Hugging Face's logo

CausalLM
/
8x7B-MoE-test-NOT-MIXTRAL

Text Generation
Transformers
PyTorch
English
Chinese
mixtral
qwen
Inference Endpoints
text-generation-inference
Model card Files Community
JosephusCheung commited on
Commit
f2cbd96
•
1 Parent(s): e3ce4d1

Upload folder using huggingface_hub

Browse files
Files changed (17) hide show
  1. config.json +33 -0
  2. generation_config.json +8 -0
  3. pytorch_model-00001-of-00008.bin +3 -0
  4. pytorch_model-00002-of-00008.bin +3 -0
  5. pytorch_model-00003-of-00008.bin +3 -0
  6. pytorch_model-00004-of-00008.bin +3 -0
  7. pytorch_model-00005-of-00008.bin +3 -0
  8. pytorch_model-00006-of-00008.bin +3 -0
  9. pytorch_model-00007-of-00008.bin +3 -0
  10. pytorch_model-00008-of-00008.bin +3 -0
  11. pytorch_model.bin.index.json +1002 -0
  12. qwen.tiktoken +0 -0
  13. special_tokens_map.json +9 -0
  14. tokenization_qwen.py +230 -0
  15. tokenizer_config.json +10 -0
  16. visual.bin +3 -0
  17. visual.py +428 -0
config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/notebooks/moe-8x7b",
3
+ "architectures": [
4
+ "MixtralForCausalLM"
5
+ ],
6
+ "attention_bias": true,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "eos_token_id": 2,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 4096,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 11008,
14
+ "max_position_embeddings": 32768,
15
+ "model_type": "mixtral",
16
+ "num_attention_heads": 32,
17
+ "num_experts_per_tok": 2,
18
+ "num_hidden_layers": 32,
19
+ "num_key_value_heads": 32,
20
+ "num_local_experts": 8,
21
+ "output_router_logits": false,
22
+ "pretraining_tp": 1,
23
+ "rms_norm_eps": 1e-06,
24
+ "rope_scaling": null,
25
+ "rope_theta": 10000.0,
26
+ "router_aux_loss_coef": 0.001,
27
+ "sliding_window": 8192,
28
+ "tie_word_embeddings": false,
29
+ "torch_dtype": "bfloat16",
30
+ "transformers_version": "4.36.1",
31
+ "use_cache": true,
32
+ "vocab_size": 151936
33
+ }
generation_config.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151643,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.36.1",
7
+ "use_cache": false
8
+ }
pytorch_model-00001-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d29cb4071cb88b84138ce94f3775aeb5c94eee8a1cde0f8699e0065238c4748f
3
+ size 9988040244
pytorch_model-00002-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4d72fc0d1eb1dc67f3d3b824831f513121f16ef77cfe9b4987a7dacdf5e046b
3
+ size 9959835736
pytorch_model-00003-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44db8160126b04f1f01f22e231799096f539304cf7dc33f1db92f7872b4cd824
3
+ size 9915711646
pytorch_model-00004-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8776cf7902c73fc184702e89feaeaf6a326fb5145cd8671483e79c7d97c8d68e
3
+ size 9915711710
pytorch_model-00005-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4dcc601734371e2416e6d8b885327ff790b6da8f02b33f122f3310853fbfd26
3
+ size 9959835980
pytorch_model-00006-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e697da58feb498426d4962850a909649807f7d29bce7e778ae6ce216cc7b8a5e
3
+ size 9915711710
pytorch_model-00007-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ef40e6681e96ac38612c6495aceba0628b2d705539906c87ba73e862562e5d4
3
+ size 9915711710
pytorch_model-00008-of-00008.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5f34090895bf1372efef56be4d788ed55e89d975bcc9c1229008856f8a99d97
3
+ size 6473075116
pytorch_model.bin.index.json ADDED
@@ -0,0 +1,1002 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 76043264000
4
+ },
5
+ "weight_map": {
6
+ "lm_head.weight": "pytorch_model-00008-of-00008.bin",
7
+ "model.embed_tokens.weight": "pytorch_model-00001-of-00008.bin",
8
+ "model.layers.0.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00001-of-00008.bin",
9
+ "model.layers.0.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00001-of-00008.bin",
10
+ "model.layers.0.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00001-of-00008.bin",
11
+ "model.layers.0.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00001-of-00008.bin",
12
+ "model.layers.0.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00001-of-00008.bin",
13
+ "model.layers.0.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00001-of-00008.bin",
14
+ "model.layers.0.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00001-of-00008.bin",
15
+ "model.layers.0.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00001-of-00008.bin",
16
+ "model.layers.0.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00001-of-00008.bin",
17
+ "model.layers.0.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00001-of-00008.bin",
18
+ "model.layers.0.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00001-of-00008.bin",
19
+ "model.layers.0.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00001-of-00008.bin",
20
+ "model.layers.0.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00001-of-00008.bin",
21
+ "model.layers.0.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00001-of-00008.bin",
22
+ "model.layers.0.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00001-of-00008.bin",
23
+ "model.layers.0.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00001-of-00008.bin",
24
+ "model.layers.0.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00001-of-00008.bin",
25
+ "model.layers.0.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00001-of-00008.bin",
26
+ "model.layers.0.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00001-of-00008.bin",
27
+ "model.layers.0.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00001-of-00008.bin",
28
+ "model.layers.0.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00001-of-00008.bin",
29
+ "model.layers.0.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00001-of-00008.bin",
30
+ "model.layers.0.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00001-of-00008.bin",
31
+ "model.layers.0.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00001-of-00008.bin",
32
+ "model.layers.0.block_sparse_moe.gate.weight": "pytorch_model-00001-of-00008.bin",
33
+ "model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00008.bin",
34
+ "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00008.bin",
35
+ "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00008.bin",
36
+ "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00008.bin",
37
+ "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00008.bin",
38
+ "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00008.bin",
39
+ "model.layers.1.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00001-of-00008.bin",
40
+ "model.layers.1.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00001-of-00008.bin",
41
+ "model.layers.1.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00001-of-00008.bin",
42
+ "model.layers.1.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00001-of-00008.bin",
43
+ "model.layers.1.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00001-of-00008.bin",
44
+ "model.layers.1.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00001-of-00008.bin",
45
+ "model.layers.1.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00001-of-00008.bin",
46
+ "model.layers.1.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00001-of-00008.bin",
47
+ "model.layers.1.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00001-of-00008.bin",
48
+ "model.layers.1.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00001-of-00008.bin",
49
+ "model.layers.1.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00001-of-00008.bin",
50
+ "model.layers.1.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00001-of-00008.bin",
51
+ "model.layers.1.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00001-of-00008.bin",
52
+ "model.layers.1.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00001-of-00008.bin",
53
+ "model.layers.1.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00001-of-00008.bin",
54
+ "model.layers.1.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00001-of-00008.bin",
55
+ "model.layers.1.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00001-of-00008.bin",
56
+ "model.layers.1.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00001-of-00008.bin",
57
+ "model.layers.1.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00001-of-00008.bin",
58
+ "model.layers.1.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00001-of-00008.bin",
59
+ "model.layers.1.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00001-of-00008.bin",
60
+ "model.layers.1.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00001-of-00008.bin",
61
+ "model.layers.1.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00001-of-00008.bin",
62
+ "model.layers.1.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00001-of-00008.bin",
63
+ "model.layers.1.block_sparse_moe.gate.weight": "pytorch_model-00001-of-00008.bin",
64
+ "model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00008.bin",
65
+ "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00008.bin",
66
+ "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00001-of-00008.bin",
67
+ "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00008.bin",
68
+ "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00001-of-00008.bin",
69
+ "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00001-of-00008.bin",
70
+ "model.layers.10.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00003-of-00008.bin",
71
+ "model.layers.10.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00003-of-00008.bin",
72
+ "model.layers.10.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00003-of-00008.bin",
73
+ "model.layers.10.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00003-of-00008.bin",
74
+ "model.layers.10.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00003-of-00008.bin",
75
+ "model.layers.10.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00003-of-00008.bin",
76
+ "model.layers.10.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00003-of-00008.bin",
77
+ "model.layers.10.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00003-of-00008.bin",
78
+ "model.layers.10.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00003-of-00008.bin",
79
+ "model.layers.10.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00003-of-00008.bin",
80
+ "model.layers.10.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00003-of-00008.bin",
81
+ "model.layers.10.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00003-of-00008.bin",
82
+ "model.layers.10.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00003-of-00008.bin",
83
+ "model.layers.10.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00003-of-00008.bin",
84
+ "model.layers.10.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00003-of-00008.bin",
85
+ "model.layers.10.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00003-of-00008.bin",
86
+ "model.layers.10.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00003-of-00008.bin",
87
+ "model.layers.10.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00003-of-00008.bin",
88
+ "model.layers.10.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00003-of-00008.bin",
89
+ "model.layers.10.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00003-of-00008.bin",
90
+ "model.layers.10.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00003-of-00008.bin",
91
+ "model.layers.10.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00003-of-00008.bin",
92
+ "model.layers.10.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00003-of-00008.bin",
93
+ "model.layers.10.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00003-of-00008.bin",
94
+ "model.layers.10.block_sparse_moe.gate.weight": "pytorch_model-00003-of-00008.bin",
95
+ "model.layers.10.input_layernorm.weight": "pytorch_model-00003-of-00008.bin",
96
+ "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00003-of-00008.bin",
97
+ "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00003-of-00008.bin",
98
+ "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00003-of-00008.bin",
99
+ "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00003-of-00008.bin",
100
+ "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00003-of-00008.bin",
101
+ "model.layers.11.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00003-of-00008.bin",
102
+ "model.layers.11.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00003-of-00008.bin",
103
+ "model.layers.11.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00003-of-00008.bin",
104
+ "model.layers.11.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00003-of-00008.bin",
105
+ "model.layers.11.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00003-of-00008.bin",
106
+ "model.layers.11.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00003-of-00008.bin",
107
+ "model.layers.11.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00003-of-00008.bin",
108
+ "model.layers.11.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00003-of-00008.bin",
109
+ "model.layers.11.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00003-of-00008.bin",
110
+ "model.layers.11.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00003-of-00008.bin",
111
+ "model.layers.11.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00003-of-00008.bin",
112
+ "model.layers.11.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00003-of-00008.bin",
113
+ "model.layers.11.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00003-of-00008.bin",
114
+ "model.layers.11.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00003-of-00008.bin",
115
+ "model.layers.11.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00003-of-00008.bin",
116
+ "model.layers.11.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00003-of-00008.bin",
117
+ "model.layers.11.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00003-of-00008.bin",
118
+ "model.layers.11.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00003-of-00008.bin",
119
+ "model.layers.11.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00003-of-00008.bin",
120
+ "model.layers.11.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00003-of-00008.bin",
121
+ "model.layers.11.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00003-of-00008.bin",
122
+ "model.layers.11.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00003-of-00008.bin",
123
+ "model.layers.11.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00003-of-00008.bin",
124
+ "model.layers.11.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00003-of-00008.bin",
125
+ "model.layers.11.block_sparse_moe.gate.weight": "pytorch_model-00003-of-00008.bin",
126
+ "model.layers.11.input_layernorm.weight": "pytorch_model-00003-of-00008.bin",
127
+ "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00003-of-00008.bin",
128
+ "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00003-of-00008.bin",
129
+ "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00003-of-00008.bin",
130
+ "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00003-of-00008.bin",
131
+ "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00003-of-00008.bin",
132
+ "model.layers.12.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00003-of-00008.bin",
133
+ "model.layers.12.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00003-of-00008.bin",
134
+ "model.layers.12.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00003-of-00008.bin",
135
+ "model.layers.12.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00003-of-00008.bin",
136
+ "model.layers.12.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00003-of-00008.bin",
137
+ "model.layers.12.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00003-of-00008.bin",
138
+ "model.layers.12.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00003-of-00008.bin",
139
+ "model.layers.12.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00003-of-00008.bin",
140
+ "model.layers.12.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00003-of-00008.bin",
141
+ "model.layers.12.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00003-of-00008.bin",
142
+ "model.layers.12.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00004-of-00008.bin",
143
+ "model.layers.12.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00004-of-00008.bin",
144
+ "model.layers.12.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00004-of-00008.bin",
145
+ "model.layers.12.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00004-of-00008.bin",
146
+ "model.layers.12.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00004-of-00008.bin",
147
+ "model.layers.12.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00004-of-00008.bin",
148
+ "model.layers.12.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00004-of-00008.bin",
149
+ "model.layers.12.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00004-of-00008.bin",
150
+ "model.layers.12.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00004-of-00008.bin",
151
+ "model.layers.12.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00004-of-00008.bin",
152
+ "model.layers.12.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00004-of-00008.bin",
153
+ "model.layers.12.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00004-of-00008.bin",
154
+ "model.layers.12.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00004-of-00008.bin",
155
+ "model.layers.12.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00004-of-00008.bin",
156
+ "model.layers.12.block_sparse_moe.gate.weight": "pytorch_model-00003-of-00008.bin",
157
+ "model.layers.12.input_layernorm.weight": "pytorch_model-00004-of-00008.bin",
158
+ "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00004-of-00008.bin",
159
+ "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00003-of-00008.bin",
160
+ "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00003-of-00008.bin",
161
+ "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00003-of-00008.bin",
162
+ "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00003-of-00008.bin",
163
+ "model.layers.13.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00004-of-00008.bin",
164
+ "model.layers.13.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00004-of-00008.bin",
165
+ "model.layers.13.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00004-of-00008.bin",
166
+ "model.layers.13.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00004-of-00008.bin",
167
+ "model.layers.13.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00004-of-00008.bin",
168
+ "model.layers.13.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00004-of-00008.bin",
169
+ "model.layers.13.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00004-of-00008.bin",
170
+ "model.layers.13.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00004-of-00008.bin",
171
+ "model.layers.13.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00004-of-00008.bin",
172
+ "model.layers.13.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00004-of-00008.bin",
173
+ "model.layers.13.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00004-of-00008.bin",
174
+ "model.layers.13.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00004-of-00008.bin",
175
+ "model.layers.13.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00004-of-00008.bin",
176
+ "model.layers.13.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00004-of-00008.bin",
177
+ "model.layers.13.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00004-of-00008.bin",
178
+ "model.layers.13.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00004-of-00008.bin",
179
+ "model.layers.13.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00004-of-00008.bin",
180
+ "model.layers.13.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00004-of-00008.bin",
181
+ "model.layers.13.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00004-of-00008.bin",
182
+ "model.layers.13.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00004-of-00008.bin",
183
+ "model.layers.13.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00004-of-00008.bin",
184
+ "model.layers.13.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00004-of-00008.bin",
185
+ "model.layers.13.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00004-of-00008.bin",
186
+ "model.layers.13.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00004-of-00008.bin",
187
+ "model.layers.13.block_sparse_moe.gate.weight": "pytorch_model-00004-of-00008.bin",
188
+ "model.layers.13.input_layernorm.weight": "pytorch_model-00004-of-00008.bin",
189
+ "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00004-of-00008.bin",
190
+ "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00004-of-00008.bin",
191
+ "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00004-of-00008.bin",
192
+ "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00004-of-00008.bin",
193
+ "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00004-of-00008.bin",
194
+ "model.layers.14.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00004-of-00008.bin",
195
+ "model.layers.14.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00004-of-00008.bin",
196
+ "model.layers.14.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00004-of-00008.bin",
197
+ "model.layers.14.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00004-of-00008.bin",
198
+ "model.layers.14.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00004-of-00008.bin",
199
+ "model.layers.14.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00004-of-00008.bin",
200
+ "model.layers.14.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00004-of-00008.bin",
201
+ "model.layers.14.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00004-of-00008.bin",
202
+ "model.layers.14.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00004-of-00008.bin",
203
+ "model.layers.14.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00004-of-00008.bin",
204
+ "model.layers.14.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00004-of-00008.bin",
205
+ "model.layers.14.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00004-of-00008.bin",
206
+ "model.layers.14.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00004-of-00008.bin",
207
+ "model.layers.14.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00004-of-00008.bin",
208
+ "model.layers.14.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00004-of-00008.bin",
209
+ "model.layers.14.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00004-of-00008.bin",
210
+ "model.layers.14.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00004-of-00008.bin",
211
+ "model.layers.14.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00004-of-00008.bin",
212
+ "model.layers.14.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00004-of-00008.bin",
213
+ "model.layers.14.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00004-of-00008.bin",
214
+ "model.layers.14.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00004-of-00008.bin",
215
+ "model.layers.14.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00004-of-00008.bin",
216
+ "model.layers.14.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00004-of-00008.bin",
217
+ "model.layers.14.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00004-of-00008.bin",
218
+ "model.layers.14.block_sparse_moe.gate.weight": "pytorch_model-00004-of-00008.bin",
219
+ "model.layers.14.input_layernorm.weight": "pytorch_model-00004-of-00008.bin",
220
+ "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00004-of-00008.bin",
221
+ "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00004-of-00008.bin",
222
+ "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00004-of-00008.bin",
223
+ "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00004-of-00008.bin",
224
+ "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00004-of-00008.bin",
225
+ "model.layers.15.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00004-of-00008.bin",
226
+ "model.layers.15.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00004-of-00008.bin",
227
+ "model.layers.15.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00004-of-00008.bin",
228
+ "model.layers.15.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00004-of-00008.bin",
229
+ "model.layers.15.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00004-of-00008.bin",
230
+ "model.layers.15.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00004-of-00008.bin",
231
+ "model.layers.15.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00004-of-00008.bin",
232
+ "model.layers.15.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00004-of-00008.bin",
233
+ "model.layers.15.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00004-of-00008.bin",
234
+ "model.layers.15.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00004-of-00008.bin",
235
+ "model.layers.15.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00004-of-00008.bin",
236
+ "model.layers.15.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00004-of-00008.bin",
237
+ "model.layers.15.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00004-of-00008.bin",
238
+ "model.layers.15.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00004-of-00008.bin",
239
+ "model.layers.15.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00004-of-00008.bin",
240
+ "model.layers.15.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00004-of-00008.bin",
241
+ "model.layers.15.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00004-of-00008.bin",
242
+ "model.layers.15.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00004-of-00008.bin",
243
+ "model.layers.15.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00004-of-00008.bin",
244
+ "model.layers.15.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00004-of-00008.bin",
245
+ "model.layers.15.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00004-of-00008.bin",
246
+ "model.layers.15.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00004-of-00008.bin",
247
+ "model.layers.15.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00004-of-00008.bin",
248
+ "model.layers.15.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00004-of-00008.bin",
249
+ "model.layers.15.block_sparse_moe.gate.weight": "pytorch_model-00004-of-00008.bin",
250
+ "model.layers.15.input_layernorm.weight": "pytorch_model-00004-of-00008.bin",
251
+ "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00004-of-00008.bin",
252
+ "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00004-of-00008.bin",
253
+ "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00004-of-00008.bin",
254
+ "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00004-of-00008.bin",
255
+ "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00004-of-00008.bin",
256
+ "model.layers.16.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00004-of-00008.bin",
257
+ "model.layers.16.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00004-of-00008.bin",
258
+ "model.layers.16.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00004-of-00008.bin",
259
+ "model.layers.16.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00004-of-00008.bin",
260
+ "model.layers.16.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00004-of-00008.bin",
261
+ "model.layers.16.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00004-of-00008.bin",
262
+ "model.layers.16.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00004-of-00008.bin",
263
+ "model.layers.16.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00004-of-00008.bin",
264
+ "model.layers.16.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00004-of-00008.bin",
265
+ "model.layers.16.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00004-of-00008.bin",
266
+ "model.layers.16.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00004-of-00008.bin",
267
+ "model.layers.16.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00004-of-00008.bin",
268
+ "model.layers.16.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00004-of-00008.bin",
269
+ "model.layers.16.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00004-of-00008.bin",
270
+ "model.layers.16.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00004-of-00008.bin",
271
+ "model.layers.16.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00004-of-00008.bin",
272
+ "model.layers.16.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00004-of-00008.bin",
273
+ "model.layers.16.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00004-of-00008.bin",
274
+ "model.layers.16.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00005-of-00008.bin",
275
+ "model.layers.16.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00005-of-00008.bin",
276
+ "model.layers.16.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00005-of-00008.bin",
277
+ "model.layers.16.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00005-of-00008.bin",
278
+ "model.layers.16.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00005-of-00008.bin",
279
+ "model.layers.16.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00005-of-00008.bin",
280
+ "model.layers.16.block_sparse_moe.gate.weight": "pytorch_model-00004-of-00008.bin",
281
+ "model.layers.16.input_layernorm.weight": "pytorch_model-00005-of-00008.bin",
282
+ "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00005-of-00008.bin",
283
+ "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00004-of-00008.bin",
284
+ "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00004-of-00008.bin",
285
+ "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00004-of-00008.bin",
286
+ "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00004-of-00008.bin",
287
+ "model.layers.17.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00005-of-00008.bin",
288
+ "model.layers.17.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00005-of-00008.bin",
289
+ "model.layers.17.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00005-of-00008.bin",
290
+ "model.layers.17.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00005-of-00008.bin",
291
+ "model.layers.17.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00005-of-00008.bin",
292
+ "model.layers.17.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00005-of-00008.bin",
293
+ "model.layers.17.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00005-of-00008.bin",
294
+ "model.layers.17.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00005-of-00008.bin",
295
+ "model.layers.17.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00005-of-00008.bin",
296
+ "model.layers.17.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00005-of-00008.bin",
297
+ "model.layers.17.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00005-of-00008.bin",
298
+ "model.layers.17.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00005-of-00008.bin",
299
+ "model.layers.17.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00005-of-00008.bin",
300
+ "model.layers.17.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00005-of-00008.bin",
301
+ "model.layers.17.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00005-of-00008.bin",
302
+ "model.layers.17.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00005-of-00008.bin",
303
+ "model.layers.17.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00005-of-00008.bin",
304
+ "model.layers.17.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00005-of-00008.bin",
305
+ "model.layers.17.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00005-of-00008.bin",
306
+ "model.layers.17.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00005-of-00008.bin",
307
+ "model.layers.17.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00005-of-00008.bin",
308
+ "model.layers.17.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00005-of-00008.bin",
309
+ "model.layers.17.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00005-of-00008.bin",
310
+ "model.layers.17.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00005-of-00008.bin",
311
+ "model.layers.17.block_sparse_moe.gate.weight": "pytorch_model-00005-of-00008.bin",
312
+ "model.layers.17.input_layernorm.weight": "pytorch_model-00005-of-00008.bin",
313
+ "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00005-of-00008.bin",
314
+ "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00005-of-00008.bin",
315
+ "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00005-of-00008.bin",
316
+ "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00005-of-00008.bin",
317
+ "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00005-of-00008.bin",
318
+ "model.layers.18.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00005-of-00008.bin",
319
+ "model.layers.18.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00005-of-00008.bin",
320
+ "model.layers.18.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00005-of-00008.bin",
321
+ "model.layers.18.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00005-of-00008.bin",
322
+ "model.layers.18.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00005-of-00008.bin",
323
+ "model.layers.18.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00005-of-00008.bin",
324
+ "model.layers.18.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00005-of-00008.bin",
325
+ "model.layers.18.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00005-of-00008.bin",
326
+ "model.layers.18.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00005-of-00008.bin",
327
+ "model.layers.18.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00005-of-00008.bin",
328
+ "model.layers.18.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00005-of-00008.bin",
