Hugging Face's logo

sandi99
/
DeepQuant-lora-24140

Safetensors
mistral
4-bit precision
bitsandbytes
Model card Files
xet
 Community
sandi99 commited on
Commit
20b0d05
·
verified ·
1 Parent(s): c55240a

Upload folder using huggingface_hub

Browse files
Files changed (27) hide show
  1. .gitattributes +2 -0
  2. adapter_config.json +37 -0
  3. adapter_model.safetensors +3 -0
  4. checkpoint-276/README.md +202 -0
  5. checkpoint-276/adapter_config.json +37 -0
  6. checkpoint-276/adapter_model.safetensors +3 -0
  7. checkpoint-276/global_step276/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt +3 -0
  8. checkpoint-276/global_step276/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt +3 -0
  9. checkpoint-276/global_step276/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt +3 -0
  10. checkpoint-276/global_step276/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt +3 -0
  11. checkpoint-276/global_step276/mp_rank_00_model_states.pt +3 -0
  12. checkpoint-276/latest +1 -0
  13. checkpoint-276/rng_state_0.pth +3 -0
  14. checkpoint-276/rng_state_1.pth +3 -0
  15. checkpoint-276/rng_state_2.pth +3 -0
  16. checkpoint-276/rng_state_3.pth +3 -0
  17. checkpoint-276/scheduler.pt +3 -0
  18. checkpoint-276/special_tokens_map.json +1026 -0
  19. checkpoint-276/tokenizer.json +3 -0
  20. checkpoint-276/tokenizer_config.json +0 -0
  21. checkpoint-276/trainer_state.json +1973 -0
  22. checkpoint-276/training_args.bin +3 -0
  23. checkpoint-276/zero_to_fp32.py +760 -0
  24. config.json +43 -0
  25. special_tokens_map.json +1026 -0
  26. tokenizer.json +3 -0
  27. tokenizer_config.json +0 -0
.gitattributes CHANGED
@@ -33,3 +33,5 @@ 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
+ checkpoint-276/tokenizer.json filter=lfs diff=lfs merge=lfs -text
37
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
adapter_config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "/kaggle/input/mistral-small-24b/transformers/mistral-small-24b-instruct-2501/2",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": null,
8
+ "fan_in_fan_out": false,
9
+ "inference_mode": true,
10
+ "init_lora_weights": true,
11
+ "layer_replication": null,
12
+ "layers_pattern": null,
13
+ "layers_to_transform": null,
14
+ "loftq_config": {},
15
+ "lora_alpha": 32,
16
+ "lora_bias": false,
17
+ "lora_dropout": 0.05,
18
+ "megatron_config": null,
19
+ "megatron_core": "megatron.core",
20
+ "modules_to_save": null,
21
+ "peft_type": "LORA",
22
+ "r": 16,
23
+ "rank_pattern": {},
24
+ "revision": null,
25
+ "target_modules": [
26
+ "down_proj",
27
+ "k_proj",
28
+ "o_proj",
29
+ "gate_proj",
30
+ "q_proj",
31
+ "up_proj",
32
+ "v_proj"
33
+ ],
34
+ "task_type": "CAUSAL_LM",
35
+ "use_dora": false,
36
+ "use_rslora": false
37
+ }
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e7698c9a3331300482d43d1554de5b567b30f1fc50ed6a1b6a0ca90d2af1835
3
+ size 184887280
checkpoint-276/README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /kaggle/input/mistral-small-24b/transformers/mistral-small-24b-instruct-2501/2
3
+ library_name: peft
4
+ ---
5
+
6
+ # Model Card for Model ID
7
+
8
+ <!-- Provide a quick summary of what the model is/does. -->
9
+
10
+
11
+
12
+ ## Model Details
13
+
14
+ ### Model Description
15
+
16
+ <!-- Provide a longer summary of what this model is. -->
17
+
18
+
19
+
20
+ - **Developed by:** [More Information Needed]
21
+ - **Funded by [optional]:** [More Information Needed]
22
+ - **Shared by [optional]:** [More Information Needed]
23
+ - **Model type:** [More Information Needed]
24
+ - **Language(s) (NLP):** [More Information Needed]
25
+ - **License:** [More Information Needed]
26
+ - **Finetuned from model [optional]:** [More Information Needed]
27
+
28
+ ### Model Sources [optional]
29
+
30
+ <!-- Provide the basic links for the model. -->
31
+
32
+ - **Repository:** [More Information Needed]
33
+ - **Paper [optional]:** [More Information Needed]
34
+ - **Demo [optional]:** [More Information Needed]
35
+
36
+ ## Uses
37
+
38
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
+
40
+ ### Direct Use
41
+
42
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
+
44
+ [More Information Needed]
45
+
46
+ ### Downstream Use [optional]
47
+
48
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Out-of-Scope Use
53
+
54
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
+
56
+ [More Information Needed]
57
+
58
+ ## Bias, Risks, and Limitations
59
+
60
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ### Recommendations
65
+
66
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
+
68
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
+
70
+ ## How to Get Started with the Model
71
+
72
+ Use the code below to get started with the model.
73
+
74
+ [More Information Needed]
75
+
76
+ ## Training Details
77
+
78
+ ### Training Data
79
+
80
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
+
82
+ [More Information Needed]
83
+
84
+ ### Training Procedure
85
+
86
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
+
88
+ #### Preprocessing [optional]
89
+
90
+ [More Information Needed]
91
+
92
+
93
+ #### Training Hyperparameters
94
+
95
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
+
97
+ #### Speeds, Sizes, Times [optional]
98
+
99
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
+
101
+ [More Information Needed]
102
+
103
+ ## Evaluation
104
+
105
+ <!-- This section describes the evaluation protocols and provides the results. -->
106
+
107
+ ### Testing Data, Factors & Metrics
108
+
109
+ #### Testing Data
110
+
111
+ <!-- This should link to a Dataset Card if possible. -->
112
+
113
+ [More Information Needed]
114
+
115
+ #### Factors
116
+
117
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Metrics
122
+
123
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
+
125
+ [More Information Needed]
126
+
127
+ ### Results
128
+
129
+ [More Information Needed]
130
+
131
+ #### Summary
132
+
133
+
134
+
135
+ ## Model Examination [optional]
136
+
137
+ <!-- Relevant interpretability work for the model goes here -->
138
+
139
+ [More Information Needed]
140
+
141
+ ## Environmental Impact
142
+
143
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
+
145
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
+
147
+ - **Hardware Type:** [More Information Needed]
148
+ - **Hours used:** [More Information Needed]
149
+ - **Cloud Provider:** [More Information Needed]
150
+ - **Compute Region:** [More Information Needed]
151
+ - **Carbon Emitted:** [More Information Needed]
152
+
153
+ ## Technical Specifications [optional]
154
+
155
+ ### Model Architecture and Objective
156
+
157
+ [More Information Needed]
158
+
159
+ ### Compute Infrastructure
160
+
161
+ [More Information Needed]
162
+
163
+ #### Hardware
164
+
165
+ [More Information Needed]
166
+
167
+ #### Software
168
+
169
+ [More Information Needed]
170
+
171
+ ## Citation [optional]
172
+
173
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
+
175
+ **BibTeX:**
176
+
177
+ [More Information Needed]
178
+
179
+ **APA:**
180
+
181
+ [More Information Needed]
182
+
183
+ ## Glossary [optional]
184
+
185
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
+
187
+ [More Information Needed]
188
+
189
+ ## More Information [optional]
190
+
191
+ [More Information Needed]
192
+
193
+ ## Model Card Authors [optional]
194
+
195
+ [More Information Needed]
196
+
197
+ ## Model Card Contact
198
+
199
+ [More Information Needed]
200
+ ### Framework versions
201
+
202
+ - PEFT 0.14.0
checkpoint-276/adapter_config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "/kaggle/input/mistral-small-24b/transformers/mistral-small-24b-instruct-2501/2",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": null,
8
+ "fan_in_fan_out": false,
9
+ "inference_mode": true,
10
+ "init_lora_weights": true,
11
+ "layer_replication": null,
12
+ "layers_pattern": null,
13
+ "layers_to_transform": null,
14
+ "loftq_config": {},
15
+ "lora_alpha": 32,
16
+ "lora_bias": false,
17
+ "lora_dropout": 0.05,
18
+ "megatron_config": null,
19
+ "megatron_core": "megatron.core",
20
+ "modules_to_save": null,
21
+ "peft_type": "LORA",
22
+ "r": 16,
23
+ "rank_pattern": {},
24
+ "revision": null,
25
+ "target_modules": [
26
+ "down_proj",
27
+ "k_proj",
28
+ "o_proj",
29
+ "gate_proj",
