Hugging Face's logo

OpenVINO
/
bloomz-3b-int4-ov

OpenVINO
bloom
Model card Files Community
openvino-ci commited on
Commit
a5fa0f1
•
1 Parent(s): 12153ab

Upload folder using huggingface_hub

Browse files
Files changed (13) hide show
  1. .gitattributes +1 -0
  2. README.md +67 -0
  3. config.json +32 -0
  4. generation_config.json +7 -0
  5. openvino_detokenizer.bin +3 -0
  6. openvino_detokenizer.xml +163 -0
  7. openvino_model.bin +3 -0
  8. openvino_model.xml +0 -0
  9. openvino_tokenizer.bin +3 -0
  10. openvino_tokenizer.xml +681 -0
  11. special_tokens_map.json +30 -0
  12. tokenizer.json +3 -0
  13. tokenizer_config.json +44 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: bigscience-bloom-rail-1.0
3
+ license_link: https://choosealicense.com/licenses/bigscience-bloom-rail-1.0/
4
+ ---
5
+ # bloomz-3b-int4-ov
6
+ * Model creator: [Bigscience](https://huggingface.co/bigscience)
7
+ * Original model: [bloomz-3b](https://huggingface.co/bigscience/bloomz-3b)
8
+
9
+ ## Description
10
+ This is [bloomz-3b](https://huggingface.co/bigscience/bloomz-3b) model converted to the [OpenVINO™ IR](https://docs.openvino.ai/2024/documentation/openvino-ir-format.html) (Intermediate Representation) format with weights compressed to INT4 by [NNCF](https://github.com/openvinotoolkit/nncf).
11
+
12
+ ## Quantization Parameters
13
+
14
+ Weight compression was performed using `nncf.compress_weights` with the following parameters:
15
+
16
+ * mode: **int4_asym**
17
+ * ratio: **1**
18
+ * group_size: **128**
19
+
20
+ For more information on quantization, check the [OpenVINO model optimization guide](https://docs.openvino.ai/2024/openvino-workflow/model-optimization-guide/weight-compression.html).
21
+
22
+
23
+ ## Compatibility
24
+
25
+ The provided OpenVINO™ IR model is compatible with:
26
+
27
+ * OpenVINO version 2024.4.0 and higher
28
+ * Optimum Intel 1.20.0 and higher
29
+
30
+ ## Running Model Inference
31
+
32
+ 1. Install packages required for using [Optimum Intel](https://huggingface.co/docs/optimum/intel/index) integration with the OpenVINO backend:
33
+
34
+ ```
35
+ pip install optimum[openvino]
36
+ ```
37
+
38
+ 2. Run model inference:
39
+
40
+ ```
41
+ from transformers import AutoTokenizer
42
+ from optimum.intel.openvino import OVModelForCausalLM
43
+
44
+ model_id = "OpenVINO/bloomz-3b-int4-ov"
45
+ tokenizer = AutoTokenizer.from_pretrained(model_id)
46
+ model = OVModelForCausalLM.from_pretrained(model_id)
47
+
48
+ inputs = tokenizer("What is OpenVINO?", return_tensors="pt")
49
+
50
+ outputs = model.generate(**inputs, max_length=200)
51
+ text = tokenizer.batch_decode(outputs)[0]
52
+ print(text)
53
+ ```
54
+
55
+ For more examples and possible optimizations, refer to the [OpenVINO Large Language Model Inference Guide](https://docs.openvino.ai/2024/learn-openvino/llm_inference_guide.html).
56
+
57
+ ## Limitations
58
+
59
+ Check the original model card for [original model card](https://huggingface.co/bigscience/bloomz-3b) for limitations.
60
+
61
+ ## Legal information
62
+
63
+ The original model is distributed under [bigscience-bloom-rail-1.0](https://huggingface.co/spaces/bigscience/license) license. More details can be found in [original model card](https://huggingface.co/bigscience/bloomz-3b).
64
+
65
+ ## Disclaimer
66
+
67
+ Intel is committed to respecting human rights and avoiding causing or contributing to adverse impacts on human rights. See [Intel’s Global Human Rights Principles](https://www.intel.com/content/dam/www/central-libraries/us/en/documents/policy-human-rights.pdf). Intel’s products and software are intended only to be used in applications that do not cause or contribute to adverse impacts on human rights.
