Hugging Face's logo

Fsoft-AIC
/
Phi3-CLIP-MoE

Image-Text-to-Text
Safetensors
English
Model card Files
xet
Community
14
DavidNguyen commited on
Commit
59e1011
·
verified ·
1 Parent(s): 791370f

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. pretrain/config.json +169 -0
  2. pretrain/mm_projector.bin +3 -0
  3. pretrain/trainer_state.json +0 -0
pretrain/config.json ADDED
@@ -0,0 +1,169 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "microsoft/Phi-3-mini-128k-instruct",
3
+ "architectures": [
4
+ "Phi3ForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "auto_map": {
9
+ "AutoConfig": "microsoft/Phi-3-mini-128k-instruct--configuration_phi3.Phi3Config",
10
+ "AutoModelForCausalLM": "microsoft/Phi-3-mini-128k-instruct--modeling_phi3.Phi3ForCausalLM",
11
+ "AutoModelForSequenceClassification": "microsoft/Phi-3-mini-128k-instruct--modeling_phi3.Phi3ForSequenceClassification",
12
+ "AutoModelForTokenClassification": "microsoft/Phi-3-mini-128k-instruct--modeling_phi3.Phi3ForTokenClassification"
13
+ },
14
+ "balance_loss_coef": 0.0,
15
+ "bos_token_id": 1,
16
+ "clip_smoe": false,
17
+ "dropout": false,
18
+ "embd_pdrop": 0.0,
19
+ "eos_token_id": 32000,
20
+ "freeze_mm_mlp_adapter": false,
21
+ "hidden_act": "silu",
22
+ "hidden_size": 3072,
23
+ "image_aspect_ratio": "square",
24
+ "initializer_range": 0.02,
25
+ "intermediate_size": 8192,
26
+ "local_rank": 0,
27
+ "max_position_embeddings": 131072,
28
+ "mlp_smoe": false,
29
+ "mm_hidden_size": 1024,
30
+ "mm_patch_merge_type": "flat",
31
+ "mm_projector_lr": null,
32
+ "mm_projector_type": "mlp2x_gelu",
33
+ "mm_use_im_patch_token": false,
34
+ "mm_use_im_start_end": false,
35
+ "mm_vision_select_feature": "patch",
36
+ "mm_vision_select_layer": -2,
37
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
38
+ "model_type": "llava_phi",
39
+ "num_attention_heads": 32,
40
+ "num_experts": 1,
41
+ "num_hidden_layers": 32,
42
+ "num_key_value_heads": 32,
43
+ "num_layers": 3,
44
+ "num_selected": 1,
45
+ "original_max_position_embeddings": 4096,
46
+ "pad_token_id": 32000,
47
+ "resid_pdrop": 0.0,
48
+ "rms_norm_eps": 1e-05,
49
+ "rope_scaling": {
50
+ "long_factor": [
51
+ 1.0299999713897705,
52
+ 1.0499999523162842,
53
+ 1.0499999523162842,
54
+ 1.0799999237060547,
55
+ 1.2299998998641968,
56
+ 1.2299998998641968,
57
+ 1.2999999523162842,
58
+ 1.4499999284744263,
59
+ 1.5999999046325684,
60
+ 1.6499998569488525,
61
+ 1.8999998569488525,
62
+ 2.859999895095825,
63
+ 3.68999981880188,
64
+ 5.419999599456787,
65
+ 5.489999771118164,
66
+ 5.489999771118164,
67
+ 9.09000015258789,
68
+ 11.579999923706055,
69
+ 15.65999984741211,
70
+ 15.769999504089355,
71
+ 15.789999961853027,
72
+ 18.360000610351562,
73
+ 21.989999771118164,
74
+ 23.079999923706055,
75
+ 30.009998321533203,
76
+ 32.35000228881836,
77
+ 32.590003967285156,
78
+ 35.56000518798828,
79
+ 39.95000457763672,
80
+ 53.840003967285156,
81
+ 56.20000457763672,
82
+ 57.95000457763672,
83
+ 59.29000473022461,
84
+ 59.77000427246094,
85
+ 59.920005798339844,
86
+ 61.190006256103516,
87
+ 61.96000671386719,
88
+ 62.50000762939453,
89
+ 63.3700065612793,
90
+ 63.48000717163086,
91
+ 63.48000717163086,
92
+ 63.66000747680664,
93
+ 63.850006103515625,
94
+ 64.08000946044922,
95
+ 64.760009765625,
96
+ 64.80001068115234,
97
+ 64.81001281738281,
98
+ 64.81001281738281
99
+ ],
100
+ "short_factor": [
101
+ 1.05,
102
+ 1.05,
103
+ 1.05,
104
+ 1.1,
105
+ 1.1,
106
+ 1.1500000000000001,
107
+ 1.2000000000000002,
108
+ 1.2500000000000002,
109
+ 1.3000000000000003,
110
+ 1.3500000000000003,
111
+ 1.5000000000000004,
112
+ 2.000000000000001,
113
+ 2.000000000000001,
114
+ 2.000000000000001,
115
+ 2.000000000000001,
116
+ 2.000000000000001,
117
+ 2.000000000000001,
118
+ 2.000000000000001,
119
+ 2.000000000000001,
120
+ 2.000000000000001,
121
+ 2.000000000000001,
122
+ 2.000000000000001,
123
+ 2.000000000000001,
124
+ 2.000000000000001,
125
+ 2.000000000000001,
126
+ 2.000000000000001,
127
+ 2.000000000000001,
128
+ 2.000000000000001,
129
+ 2.000000000000001,
130
+ 2.000000000000001,
131
+ 2.000000000000001,
132
+ 2.000000000000001,
133
+ 2.0500000000000007,
134
+ 2.0500000000000007,
135
+ 2.0500000000000007,
136
+ 2.1000000000000005,
137
+ 2.1000000000000005,
138
+ 2.1000000000000005,
139
+ 2.1500000000000004,
140
+ 2.1500000000000004,
141
+ 2.3499999999999996,
142
+ 2.549999999999999,
143
+ 2.5999999999999988,
144
+ 2.5999999999999988,
145
+ 2.7499999999999982,
146
+ 2.849999999999998,
147
+ 2.849999999999998,
148
+ 2.9499999999999975
149
+ ],
150
+ "type": "su"
151
+ },
152
+ "rope_theta": 10000.0,
153
+ "router_z_loss_coef": 0.0,
154
+ "scales": [
155
+ 1,
156
+ 3
157
+ ],
158
+ "sliding_window": 262144,
159
+ "tie_word_embeddings": false,
160
+ "tokenizer_model_max_length": 2048,
161
+ "tokenizer_padding_side": "right",
162
+ "torch_dtype": "bfloat16",
163
+ "training": true,
164
+ "transformers_version": "4.41.2",
165
+ "tune_mm_mlp_adapter": true,
166
+ "use_cache": true,
167
+ "use_mm_proj": true,
168
+ "vocab_size": 32064
169
+ }
pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe20cb96aaeec81bb87f3fe1b31c45a5656f7b8d1f954a46242b0b15b86f327b
3
+ size 31471672
pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff