Hugging Face's logo

ndkhanh95
/
florence-vl-3b-pretrain

Safetensors
llava_phi
custom_code
Model card Files Community
ndkhanh95 commited on
Commit
8e56b50
·
verified ·
1 Parent(s): c5a0424

Upload adapter_config.json

Browse files
Files changed (1) hide show
  1. adapter_config.json +175 -0
adapter_config.json ADDED
@@ -0,0 +1,175 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": true,
3
+ "_name_or_path": "jiuhai/florence-vl-3b-pretrain",
4
+ "architectures": [
5
+ "LlavaPhiForCausalLM"
6
+ ],
7
+ "attention_bias": false,
8
+ "attention_dropout": 0.0,
9
+ "auto_map": {
10
+ "AutoConfig": "jiuhai/florence-vl-3b-pretrain--configuration_phi3.Phi3Config",
11
+ "AutoModelForCausalLM": "jiuhai/florence-vl-3b-pretrain--modeling_phi3.Phi3ForCausalLM"
12
+ },
13
+ "bos_token_id": 1,
14
+ "embd_pdrop": 0.0,
15
+ "eos_token_id": 32000,
16
+ "freeze_mm_mlp_adapter": false,
17
+ "head_dim": 96,
18
+ "hidden_act": "silu",
19
+ "hidden_size": 3072,
20
+ "image_aspect_ratio": "square",
21
+ "initializer_range": 0.02,
22
+ "intermediate_size": 8192,
23
+ "max_position_embeddings": 131072,
24
+ "mlp_bias": false,
25
+ "mm_hidden_size": 1024,
26
+ "mm_patch_merge_type": "flat",
27
+ "mm_projector_lr": null,
28
+ "mm_projector_type": "mlp3x_gelu",
29
+ "mm_use_im_patch_token": false,
30
+ "mm_use_im_start_end": false,
31
+ "mm_vision_select_feature": "patch",
32
+ "mm_vision_select_layer": -2,
33
+ "mm_vision_tower": "microsoft/Florence-2-large-ft",
34
+ "model_type": "llama",
35
+ "num_attention_heads": 32,
36
+ "num_hidden_layers": 32,
37
+ "num_key_value_heads": 32,
38
+ "original_max_position_embeddings": 4096,
39
+ "pad_token_id": 32000,
40
+ "pretraining_tp": 1,
41
+ "quantization_config": {
42
+ "_load_in_4bit": true,
43
+ "_load_in_8bit": false,
44
+ "bnb_4bit_compute_dtype": "float32",
45
+ "bnb_4bit_quant_storage": "uint8",
46
+ "bnb_4bit_quant_type": "fp4",
47
+ "bnb_4bit_use_double_quant": false,
48
+ "llm_int8_enable_fp32_cpu_offload": false,
49
+ "llm_int8_has_fp16_weight": false,
50
+ "llm_int8_skip_modules": null,
51
+ "llm_int8_threshold": 6.0,
52
+ "load_in_4bit": true,
53
+ "load_in_8bit": false,
54
+ "quant_method": "bitsandbytes"
55
+ },
56
+ "resid_pdrop": 0.0,
57
+ "rms_norm_eps": 1e-05,
58
+ "rope_scaling": {
59
+ "long_factor": [
60
+ 1.0800000429153442,
61
+ 1.1100000143051147,
62
+ 1.1399999856948853,
63
+ 1.340000033378601,
64
+ 1.5899999141693115,
65
+ 1.600000023841858,
66
+ 1.6200000047683716,
67
+ 2.620000123977661,
68
+ 3.2300000190734863,
69
+ 3.2300000190734863,
70
+ 4.789999961853027,
71
+ 7.400000095367432,
72
+ 7.700000286102295,
73
+ 9.09000015258789,
74
+ 12.199999809265137,
75
+ 17.670000076293945,
76
+ 24.46000099182129,
77
+ 28.57000160217285,
78
+ 30.420001983642578,
79
+ 30.840002059936523,
80
+ 32.590003967285156,
81
+ 32.93000411987305,
82
+ 42.320003509521484,
83
+ 44.96000289916992,
84
+ 50.340003967285156,
85
+ 50.45000457763672,
86
+ 57.55000305175781,
87
+ 57.93000411987305,
88
+ 58.21000289916992,
89
+ 60.1400032043457,
90
+ 62.61000442504883,
91
+ 62.62000274658203,
92
+ 62.71000289916992,
93
+ 63.1400032043457,
94
+ 63.1400032043457,
95
+ 63.77000427246094,
96
+ 63.93000411987305,
97
+ 63.96000289916992,
98
+ 63.970001220703125,
99
+ 64.02999877929688,
100
+ 64.06999969482422,
101
+ 64.08000183105469,
102
+ 64.12000274658203,
103
+ 64.41000366210938,
104
+ 64.4800033569336,
105
+ 64.51000213623047,
106
+ 64.52999877929688,
107
+ 64.83999633789062
108
+ ],
109
+ "rope_type": "longrope",
110
+ "short_factor": [
111
+ 1.0,
112
+ 1.0199999809265137,
113
+ 1.0299999713897705,
114
+ 1.0299999713897705,
115
+ 1.0499999523162842,
116
+ 1.0499999523162842,
117
+ 1.0499999523162842,
118
+ 1.0499999523162842,
119
+ 1.0499999523162842,
120
+ 1.0699999332427979,
121
+ 1.0999999046325684,
122
+ 1.1099998950958252,
123
+ 1.1599998474121094,
124
+ 1.1599998474121094,
125
+ 1.1699998378753662,
126
+ 1.2899998426437378,
127
+ 1.339999794960022,
128
+ 1.679999828338623,
129
+ 1.7899998426437378,
130
+ 1.8199998140335083,
131
+ 1.8499997854232788,
132
+ 1.8799997568130493,
133
+ 1.9099997282028198,
134
+ 1.9399996995925903,
135
+ 1.9899996519088745,
136
+ 2.0199997425079346,
137
+ 2.0199997425079346,
138
+ 2.0199997425079346,
139
+ 2.0199997425079346,
140
+ 2.0199997425079346,
141
+ 2.0199997425079346,
142
+ 2.0299997329711914,
143
+ 2.0299997329711914,
144
+ 2.0299997329711914,
145
+ 2.0299997329711914,
146
+ 2.0299997329711914,
147
+ 2.0299997329711914,
148
+ 2.0299997329711914,
149
+ 2.0299997329711914,
150
+ 2.0299997329711914,
151
+ 2.0799996852874756,
152
+ 2.0899996757507324,
153
+ 2.189999580383301,
154
+ 2.2199995517730713,
155
+ 2.5899994373321533,
156
+ 2.729999542236328,
157
+ 2.749999523162842,
158
+ 2.8399994373321533
159
+ ],
160
+ "type": "longrope"
161
+ },
162
+ "rope_theta": 10000.0,
163
+ "sliding_window": 262144,
164
+ "tie_word_embeddings": false,
165
+ "tokenizer_model_max_length": 2048,
166
+ "tokenizer_padding_side": "right",
167
+ "torch_dtype": "bfloat16",
168
+ "transformers_version": "4.48.0.dev0",
169
+ "tune_entire_model": true,
170
+ "tune_mm_mlp_adapter": false,
171
+ "use_cache": true,
172
+ "use_mm_proj": true,
173
+ "vision_tower_lr": null,
174
+ "vocab_size": 32064
175
+ }