329
+ "model.layers.18.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00005-of-00008.bin",
330
+ "model.layers.18.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00005-of-00008.bin",
331
+ "model.layers.18.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00005-of-00008.bin",
332
+ "model.layers.18.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00005-of-00008.bin",
333
+ "model.layers.18.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00005-of-00008.bin",
334
+ "model.layers.18.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00005-of-00008.bin",
335
+ "model.layers.18.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00005-of-00008.bin",
336
+ "model.layers.18.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00005-of-00008.bin",
337
+ "model.layers.18.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00005-of-00008.bin",
338
+ "model.layers.18.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00005-of-00008.bin",
339
+ "model.layers.18.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00005-of-00008.bin",
340
+ "model.layers.18.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00005-of-00008.bin",
341
+ "model.layers.18.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00005-of-00008.bin",
342
+ "model.layers.18.block_sparse_moe.gate.weight": "pytorch_model-00005-of-00008.bin",
343
+ "model.layers.18.input_layernorm.weight": "pytorch_model-00005-of-00008.bin",
344
+ "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00005-of-00008.bin",
345
+ "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00005-of-00008.bin",
346
+ "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00005-of-00008.bin",
347
+ "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00005-of-00008.bin",
348
+ "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00005-of-00008.bin",
349
+ "model.layers.19.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00005-of-00008.bin",
350
+ "model.layers.19.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00005-of-00008.bin",
351
+ "model.layers.19.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00005-of-00008.bin",
352
+ "model.layers.19.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00005-of-00008.bin",
353
+ "model.layers.19.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00005-of-00008.bin",
354
+ "model.layers.19.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00005-of-00008.bin",
355
+ "model.layers.19.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00005-of-00008.bin",
356
+ "model.layers.19.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00005-of-00008.bin",
357
+ "model.layers.19.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00005-of-00008.bin",
358
+ "model.layers.19.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00005-of-00008.bin",
359
+ "model.layers.19.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00005-of-00008.bin",
360
+ "model.layers.19.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00005-of-00008.bin",
361
+ "model.layers.19.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00005-of-00008.bin",
362
+ "model.layers.19.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00005-of-00008.bin",
363
+ "model.layers.19.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00005-of-00008.bin",
364
+ "model.layers.19.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00005-of-00008.bin",
365
+ "model.layers.19.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00005-of-00008.bin",
366
+ "model.layers.19.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00005-of-00008.bin",
367
+ "model.layers.19.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00005-of-00008.bin",
368
+ "model.layers.19.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00005-of-00008.bin",
369
+ "model.layers.19.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00005-of-00008.bin",
370
+ "model.layers.19.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00005-of-00008.bin",
371
+ "model.layers.19.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00005-of-00008.bin",
372
+ "model.layers.19.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00005-of-00008.bin",
373
+ "model.layers.19.block_sparse_moe.gate.weight": "pytorch_model-00005-of-00008.bin",
374
+ "model.layers.19.input_layernorm.weight": "pytorch_model-00005-of-00008.bin",
375
+ "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00005-of-00008.bin",
376
+ "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00005-of-00008.bin",
377
+ "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00005-of-00008.bin",
378
+ "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00005-of-00008.bin",
379
+ "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00005-of-00008.bin",
380
+ "model.layers.2.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00001-of-00008.bin",
381
+ "model.layers.2.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00001-of-00008.bin",
382
+ "model.layers.2.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00001-of-00008.bin",
383
+ "model.layers.2.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00001-of-00008.bin",
384
+ "model.layers.2.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00001-of-00008.bin",
385
+ "model.layers.2.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00001-of-00008.bin",
386
+ "model.layers.2.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00001-of-00008.bin",
387
+ "model.layers.2.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00001-of-00008.bin",
388
+ "model.layers.2.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00001-of-00008.bin",
389
+ "model.layers.2.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00001-of-00008.bin",
390
+ "model.layers.2.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00001-of-00008.bin",
391
+ "model.layers.2.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00001-of-00008.bin",
392
+ "model.layers.2.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00001-of-00008.bin",
393
+ "model.layers.2.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00001-of-00008.bin",
394
+ "model.layers.2.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00001-of-00008.bin",
395
+ "model.layers.2.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00001-of-00008.bin",
396
+ "model.layers.2.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00001-of-00008.bin",
397
+ "model.layers.2.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00001-of-00008.bin",
398
+ "model.layers.2.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00001-of-00008.bin",
399
+ "model.layers.2.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00001-of-00008.bin",
400
+ "model.layers.2.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00001-of-00008.bin",
401
+ "model.layers.2.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00001-of-00008.bin",
402
+ "model.layers.2.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00001-of-00008.bin",
403
+ "model.layers.2.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00001-of-00008.bin",
404
+ "model.layers.2.block_sparse_moe.gate.weight": "pytorch_model-00001-of-00008.bin",
405
+ "model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00008.bin",
406
+ "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00008.bin",
407
+ "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00001-of-00008.bin",
408
+ "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00008.bin",
409
+ "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00001-of-00008.bin",
410
+ "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00001-of-00008.bin",
411
+ "model.layers.20.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00005-of-00008.bin",
412
+ "model.layers.20.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00005-of-00008.bin",
413
+ "model.layers.20.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00005-of-00008.bin",
414
+ "model.layers.20.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00005-of-00008.bin",
415
+ "model.layers.20.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00005-of-00008.bin",
416
+ "model.layers.20.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00005-of-00008.bin",
417
+ "model.layers.20.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00005-of-00008.bin",
418
+ "model.layers.20.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00005-of-00008.bin",
419
+ "model.layers.20.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00005-of-00008.bin",
420
+ "model.layers.20.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00005-of-00008.bin",
421
+ "model.layers.20.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00005-of-00008.bin",
422
+ "model.layers.20.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00005-of-00008.bin",
423
+ "model.layers.20.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00005-of-00008.bin",
424
+ "model.layers.20.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00005-of-00008.bin",
425
+ "model.layers.20.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00005-of-00008.bin",
426
+ "model.layers.20.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00005-of-00008.bin",
427
+ "model.layers.20.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00005-of-00008.bin",
428
+ "model.layers.20.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00005-of-00008.bin",
429
+ "model.layers.20.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00005-of-00008.bin",
430
+ "model.layers.20.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00005-of-00008.bin",
431
+ "model.layers.20.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00005-of-00008.bin",
432
+ "model.layers.20.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00005-of-00008.bin",
433
+ "model.layers.20.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00005-of-00008.bin",
434
+ "model.layers.20.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00005-of-00008.bin",
435
+ "model.layers.20.block_sparse_moe.gate.weight": "pytorch_model-00005-of-00008.bin",
436
+ "model.layers.20.input_layernorm.weight": "pytorch_model-00005-of-00008.bin",
437
+ "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00005-of-00008.bin",
438
+ "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00005-of-00008.bin",
439
+ "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00005-of-00008.bin",
440
+ "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00005-of-00008.bin",
441
+ "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00005-of-00008.bin",
442
+ "model.layers.21.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00005-of-00008.bin",
443
+ "model.layers.21.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00006-of-00008.bin",
444
+ "model.layers.21.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00006-of-00008.bin",
445
+ "model.layers.21.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00006-of-00008.bin",
446
+ "model.layers.21.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00006-of-00008.bin",
447
+ "model.layers.21.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00006-of-00008.bin",
448
+ "model.layers.21.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00006-of-00008.bin",
449
+ "model.layers.21.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00006-of-00008.bin",
450
+ "model.layers.21.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00006-of-00008.bin",
451
+ "model.layers.21.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00006-of-00008.bin",
452
+ "model.layers.21.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00006-of-00008.bin",
453
+ "model.layers.21.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00006-of-00008.bin",
454
+ "model.layers.21.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00006-of-00008.bin",
455
+ "model.layers.21.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00006-of-00008.bin",
456
+ "model.layers.21.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00006-of-00008.bin",
457
+ "model.layers.21.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00006-of-00008.bin",
458
+ "model.layers.21.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00006-of-00008.bin",
459
+ "model.layers.21.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00006-of-00008.bin",
460
+ "model.layers.21.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00006-of-00008.bin",
461
+ "model.layers.21.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00006-of-00008.bin",
462
+ "model.layers.21.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00006-of-00008.bin",
463
+ "model.layers.21.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00006-of-00008.bin",
464
+ "model.layers.21.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00006-of-00008.bin",
465
+ "model.layers.21.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00006-of-00008.bin",
466
+ "model.layers.21.block_sparse_moe.gate.weight": "pytorch_model-00005-of-00008.bin",
467
+ "model.layers.21.input_layernorm.weight": "pytorch_model-00006-of-00008.bin",
468
+ "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00006-of-00008.bin",
469
+ "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00005-of-00008.bin",
470
+ "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00005-of-00008.bin",
471
+ "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00005-of-00008.bin",
472
+ "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00005-of-00008.bin",
473
+ "model.layers.22.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00006-of-00008.bin",
474
+ "model.layers.22.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00006-of-00008.bin",
475
+ "model.layers.22.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00006-of-00008.bin",
476
+ "model.layers.22.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00006-of-00008.bin",
477
+ "model.layers.22.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00006-of-00008.bin",
478
+ "model.layers.22.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00006-of-00008.bin",
479
+ "model.layers.22.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00006-of-00008.bin",
480
+ "model.layers.22.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00006-of-00008.bin",
481
+ "model.layers.22.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00006-of-00008.bin",
482
+ "model.layers.22.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00006-of-00008.bin",
483
+ "model.layers.22.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00006-of-00008.bin",
484
+ "model.layers.22.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00006-of-00008.bin",
485
+ "model.layers.22.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00006-of-00008.bin",
486
+ "model.layers.22.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00006-of-00008.bin",
487
+ "model.layers.22.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00006-of-00008.bin",
488
+ "model.layers.22.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00006-of-00008.bin",
489
+ "model.layers.22.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00006-of-00008.bin",
490
+ "model.layers.22.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00006-of-00008.bin",
491
+ "model.layers.22.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00006-of-00008.bin",
492
+ "model.layers.22.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00006-of-00008.bin",
493
+ "model.layers.22.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00006-of-00008.bin",
494
+ "model.layers.22.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00006-of-00008.bin",
495
+ "model.layers.22.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00006-of-00008.bin",
496
+ "model.layers.22.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00006-of-00008.bin",
497
+ "model.layers.22.block_sparse_moe.gate.weight": "pytorch_model-00006-of-00008.bin",
498
+ "model.layers.22.input_layernorm.weight": "pytorch_model-00006-of-00008.bin",
499
+ "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00006-of-00008.bin",
500
+ "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00006-of-00008.bin",
501
+ "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00006-of-00008.bin",
502
+ "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00006-of-00008.bin",
503
+ "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00006-of-00008.bin",
504
+ "model.layers.23.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00006-of-00008.bin",
505
+ "model.layers.23.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00006-of-00008.bin",
506
+ "model.layers.23.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00006-of-00008.bin",
507
+ "model.layers.23.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00006-of-00008.bin",
508
+ "model.layers.23.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00006-of-00008.bin",
509
+ "model.layers.23.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00006-of-00008.bin",
510
+ "model.layers.23.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00006-of-00008.bin",
511
+ "model.layers.23.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00006-of-00008.bin",
512
+ "model.layers.23.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00006-of-00008.bin",
513
+ "model.layers.23.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00006-of-00008.bin",
514
+ "model.layers.23.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00006-of-00008.bin",
515
+ "model.layers.23.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00006-of-00008.bin",
516
+ "model.layers.23.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00006-of-00008.bin",
517
+ "model.layers.23.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00006-of-00008.bin",
518
+ "model.layers.23.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00006-of-00008.bin",
519
+ "model.layers.23.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00006-of-00008.bin",
520
+ "model.layers.23.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00006-of-00008.bin",
521
+ "model.layers.23.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00006-of-00008.bin",
522
+ "model.layers.23.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00006-of-00008.bin",
523
+ "model.layers.23.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00006-of-00008.bin",
524
+ "model.layers.23.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00006-of-00008.bin",
525
+ "model.layers.23.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00006-of-00008.bin",
526
+ "model.layers.23.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00006-of-00008.bin",
527
+ "model.layers.23.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00006-of-00008.bin",
528
+ "model.layers.23.block_sparse_moe.gate.weight": "pytorch_model-00006-of-00008.bin",
529
+ "model.layers.23.input_layernorm.weight": "pytorch_model-00006-of-00008.bin",
530
+ "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00006-of-00008.bin",
531
+ "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00006-of-00008.bin",
532
+ "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00006-of-00008.bin",
533
+ "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00006-of-00008.bin",
534
+ "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00006-of-00008.bin",
535
+ "model.layers.24.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00006-of-00008.bin",
536
+ "model.layers.24.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00006-of-00008.bin",
537
+ "model.layers.24.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00006-of-00008.bin",
538
+ "model.layers.24.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00006-of-00008.bin",
539
+ "model.layers.24.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00006-of-00008.bin",
540
+ "model.layers.24.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00006-of-00008.bin",
541
+ "model.layers.24.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00006-of-00008.bin",
542
+ "model.layers.24.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00006-of-00008.bin",
543
+ "model.layers.24.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00006-of-00008.bin",
544
+ "model.layers.24.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00006-of-00008.bin",
545
+ "model.layers.24.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00006-of-00008.bin",
546
+ "model.layers.24.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00006-of-00008.bin",
547
+ "model.layers.24.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00006-of-00008.bin",
548
+ "model.layers.24.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00006-of-00008.bin",
549
+ "model.layers.24.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00006-of-00008.bin",
550
+ "model.layers.24.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00006-of-00008.bin",
551
+ "model.layers.24.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00006-of-00008.bin",
552
+ "model.layers.24.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00006-of-00008.bin",
553
+ "model.layers.24.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00006-of-00008.bin",
554
+ "model.layers.24.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00006-of-00008.bin",
555
+ "model.layers.24.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00006-of-00008.bin",
556
+ "model.layers.24.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00006-of-00008.bin",
557
+ "model.layers.24.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00006-of-00008.bin",
558
+ "model.layers.24.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00006-of-00008.bin",
559
+ "model.layers.24.block_sparse_moe.gate.weight": "pytorch_model-00006-of-00008.bin",
560
+ "model.layers.24.input_layernorm.weight": "pytorch_model-00006-of-00008.bin",
561
+ "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00006-of-00008.bin",
562
+ "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00006-of-00008.bin",
563
+ "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00006-of-00008.bin",
564
+ "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00006-of-00008.bin",
565
+ "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00006-of-00008.bin",
566
+ "model.layers.25.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00006-of-00008.bin",
567
+ "model.layers.25.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00006-of-00008.bin",
568
+ "model.layers.25.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00006-of-00008.bin",
569
+ "model.layers.25.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00006-of-00008.bin",
570
+ "model.layers.25.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00006-of-00008.bin",
571
+ "model.layers.25.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00006-of-00008.bin",
572
+ "model.layers.25.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00006-of-00008.bin",
573
+ "model.layers.25.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00006-of-00008.bin",
574
+ "model.layers.25.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00006-of-00008.bin",
575
+ "model.layers.25.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00007-of-00008.bin",
576
+ "model.layers.25.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00007-of-00008.bin",
577
+ "model.layers.25.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00007-of-00008.bin",
578
+ "model.layers.25.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00007-of-00008.bin",
579
+ "model.layers.25.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00007-of-00008.bin",
580
+ "model.layers.25.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00007-of-00008.bin",
581
+ "model.layers.25.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00007-of-00008.bin",
582
+ "model.layers.25.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00007-of-00008.bin",
583
+ "model.layers.25.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00007-of-00008.bin",
584
+ "model.layers.25.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00007-of-00008.bin",
585
+ "model.layers.25.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00007-of-00008.bin",
586
+ "model.layers.25.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00007-of-00008.bin",
587
+ "model.layers.25.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00007-of-00008.bin",
588
+ "model.layers.25.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00007-of-00008.bin",
589
+ "model.layers.25.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00007-of-00008.bin",
590
+ "model.layers.25.block_sparse_moe.gate.weight": "pytorch_model-00006-of-00008.bin",
591
+ "model.layers.25.input_layernorm.weight": "pytorch_model-00007-of-00008.bin",
592
+ "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00007-of-00008.bin",
593
+ "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00006-of-00008.bin",
594
+ "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00006-of-00008.bin",
595
+ "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00006-of-00008.bin",
596
+ "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00006-of-00008.bin",
597
+ "model.layers.26.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00007-of-00008.bin",
598
+ "model.layers.26.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00007-of-00008.bin",
599
+ "model.layers.26.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00007-of-00008.bin",
600
+ "model.layers.26.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00007-of-00008.bin",
601
+ "model.layers.26.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00007-of-00008.bin",
602
+ "model.layers.26.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00007-of-00008.bin",
603
+ "model.layers.26.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00007-of-00008.bin",
604
+ "model.layers.26.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00007-of-00008.bin",
605
+ "model.layers.26.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00007-of-00008.bin",
606
+ "model.layers.26.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00007-of-00008.bin",
607
+ "model.layers.26.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00007-of-00008.bin",
608
+ "model.layers.26.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00007-of-00008.bin",
609
+ "model.layers.26.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00007-of-00008.bin",
610
+ "model.layers.26.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00007-of-00008.bin",
611
+ "model.layers.26.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00007-of-00008.bin",
612
+ "model.layers.26.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00007-of-00008.bin",
613
+ "model.layers.26.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00007-of-00008.bin",
614
+ "model.layers.26.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00007-of-00008.bin",
615
+ "model.layers.26.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00007-of-00008.bin",
616
+ "model.layers.26.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00007-of-00008.bin",
617
+ "model.layers.26.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00007-of-00008.bin",
618
+ "model.layers.26.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00007-of-00008.bin",
619
+ "model.layers.26.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00007-of-00008.bin",
620
+ "model.layers.26.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00007-of-00008.bin",
621
+ "model.layers.26.block_sparse_moe.gate.weight": "pytorch_model-00007-of-00008.bin",
622
+ "model.layers.26.input_layernorm.weight": "pytorch_model-00007-of-00008.bin",
623
+ "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00007-of-00008.bin",
624
+ "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00007-of-00008.bin",
625
+ "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00007-of-00008.bin",
626
+ "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00007-of-00008.bin",
627
+ "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00007-of-00008.bin",
628
+ "model.layers.27.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00007-of-00008.bin",
629
+ "model.layers.27.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00007-of-00008.bin",
630
+ "model.layers.27.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00007-of-00008.bin",
631
+ "model.layers.27.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00007-of-00008.bin",
632
+ "model.layers.27.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00007-of-00008.bin",
633
+ "model.layers.27.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00007-of-00008.bin",
634
+ "model.layers.27.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00007-of-00008.bin",
635
+ "model.layers.27.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00007-of-00008.bin",
636
+ "model.layers.27.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00007-of-00008.bin",
637
+ "model.layers.27.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00007-of-00008.bin",
638
+ "model.layers.27.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00007-of-00008.bin",
639
+ "model.layers.27.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00007-of-00008.bin",
640
+ "model.layers.27.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00007-of-00008.bin",
641
+ "model.layers.27.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00007-of-00008.bin",
642
+ "model.layers.27.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00007-of-00008.bin",
643
+ "model.layers.27.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00007-of-00008.bin",
644
+ "model.layers.27.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00007-of-00008.bin",
645
+ "model.layers.27.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00007-of-00008.bin",
646
+ "model.layers.27.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00007-of-00008.bin",
647
+ "model.layers.27.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00007-of-00008.bin",
648
+ "model.layers.27.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00007-of-00008.bin",
649
+ "model.layers.27.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00007-of-00008.bin",
650
+ "model.layers.27.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00007-of-00008.bin",
651
+ "model.layers.27.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00007-of-00008.bin",
652
+ "model.layers.27.block_sparse_moe.gate.weight": "pytorch_model-00007-of-00008.bin",
653
+ "model.layers.27.input_layernorm.weight": "pytorch_model-00007-of-00008.bin",
654
+ "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00007-of-00008.bin",
655
+ "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00007-of-00008.bin",
656
+ "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00007-of-00008.bin",
657
+ "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00007-of-00008.bin",
658
+ "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00007-of-00008.bin",
659
+ "model.layers.28.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00007-of-00008.bin",
660
+ "model.layers.28.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00007-of-00008.bin",
661
+ "model.layers.28.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00007-of-00008.bin",
662
+ "model.layers.28.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00007-of-00008.bin",
663
+ "model.layers.28.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00007-of-00008.bin",
664
+ "model.layers.28.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00007-of-00008.bin",
665
+ "model.layers.28.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00007-of-00008.bin",
666
+ "model.layers.28.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00007-of-00008.bin",
667
+ "model.layers.28.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00007-of-00008.bin",
668
+ "model.layers.28.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00007-of-00008.bin",
669
+ "model.layers.28.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00007-of-00008.bin",
670
+ "model.layers.28.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00007-of-00008.bin",
671
+ "model.layers.28.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00007-of-00008.bin",
672
+ "model.layers.28.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00007-of-00008.bin",
673
+ "model.layers.28.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00007-of-00008.bin",
674
+ "model.layers.28.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00007-of-00008.bin",
675
+ "model.layers.28.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00007-of-00008.bin",
676
+ "model.layers.28.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00007-of-00008.bin",
677
+ "model.layers.28.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00007-of-00008.bin",
678
+ "model.layers.28.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00007-of-00008.bin",
679
+ "model.layers.28.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00007-of-00008.bin",
680
+ "model.layers.28.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00007-of-00008.bin",
681
+ "model.layers.28.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00007-of-00008.bin",
682
+ "model.layers.28.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00007-of-00008.bin",
683
+ "model.layers.28.block_sparse_moe.gate.weight": "pytorch_model-00007-of-00008.bin",
684
+ "model.layers.28.input_layernorm.weight": "pytorch_model-00007-of-00008.bin",
685
+ "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00007-of-00008.bin",
686
+ "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00007-of-00008.bin",
687
+ "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00007-of-00008.bin",
688
+ "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00007-of-00008.bin",
689
+ "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00007-of-00008.bin",
690
+ "model.layers.29.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00007-of-00008.bin",
691
+ "model.layers.29.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00007-of-00008.bin",
692
+ "model.layers.29.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00007-of-00008.bin",
693
+ "model.layers.29.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00007-of-00008.bin",
694
+ "model.layers.29.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00007-of-00008.bin",
695
+ "model.layers.29.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00007-of-00008.bin",
696
+ "model.layers.29.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00007-of-00008.bin",
697
+ "model.layers.29.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00007-of-00008.bin",
698
+ "model.layers.29.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00007-of-00008.bin",
699
+ "model.layers.29.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00007-of-00008.bin",
700
+ "model.layers.29.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00007-of-00008.bin",
701
+ "model.layers.29.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00007-of-00008.bin",
702
+ "model.layers.29.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00007-of-00008.bin",
703
+ "model.layers.29.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00007-of-00008.bin",
704
+ "model.layers.29.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00007-of-00008.bin",
705
+ "model.layers.29.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00007-of-00008.bin",
706
+ "model.layers.29.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00007-of-00008.bin",
707
+ "model.layers.29.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00008-of-00008.bin",
708
+ "model.layers.29.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00008-of-00008.bin",
709
+ "model.layers.29.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00008-of-00008.bin",
710
+ "model.layers.29.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00008-of-00008.bin",
711
+ "model.layers.29.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00008-of-00008.bin",
712
+ "model.layers.29.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00008-of-00008.bin",
713
+ "model.layers.29.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00008-of-00008.bin",
714
+ "model.layers.29.block_sparse_moe.gate.weight": "pytorch_model-00007-of-00008.bin",
715
+ "model.layers.29.input_layernorm.weight": "pytorch_model-00008-of-00008.bin",
716
+ "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00008-of-00008.bin",
717
+ "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00007-of-00008.bin",
718
+ "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00007-of-00008.bin",
719
+ "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00007-of-00008.bin",
720
+ "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00007-of-00008.bin",
721
+ "model.layers.3.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00001-of-00008.bin",
722
+ "model.layers.3.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00001-of-00008.bin",
723
+ "model.layers.3.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00001-of-00008.bin",
724
+ "model.layers.3.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00001-of-00008.bin",
725
+ "model.layers.3.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00001-of-00008.bin",
726
+ "model.layers.3.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00001-of-00008.bin",
727
+ "model.layers.3.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00001-of-00008.bin",
728
+ "model.layers.3.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00001-of-00008.bin",
729
+ "model.layers.3.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00001-of-00008.bin",
730
+ "model.layers.3.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00001-of-00008.bin",
731
+ "model.layers.3.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00001-of-00008.bin",
732
+ "model.layers.3.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00001-of-00008.bin",
733
+ "model.layers.3.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00001-of-00008.bin",
734
+ "model.layers.3.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00001-of-00008.bin",
735
+ "model.layers.3.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00001-of-00008.bin",
736
+ "model.layers.3.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00001-of-00008.bin",
737
+ "model.layers.3.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00001-of-00008.bin",
738
+ "model.layers.3.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00001-of-00008.bin",
739
+ "model.layers.3.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00001-of-00008.bin",
740
+ "model.layers.3.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00002-of-00008.bin",
741
+ "model.layers.3.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00002-of-00008.bin",
742
+ "model.layers.3.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00002-of-00008.bin",
743
+ "model.layers.3.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00002-of-00008.bin",
744
+ "model.layers.3.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00002-of-00008.bin",
745
+ "model.layers.3.block_sparse_moe.gate.weight": "pytorch_model-00001-of-00008.bin",
746
+ "model.layers.3.input_layernorm.weight": "pytorch_model-00002-of-00008.bin",
747
+ "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00002-of-00008.bin",
748
+ "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00001-of-00008.bin",
749
+ "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00008.bin",
750
+ "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00001-of-00008.bin",
751
+ "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00001-of-00008.bin",
752
+ "model.layers.30.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00008-of-00008.bin",
753
+ "model.layers.30.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00008-of-00008.bin",
754
+ "model.layers.30.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00008-of-00008.bin",
755
+ "model.layers.30.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00008-of-00008.bin",
756
+ "model.layers.30.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00008-of-00008.bin",
757
+ "model.layers.30.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00008-of-00008.bin",
758
+ "model.layers.30.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00008-of-00008.bin",
759
+ "model.layers.30.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00008-of-00008.bin",
760
+ "model.layers.30.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00008-of-00008.bin",
761
+ "model.layers.30.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00008-of-00008.bin",
762
+ "model.layers.30.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00008-of-00008.bin",
763
+ "model.layers.30.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00008-of-00008.bin",
764
+ "model.layers.30.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00008-of-00008.bin",
765
+ "model.layers.30.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00008-of-00008.bin",
766
+ "model.layers.30.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00008-of-00008.bin",
767
+ "model.layers.30.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00008-of-00008.bin",
768
+ "model.layers.30.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00008-of-00008.bin",
769
+ "model.layers.30.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00008-of-00008.bin",
770
+ "model.layers.30.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00008-of-00008.bin",
771
+ "model.layers.30.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00008-of-00008.bin",
772
+ "model.layers.30.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00008-of-00008.bin",
773
+ "model.layers.30.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00008-of-00008.bin",
774
+ "model.layers.30.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00008-of-00008.bin",
775
+ "model.layers.30.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00008-of-00008.bin",
776
+ "model.layers.30.block_sparse_moe.gate.weight": "pytorch_model-00008-of-00008.bin",
777
+ "model.layers.30.input_layernorm.weight": "pytorch_model-00008-of-00008.bin",
778
+ "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00008-of-00008.bin",
779
+ "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00008-of-00008.bin",
780
+ "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00008-of-00008.bin",
781
+ "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00008-of-00008.bin",
782
+ "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00008-of-00008.bin",
783
+ "model.layers.31.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00008-of-00008.bin",
784
+ "model.layers.31.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00008-of-00008.bin",
785
+ "model.layers.31.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00008-of-00008.bin",
786
+ "model.layers.31.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00008-of-00008.bin",
787
+ "model.layers.31.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00008-of-00008.bin",
788
+ "model.layers.31.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00008-of-00008.bin",
789
+ "model.layers.31.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00008-of-00008.bin",
790
+ "model.layers.31.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00008-of-00008.bin",
791
+ "model.layers.31.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00008-of-00008.bin",
792
+ "model.layers.31.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00008-of-00008.bin",
793
+ "model.layers.31.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00008-of-00008.bin",
794
+ "model.layers.31.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00008-of-00008.bin",
795
+ "model.layers.31.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00008-of-00008.bin",
796
+ "model.layers.31.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00008-of-00008.bin",
797
+ "model.layers.31.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00008-of-00008.bin",
798
+ "model.layers.31.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00008-of-00008.bin",
799
+ "model.layers.31.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00008-of-00008.bin",
800
+ "model.layers.31.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00008-of-00008.bin",
801
+ "model.layers.31.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00008-of-00008.bin",
802
+ "model.layers.31.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00008-of-00008.bin",
803
+ "model.layers.31.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00008-of-00008.bin",
804
+ "model.layers.31.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00008-of-00008.bin",
805
+ "model.layers.31.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00008-of-00008.bin",
806
+ "model.layers.31.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00008-of-00008.bin",
807
+ "model.layers.31.block_sparse_moe.gate.weight": "pytorch_model-00008-of-00008.bin",
808
+ "model.layers.31.input_layernorm.weight": "pytorch_model-00008-of-00008.bin",
809
+ "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00008-of-00008.bin",
810
+ "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00008-of-00008.bin",
811
+ "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00008-of-00008.bin",
812
+ "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00008-of-00008.bin",
813
+ "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00008-of-00008.bin",
814
+ "model.layers.4.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00002-of-00008.bin",
815
+ "model.layers.4.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00002-of-00008.bin",
816
+ "model.layers.4.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00002-of-00008.bin",
817
+ "model.layers.4.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00002-of-00008.bin",
818
+ "model.layers.4.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00002-of-00008.bin",
819
+ "model.layers.4.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00002-of-00008.bin",
820
+ "model.layers.4.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00002-of-00008.bin",
821
+ "model.layers.4.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00002-of-00008.bin",
822
+ "model.layers.4.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00002-of-00008.bin",
823
+ "model.layers.4.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00002-of-00008.bin",
824
+ "model.layers.4.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00002-of-00008.bin",
825
+ "model.layers.4.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00002-of-00008.bin",
826
+ "model.layers.4.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00002-of-00008.bin",
827
+ "model.layers.4.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00002-of-00008.bin",
828
+ "model.layers.4.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00002-of-00008.bin",
829
+ "model.layers.4.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00002-of-00008.bin",
830
+ "model.layers.4.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00002-of-00008.bin",
831
+ "model.layers.4.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00002-of-00008.bin",
832
+ "model.layers.4.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00002-of-00008.bin",
833
+ "model.layers.4.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00002-of-00008.bin",
834
+ "model.layers.4.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00002-of-00008.bin",
835
+ "model.layers.4.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00002-of-00008.bin",
836
+ "model.layers.4.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00002-of-00008.bin",
837
+ "model.layers.4.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00002-of-00008.bin",
838
+ "model.layers.4.block_sparse_moe.gate.weight": "pytorch_model-00002-of-00008.bin",
839
+ "model.layers.4.input_layernorm.weight": "pytorch_model-00002-of-00008.bin",
840
+ "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00002-of-00008.bin",
841
+ "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00002-of-00008.bin",
842
+ "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00002-of-00008.bin",
843
+ "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00002-of-00008.bin",
844
+ "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00002-of-00008.bin",
845
+ "model.layers.5.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00002-of-00008.bin",
846
+ "model.layers.5.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00002-of-00008.bin",
847
+ "model.layers.5.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00002-of-00008.bin",
848
+ "model.layers.5.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00002-of-00008.bin",
849
+ "model.layers.5.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00002-of-00008.bin",
850
+ "model.layers.5.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00002-of-00008.bin",
851
+ "model.layers.5.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00002-of-00008.bin",
852
+ "model.layers.5.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00002-of-00008.bin",
853
+ "model.layers.5.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00002-of-00008.bin",
854
+ "model.layers.5.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00002-of-00008.bin",
855
+ "model.layers.5.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00002-of-00008.bin",
856
+ "model.layers.5.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00002-of-00008.bin",
857
+ "model.layers.5.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00002-of-00008.bin",
858
+ "model.layers.5.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00002-of-00008.bin",
859
+ "model.layers.5.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00002-of-00008.bin",
860
+ "model.layers.5.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00002-of-00008.bin",
861
+ "model.layers.5.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00002-of-00008.bin",
862
+ "model.layers.5.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00002-of-00008.bin",
863
+ "model.layers.5.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00002-of-00008.bin",
864
+ "model.layers.5.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00002-of-00008.bin",
865
+ "model.layers.5.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00002-of-00008.bin",
866
