Isaak Carter Augustus
commited on
Commit
•
f238cd4
1
Parent(s):
f3c4b35
Delete Version4.7-architecture.txt
Browse files- Version4.7-architecture.txt +0 -105
Version4.7-architecture.txt
DELETED
@@ -1,105 +0,0 @@
|
|
1 |
-
Josiev47(
|
2 |
-
(encoder): EncoderModel(
|
3 |
-
(modality_preprocessors): ModuleDict(
|
4 |
-
(vision): RGBDTPreprocessor(
|
5 |
-
(rgbt_stem): PatchEmbedGeneric(
|
6 |
-
(proj): Sequential(
|
7 |
-
(0): make_image_to_video()
|
8 |
-
(1): Conv3d(3, {llm_in_embedding}, kernel_size=(2, 14, 14), stride=(2, 14, 14), bias=False)
|
9 |
-
)
|
10 |
-
(norm_layer): RMSNorm()
|
11 |
-
)
|
12 |
-
(pos_embedding_helper): SpatioTemporalPosEmbeddingHelper()
|
13 |
-
)
|
14 |
-
(audio): AudioPreprocessor(
|
15 |
-
(rgbt_stem): PatchEmbedGeneric(
|
16 |
-
(proj): Conv2d(1, {llm_in_embedding}, kernel_size=(16, 16), stride=(10, 10), bias=False)
|
17 |
-
(norm_layer): RMSNorm()
|
18 |
-
)
|
19 |
-
(pos_embedding_helper): SpatioTemporalPosEmbeddingHelper()
|
20 |
-
)
|
21 |
-
)
|
22 |
-
|
23 |
-
(modality_transformers): ModuleDict(
|
24 |
-
(vision): EncoderTransformer(
|
25 |
-
(pre_transformer_layer_norm): RMSNorm()
|
26 |
-
(layers): ModuleList(
|
27 |
-
(0 - n): EncoderTransformerLayer(
|
28 |
-
(attn): EncoderTransformerAttention(
|
29 |
-
(qkv): Linear(in_features={llm_in_embedding}, out_features=3072, bias=True)
|
30 |
-
(attn_drop): Dropout(p=0.0, inplace=False)
|
31 |
-
(proj): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=True)
|
32 |
-
(proj_drop): Dropout(p=0.0, inplace=False)
|
33 |
-
)
|
34 |
-
(drop_path): Identity()
|
35 |
-
(norm_1): RMSNorm()
|
36 |
-
(mlp): EncoderMLP(
|
37 |
-
(fc1): Linear(in_features={llm_in_embedding}, out_features=4096, bias=True)
|
38 |
-
(fc2): Linear(in_features=4096, out_features={llm_in_embedding}, bias=True)
|
39 |
-
(drop): Dropout(p=0.0, inplace=False)
|
40 |
-
(act): SiLU()
|
41 |
-
)
|
42 |
-
(norm_2): RMSNorm()
|
43 |
-
)
|
44 |
-
)
|
45 |
-
(head): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=False)
|
46 |
-
(post_transformer_layer_norm): RMSNorm()
|
47 |
-
)
|
48 |
-
(audio): EncoderTransformer(
|
49 |
-
(pre_transformer_layer_norm): RMSNorm()
|
50 |
-
(layers): ModuleList(
|
51 |
-
(0 - n): EncoderTransformerLayer(
|
52 |
-
(attn): EncoderTransformerAttention(
|
53 |
-
(qkv): Linear(in_features={llm_in_embedding}, out_features=3072, bias=True)
|
54 |
-
(attn_drop): Dropout(p=0.0, inplace=False)
|
55 |
-
(proj): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=True)
|
56 |
-
(proj_drop): Dropout(p=0.0, inplace=False)
|
57 |
-
)
|
58 |
-
(drop_path): DropPath()
|
59 |
-
(norm_1): RMSNorm()
|
60 |
-
(mlp): EncoderMLP(
|
61 |
-
(fc1): Linear(in_features={llm_in_embedding}, out_features=4096, bias=True)
|
62 |
-
(fc2): Linear(in_features=4096, out_features={llm_in_embedding}, bias=True)
|
63 |
-
(drop): Dropout(p=0.0, inplace=False)
|
64 |
-
(act): SiLU()
|
65 |
-
)
|
66 |
-
(norm_2): RMSNorm()
|
67 |
-
)
|
68 |
-
)
|
69 |
-
(head): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=False)
|
70 |
-
(post_transformer_layer_norm): RMSNorm()
|
71 |
-
)
|
72 |
-
)
|
73 |
-
)
|
74 |
-
|
75 |
-
|
76 |
-
(llm): LlamaForCausalLM(
|
77 |
-
(model): LlamaModel(
|
78 |
-
(embed_tokens): Embedding({vocab_size}, {llm_in_embedding}, {padding_idx})
|
79 |
-
|
80 |
-
(layers): ModuleList(
|
81 |
-
(0 - n): LlamaDecoderLayer(
|
82 |
-
(self_attn): LlamaAttention(
|
83 |
-
(q_proj): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=False)
|
84 |
-
(k_proj): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=False)
|
85 |
-
(v_proj): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=False)
|
86 |
-
(o_proj): Linear(in_features={llm_in_embedding}, out_features={llm_in_embedding}, bias=False)
|
87 |
-
(rotary_emb): LlamaRotaryEmbedding()
|
88 |
-
)
|
89 |
-
(mlp): LlamaMLP(
|
90 |
-
(gate_proj): Linear(in_features={llm_in_embedding}, out_features=4096, bias=False)
|
91 |
-
(down_proj): Linear(in_features=4096, out_features={llm_in_embedding}, bias=False)
|
92 |
-
(up_proj): Linear(in_features={llm_in_embedding}, out_features=4096, bias=False)
|
93 |
-
(act_fn): SiLU()
|
94 |
-
)
|
95 |
-
(input_layernorm): RMSNorm()
|
96 |
-
(post_attention_layernorm): RMSNorm()
|
97 |
-
)
|
98 |
-
)
|
99 |
-
(norm): RMSNorm()
|
100 |
-
)
|
101 |
-
(lm_head): Linear(in_features={llm_in_embedding}, out_features={vocab_size}, bias=False)
|
102 |
-
)
|
103 |
-
|
104 |
-
(input_embeddings): Embedding({vocab_size}, {llm_in_embedding}, {padding_idx})
|
105 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|