sharpenb commited on
Commit
179db39
1 Parent(s): 5b4a895

Upload folder using huggingface_hub (#3)

Browse files

- 192568bac6f8bde59f871d417c340b64b67257e03f454be32fee737dd4f90ff4 (fd71b568ad4653b15c7e48f187c8955f4c1b51ce)

Files changed (6) hide show
  1. README.md +14 -10
  2. config.json +71 -70
  3. qmodel.pt +1 -1
  4. smash_config.json +1 -1
  5. tokenizer.json +13 -41
  6. tokenizer_config.json +15 -14
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
3
- base_model: PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed
4
  metrics:
5
  - memory_disk
6
  - memory_inference
@@ -31,14 +31,14 @@ tags:
31
  - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
32
  - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
33
  - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
34
- - Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
35
 
36
  ## Results
37
 
38
  ![image info](./plots.png)
39
 
40
  **Frequently Asked Questions**
41
- - ***How does the compression work?*** The model is compressed with [.
42
  - ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
43
  - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
44
  - ***What is the model format?*** We use safetensors.
@@ -52,18 +52,22 @@ tags:
52
 
53
  You can run the smashed model with these steps:
54
 
55
- 0. Check requirements from the original repo PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed installed. In particular, check python, cuda, and transformers versions.
56
  1. Make sure that you have installed quantization related packages.
57
  ```bash
58
- REQUIREMENTS_INSTRUCTIONS
59
  ```
60
  2. Load & run the model.
61
  ```python
62
  from transformers import AutoModelForCausalLM, AutoTokenizer
63
- IMPORTS
64
-
65
- MODEL_LOAD
66
- tokenizer = AutoTokenizer.from_pretrained("PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed")
 
 
 
 
67
 
68
  input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
69
 
@@ -77,7 +81,7 @@ The configuration info are in `smash_config.json`.
77
 
78
  ## Credits & License
79
 
80
- The license of the smashed model follows the license of the original model. Please check the license of the original model PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
81
 
82
  ## Want to compress other models?
83
 
 
1
  ---
2
  thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
3
+ base_model: microsoft/Phi-3-mini-128k-instruct
4
  metrics:
5
  - memory_disk
6
  - memory_inference
 
31
  - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
32
  - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
33
  - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
34
+ - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
35
 
36
  ## Results
37
 
38
  ![image info](./plots.png)
39
 
40
  **Frequently Asked Questions**
41
+ - ***How does the compression work?*** The model is compressed with hqq.
42
  - ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
43
  - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
44
  - ***What is the model format?*** We use safetensors.
 
52
 
53
  You can run the smashed model with these steps:
54
 
55
+ 0. Check requirements from the original repo microsoft/Phi-3-mini-128k-instruct installed. In particular, check python, cuda, and transformers versions.
56
  1. Make sure that you have installed quantization related packages.
57
  ```bash
58
+ pip install hqq
59
  ```
60
  2. Load & run the model.
61
  ```python
62
  from transformers import AutoModelForCausalLM, AutoTokenizer
63
+ from hqq.engine.hf import HQQModelForCausalLM
64
+ from hqq.models.hf.base import AutoHQQHFModel
65
+
66
+ try:
67
+ model = HQQModelForCausalLM.from_quantized("PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed", device_map='auto')
68
+ except:
69
+ model = AutoHQQHFModel.from_quantized("PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed")
70
+ tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-128k-instruct")
71
 
72
  input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
73
 
 
81
 
82
  ## Credits & License
83
 
84
+ The license of the smashed model follows the license of the original model. Please check the license of the original model microsoft/Phi-3-mini-128k-instruct before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
85
 
86
  ## Want to compress other models?
87
 
config.json CHANGED
@@ -3,6 +3,7 @@
3
  "architectures": [
4
  "Phi3ForCausalLM"
5
  ],
 
6
  "attention_dropout": 0.0,
7
  "auto_map": {
8
  "AutoConfig": "microsoft/Phi-3-mini-128k-instruct--configuration_phi3.Phi3Config",
@@ -26,71 +27,64 @@
26
  "rms_norm_eps": 1e-05,
27
  "rope_scaling": {
28
  "long_factor": [
29
- 1.0299999713897705,
30
- 1.0499999523162842,
31
- 1.0499999523162842,
32
- 1.0799999237060547,
33
- 1.2299998998641968,
34
- 1.2299998998641968,
35
- 1.2999999523162842,
36
- 1.4499999284744263,
37
- 1.5999999046325684,
38
- 1.6499998569488525,
39
- 1.8999998569488525,
40
- 2.859999895095825,
41
- 3.68999981880188,
42
- 5.419999599456787,
43
- 5.489999771118164,
44
- 5.489999771118164,
45
- 9.09000015258789,
46
- 11.579999923706055,
47
- 15.65999984741211,
48
- 15.769999504089355,
49
- 15.789999961853027,
50
- 18.360000610351562,
51
- 21.989999771118164,
52
- 23.079999923706055,
53
- 30.009998321533203,
54
- 32.35000228881836,
55
- 32.590003967285156,
56
- 35.56000518798828,
57
- 39.95000457763672,
58
- 53.840003967285156,
59
- 56.20000457763672,
60
- 57.95000457763672,
61
- 59.29000473022461,
62
- 59.77000427246094,
63
- 59.920005798339844,
64
- 61.190006256103516,
65
- 61.96000671386719,
66
- 62.50000762939453,
67
- 63.3700065612793,
68
- 63.48000717163086,
69
- 63.48000717163086,
70
- 63.66000747680664,
71
- 63.850006103515625,
72
- 64.08000946044922,
73
- 64.760009765625,
74
- 64.80001068115234,
75
- 64.81001281738281,
76
- 64.81001281738281
77
  ],
78
  "short_factor": [
79
- 1.05,
80
- 1.05,
81
- 1.05,
82
  1.1,
83
  1.1,
84
- 1.1500000000000001,
85
- 1.2000000000000002,
86
- 1.2500000000000002,
87
  1.3000000000000003,
88
  1.3500000000000003,
89
- 1.5000000000000004,
90
- 2.000000000000001,
91
- 2.000000000000001,
92
- 2.000000000000001,
93
- 2.000000000000001,
94
  2.000000000000001,
95
  2.000000000000001,
96
  2.000000000000001,
@@ -111,27 +105,34 @@
111
  2.0500000000000007,
112
  2.0500000000000007,
113
  2.0500000000000007,
 
 
 
114
  2.1000000000000005,
115
  2.1000000000000005,
116
- 2.1000000000000005,
117
- 2.1500000000000004,
118
  2.1500000000000004,
119
- 2.3499999999999996,
120
- 2.549999999999999,
121
- 2.5999999999999988,
122
- 2.5999999999999988,
 
 
 
 
 
 
123
  2.7499999999999982,
124
- 2.849999999999998,
125
- 2.849999999999998,
126
- 2.9499999999999975
127
  ],
128
- "type": "su"
129
  },
130
  "rope_theta": 10000.0,
131
  "sliding_window": 262144,
132
  "tie_word_embeddings": false,
133
  "torch_dtype": "bfloat16",
134
- "transformers_version": "4.40.0",
135
  "use_cache": true,
136
  "vocab_size": 32064
137
  }
 
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",
 
27
  "rms_norm_eps": 1e-05,
28
  "rope_scaling": {
29
  "long_factor": [
30
+ 1.0700000524520874,
31
+ 1.1200000047683716,
32
+ 1.149999976158142,
33
+ 1.4199999570846558,
34
+ 1.5699999332427979,
35
+ 1.7999999523162842,
36
+ 2.129999876022339,
37
+ 2.129999876022339,
38
+ 3.009999990463257,
39
+ 5.910000324249268,
40
+ 6.950000286102295,
41
+ 9.070000648498535,
42
+ 9.930000305175781,
43
+ 10.710000038146973,
44
+ 11.130000114440918,
45
+ 14.609999656677246,
46
+ 15.409998893737793,
47
+ 19.809999465942383,
48
+ 37.279998779296875,
49
+ 38.279998779296875,
50
+ 38.599998474121094,
51
+ 40.12000274658203,
52
+ 46.20000457763672,
53
+ 50.940006256103516,
54
+ 53.66000747680664,
55
+ 54.9373893737793,
56
+ 56.89738845825195,
57
+ 57.28738784790039,
58
+ 59.98738479614258,
59
+ 60.86738586425781,
60
+ 60.887386322021484,
61
+ 61.71739196777344,
62
+ 62.91739273071289,
63
+ 62.957393646240234,
64
+ 63.41739273071289,
65
+ 63.8173942565918,
66
+ 63.83739471435547,
67
+ 63.897396087646484,
68
+ 63.93739700317383,
69
+ 64.06739807128906,
70
+ 64.11434936523438,
71
+ 64.12435150146484,
72
+ 64.15435028076172,
73
+ 64.19435119628906,
74
+ 64.24435424804688,
75
+ 64.57435607910156,
76
+ 64.69000244140625,
77
+ 64.76000213623047
78
  ],
79
  "short_factor": [
 
 
 
80
  1.1,
81
  1.1,
82
+ 1.1,
 
 
83
  1.3000000000000003,
84
  1.3500000000000003,
85
+ 1.3500000000000003,
86
+ 1.4000000000000004,
87
+ 1.5500000000000005,
 
 
88
  2.000000000000001,
89
  2.000000000000001,
90
  2.000000000000001,
 
105
  2.0500000000000007,
106
  2.0500000000000007,
107
  2.0500000000000007,
108
+ 2.0500000000000007,
109
+ 2.0500000000000007,
110
+ 2.0500000000000007,
111
  2.1000000000000005,
112
  2.1000000000000005,
 
 
113
  2.1500000000000004,
114
+ 2.25,
115
+ 2.25,
116
+ 2.25,
117
+ 2.25,
118
+ 2.25,
119
+ 2.3999999999999995,
120
+ 2.4499999999999993,
121
+ 2.499999999999999,
122
+ 2.6999999999999984,
123
+ 2.6999999999999984,
124
  2.7499999999999982,
125
+ 2.799999999999998,
126
+ 2.8999999999999977,
127
+ 3.049999999999997
128
  ],
129
+ "type": "longrope"
130
  },
131
  "rope_theta": 10000.0,
132
  "sliding_window": 262144,
133
  "tie_word_embeddings": false,
134
  "torch_dtype": "bfloat16",
135
+ "transformers_version": "4.42.4",
136
  "use_cache": true,
137
  "vocab_size": 32064
138
  }
qmodel.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:35990b2920da3a31e29488e6cd3c3322b186ac33a8806c5fcb2f01caeaea000c
3
  size 1385530405
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90511a5d2fcdfd0211978119553af91c7feccc925b2fd57479a823cf9491792f
3
  size 1385530405
smash_config.json CHANGED
@@ -14,7 +14,7 @@
14
  "controlnet": "None",
15
  "unet_dim": 4,
16
  "device": "cuda",
17
- "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsni1858vt",
18
  "batch_size": 1,
19
  "model_name": "microsoft/Phi-3-mini-128k-instruct",
20
  "task": "text_text_generation",
 
14
  "controlnet": "None",
15
  "unet_dim": 4,
16
  "device": "cuda",
17
+ "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsx7u5os_k",
18
  "batch_size": 1,
19
  "model_name": "microsoft/Phi-3-mini-128k-instruct",
20
  "task": "text_text_generation",
tokenizer.json CHANGED
@@ -26,9 +26,9 @@
26
  "content": "</s>",
27
  "single_word": false,
28
  "lstrip": false,
29
- "rstrip": false,
30
  "normalized": false,
31
- "special": true
32
  },
33
  {
34
  "id": 32000,
@@ -44,7 +44,7 @@
44
  "content": "<|assistant|>",
45
  "single_word": false,
46
  "lstrip": false,
47
- "rstrip": false,
48
  "normalized": false,
49
  "special": true
50
  },
@@ -53,7 +53,7 @@
53
  "content": "<|placeholder1|>",
54
  "single_word": false,
55
  "lstrip": false,
56
- "rstrip": false,
57
  "normalized": false,
58
  "special": true
59
  },
@@ -62,7 +62,7 @@
62
  "content": "<|placeholder2|>",
63
  "single_word": false,
64
  "lstrip": false,
65
- "rstrip": false,
66
  "normalized": false,
67
  "special": true
68
  },
@@ -71,7 +71,7 @@
71
  "content": "<|placeholder3|>",
72
  "single_word": false,
73
  "lstrip": false,
74
- "rstrip": false,
75
  "normalized": false,
76
  "special": true
77
  },
@@ -80,7 +80,7 @@
80
  "content": "<|placeholder4|>",
81
  "single_word": false,
82
  "lstrip": false,
83
- "rstrip": false,
84
  "normalized": false,
85
  "special": true
86
  },
@@ -89,7 +89,7 @@
89
  "content": "<|system|>",
90
  "single_word": false,
91
  "lstrip": false,
92
- "rstrip": false,
93
  "normalized": false,
94
  "special": true
95
  },
@@ -98,7 +98,7 @@
98
  "content": "<|end|>",
99
  "single_word": false,
100
  "lstrip": false,
101
- "rstrip": false,
102
  "normalized": false,
103
  "special": true
104
  },
@@ -107,7 +107,7 @@
107
  "content": "<|placeholder5|>",
108
  "single_word": false,
109
  "lstrip": false,
110
- "rstrip": false,
111
  "normalized": false,
112
  "special": true
113
  },
@@ -116,7 +116,7 @@
116
  "content": "<|placeholder6|>",
117
  "single_word": false,
118
  "lstrip": false,
119
- "rstrip": false,
120
  "normalized": false,
121
  "special": true
122
  },
@@ -125,7 +125,7 @@
125
  "content": "<|user|>",
126
  "single_word": false,
127
  "lstrip": false,
128
- "rstrip": false,
129
  "normalized": false,
130
  "special": true
131
  }
@@ -150,12 +150,6 @@
150
  "post_processor": {
151
  "type": "TemplateProcessing",
152
  "single": [
153
- {
154
- "SpecialToken": {
155
- "id": "<s>",
156
- "type_id": 0
157
- }
158
- },
159
  {
160
  "Sequence": {
161
  "id": "A",
@@ -164,24 +158,12 @@
164
  }
165
  ],
166
  "pair": [
167
- {
168
- "SpecialToken": {
169
- "id": "<s>",
170
- "type_id": 0
171
- }
172
- },
173
  {
174
  "Sequence": {
175
  "id": "A",
176
  "type_id": 0
177
  }
178
  },
179
- {
180
- "SpecialToken": {
181
- "id": "<s>",
182
- "type_id": 1
183
- }
184
- },
185
  {
186
  "Sequence": {
187
  "id": "B",
@@ -189,17 +171,7 @@
189
  }
190
  }
191
  ],
192
- "special_tokens": {
193
- "<s>": {
194
- "id": "<s>",
195
- "ids": [
196
- 1
197
- ],
198
- "tokens": [
199
- "<s>"
200
- ]
201
- }
202
- }
203
  },
204
  "decoder": {
205
  "type": "Sequence",
 
26
  "content": "</s>",
27
  "single_word": false,
28
  "lstrip": false,
29
+ "rstrip": true,
30
  "normalized": false,
31
+ "special": false
32
  },
33
  {
34
  "id": 32000,
 
44
  "content": "<|assistant|>",
45
  "single_word": false,
46
  "lstrip": false,
47
+ "rstrip": true,
48
  "normalized": false,
49
  "special": true
50
  },
 
53
  "content": "<|placeholder1|>",
54
  "single_word": false,
55
  "lstrip": false,
56
+ "rstrip": true,
57
  "normalized": false,
58
  "special": true
59
  },
 
62
  "content": "<|placeholder2|>",
63
  "single_word": false,
64
  "lstrip": false,
65
+ "rstrip": true,
66
  "normalized": false,
67
  "special": true
68
  },
 
71
  "content": "<|placeholder3|>",
72
  "single_word": false,
73
  "lstrip": false,
74
+ "rstrip": true,
75
  "normalized": false,
76
  "special": true
77
  },
 