+ "model.layers.5.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00002-of-00008.bin",
867
+ "model.layers.5.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00002-of-00008.bin",
868
+ "model.layers.5.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00002-of-00008.bin",
869
+ "model.layers.5.block_sparse_moe.gate.weight": "pytorch_model-00002-of-00008.bin",
870
+ "model.layers.5.input_layernorm.weight": "pytorch_model-00002-of-00008.bin",
871
+ "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00002-of-00008.bin",
872
+ "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00002-of-00008.bin",
873
+ "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00002-of-00008.bin",
874
+ "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00002-of-00008.bin",
875
+ "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00002-of-00008.bin",
876
+ "model.layers.6.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00002-of-00008.bin",
877
+ "model.layers.6.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00002-of-00008.bin",
878
+ "model.layers.6.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00002-of-00008.bin",
879
+ "model.layers.6.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00002-of-00008.bin",
880
+ "model.layers.6.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00002-of-00008.bin",
881
+ "model.layers.6.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00002-of-00008.bin",
882
+ "model.layers.6.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00002-of-00008.bin",
883
+ "model.layers.6.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00002-of-00008.bin",
884
+ "model.layers.6.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00002-of-00008.bin",
885
+ "model.layers.6.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00002-of-00008.bin",
886
+ "model.layers.6.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00002-of-00008.bin",
887
+ "model.layers.6.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00002-of-00008.bin",
888
+ "model.layers.6.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00002-of-00008.bin",
889
+ "model.layers.6.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00002-of-00008.bin",
890
+ "model.layers.6.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00002-of-00008.bin",
891
+ "model.layers.6.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00002-of-00008.bin",
892
+ "model.layers.6.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00002-of-00008.bin",
893
+ "model.layers.6.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00002-of-00008.bin",
894
+ "model.layers.6.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00002-of-00008.bin",
895
+ "model.layers.6.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00002-of-00008.bin",
896
+ "model.layers.6.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00002-of-00008.bin",
897
+ "model.layers.6.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00002-of-00008.bin",
898
+ "model.layers.6.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00002-of-00008.bin",
899
+ "model.layers.6.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00002-of-00008.bin",
900
+ "model.layers.6.block_sparse_moe.gate.weight": "pytorch_model-00002-of-00008.bin",
901
+ "model.layers.6.input_layernorm.weight": "pytorch_model-00002-of-00008.bin",
902
+ "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00002-of-00008.bin",
903
+ "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00002-of-00008.bin",
904
+ "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00002-of-00008.bin",
905
+ "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00002-of-00008.bin",
906
+ "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00002-of-00008.bin",
907
+ "model.layers.7.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00002-of-00008.bin",
908
+ "model.layers.7.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00002-of-00008.bin",
909
+ "model.layers.7.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00002-of-00008.bin",
910
+ "model.layers.7.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00002-of-00008.bin",
911
+ "model.layers.7.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00002-of-00008.bin",
912
+ "model.layers.7.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00002-of-00008.bin",
913
+ "model.layers.7.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00002-of-00008.bin",
914
+ "model.layers.7.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00002-of-00008.bin",
915
+ "model.layers.7.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00002-of-00008.bin",
916
+ "model.layers.7.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00002-of-00008.bin",
917
+ "model.layers.7.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00002-of-00008.bin",
918
+ "model.layers.7.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00002-of-00008.bin",
919
+ "model.layers.7.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00002-of-00008.bin",
920
+ "model.layers.7.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00002-of-00008.bin",
921
+ "model.layers.7.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00002-of-00008.bin",
922
+ "model.layers.7.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00002-of-00008.bin",
923
+ "model.layers.7.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00002-of-00008.bin",
924
+ "model.layers.7.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00002-of-00008.bin",
925
+ "model.layers.7.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00002-of-00008.bin",
926
+ "model.layers.7.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00002-of-00008.bin",
927
+ "model.layers.7.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00002-of-00008.bin",
928
+ "model.layers.7.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00002-of-00008.bin",
929
+ "model.layers.7.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00002-of-00008.bin",
930
+ "model.layers.7.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00002-of-00008.bin",
931
+ "model.layers.7.block_sparse_moe.gate.weight": "pytorch_model-00002-of-00008.bin",
932
+ "model.layers.7.input_layernorm.weight": "pytorch_model-00002-of-00008.bin",
933
+ "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00002-of-00008.bin",
934
+ "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00002-of-00008.bin",
935
+ "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00002-of-00008.bin",
936
+ "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00002-of-00008.bin",
937
+ "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00002-of-00008.bin",
938
+ "model.layers.8.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00002-of-00008.bin",
939
+ "model.layers.8.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00002-of-00008.bin",
940
+ "model.layers.8.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00003-of-00008.bin",
941
+ "model.layers.8.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00003-of-00008.bin",
942
+ "model.layers.8.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00003-of-00008.bin",
943
+ "model.layers.8.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00003-of-00008.bin",
944
+ "model.layers.8.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00003-of-00008.bin",
945
+ "model.layers.8.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00003-of-00008.bin",
946
+ "model.layers.8.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00003-of-00008.bin",
947
+ "model.layers.8.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00003-of-00008.bin",
948
+ "model.layers.8.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00003-of-00008.bin",
949
+ "model.layers.8.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00003-of-00008.bin",
950
+ "model.layers.8.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00003-of-00008.bin",
951
+ "model.layers.8.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00003-of-00008.bin",
952
+ "model.layers.8.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00003-of-00008.bin",
953
+ "model.layers.8.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00003-of-00008.bin",
954
+ "model.layers.8.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00003-of-00008.bin",
955
+ "model.layers.8.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00003-of-00008.bin",
956
+ "model.layers.8.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00003-of-00008.bin",
957
+ "model.layers.8.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00003-of-00008.bin",
958
+ "model.layers.8.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00003-of-00008.bin",
959
+ "model.layers.8.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00003-of-00008.bin",
960
+ "model.layers.8.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00003-of-00008.bin",
961
+ "model.layers.8.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00003-of-00008.bin",
962
+ "model.layers.8.block_sparse_moe.gate.weight": "pytorch_model-00002-of-00008.bin",
963
+ "model.layers.8.input_layernorm.weight": "pytorch_model-00003-of-00008.bin",
964
+ "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00003-of-00008.bin",
965
+ "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00002-of-00008.bin",
966
+ "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00002-of-00008.bin",
967
+ "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00002-of-00008.bin",
968
+ "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00002-of-00008.bin",
969
+ "model.layers.9.block_sparse_moe.experts.0.w1.weight": "pytorch_model-00003-of-00008.bin",
970
+ "model.layers.9.block_sparse_moe.experts.0.w2.weight": "pytorch_model-00003-of-00008.bin",
971
+ "model.layers.9.block_sparse_moe.experts.0.w3.weight": "pytorch_model-00003-of-00008.bin",
972
+ "model.layers.9.block_sparse_moe.experts.1.w1.weight": "pytorch_model-00003-of-00008.bin",
973
+ "model.layers.9.block_sparse_moe.experts.1.w2.weight": "pytorch_model-00003-of-00008.bin",
974
+ "model.layers.9.block_sparse_moe.experts.1.w3.weight": "pytorch_model-00003-of-00008.bin",
975
+ "model.layers.9.block_sparse_moe.experts.2.w1.weight": "pytorch_model-00003-of-00008.bin",
976
+ "model.layers.9.block_sparse_moe.experts.2.w2.weight": "pytorch_model-00003-of-00008.bin",
977
+ "model.layers.9.block_sparse_moe.experts.2.w3.weight": "pytorch_model-00003-of-00008.bin",
978
+ "model.layers.9.block_sparse_moe.experts.3.w1.weight": "pytorch_model-00003-of-00008.bin",
979
+ "model.layers.9.block_sparse_moe.experts.3.w2.weight": "pytorch_model-00003-of-00008.bin",
980
+ "model.layers.9.block_sparse_moe.experts.3.w3.weight": "pytorch_model-00003-of-00008.bin",
981
+ "model.layers.9.block_sparse_moe.experts.4.w1.weight": "pytorch_model-00003-of-00008.bin",
982
+ "model.layers.9.block_sparse_moe.experts.4.w2.weight": "pytorch_model-00003-of-00008.bin",
983
+ "model.layers.9.block_sparse_moe.experts.4.w3.weight": "pytorch_model-00003-of-00008.bin",
984
+ "model.layers.9.block_sparse_moe.experts.5.w1.weight": "pytorch_model-00003-of-00008.bin",
985
+ "model.layers.9.block_sparse_moe.experts.5.w2.weight": "pytorch_model-00003-of-00008.bin",
986
+ "model.layers.9.block_sparse_moe.experts.5.w3.weight": "pytorch_model-00003-of-00008.bin",
987
+ "model.layers.9.block_sparse_moe.experts.6.w1.weight": "pytorch_model-00003-of-00008.bin",
988
+ "model.layers.9.block_sparse_moe.experts.6.w2.weight": "pytorch_model-00003-of-00008.bin",
989
+ "model.layers.9.block_sparse_moe.experts.6.w3.weight": "pytorch_model-00003-of-00008.bin",
990
+ "model.layers.9.block_sparse_moe.experts.7.w1.weight": "pytorch_model-00003-of-00008.bin",
991
+ "model.layers.9.block_sparse_moe.experts.7.w2.weight": "pytorch_model-00003-of-00008.bin",
992
+ "model.layers.9.block_sparse_moe.experts.7.w3.weight": "pytorch_model-00003-of-00008.bin",
993
+ "model.layers.9.block_sparse_moe.gate.weight": "pytorch_model-00003-of-00008.bin",
994
+ "model.layers.9.input_layernorm.weight": "pytorch_model-00003-of-00008.bin",
995
+ "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00003-of-00008.bin",
996
+ "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00003-of-00008.bin",
997
+ "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00003-of-00008.bin",
998
+ "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00003-of-00008.bin",
999
+ "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00003-of-00008.bin",
1000
+ "model.norm.weight": "pytorch_model-00008-of-00008.bin"
1001
+ }
1002
+ }
qwen.tiktoken ADDED
The diff for this file is too large to render. See raw diff
 
special_tokens_map.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>"
5
+ ],
6
+ "bos_token": "<|endoftext|>",
7
+ "eos_token": "<|endoftext|>",
8
+ "unk_token": "<|endoftext|>"
9
+ }
tokenization_qwen.py ADDED
@@ -0,0 +1,230 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Copyright (c) Alibaba Cloud.
2
+ #
3
+ # This source code is licensed under the license found in the
4
+ # LICENSE file in the root directory of this source tree.
5
+
6
+ """Tokenization classes for QWen."""
7
+
8
+ import base64
9
+ import logging
10
+ import os
11
+ import unicodedata
12
+ from typing import Collection, Dict, List, Set, Tuple, Union
13
+
14
+ import tiktoken
15
+ from transformers import PreTrainedTokenizer, AddedToken
16
+
17
+ logger = logging.getLogger(__name__)
18
+
19
+
20
+ VOCAB_FILES_NAMES = {"vocab_file": "qwen.tiktoken"}
21
+
22
+ PAT_STR = r"""(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"""
23
+ ENDOFTEXT = "<|endoftext|>"
24
+ IMSTART = "<|im_start|>"
25
+ IMEND = "<|im_end|>"
26
+ # as the default behavior is changed to allow special tokens in
27
+ # regular texts, the surface forms of special tokens need to be
28
+ # as different as possible to minimize the impact
29
+ EXTRAS = tuple((f"<|extra_{i}|>" for i in range(205)))
30
+ SPECIAL_TOKENS = (
31
+ ENDOFTEXT,
32
+ IMSTART,
33
+ IMEND,
34
+ ) + EXTRAS
35
+
36
+
37
+ def _load_tiktoken_bpe(tiktoken_bpe_file: str) -> Dict[bytes, int]:
38
+ with open(tiktoken_bpe_file, "rb") as f:
39
+ contents = f.read()
40
+ return {
41
+ base64.b64decode(token): int(rank)
42
+ for token, rank in (line.split() for line in contents.splitlines() if line)
43
+ }
44
+
45
+ class QWenTokenizer(PreTrainedTokenizer):
46
+ """QWen tokenizer."""
47
+
48
+ vocab_files_names = VOCAB_FILES_NAMES
49
+
50
+ def __init__(
51
+ self,
52
+ vocab_file,
53
+ errors="replace",
54
+ **kwargs,
55
+ ):
56
+ super().__init__(**kwargs)
57
+
58
+ self.errors = errors # how to handle errors in decoding
59
+
60
+ self.mergeable_ranks = _load_tiktoken_bpe(vocab_file) # type: dict[bytes, int]
61
+ self.special_tokens = {
62
+ token: index
63
+ for index, token in enumerate(
64
+ SPECIAL_TOKENS, start=len(self.mergeable_ranks)
65
+ )
66
+ }
67
+
68
+ enc = tiktoken.Encoding(
69
+ "Qwen",
70
+ pat_str=PAT_STR,
71
+ mergeable_ranks=self.mergeable_ranks,
72
+ special_tokens=self.special_tokens,
73
+ )
74
+ assert (
75
+ len(self.mergeable_ranks) + len(self.special_tokens) == enc.n_vocab
76
+ ), f"{len(self.mergeable_ranks) + len(self.special_tokens)} != {enc.n_vocab} in encoding"
77
+
78
+ self.decoder = {
79
+ v: k for k, v in self.mergeable_ranks.items()
80
+ } # type: dict[int, bytes|str]
81
+ self.decoder.update({v: k for k, v in self.special_tokens.items()})
82
+
83
+ self.tokenizer = enc # type: tiktoken.Encoding
84
+
85
+ self.eod_id = self.tokenizer.eot_token
86
+ self.im_start_id = self.special_tokens[IMSTART]
87
+ self.im_end_id = self.special_tokens[IMEND]
88
+
89
+ def __len__(self) -> int:
90
+ return self.tokenizer.n_vocab
91
+
92
+ def get_vocab(self) -> Dict[bytes, int]:
93
+ return self.mergeable_ranks
94
+
95
+ def convert_tokens_to_ids(
96
+ self, tokens: Union[bytes, str, List[Union[bytes, str]]]
97
+ ) -> List[int]:
98
+ ids = []
99
+ if isinstance(tokens, (str, bytes)):
100
+ if tokens in self.special_tokens:
101
+ return self.special_tokens[tokens]
102
+ else:
103
+ return self.mergeable_ranks.get(tokens)
104
+ for token in tokens:
105
+ if token in self.special_tokens:
106
+ ids.append(self.special_tokens[token])
107
+ else:
108
+ ids.append(self.mergeable_ranks.get(token))
109
+ return ids
110
+
111
+ def _add_tokens(self, new_tokens: Union[List[str], List[AddedToken]], special_tokens: bool = False) -> int:
112
+ if not special_tokens and new_tokens:
113
+ raise ValueError('Adding regular tokens is not supported')
114
+ for token in new_tokens:
115
+ surface_form = token.content if isinstance(token, AddedToken) else token
116
+ if surface_form not in SPECIAL_TOKENS:
117
+ raise ValueError('Adding unknown special tokens is not supported')
118
+ return 0
119
+
120
+ def save_vocabulary(self, save_directory: str, **kwargs) -> Tuple[str]:
121
+ """
122
+ Save only the vocabulary of the tokenizer (vocabulary).
