att_11_wkv_out_flatten 0=768
|
1173 |
+
GroupNorm groupnorm_11 1 1 att_11_wkv_out_flatten att_11_x_gn 0=12 1=768 2=0.00064 3=1
|
1174 |
+
BinaryOp mul_150 2 1 att_11_receptance_1 att_11_key_a_1 att_11_receptance_key_a 0=2
|
1175 |
+
BinaryOp mul_151 2 1 att_11_receptance_key_a att_11_r_k att_11_rk 0=2
|
1176 |
+
Reshape reshape_142 1 1 att_11_rk att_11_rk_reshape 0=64 1=12
|
1177 |
+
Reduction sum_11 1 1 att_11_rk_reshape att_11_rk_sum 1=0 -23303=1,-1 4=1 5=1
|
1178 |
+
BinaryOp mul_152 2 1 att_11_value_final_0_reshape att_11_rk_sum att_11_value_final_0_rk 0=2
|
1179 |
+
Reshape reshape_143 1 1 att_11_value_final_0_rk att_11_value_final_0_rk_reshape 0=768
|
1180 |
+
BinaryOp add_150 2 1 att_11_value_final_0_rk_reshape att_11_x_gn att_11_x_gn_rkv 0=0
|
1181 |
+
BinaryOp mul_153 2 1 att_11_x_gn_rkv att_11_gate att_11_x_gate 0=2
|
1182 |
+
Gemm gemm_att_11_x_out 1 1 att_11_x_gate att_11_x_out 4=0 5=1 6=0 7=0 8=768 9=768 10=-1
|
1183 |
+
BinaryOp add_151 2 1 att_11_x_out att_11_x_last time_mixing_11_out 0=0
|
1184 |
+
Split split_time_mixing_11_out 1 2 time_mixing_11_out ffn_11_x_last ffn_11_x
|
1185 |
+
LayerNorm layernorm_23 1 1 ffn_11_x ffn_11_xx 0=768 1=0.00001 2=1
|
1186 |
+
Split split_ffn_11_xx 1 3 ffn_11_xx ffn_11_xx_0 ffn_11_xx_1 state_35_out
|
1187 |
+
BinaryOp sub_23 2 1 state_35_in ffn_11_xx_0 ffn_11_sx 0=1
|
1188 |
+
Split split_ffn_11_sx 1 1 ffn_11_sx ffn_11_sx_0
|
1189 |
+
MemoryData data_ffn_11_x_k 0 1 ffn_11_x_k 0=768 21=0
|
1190 |
+
BinaryOp mul_154 2 1 ffn_11_sx_0 ffn_11_x_k ffn_11_xk 0=2
|
1191 |
+
BinaryOp add_152 2 1 ffn_11_xk ffn_11_xx_1 ffn_11_xxk 0=0
|
1192 |
+
Gemm gemm_ffn_11_key 1 1 ffn_11_xxk ffn_11_key 4=0 5=1 6=0 7=0 8=3072 9=768 10=-1
|
1193 |
+
ReLU relu_11 1 1 ffn_11_key ffn_11_key_relu
|
1194 |
+
UnaryOp square_ffn_11_key_relu 1 1 ffn_11_key_relu ffn_11_key_relu_square 0=4
|
1195 |
+
Gemm gemm_ffn_11_value 1 1 ffn_11_key_relu_square ffn_11_value 4=0 5=1 6=0 7=0 8=768 9=3072 10=-1
|
1196 |
+
BinaryOp add_153 2 1 ffn_11_value ffn_11_x_last channel_mixing_11_out 0=0
|
1197 |
+
LayerNorm layernorm_24 1 1 channel_mixing_11_out norm_head 0=768 1=0.00001 2=1
|
1198 |
+
Gemm gemm_logits 1 1 norm_head logits 4=0 5=1 6=0 7=0 8=65536 9=768 10=-1
|