80
  "content": "<|placeholder4|>",
81
  "single_word": false,
82
  "lstrip": false,
83
+ "rstrip": true,
84
  "normalized": false,
85
  "special": true
86
  },
 
89
  "content": "<|system|>",
90
  "single_word": false,
91
  "lstrip": false,
92
+ "rstrip": true,
93
  "normalized": false,
94
  "special": true
95
  },
 
98
  "content": "<|end|>",
99
  "single_word": false,
100
  "lstrip": false,
101
+ "rstrip": true,
102
  "normalized": false,
103
  "special": true
104
  },
 
107
  "content": "<|placeholder5|>",
108
  "single_word": false,
109
  "lstrip": false,
110
+ "rstrip": true,
111
  "normalized": false,
112
  "special": true
113
  },
 
116
  "content": "<|placeholder6|>",
117
  "single_word": false,
118
  "lstrip": false,
119
+ "rstrip": true,
120
  "normalized": false,
121
  "special": true
122
  },
 
125
  "content": "<|user|>",
126
  "single_word": false,
127
  "lstrip": false,
128
+ "rstrip": true,
129
  "normalized": false,
130
  "special": true
131
  }
 
150
  "post_processor": {
151
  "type": "TemplateProcessing",
152
  "single": [
 
 
 
 
 
 
153
  {
154
  "Sequence": {
155
  "id": "A",
 
158
  }
159
  ],
160
  "pair": [
 
 
 
 
 
 
161
  {
162
  "Sequence": {
163
  "id": "A",
164
  "type_id": 0
165
  }
166
  },
 
 
 
 
 
 
167
  {
168
  "Sequence": {
169
  "id": "B",
 
171
  }
172
  }
173
  ],
174
+ "special_tokens": {}
 
 
 
 
 
 
 
 
 
 
175
  },
176
  "decoder": {
177
  "type": "Sequence",
tokenizer_config.json CHANGED
@@ -1,6 +1,7 @@
1
  {
2
- "add_bos_token": true,
3
  "add_eos_token": false,
 
4
  "added_tokens_decoder": {
5
  "0": {
6
  "content": "<unk>",
@@ -22,9 +23,9 @@
22
  "content": "</s>",
23
  "lstrip": false,
24
  "normalized": false,
25
- "rstrip": false,
26
  "single_word": false,
27
- "special": true
28
  },
29
  "32000": {
30
  "content": "<|endoftext|>",
@@ -38,7 +39,7 @@
38
  "content": "<|assistant|>",
39
  "lstrip": false,
40
  "normalized": false,
41
- "rstrip": false,
42
  "single_word": false,
43
  "special": true
44
  },
@@ -46,7 +47,7 @@
46
  "content": "<|placeholder1|>",
47
  "lstrip": false,
48
  "normalized": false,
49
- "rstrip": false,
50
  "single_word": false,
51
  "special": true
52
  },
@@ -54,7 +55,7 @@
54
  "content": "<|placeholder2|>",
55
  "lstrip": false,
56
  "normalized": false,
57
- "rstrip": false,
58
  "single_word": false,
59
  "special": true
60
  },
@@ -62,7 +63,7 @@
62
  "content": "<|placeholder3|>",
63
  "lstrip": false,
64
  "normalized": false,
65
- "rstrip": false,
66
  "single_word": false,
67
  "special": true
68
  },
@@ -70,7 +71,7 @@
70
  "content": "<|placeholder4|>",
71
  "lstrip": false,
72
  "normalized": false,
73
- "rstrip": false,
74
  "single_word": false,
75
  "special": true
76
  },
@@ -78,7 +79,7 @@
78
  "content": "<|system|>",
79
  "lstrip": false,
80
  "normalized": false,
81
- "rstrip": false,
82
  "single_word": false,
83
  "special": true
84
  },
@@ -86,7 +87,7 @@
86
  "content": "<|end|>",
87
  "lstrip": false,
88
  "normalized": false,
89
- "rstrip": false,
90
  "single_word": false,
91
  "special": true
92
  },
@@ -94,7 +95,7 @@
94
  "content": "<|placeholder5|>",
95
  "lstrip": false,
96
  "normalized": false,
97
- "rstrip": false,
98
  "single_word": false,
99
  "special": true
100
  },
@@ -102,7 +103,7 @@
102
  "content": "<|placeholder6|>",
103
  "lstrip": false,
104
  "normalized": false,
105
- "rstrip": false,
106
  "single_word": false,
107
  "special": true
108
  },
@@ -110,13 +111,13 @@
110
  "content": "<|user|>",
111
  "lstrip": false,
112
  "normalized": false,
113
- "rstrip": false,
114
  "single_word": false,
115
  "special": true
116
  }
117
  },
118
  "bos_token": "<s>",
119
- "chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] in ['user', 'system']) %}{{'<|user|>' + '\n' + message['content'] + '<|end|>' + '\n' + '<|assistant|>' + '\n'}}{% elif message['role'] == 'assistant' %}{{message['content'] + '<|end|>' + '\n'}}{% endif %}{% endfor %}",
120
  "clean_up_tokenization_spaces": false,
121
  "eos_token": "<|endoftext|>",
122
  "legacy": false,
 