123
+
124
+ Returns:
125
+ `Tuple(str)`: Paths to the files saved.
126
+ """
127
+ file_path = os.path.join(save_directory, "qwen.tiktoken")
128
+ with open(file_path, "w", encoding="utf8") as w:
129
+ for k, v in self.mergeable_ranks.items():
130
+ line = base64.b64encode(k).decode("utf8") + " " + str(v) + "\n"
131
+ w.write(line)
132
+ return (file_path,)
133
+
134
+ def tokenize(
135
+ self,
136
+ text: str,
137
+ allowed_special: Union[Set, str] = "all",
138
+ disallowed_special: Union[Collection, str] = (),
139
+ **kwargs,
140
+ ) -> List[Union[bytes, str]]:
141
+ """
142
+ Converts a string in a sequence of tokens.
143
+
144
+ Args:
145
+ text (`str`):
146
+ The sequence to be encoded.
147
+ allowed_special (`Literal["all"]` or `set`):
148
+ The surface forms of the tokens to be encoded as special tokens in regular texts.
149
+ Default to "all".
150
+ disallowed_special (`Literal["all"]` or `Collection`):
151
+ The surface forms of the tokens that should not be in regular texts and trigger errors.
152
+ Default to an empty tuple.
153
+
154
+ kwargs (additional keyword arguments, *optional*):
155
+ Will be passed to the underlying model specific encode method.
156
+
157
+ Returns:
158
+ `List[bytes|str]`: The list of tokens.
159
+ """
160
+
161
+
162
+ tokens = []
163
+ text = unicodedata.normalize("NFC", text)
164
+
165
+ # this implementation takes a detour: text -> token id -> token surface forms
166
+ for t in self.tokenizer.encode(
167
+ text, allowed_special=allowed_special, disallowed_special=disallowed_special
168
+ ):
169
+ tokens.append(self.decoder[t])
170
+ return tokens
171
+
172
+ def convert_tokens_to_string(self, tokens: List[Union[bytes, str]]) -> str:
173
+ """
174
+ Converts a sequence of tokens in a single string.
175
+ """
176
+ text = ""
177
+ temp = b""
178
+ for t in tokens:
179
+ if isinstance(t, str):
180
+ if temp:
181
+ text += temp.decode("utf-8", errors=self.errors)
182
+ temp = b""
183
+ text += t
184
+ elif isinstance(t, bytes):
185
+ temp += t
186
+ else:
187
+ raise TypeError("token should only be of type types or str")
188
+ if temp:
189
+ text += temp.decode("utf-8", errors=self.errors)
190
+ return text
191
+
192
+ @property
193
+ def vocab_size(self):
194
+ return self.tokenizer.n_vocab
195
+
196
+ def _convert_id_to_token(self, index: int) -> Union[bytes, str]:
197
+ """Converts an id to a token, special tokens included"""
198
+ if index in self.decoder:
199
+ return self.decoder[index]
200
+ raise ValueError("unknown ids")
201
+
202
+ def _convert_token_to_id(self, token: Union[bytes, str]) -> int:
203
+ """Converts a token to an id using the vocab, special tokens included"""
204
+ if token in self.special_tokens:
205
+ return self.special_tokens[token]
206
+ if token in self.mergeable_ranks:
207
+ return self.mergeable_ranks[token]
208
+ raise ValueError("unknown token")
209
+
210
+ def _tokenize(self, text: str, **kwargs):
211
+ """
212
+ Converts a string in a sequence of tokens (string), using the tokenizer. Split in words for word-based
213
+ vocabulary or sub-words for sub-word-based vocabularies (BPE/SentencePieces/WordPieces).
214
+
215
+ Do NOT take care of added tokens.
216
+ """
217
+ raise NotImplementedError
218
+
219
+ def _decode(
220
+ self,
221
+ token_ids: Union[int, List[int]],
222
+ skip_special_tokens: bool = False,
223
+ errors: str = None,
224
+ **kwargs,
225
+ ) -> str:
226
+ if isinstance(token_ids, int):
227
+ token_ids = [token_ids]
228
+ if skip_special_tokens:
229
+ token_ids = [i for i in token_ids if i < self.eod_id]
230
+ return self.tokenizer.decode(token_ids, errors=errors or self.errors)
tokenizer_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_max_length": 999999999999999999,
3
+ "tokenizer_class": "QWenTokenizer",
4
+ "auto_map": {
5
+ "AutoTokenizer": [
6
+ "tokenization_qwen.QWenTokenizer",
7
+ null
8
+ ]
9
+ }
10
+ }
visual.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:219ab65695072fc179a29903587e9744b124aee1bd4d09ec960d22d81f207450
3
+ size 3871401097
visual.py ADDED
@@ -0,0 +1,428 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Copyright (c) Alibaba Cloud.
2
+ #
3
+ # This source code is licensed under the license found in the
4
+ # LICENSE file in the root directory of this source tree.
5
+
6
+ from collections import OrderedDict
7
+ import math
8
+ import requests
9
+ from io import BytesIO
10
+ from functools import partial
11
+ from PIL import Image
12
+ from typing import Callable, Optional, Sequence, Tuple, List
13
+ import numpy as np
14
+
15
+ import torch
16
+ from torch import nn
17
+ from torch.nn import functional as F
18
+ from torch.nn.init import trunc_normal_
19
+ from torchvision import transforms
20
+ from torchvision.transforms import InterpolationMode
21
+ from einops import rearrange
22
+
23
+ def get_abs_pos(abs_pos, tgt_size):
24
+ # abs_pos: L, C
25
+ # tgt_size: M
26
+ # return: M, C
27
+ src_size = int(math.sqrt(abs_pos.size(0)))
28
+ tgt_size = int(math.sqrt(tgt_size))
29
+ dtype = abs_pos.dtype
30
+
31
+ if src_size != tgt_size:
32
+ return F.interpolate(
33
+ abs_pos.float().reshape(1, src_size, src_size, -1).permute(0, 3, 1, 2),
34
+ size=(tgt_size, tgt_size),
35
+ mode="bicubic",
36
+ align_corners=False,
37
+ ).permute(0, 2, 3, 1).flatten(0, 2).to(dtype=dtype)
38
+ else:
39
+ return abs_pos
40
+
41
+ # https://github.com/facebookresearch/mae/blob/efb2a8062c206524e35e47d04501ed4f544c0ae8/util/pos_embed.py#L20
42
+ def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):
43
+ """
44
+ grid_size: int of the grid height and width
45
+ return:
46
+ pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
47
+ """
48
+ grid_h = np.arange(grid_size, dtype=np.float32)
49
+ grid_w = np.arange(grid_size, dtype=np.float32)
50
+ grid = np.meshgrid(grid_w, grid_h) # here w goes first
51
+ grid = np.stack(grid, axis=0)
52
+
53
+ grid = grid.reshape([2, 1, grid_size, grid_size])
54
+ pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
55
+ if cls_token:
56
+ pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)
57
+ return pos_embed
58
+
59
+
60
+ def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
61
+ assert embed_dim % 2 == 0
62
+
63
+ # use half of dimensions to encode grid_h
64
+ emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0]) # (H*W, D/2)
65
+ emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1]) # (H*W, D/2)
66
+
67
+ emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)
68
+ return emb
69
+
70
+
71
+ def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
72
+ """
73
+ embed_dim: output dimension for each position
74
+ pos: a list of positions to be encoded: size (M,)
75
+ out: (M, D)
76
+ """
77
+ assert embed_dim % 2 == 0
78
+ omega = np.arange(embed_dim // 2, dtype=np.float32)
79
+ omega /= embed_dim / 2.
80
+ omega = 1. / 10000**omega # (D/2,)
81
+
82
+ pos = pos.reshape(-1) # (M,)
83
+ out = np.einsum('m,d->md', pos, omega) # (M, D/2), outer product
84
+
85
+ emb_sin = np.sin(out) # (M, D/2)
86
+ emb_cos = np.cos(out) # (M, D/2)
87
+
88
+ emb = np.concatenate([emb_sin, emb_cos], axis=1) # (M, D)
89
+ return emb
90
+
91
+
92
+ class Resampler(nn.Module):
93
+ """
94
+ A 2D perceiver-resampler network with one cross attention layers by
95
+ (grid_size**2) learnable queries and 2d sincos pos_emb
96
+ Outputs:
97
+ A tensor with the shape of (grid_size**2, embed_dim)
98
+ """
99
+ def __init__(
100
+ self,
101
+ grid_size,
102
+ embed_dim,
103
+ num_heads,
104
+ kv_dim=None,
105
+ norm_layer=nn.LayerNorm
106
+ ):
107
+ super().__init__()
108
+ self.num_queries = grid_size ** 2
109
+ self.embed_dim = embed_dim
110
+ self.num_heads = num_heads
111
+
112
+ self.pos_embed = nn.Parameter(
113
+ torch.from_numpy(get_2d_sincos_pos_embed(embed_dim, grid_size)).float()
114
+ ).requires_grad_(False)
115
+
116
+ self.query = nn.Parameter(torch.zeros(self.num_queries, embed_dim))
117
+ trunc_normal_(self.query, std=.02)
118
+
119
+ if kv_dim is not None and kv_dim != embed_dim:
120
+ self.kv_proj = nn.Linear(kv_dim, embed_dim, bias=False)
121
+ else:
122
+ self.kv_proj = nn.Identity()
123
+
124
+ self.attn = nn.MultiheadAttention(embed_dim, num_heads)
125
+ self.ln_q = norm_layer(embed_dim)
126
+ self.ln_kv = norm_layer(embed_dim)
127
+
128
+ self.apply(self._init_weights)
129
+
130
+ def _init_weights(self, m):
131
+ if isinstance(m, nn.Linear):
132
+ trunc_normal_(m.weight, std=.02)
133
+ if isinstance(m, nn.Linear) and m.bias is not None:
134
+ nn.init.constant_(m.bias, 0)
135
+ elif isinstance(m, nn.LayerNorm):
136
+ nn.init.constant_(m.bias, 0)
137
+ nn.init.constant_(m.weight, 1.0)
138
+
139
+ def forward(self, x, attn_mask=None):
140
+
141
+ pos_embed = get_abs_pos(self.pos_embed, x.size(1))
142
+
143
+ x = self.kv_proj(x)
144
+ x = self.ln_kv(x).permute(1, 0, 2)
145
+
146
+ N = x.shape[1]
147
+ q = self.ln_q(self.query)
148
+ out = self.attn(
149
+ self._repeat(q, N) + self.pos_embed.unsqueeze(1),
150
+ x + pos_embed.unsqueeze(1),
151
+ x,
152
+ attn_mask=attn_mask)[0]
153
+ return out.permute(1, 0, 2)
154
+
155
+ def _repeat(self, query, N: int):
156
+ return query.unsqueeze(1).repeat(1, N, 1)
157
+
158
+
159
+ class VisualAttention(nn.Module):
160
+ """self-attention layer class.