config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "bigscience/bloomz-3b",
3
+ "apply_residual_connection_post_layernorm": false,
4
+ "architectures": [
5
+ "BloomForCausalLM"
6
+ ],
7
+ "attention_dropout": 0.0,
8
+ "attention_softmax_in_fp32": true,
9
+ "bias_dropout_fusion": true,
10
+ "bos_token_id": 1,
11
+ "eos_token_id": 2,
12
+ "hidden_dropout": 0.0,
13
+ "hidden_size": 2560,
14
+ "initializer_range": 0.02,
15
+ "layer_norm_epsilon": 1e-05,
16
+ "masked_softmax_fusion": true,
17
+ "model_type": "bloom",
18
+ "n_head": 32,
19
+ "n_inner": null,
20
+ "n_layer": 30,
21
+ "offset_alibi": 100,
22
+ "pad_token_id": 3,
23
+ "pretraining_tp": 4,
24
+ "seq_length": 2048,
25
+ "skip_bias_add": true,
26
+ "skip_bias_add_qkv": false,
27
+ "slow_but_exact": false,
28
+ "transformers_version": "4.45.2",
29
+ "unk_token_id": 0,
30
+ "use_cache": true,
31
+ "vocab_size": 250880
32
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "pad_token_id": 3,
6
+ "transformers_version": "4.45.2"
7
+ }
openvino_detokenizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9db271ab1e515856fe1fdb70a2999d1308ade18550a1aacbe5c2630c36027f8
3
+ size 3189816
openvino_detokenizer.xml ADDED
@@ -0,0 +1,163 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="detokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_52019" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="i64" />
6
+ <output>
7
+ <port id="0" precision="I64" names="Parameter_52019">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="1" name="Convert_52030" type="Convert" version="opset1">
14
+ <data destination_type="i32" />
15
+ <input>
16
+ <port id="0" precision="I64">
17
+ <dim>-1</dim>
18
+ <dim>-1</dim>
19
+ </port>
20
+ </input>
21
+ <output>
22
+ <port id="1" precision="I32">
23
+ <dim>-1</dim>
24
+ <dim>-1</dim>
25
+ </port>
26
+ </output>
27
+ </layer>
28
+ <layer id="2" name="Constant_51994" type="Const" version="opset1">
29
+ <data element_type="u8" shape="3189816" offset="0" size="3189816" />
30
+ <output>
31
+ <port id="0" precision="U8">
32
+ <dim>3189816</dim>
33
+ </port>
34
+ </output>
35
+ </layer>
36
+ <layer id="3" name="StringTensorUnpack_51995" type="StringTensorUnpack" version="extension">
37
+ <data mode="begins_ends" />
38
+ <input>
39
+ <port id="0" precision="U8">
40
+ <dim>3189816</dim>
41
+ </port>
42
+ </input>
43
+ <output>
44
+ <port id="1" precision="I32">
45
+ <dim>-1</dim>
46
+ </port>
47
+ <port id="2" precision="I32">
48
+ <dim>-1</dim>
49
+ </port>
50
+ <port id="3" precision="U8">
51
+ <dim>-1</dim>
52
+ </port>
53
+ </output>
54
+ </layer>
55
+ <layer id="4" name="VocabDecoder_52020" type="VocabDecoder" version="extension">
56
+ <data skip_tokens="0, 1, 2, 3" />
57
+ <input>
58
+ <port id="0" precision="I32">
59
+ <dim>-1</dim>
60
+ <dim>-1</dim>
61
+ </port>
62
+ <port id="1" precision="I32">
63
+ <dim>-1</dim>
64
+ </port>
65
+ <port id="2" precision="I32">
66
+ <dim>-1</dim>
67
+ </port>
68
+ <port id="3" precision="U8">
69
+ <dim>-1</dim>
70
+ </port>
71
+ </input>
72
+ <output>
73
+ <port id="4" precision="I32">
74
+ <dim>-1</dim>
75
+ </port>
76
+ <port id="5" precision="I32">
77
+ <dim>-1</dim>
78
+ </port>
79
+ <port id="6" precision="I32">
80
+ <dim>-1</dim>
81
+ </port>
82
+ <port id="7" precision="I32">
83
+ <dim>-1</dim>
84
+ </port>
85
+ <port id="8" precision="U8">
86
+ <dim>-1</dim>
87
+ </port>
88
+ </output>
89
+ </layer>
90
+ <layer id="5" name="FuzeRagged_52021" type="FuzeRagged" version="extension">
91
+ <input>
92
+ <port id="0" precision="I32">
93
+ <dim>-1</dim>
94
+ </port>
95
+ <port id="1" precision="I32">
96
+ <dim>-1</dim>
97
+ </port>
98
+ <port id="2" precision="I32">
99
+ <dim>-1</dim>
100
+ </port>
101
+ <port id="3" precision="I32">
102
+ <dim>-1</dim>
103
+ </port>
104
+ </input>
105
+ <output>
106
+ <port id="4" precision="I32">
107
+ <dim>-1</dim>
108
+ </port>
109
+ <port id="5" precision="I32">
110
+ <dim>-1</dim>
111
+ </port>
112
+ </output>
113
+ </layer>
114
+ <layer id="6" name="StringTensorPack_52022" type="StringTensorPack" version="extension">
115
+ <data mode="begins_ends" />
116
+ <input>
117
+ <port id="0" precision="I32">
118
+ <dim>-1</dim>
119
+ </port>
120
+ <port id="1" precision="I32">
121
+ <dim>-1</dim>
122
+ </port>
123
+ <port id="2" precision="U8">
124
+ <dim>-1</dim>
125
+ </port>
126
+ </input>
127
+ <output>
128
+ <port id="3" precision="STRING" names="string_output">
129
+ <dim>-1</dim>
130
+ </port>
131
+ </output>
132
+ </layer>
133
+ <layer id="7" name="Result_52023" type="Result" version="opset1">
134
+ <input>
135
+ <port id="0" precision="STRING">
136
+ <dim>-1</dim>
137
+ </port>
138
+ </input>
139
+ </layer>
140
+ </layers>
141
+ <edges>
142
+ <edge from-layer="0" from-port="0" to-layer="1" to-port="0" />
143
+ <edge from-layer="1" from-port="1" to-layer="4" to-port="0" />
144
+ <edge from-layer="2" from-port="0" to-layer="3" to-port="0" />
145
+ <edge from-layer="3" from-port="1" to-layer="4" to-port="1" />
146
+ <edge from-layer="3" from-port="2" to-layer="4" to-port="2" />
147
+ <edge from-layer="3" from-port="3" to-layer="4" to-port="3" />
148
+ <edge from-layer="4" from-port="4" to-layer="5" to-port="0" />
149
+ <edge from-layer="4" from-port="5" to-layer="5" to-port="1" />
150
+ <edge from-layer="4" from-port="6" to-layer="5" to-port="2" />
151
+ <edge from-layer="4" from-port="7" to-layer="5" to-port="3" />
152
+ <edge from-layer="4" from-port="8" to-layer="6" to-port="2" />
153
+ <edge from-layer="5" from-port="4" to-layer="6" to-port="0" />
154
+ <edge from-layer="5" from-port="5" to-layer="6" to-port="1" />
155
+ <edge from-layer="6" from-port="3" to-layer="7" to-port="0" />
156
+ </edges>
157
+ <rt_info>
158
+ <bos_token_id value="1" />
159
+ <eos_token_id value="2" />
160
+ <original_tokenizer_class value="&lt;class 'transformers.models.bloom.tokenization_bloom_fast.BloomTokenizerFast'>" />
161
+ <pad_token_id value="3" />
162
+ </rt_info>
163
+ </net>
openvino_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18b028ed2dbdeab13374b74e89fb4bcec296797d6b431953ceec0de619c9dafa
3
+ size 1872768388
openvino_model.xml ADDED
The diff for this file is too large to render. See raw diff
 
openvino_tokenizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1339724674d23d50427026fff2db9e17f06b100fbdaf67fdcb6b268e87e2e35c
3
+ size 7380278
openvino_tokenizer.xml ADDED
@@ -0,0 +1,681 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="tokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_51913" type="Parameter" version="opset1">
5
+ <data shape="?" element_type="string" />
6
+ <output>
7
+ <port id="0" precision="STRING" names="Parameter_51913">
8
+ <dim>-1</dim>
9
+ </port>
10
+ </output>
11
+ </layer>
12
+ <layer id="1" name="Constant_51919" type="Const" version="opset1">
13
+ <data element_type="i64" shape="" offset="0" size="8" />
14
+ <output>
15
+ <port id="0" precision="I64" />
16
+ </output>
17
+ </layer>
18
+ <layer id="2" name="StringTensorUnpack_51914" type="StringTensorUnpack" version="extension">
19
+ <data mode="begins_ends" />
20
+ <input>
21
+ <port id="0" precision="STRING">
22
+ <dim>-1</dim>
23
+ </port>
24
+ </input>
25
+ <output>
26
+ <port id="1" precision="I32">
27
+ <dim>-1</dim>
28
+ </port>
29
+ <port id="2" precision="I32">
30
+ <dim>-1</dim>
31
+ </port>
32
+ <port id="3" precision="U8">
33
+ <dim>-1</dim>
34
+ </port>
35
+ </output>
36
+ </layer>
37
+ <layer id="3" name="ShapeOf_51915" type="ShapeOf" version="opset3">
38
+ <data output_type="i64" />
39
+ <input>
40
+ <port id="0" precision="I32">
41
+ <dim>-1</dim>
42
+ </port>
43
+ </input>
44
+ <output>
45
+ <port id="1" precision="I64">
46
+ <dim>1</dim>
47
+ </port>
48
+ </output>
49
+ </layer>
50
+ <layer id="4" name="Constant_51916" type="Const" version="opset1">
51
+ <data element_type="i64" shape="" offset="0" size="8" />
52
+ <output>
53
+ <port id="0" precision="I64" />
54
+ </output>
55
+ </layer>
56
+ <layer id="5" name="Constant_51917" type="Const" version="opset1">
57
+ <data element_type="i64" shape="" offset="0" size="8" />
58
+ <output>
59
+ <port id="0" precision="I64" />
60
+ </output>
61
+ </layer>
62
+ <layer id="6" name="Gather_51918" type="Gather" version="opset8">
63
+ <data batch_dims="0" />
64
+ <input>
65
+ <port id="0" precision="I64">
66
+ <dim>1</dim>
67
+ </port>
68
+ <port id="1" precision="I64" />
69
+ <port id="2" precision="I64" />
70
+ </input>
71
+ <output>
72
+ <port id="3" precision="I64" />
73
+ </output>
74
+ </layer>
75
+ <layer id="7" name="Constant_51920" type="Const" version="opset1">
76
+ <data element_type="i64" shape="" offset="8" size="8" />
77
+ <output>
78
+ <port id="0" precision="I64" />
79
+ </output>
80
+ </layer>
81
+ <layer id="8" name="Range_51921" type="Range" version="opset4">
82
+ <data output_type="i32" />
83
+ <input>
84
+ <port id="0" precision="I64" />
85
+ <port id="1" precision="I64" />
86
+ <port id="2" precision="I64" />
87
+ </input>
88
+ <output>
89
+ <port id="3" precision="I32">
90
+ <dim>-1</dim>
91
+ </port>
92
+ </output>
93
+ </layer>
94
+ <layer id="9" name="Constant_51922" type="Const" version="opset1">
95
+ <data element_type="i64" shape="" offset="8" size="8" />
96
+ <output>
97
+ <port id="0" precision="I64" />
98
+ </output>
99
+ </layer>
100
+ <layer id="10" name="Constant_51923" type="Const" version="opset1">
101
+ <data element_type="i64" shape="" offset="8" size="8" />
102
+ <output>
103
+ <port id="0" precision="I64" />
104
+ </output>
105
+ </layer>
106
+ <layer id="11" name="Add_51924" type="Add" version="opset1">
107
+ <data auto_broadcast="numpy" />
108
+ <input>
109
+ <port id="0" precision="I64" />
110
+ <port id="1" precision="I64" />
111
+ </input>
112
+ <output>
113
+ <port id="2" precision="I64" />
114
+ </output>
115
+ </layer>
116
+ <layer id="12" name="Constant_51925" type="Const" version="opset1">
117
+ <data element_type="i64" shape="" offset="8" size="8" />
118
+ <output>
119
+ <port id="0" precision="I64" />
120
+ </output>
121
+ </layer>
122
+ <layer id="13" name="Range_51926" type="Range" version="opset4">
123
+ <data output_type="i32" />
124
+ <input>
125
+ <port id="0" precision="I64" />
126
+ <port id="1" precision="I64" />
127
+ <port id="2" precision="I64" />
128
+ </input>
129
+ <output>
130
+ <port id="3" precision="I32">
131
+ <dim>-1</dim>
132
+ </port>
133
+ </output>
134
+ </layer>
135
+ <layer id="14" name="Constant_51988" type="Const" version="opset1">
136
+ <data element_type="u8" shape="37" offset="16" size="37" />
137
+ <output>
138
+ <port id="0" precision="U8">
139
+ <dim>37</dim>
140
+ </port>
141
+ </output>
142
+ </layer>
143
+ <layer id="15" name="SpecialTokensSplit_51989" type="SpecialTokensSplit" version="extension">
144
+ <input>
145
+ <port id="0" precision="I32">
146
+ <dim>-1</dim>
147
+ </port>
148
+ <port id="1" precision="I32">
149
+ <dim>-1</dim>
150
+ </port>
151
+ <port id="2" precision="I32">
152
+ <dim>-1</dim>
153
+ </port>
154
+ <port id="3" precision="I32">
155
+ <dim>-1</dim>
156
+ </port>
157
+ <port id="4" precision="U8">
158
+ <dim>-1</dim>
159
+ </port>
160
+ <port id="5" precision="U8">
161
+ <dim>37</dim>
162
+ </port>
163
+ </input>
164
+ <output>
165
+ <port id="6" precision="I32">
166
+ <dim>-1</dim>
167
+ </port>
168
+ <port id="7" precision="I32">
169
+ <dim>-1</dim>
170
+ </port>
171
+ <port id="8" precision="I32">
172
+ <dim>-1</dim>
173
+ </port>
174
+ <port id="9" precision="I32">
175
+ <dim>-1</dim>
176
+ </port>
177
+ <port id="10" precision="U8">
178
+ <dim>-1</dim>
179
+ </port>
180
+ <port id="11" precision="BOOL">
181
+ <dim>-1</dim>
182
+ </port>
183
+ </output>
184
+ </layer>
185
+ <layer id="16" name="Constant_51991" type="Const" version="opset1">
186
+ <data element_type="u8" shape="36" offset="53" size="36" />
187
+ <output>
188
+ <port id="0" precision="U8">
189
+ <dim>36</dim>
190
+ </port>
191
+ </output>
192
+ </layer>
193
+ <layer id="17" name="RegexSplit_51992" type="RegexSplit" version="extension">
194
+ <data behaviour="isolate" invert="false" max_splits="-1" />
195
+ <input>
196
+ <port id="0" precision="I32">
197
+ <dim>-1</dim>
198
+ </port>
199
+ <port id="1" precision="I32">
200
+ <dim>-1</dim>
201
+ </port>
202
+ <port id="2" precision="I32">
203
+ <dim>-1</dim>
204
+ </port>
205
+ <port id="3" precision="I32">
206
+ <dim>-1</dim>
207
+ </port>
208
+ <port id="4" precision="U8">
209
+ <dim>-1</dim>
210
+ </port>
211
+ <port id="5" precision="BOOL">
212
+ <dim>-1</dim>
213
+ </port>
214
+ <port id="6" precision="U8">
215
+ <dim>36</dim>
216
+ </port>
217
+ </input>
218
+ <output>
219
+ <port id="7" precision="I32">
220
+ <dim>-1</dim>
221
+ </port>
222
+ <port id="8" precision="I32">
223
+ <dim>-1</dim>
224
+ </port>
225
+ <port id="9" precision="I32">
226
+ <dim>-1</dim>
227
+ </port>
228
+ <port id="10" precision="I32">
229
+ <dim>-1</dim>
230
+ </port>
231
+ <port id="11" precision="U8">
232
+ <dim>-1</dim>
233
+ </port>
234
+ <port id="12" precision="BOOL">
235
+ <dim>-1</dim>
236
+ </port>
237
+ </output>
238
+ </layer>
239
+ <layer id="18" name="Constant_51994" type="Const" version="opset1">
240
+ <data element_type="u8" shape="3189816" offset="89" size="3189816" />
241
+ <output>
242
+ <port id="0" precision="U8">
243
+ <dim>3189816</dim>
244
+ </port>
245
+ </output>
246
+ </layer>
247
+ <layer id="19" name="StringTensorUnpack_51995" type="StringTensorUnpack" version="extension">
248
+ <data mode="begins_ends" />
249
+ <input>
250
+ <port id="0" precision="U8">
251
+ <dim>3189816</dim>
252
+ </port>
253
+ </input>
254
+ <output>
255
+ <port id="1" precision="I32">
256
+ <dim>-1</dim>
257
+ </port>
258
+ <port id="2" precision="I32">
259
+ <dim>-1</dim>
260
+ </port>
261
+ <port id="3" precision="U8">
262
+ <dim>-1</dim>
263
+ </port>
264
+ </output>
265
+ </layer>
266
+ <layer id="20" name="Constant_52000" type="Const" version="opset1">
267
+ <data element_type="u8" shape="2148125" offset="3189905" size="2148125" />
268
+ <output>
269
+ <port id="0" precision="U8">
270
+ <dim>2148125</dim>
271
+ </port>
272
+ </output>
273
+ </layer>
274
+ <layer id="21" name="StringTensorUnpack_52001" type="StringTensorUnpack" version="extension">
275
+ <data mode="begins_ends" />
276
+ <input>
277
+ <port id="0" precision="U8">
278
+ <dim>2148125</dim>
279
+ </port>
280
+ </input>
281
+ <output>
282
+ <port id="1" precision="I32">
283
+ <dim>-1</dim>
284
+ </port>
285
+ <port id="2" precision="I32">
286
+ <dim>-1</dim>
287
+ </port>
288
+ <port id="3" precision="U8">
289
+ <dim>-1</dim>
290
+ </port>
291
+ </output>
292
+ </layer>
293
+ <layer id="22" name="Constant_52003" type="Const" version="opset1">
294
+ <data element_type="u8" shape="2042192" offset="5338030" size="2042192" />
295
+ <output>
296
+ <port id="0" precision="U8">
297
+ <dim>2042192</dim>
298
+ </port>
299
+ </output>
300
+ </layer>
301
+ <layer id="23" name="StringTensorUnpack_52004" type="StringTensorUnpack" version="extension">
302
+ <data mode="begins_ends" />
303
+ <input>
304
+ <port id="0" precision="U8">
305
+ <dim>2042192</dim>
306
+ </port>
307
+ </input>
308
+ <output>
309
+ <port id="1" precision="I32">
310
+ <dim>-1</dim>
311
+ </port>
312
+ <port id="2" precision="I32">
313
+ <dim>-1</dim>
314
+ </port>
315
+ <port id="3" precision="U8">
316
+ <dim>-1</dim>
317
+ </port>
318
+ </output>
319
+ </layer>
320
+ <layer id="24" name="Constant_51997" type="Const" version="opset1">
321
+ <data element_type="u8" shape="32" offset="7380222" size="32" />
322
+ <output>
323
+ <port id="0" precision="U8">
324
+ <dim>32</dim>
325
+ </port>
326
+ </output>
327
+ </layer>
328
+ <layer id="25" name="StringTensorUnpack_51998" type="StringTensorUnpack" version="extension">
329
+ <data mode="begins_ends" />
330
+ <input>
331
+ <port id="0" precision="U8">
332
+ <dim>32</dim>
333
+ </port>
334
+ </input>
335
+ <output>
336
+ <port id="1" precision="I32">
337
+ <dim>-1</dim>
338
+ </port>
339
+ <port id="2" precision="I32">
340
+ <dim>-1</dim>
341
+ </port>
342
+ <port id="3" precision="U8">
343
+ <dim>-1</dim>
344
+ </port>
345
+ </output>
346
+ </layer>
347
+ <layer id="26" name="Constant_52005" type="Const" version="opset1">
348
+ <data element_type="i32" shape="3" offset="7380254" size="12" />
349
+ <output>
350
+ <port id="0" precision="I32">
351
+ <dim>3</dim>
352
+ </port>
353
+ </output>
354
+ </layer>
355
+ <layer id="27" name="BPETokenizer_52006" type="BPETokenizer" version="extension">
356
+ <data unk_token="" fuse_unk="false" suffix_indicator="" end_suffix="" byte_fallback="false" cache_capacity="50136" />
357
+ <input>
358
+ <port id="0" precision="I32">
359
+ <dim>-1</dim>
360
+ </port>
361
+ <port id="1" precision="I32">
362
+ <dim>-1</dim>
363
+ </port>
364
+ <port id="2" precision="I32">
365
+ <dim>-1</dim>
366
+ </port>
367
+ <port id="3" precision="I32">
368
+ <dim>-1</dim>
369
+ </port>
370
+ <port id="4" precision="U8">
371
+ <dim>-1</dim>
372
+ </port>
373
+ <port id="5" precision="I32">
374
+ <dim>-1</dim>
375
+ </port>
376
+ <port id="6" precision="I32">
377
+ <dim>-1</dim>
378
+ </port>
379
+ <port id="7" precision="U8">
380
+ <dim>-1</dim>
381
+ </port>
382
+ <port id="8" precision="I32">
383
+ <dim>-1</dim>
384
+ </port>
385
+ <port id="9" precision="I32">
386
+ <dim>-1</dim>
387
+ </port>
388
+ <port id="10" precision="U8">
389
+ <dim>-1</dim>
390
+ </port>
391
+ <port id="11" precision="I32">
392
+ <dim>-1</dim>
393
+ </port>
394
+ <port id="12" precision="I32">
395
+ <dim>-1</dim>
396
+ </port>
397
+ <port id="13" precision="U8">
398
+ <dim>-1</dim>
399
+ </port>
400
+ <port id="14" precision="I32">
401
+ <dim>-1</dim>
402
+ </port>
403
+ <port id="15" precision="I32">
404
+ <dim>-1</dim>
405
+ </port>
406
+ <port id="16" precision="U8">
407
+ <dim>-1</dim>
408
+ </port>
409
+ <port id="17" precision="I32">
410
+ <dim>3</dim>
411
+ </port>
412
+ </input>
413
+ <output>
414
+ <port id="18" precision="I32">
415
+ <dim>-1</dim>
416
+ </port>
417
+ <port id="19" precision="I32">
418
+ <dim>-1</dim>
419
+ </port>
420
+ <port id="20" precision="I32">
421
+ <dim>-1</dim>
422
+ </port>
423
+ </output>
424
+ </layer>
425
+ <layer id="28" name="Subtract_52007" type="Subtract" version="opset1">
426
+ <data auto_broadcast="numpy" />
427
+ <input>
428
+ <port id="0" precision="I32">
429
+ <dim>-1</dim>
430
+ </port>
431
+ <port id="1" precision="I32">
432
+ <dim>-1</dim>
433
+ </port>
434
+ </input>
435
+ <output>
436
+ <port id="2" precision="I32">
437
+ <dim>-1</dim>
438
+ </port>
439
+ </output>
440
+ </layer>
441
+ <layer id="29" name="Constant_52008" type="Const" version="opset1">
442
+ <data element_type="i32" shape="" offset="7380266" size="4" />
443
+ <output>
444
+ <port id="0" precision="I32" />
445
+ </output>
446
+ </layer>
447
+ <layer id="30" name="Minimum_52009" type="Minimum" version="opset1">
448
+ <data auto_broadcast="numpy" />
449
+ <input>
450
+ <port id="0" precision="I32">
451
+ <dim>-1</dim>
452
+ </port>
453
+ <port id="1" precision="I32" />
454
+ </input>
455
+ <output>
456
+ <port id="2" precision="I32">
457
+ <dim>-1</dim>
458
+ </port>
459
+ </output>
460
+ </layer>
461
+ <layer id="31" name="Subtract_52010" type="Subtract" version="opset1">
462
+ <data auto_broadcast="numpy" />
463
+ <input>
464
+ <port id="0" precision="I32">
465
+ <dim>-1</dim>
466
+ </port>
467
+ <port id="1" precision="I32">
468
+ <dim>-1</dim>
469
+ </port>
470
+ </input>
471
+ <output>
472
+ <port id="2" precision="I32">
473
+ <dim>-1</dim>
474
+ </port>
475
+ </output>
476
+ </layer>
477
+ <layer id="32" name="Subtract_52011" type="Subtract" version="opset1">
478
+ <data auto_broadcast="numpy" />
479
+ <input>
480
+ <port id="0" precision="I32">
481
+ <dim>-1</dim>
482
+ </port>
483
+ <port id="1" precision="I32">
484
+ <dim>-1</dim>
485
+ </port>
486
+ </input>
487
+ <output>
488
+ <port id="2" precision="I32">
489
+ <dim>-1</dim>
490
+ </port>
491
+ </output>
492
+ </layer>
493
+ <layer id="33" name="Constant_52012" type="Const" version="opset1">
494
+ <data element_type="i32" shape="" offset="7380270" size="4" />
495
+ <output>
496
+ <port id="0" precision="I32" />
497
+ </output>
498
+ </layer>
499
+ <layer id="34" name="ReduceMax_52013" type="ReduceMax" version="opset1">
500
+ <data keep_dims="false" />
501
+ <input>
502
+ <port id="0" precision="I32">
503
+ <dim>-1</dim>
504
+ </port>
505
+ <port id="1" precision="I32" />
506
+ </input>
507
+ <output>
508
+ <port id="2" precision="I32" />
509
+ </output>
510
+ </layer>
511
+ <layer id="35" name="Constant_52014" type="Const" version="opset1">
512
+ <data element_type="i32" shape="" offset="7380274" size="4" />
513
+ <output>
514
+ <port id="0" precision="I32" />
515
+ </output>
516
+ </layer>
517
+ <layer id="36" name="RaggedToDense_52015" type="RaggedToDense" version="extension">
518
+ <data pad_right="false" />
519
+ <input>
520
+ <port id="0" precision="I32">
521
+ <dim>-1</dim>
522
+ </port>
523
+ <port id="1" precision="I32">
524
+ <dim>-1</dim>
525
+ </port>
526
+ <port id="2" precision="I32">
527
+ <dim>-1</dim>
528
+ </port>
529
+ <port id="3" precision="I32" />
530
+ <port id="4" precision="I32" />
531
+ </input>
532
+ <output>
533
+ <port id="5" precision="I32">
534
+ <dim>-1</dim>
535
+ <dim>-1</dim>
536
+ </port>
537
+ <port id="6" precision="BOOL">
538
+ <dim>-1</dim>
539
+ <dim>-1</dim>
540
+ </port>
541
+ </output>
542
+ </layer>
543
+ <layer id="37" name="Convert_52016" type="Convert" version="opset1">
544
+ <data destination_type="i32" />
545
+ <input>
546
+ <port id="0" precision="BOOL">
547
+ <dim>-1</dim>
548
+ <dim>-1</dim>
549
+ </port>
550
+ </input>
551
+ <output>
552
+ <port id="1" precision="I32">
553
+ <dim>-1</dim>
554
+ <dim>-1</dim>
555
+ </port>
556
+ </output>
557
+ </layer>
558
+ <layer id="38" name="Convert_52016" type="Convert" version="opset1">
559
+ <data destination_type="i64" />
560
+ <input>
561
+ <port id="0" precision="I32">
562
+ <dim>-1</dim>
563
+ <dim>-1</dim>
564
+ </port>
565
+ </input>
566
+ <output>
567
+ <port id="1" precision="I64" names="attention_mask">
568
+ <dim>-1</dim>
569
+ <dim>-1</dim>
570
+ </port>
571
+ </output>
572
+ </layer>
573
+ <layer id="40" name="RaggedToDense_52015.0" type="Convert" version="opset1">
574
+ <data destination_type="i64" />
575
+ <input>
576
+ <port id="0" precision="I32">
577
+ <dim>-1</dim>
578
+ <dim>-1</dim>
579
+ </port>
580
+ </input>
581
+ <output>
582
+ <port id="1" precision="I64" names="input_ids">
583
+ <dim>-1</dim>
584
+ <dim>-1</dim>
585
+ </port>
586
+ </output>
587
+ </layer>
588
+ <layer id="41" name="Result_52017" type="Result" version="opset1">
589
+ <input>
590
+ <port id="0" precision="I64">
591
+ <dim>-1</dim>
592
+ <dim>-1</dim>
593
+ </port>
594
+ </input>
595
+ </layer>
596
+ <layer id="39" name="Result_52018" type="Result" version="opset1">
597
+ <input>
598
+ <port id="0" precision="I64">
599
+ <dim>-1</dim>
600
+ <dim>-1</dim>
601
+ </port>
602
+ </input>
603
+ </layer>
604
+ </layers>
605
+ <edges>
606
+ <edge from-layer="0" from-port="0" to-layer="2" to-port="0" />
607
+ <edge from-layer="1" from-port="0" to-layer="8" to-port="0" />
608
+ <edge from-layer="2" from-port="1" to-layer="3" to-port="0" />
609
+ <edge from-layer="2" from-port="3" to-layer="15" to-port="4" />
610
+ <edge from-layer="2" from-port="2" to-layer="15" to-port="3" />
611
+ <edge from-layer="2" from-port="1" to-layer="15" to-port="2" />
612
+ <edge from-layer="3" from-port="1" to-layer="6" to-port="0" />
613
+ <edge from-layer="4" from-port="0" to-layer="6" to-port="1" />
614
+ <edge from-layer="5" from-port="0" to-layer="6" to-port="2" />
615
+ <edge from-layer="6" from-port="3" to-layer="11" to-port="0" />
616
+ <edge from-layer="6" from-port="3" to-layer="8" to-port="1" />
617
+ <edge from-layer="7" from-port="0" to-layer="8" to-port="2" />
618
+ <edge from-layer="8" from-port="3" to-layer="15" to-port="0" />
619
+ <edge from-layer="9" from-port="0" to-layer="13" to-port="0" />
620
+ <edge from-layer="10" from-port="0" to-layer="11" to-port="1" />
621
+ <edge from-layer="11" from-port="2" to-layer="13" to-port="1" />
622
+ <edge from-layer="12" from-port="0" to-layer="13" to-port="2" />
623
+ <edge from-layer="13" from-port="3" to-layer="15" to-port="1" />
624
+ <edge from-layer="14" from-port="0" to-layer="15" to-port="5" />
625
+ <edge from-layer="15" from-port="6" to-layer="17" to-port="0" />
626
+ <edge from-layer="15" from-port="7" to-layer="17" to-port="1" />
627
+ <edge from-layer="15" from-port="8" to-layer="17" to-port="2" />
628
+ <edge from-layer="15" from-port="9" to-layer="17" to-port="3" />
629
+ <edge from-layer="15" from-port="10" to-layer="17" to-port="4" />
630
+ <edge from-layer="15" from-port="11" to-layer="17" to-port="5" />
631
+ <edge from-layer="16" from-port="0" to-layer="17" to-port="6" />
632
+ <edge from-layer="17" from-port="11" to-layer="27" to-port="4" />
633
+ <edge from-layer="17" from-port="10" to-layer="27" to-port="3" />
634
+ <edge from-layer="17" from-port="9" to-layer="27" to-port="2" />
635
+ <edge from-layer="17" from-port="8" to-layer="27" to-port="1" />
636
+ <edge from-layer="17" from-port="7" to-layer="27" to-port="0" />
637
+ <edge from-layer="18" from-port="0" to-layer="19" to-port="0" />
638
+ <edge from-layer="19" from-port="1" to-layer="27" to-port="5" />
639
+ <edge from-layer="19" from-port="2" to-layer="27" to-port="6" />
640
+ <edge from-layer="19" from-port="3" to-layer="27" to-port="7" />
641
+ <edge from-layer="20" from-port="0" to-layer="21" to-port="0" />
642
+ <edge from-layer="21" from-port="1" to-layer="27" to-port="8" />
643
+ <edge from-layer="21" from-port="2" to-layer="27" to-port="9" />
644
+ <edge from-layer="21" from-port="3" to-layer="27" to-port="10" />
645
+ <edge from-layer="22" from-port="0" to-layer="23" to-port="0" />
646
+ <edge from-layer="23" from-port="1" to-layer="27" to-port="11" />
647
+ <edge from-layer="23" from-port="2" to-layer="27" to-port="12" />
648
+ <edge from-layer="23" from-port="3" to-layer="27" to-port="13" />
649
+ <edge from-layer="24" from-port="0" to-layer="25" to-port="0" />
650
+ <edge from-layer="25" from-port="1" to-layer="27" to-port="14" />
651
+ <edge from-layer="25" from-port="2" to-layer="27" to-port="15" />
652
+ <edge from-layer="25" from-port="3" to-layer="27" to-port="16" />
653
+ <edge from-layer="26" from-port="0" to-layer="27" to-port="17" />
654
+ <edge from-layer="27" from-port="20" to-layer="36" to-port="2" />
655
+ <edge from-layer="27" from-port="18" to-layer="28" to-port="1" />
656
+ <edge from-layer="27" from-port="19" to-layer="36" to-port="1" />
657
+ <edge from-layer="27" from-port="19" to-layer="32" to-port="0" />
658
+ <edge from-layer="27" from-port="19" to-layer="31" to-port="0" />
659
+ <edge from-layer="27" from-port="19" to-layer="28" to-port="0" />
660
+ <edge from-layer="28" from-port="2" to-layer="30" to-port="0" />
661
+ <edge from-layer="29" from-port="0" to-layer="30" to-port="1" />
662
+ <edge from-layer="30" from-port="2" to-layer="31" to-port="1" />
663
+ <edge from-layer="31" from-port="2" to-layer="32" to-port="1" />
664
+ <edge from-layer="31" from-port="2" to-layer="36" to-port="0" />
665
+ <edge from-layer="32" from-port="2" to-layer="34" to-port="0" />
666
+ <edge from-layer="33" from-port="0" to-layer="34" to-port="1" />
667
+ <edge from-layer="34" from-port="2" to-layer="36" to-port="3" />
668
+ <edge from-layer="35" from-port="0" to-layer="36" to-port="4" />
669
+ <edge from-layer="36" from-port="6" to-layer="37" to-port="0" />
670
+ <edge from-layer="36" from-port="5" to-layer="40" to-port="0" />
671
+ <edge from-layer="37" from-port="1" to-layer="38" to-port="0" />
672
+ <edge from-layer="38" from-port="1" to-layer="39" to-port="0" />
673
+ <edge from-layer="40" from-port="1" to-layer="41" to-port="0" />
674
+ </edges>
675
+ <rt_info>
676
+ <bos_token_id value="1" />
677
+ <eos_token_id value="2" />
678
+ <original_tokenizer_class value="&lt;class 'transformers.models.bloom.tokenization_bloom_fast.BloomTokenizerFast'>" />
679
+ <pad_token_id value="3" />
680
+ </rt_info>
681
+ </net>
special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<pad>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<unk>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d963066d6adae5034a1dc114c3ac444512de09928cf14ed4562ba94d9a440e66
3
+ size 21763085
tokenizer_config.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "0": {
5
+ "content": "<unk>",
6
+ "lstrip": false,
7
+ "normalized": false,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ },
12
+ "1": {
13
+ "content": "<s>",
14
+ "lstrip": false,
15
+ "normalized": false,
16
+ "rstrip": false,
17
+ "single_word": false,
18
+ "special": true
19
+ },
20
+ "2": {
21
+ "content": "</s>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false,
26
+ "special": true
27
+ },
28
+ "3": {
29
+ "content": "<pad>",
30
+ "lstrip": false,
31
+ "normalized": false,
32
+ "rstrip": false,
33
+ "single_word": false,
34
+ "special": true
35
+ }
36
+ },
37
+ "bos_token": "<s>",
38
+ "clean_up_tokenization_spaces": false,
39
+ "eos_token": "</s>",
40
+ "model_max_length": 1000000000000000019884624838656,
41
+ "pad_token": "<pad>",
42
+ "tokenizer_class": "BloomTokenizer",
43
+ "unk_token": "<unk>"
44
+ }