1
  {
2
+ "add_bos_token": false,
3
  "add_eos_token": false,
4
+ "add_prefix_space": null,
5
  "added_tokens_decoder": {
6
  "0": {
7
  "content": "<unk>",
 
23
  "content": "</s>",
24
  "lstrip": false,
25
  "normalized": false,
26
+ "rstrip": true,
27
  "single_word": false,
28
+ "special": false
29
  },
30
  "32000": {
31
  "content": "<|endoftext|>",
 
39
  "content": "<|assistant|>",
40
  "lstrip": false,
41
  "normalized": false,
42
+ "rstrip": true,
43
  "single_word": false,
44
  "special": true
45
  },
 
47
  "content": "<|placeholder1|>",
48
  "lstrip": false,
49
  "normalized": false,
50
+ "rstrip": true,
51
  "single_word": false,
52
  "special": true
53
  },
 
55
  "content": "<|placeholder2|>",
56
  "lstrip": false,
57
  "normalized": false,
58
+ "rstrip": true,
59
  "single_word": false,
60
  "special": true
61
  },
 
63
  "content": "<|placeholder3|>",
64
  "lstrip": false,
65
  "normalized": false,
66
+ "rstrip": true,
67
  "single_word": false,
68
  "special": true
69
  },
 
71
  "content": "<|placeholder4|>",
72
  "lstrip": false,
73
  "normalized": false,
74
+ "rstrip": true,
75
  "single_word": false,
76
  "special": true
77
  },
 
79
  "content": "<|system|>",
80
  "lstrip": false,
81
  "normalized": false,
82
+ "rstrip": true,
83
  "single_word": false,
84
  "special": true
85
  },
 
87
  "content": "<|end|>",
88
  "lstrip": false,
89
  "normalized": false,
90
+ "rstrip": true,
91
  "single_word": false,
92
  "special": true
93
  },
 
95
  "content": "<|placeholder5|>",
96
  "lstrip": false,
97
  "normalized": false,
98
+ "rstrip": true,
99
  "single_word": false,
100
  "special": true
101
  },
 
103
  "content": "<|placeholder6|>",
104
  "lstrip": false,
105
  "normalized": false,
106
+ "rstrip": true,
107
  "single_word": false,
108
  "special": true
109
  },
 
111
  "content": "<|user|>",
112
  "lstrip": false,
113
  "normalized": false,
114
+ "rstrip": true,
115
  "single_word": false,
116
  "special": true
117
  }
118
  },
119
  "bos_token": "<s>",
120
+ "chat_template": "{% for message in messages %}{% if message['role'] == 'system' %}{{'<|system|>\n' + message['content'] + '<|end|>\n'}}{% elif message['role'] == 'user' %}{{'<|user|>\n' + message['content'] + '<|end|>\n'}}{% elif message['role'] == 'assistant' %}{{'<|assistant|>\n' + message['content'] + '<|end|>\n'}}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>\n' }}{% else %}{{ eos_token }}{% endif %}",
121
  "clean_up_tokenization_spaces": false,
122
  "eos_token": "<|endoftext|>",
123
  "legacy": false,