161
+
162
+ Self-attention layer takes input with size [s, b, h]
163
+ and returns output of the same size.
164
+ """
165
+
166
+ def __init__(self, embed_dim, num_heads,
167
+ bias=True, kdim=None, vdim=None):
168
+ super(VisualAttention, self).__init__()
169
+ self.embed_dim = embed_dim
170
+ self.kdim = kdim if kdim is not None else embed_dim
171
+ self.vdim = vdim if vdim is not None else embed_dim
172
+ self._qkv_same_embed_dim = self.kdim == embed_dim and self.vdim == embed_dim
173
+
174
+ self.num_heads = num_heads
175
+
176
+ # Per attention head and per partition values.
177
+ assert embed_dim % num_heads == 0
178
+ self.hidden_size_per_attention_head = embed_dim // num_heads
179
+ self.num_attention_heads_per_partition = num_heads
180
+ self.hidden_size_per_partition = embed_dim
181
+
182
+ # Strided linear layer.
183
+ assert self._qkv_same_embed_dim, 'Only Support SelfAttention Currently'
184
+ self.in_proj = nn.Linear(embed_dim, 3 * embed_dim)
185
+ self.out_proj = nn.Linear(embed_dim, embed_dim)
186
+ self.norm_factor = math.sqrt(self.hidden_size_per_attention_head)
187
+
188
+ def forward(self, query, key, value, attn_mask = None):
189
+ # query/key/value: [sq, b, h]
190
+ sq, b, _ = query.size()
191
+
192
+ assert query is key, 'Only Support Self-Attention Currently'
193
+ sk = sq
194
+ mixed_x_layer = self.in_proj(query)
195
+
196
+ # [sq, b, (np * 3 * hn)] --> [sq, b, np, 3 * hn]
197
+ new_tensor_shape = mixed_x_layer.size()[:-1] + \
198
+ (self.num_attention_heads_per_partition,
199
+ 3 * self.hidden_size_per_attention_head)
200
+ mixed_x_layer = mixed_x_layer.view(*new_tensor_shape)
201
+
202
+ # [sq, b, np, 3 * hn] --> 3 [sq, b, np, hn]
203
+ query_layer, key_layer, value_layer = mixed_x_layer.split(
204
+ self.hidden_size_per_attention_head, dim=-1)
205
+
206
+ # [sq, b, np, hn] -> [sq, b * np, hn]
207
+ query_layer = query_layer.view(sq,
208
+ b * self.num_attention_heads_per_partition,
209
+ self.hidden_size_per_attention_head).transpose(0, 1)
210
+ # [sk, b, np, hn] -> [sk, b * np, hn]
211
+ key_layer = key_layer.view(sk,
212
+ b * self.num_attention_heads_per_partition,
213
+ self.hidden_size_per_attention_head).transpose(0, 1)
214
+
215
+ q_scaled = query_layer / self.norm_factor
216
+ if attn_mask is not None:
217
+ attention_probs = torch.baddbmm(attn_mask, q_scaled, key_layer.transpose(-2, -1))
218
+ else:
219
+ attention_probs = torch.bmm(q_scaled, key_layer.transpose(-2, -1))
220
+ attention_probs = attention_probs.softmax(dim=-1)
221
+
222
+ value_layer = value_layer.view(sk,
223
+ b * self.num_attention_heads_per_partition,
224
+ self.hidden_size_per_attention_head).transpose(0, 1)
225
+
226
+ # matmul: [b * np, sq, hn]
227
+ context_layer = torch.bmm(attention_probs, value_layer)
228
+
229
+ # change view [b, np, sq, hn]
230
+ context_layer = context_layer.view(b,
231
+ self.num_attention_heads_per_partition,
232
+ sq, self.hidden_size_per_attention_head)
233
+
234
+ # [b, np, sq, hn] --> [sq, b, np, hn]
235
+ context_layer = context_layer.permute(2, 0, 1, 3).contiguous()
236
+
237
+ # [sq, b, np, hn] --> [sq, b, hp]
238
+ new_context_layer_shape = context_layer.size()[:-2] + \
239
+ (self.hidden_size_per_partition,)
240
+ context_layer = context_layer.view(*new_context_layer_shape)
241
+
242
+ output = self.out_proj(context_layer)
243
+
244
+ return output
245
+
246
+
247
+ class VisualAttentionBlock(nn.Module):
248
+ def __init__(
249
+ self,
250
+ d_model: int,
251
+ n_head: int,
252
+ mlp_ratio: float = 4.0,
253
+ act_layer: Callable = nn.GELU,
254
+ norm_layer: Callable = nn.LayerNorm,
255
+ is_cross_attention: bool = False,
256
+ ):
257
+ super().__init__()
258
+
259
+ self.ln_1 = norm_layer(d_model)
260
+ if is_cross_attention:
261
+ self.ln_1_kv = norm_layer(d_model)
262
+
263
+ self.ln_2 = norm_layer(d_model)
264
+ mlp_width = int(d_model * mlp_ratio)
265
+ self.attn = VisualAttention(d_model, n_head)
266
+ self.mlp = nn.Sequential(OrderedDict([
267
+ ("c_fc", nn.Linear(d_model, mlp_width)),
268
+ ("gelu", act_layer()),
269
+ ("c_proj", nn.Linear(mlp_width, d_model))
270
+ ]))
271
+
272
+ def attention(
273
+ self,
274
+ q_x: torch.Tensor,
275
+ k_x: Optional[torch.Tensor] = None,
276
+ v_x: Optional[torch.Tensor] = None,
277
+ attn_mask: Optional[torch.Tensor] = None,
278
+ ):
279
+ k_x = k_x if k_x is not None else q_x
280
+ v_x = v_x if v_x is not None else q_x
281
+
282
+ attn_mask = attn_mask.to(q_x.dtype) if attn_mask is not None else None
283
+ return self.attn(q_x, k_x, v_x, attn_mask=attn_mask)
284
+
285
+ def forward(
286
+ self,
287
+ q_x: torch.Tensor,
288
+ k_x: Optional[torch.Tensor] = None,
289
+ v_x: Optional[torch.Tensor] = None,
290
+ attn_mask: Optional[torch.Tensor] = None,
291
+ ):
292
+ k_x = self.ln_1_kv(k_x) if hasattr(self, "ln_1_kv") and k_x is not None else None
293
+ v_x = self.ln_1_kv(v_x) if hasattr(self, "ln_1_kv") and v_x is not None else None
294
+
295
+ x = q_x + self.attention(q_x=self.ln_1(q_x), k_x=k_x, v_x=v_x, attn_mask=attn_mask)
296
+ x = x + self.mlp(self.ln_2(x))
297
+ return x
298
+
299
+
300
+ class TransformerBlock(nn.Module):
301
+ def __init__(
302
+ self,
303
+ width: int,
304
+ layers: int,
305
+ heads: int,
306
+ mlp_ratio: float = 4.0,
307
+ act_layer: Callable = nn.GELU,
308
+ norm_layer: Callable = nn.LayerNorm,
309
+ ):
310
+ super().__init__()
311
+ self.width = width
312
+ self.layers = layers
313
+
314
+ self.resblocks = nn.ModuleList([
315
+ VisualAttentionBlock(
316
+ width, heads, mlp_ratio, act_layer=act_layer, norm_layer=norm_layer)
317
+ for _ in range(layers)
318
+ ])
319
+
320
+ def get_cast_dtype(self) -> torch.dtype:
321
+ return self.resblocks[0].mlp.c_fc.weight.dtype
322
+
323
+ def get_cast_device(self) -> torch.device:
324
+ return self.resblocks[0].mlp.c_fc.weight.device
325
+
326
+ def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
327
+ for r in self.resblocks:
328
+ x = r(x, attn_mask=attn_mask)
329
+ return x
330
+
331
+
332
+ class VisionTransformer(nn.Module):
333
+
334
+ def __init__(
335
+ self,
336
+ image_size: int = 448,
337
+ patch_size: int = 14,
338
+ width: int = 1664,
339
+ layers: int = 48,
340
+ heads: int = 16,
341
+ mlp_ratio: float = 4.9231,
342
+ n_queries: int = 256,
343
+ output_dim: int = 4096,
344
+ **kwargs
345
+ ):
346
+ super().__init__()
347
+ image_height, image_width = self.image_size = (image_size, image_size)
348
+ patch_height, patch_width = self.patch_size = (patch_size, patch_size)
349
+ self.grid_size = (image_height // patch_height, image_width // patch_width)
350
+ self.output_dim = output_dim
351
+
352
+ mean = (0.48145466, 0.4578275, 0.40821073)
353
+ std = (0.26862954, 0.26130258, 0.27577711)
354
+ self.image_transform = transforms.Compose([
355
+ transforms.Resize(
356
+ (image_size, image_size),
357
+ interpolation=InterpolationMode.BICUBIC
358
+ ),
359
+ transforms.ToTensor(),
360
+ transforms.Normalize(mean=mean, std=std),
361
+ ])
362
+
363
+ self.conv1 = nn.Conv2d(in_channels=3, out_channels=width, kernel_size=patch_size, stride=patch_size, bias=False)
364
+
365
+ # class embeddings and positional embeddings
366
+ scale = width ** -0.5
367
+ self.positional_embedding = nn.Parameter(scale * torch.randn(256, width))
368
+
369
+ norm_layer = partial(nn.LayerNorm, eps=1e-6)
370
+ act_layer = nn.GELU
371
+
372
+ self.ln_pre = norm_layer(width)
373
+ self.transformer = TransformerBlock(
374
+ width,
375
+ layers,
376
+ heads,
377
+ mlp_ratio,
378
+ act_layer=act_layer,
379
+ norm_layer=norm_layer,
380
+ )
381
+
382
+ self.attn_pool = Resampler(
383
+ grid_size=int(math.sqrt(n_queries)),
384
+ embed_dim=4096,
385
+ num_heads=4096 // 128,
386
+ kv_dim=width,
387
+ norm_layer=norm_layer,
388
+ )
389
+ self.ln_post = norm_layer(4096)
390
+ self.proj = nn.Parameter((output_dim** -0.5) * torch.randn(4096, output_dim))
391
+
392
+ def forward(self, x: torch.Tensor):
393
+ x = x.to(
394
+ dtype=self.transformer.get_cast_dtype(),
395
+ device=self.transformer.get_cast_device(),
396
+ )
397
+ # to patches
398
+ x = self.conv1(x) # shape = [*, width, grid, grid]
399
+ x = x.reshape(x.shape[0], x.shape[1], -1) # shape = [*, width, grid ** 2]
400
+ x = x.permute(0, 2, 1) # shape = [*, grid ** 2, width]
401
+
402
+ x = x + get_abs_pos(self.positional_embedding, x.size(1))
403
+
404
+ x = self.ln_pre(x)
405
+
406
+ x = x.permute(1, 0, 2) # NLD -> LND
407
+ x = self.transformer(x)
408
+ x = x.permute(1, 0, 2) # LND -> NLD
409
+
410
+ x = self.attn_pool(x)
411
+ x = self.ln_post(x)
412
+ x = x @ self.proj
413
+
414
+ return x
415
+
416
+ def encode(self, image_paths):
417
+ images = []
418
+ for image_path in image_paths:
419
+ if isinstance(image_path, Image.Image):
420
+ image = image_path
421
+ elif image_path.startswith("http://") or image_path.startswith("https://"):
422
+ image = Image.open(requests.get(image_path, stream=True).raw)
423
+ else:
424
+ image = Image.open(image_path)
425
+ image = image.convert("RGB")
426
+ images.append(self.image_transform(image))
427
+ images = torch.stack(images, dim=0)
428
+ return self(images)