30
+ "q_proj",
31
+ "up_proj",
32
+ "v_proj"
33
+ ],
34
+ "task_type": "CAUSAL_LM",
35
+ "use_dora": false,
36
+ "use_rslora": false
37
+ }
checkpoint-276/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e7698c9a3331300482d43d1554de5b567b30f1fc50ed6a1b6a0ca90d2af1835
3
+ size 184887280
checkpoint-276/global_step276/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d5f2f617eb1285fb7d3ef1844561376fc2e74f28b73e5c6e1da64abb4bd1d48
3
+ size 277235728
checkpoint-276/global_step276/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b562d35a0230e34023c6ddf55326e89c571f7a3351aecea16a0765847be46e38
3
+ size 277235856
checkpoint-276/global_step276/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a6797f05a8e5b8af765cfe99ee79f8112471c80b5b65ab4df38f1d4bdd9382
3
+ size 277235856
checkpoint-276/global_step276/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f33890f7d4012263284b91ac2113e602b8a1c7eaa7aa02772afb1ecf4a798cd3
3
+ size 277235856
checkpoint-276/global_step276/mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6cefdc716065edd3ec1229c3778b4d25f14a6630945bb0c8c8ef9e5a53fac99
3
+ size 538888289
checkpoint-276/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step276
checkpoint-276/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:029d0995b8d1ea1404fb3b1a2494020e4260275a104b6e7458c27ca9eba73cdf
3
+ size 15024
checkpoint-276/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:682455285b5af7313f212ca9282aecdfbfcceaa71632c870a42419d787a9bc97
3
+ size 15024
checkpoint-276/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:924dd636239e24e57178522974018dfb1f0cb840f66b601ae2e660d5586964f8
3
+ size 15024
checkpoint-276/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32496f63cbd70b944707ea8bee6ae4596898a56a17581e3efbed7b9e85a62252
3
+ size 15024
checkpoint-276/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d965a579421847f536cab4e3633249c0a9eecfa860f51f74761cf0c98c33651e
3
+ size 1064
checkpoint-276/special_tokens_map.json ADDED
@@ -0,0 +1,1026 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<unk>",
4
+ "<s>",
5
+ "</s>",
6
+ "[INST]",
7
+ "[/INST]",
8
+ "[AVAILABLE_TOOLS]",
9
+ "[/AVAILABLE_TOOLS]",
10
+ "[TOOL_RESULTS]",
11
+ "[/TOOL_RESULTS]",
12
+ "[TOOL_CALLS]",
13
+ "[IMG]",
14
+ "<pad>",
15
+ "[IMG_BREAK]",
16
+ "[IMG_END]",
17
+ "[PREFIX]",
18
+ "[MIDDLE]",
19
+ "[SUFFIX]",
20
+ "[SYSTEM_PROMPT]",
21
+ "[/SYSTEM_PROMPT]",
22
+ "[TOOL_CONTENT]",
23
+ "<SPECIAL_20>",
24
+ "<SPECIAL_21>",
25
+ "<SPECIAL_22>",
26
+ "<SPECIAL_23>",
27
+ "<SPECIAL_24>",
28
+ "<SPECIAL_25>",
29
+ "<SPECIAL_26>",
30
+ "<SPECIAL_27>",
31
+ "<SPECIAL_28>",
32
+ "<SPECIAL_29>",
33
+ "<SPECIAL_30>",
34
+ "<SPECIAL_31>",
35
+ "<SPECIAL_32>",
36
+ "<SPECIAL_33>",
37
+ "<SPECIAL_34>",
38
+ "<SPECIAL_35>",
39
+ "<SPECIAL_36>",
40
+ "<SPECIAL_37>",
41
+ "<SPECIAL_38>",
42
+ "<SPECIAL_39>",
43
+ "<SPECIAL_40>",
44
+ "<SPECIAL_41>",
45
+ "<SPECIAL_42>",
46
+ "<SPECIAL_43>",
47
+ "<SPECIAL_44>",
48
+ "<SPECIAL_45>",
49
+ "<SPECIAL_46>",
50
+ "<SPECIAL_47>",
51
+ "<SPECIAL_48>",
52
+ "<SPECIAL_49>",
53
+ "<SPECIAL_50>",
54
+ "<SPECIAL_51>",
55
+ "<SPECIAL_52>",
56
+ "<SPECIAL_53>",
57
+ "<SPECIAL_54>",
58
+ "<SPECIAL_55>",
59
+ "<SPECIAL_56>",
60
+ "<SPECIAL_57>",
61
+ "<SPECIAL_58>",
62
+ "<SPECIAL_59>",
63
+ "<SPECIAL_60>",
64
+ "<SPECIAL_61>",
65
+ "<SPECIAL_62>",
66
+ "<SPECIAL_63>",
67
+ "<SPECIAL_64>",
68
+ "<SPECIAL_65>",
69
+ "<SPECIAL_66>",
70
+ "<SPECIAL_67>",
71
+ "<SPECIAL_68>",
72
+ "<SPECIAL_69>",
73
+ "<SPECIAL_70>",
74
+ "<SPECIAL_71>",
75
+ "<SPECIAL_72>",
76
+ "<SPECIAL_73>",
77
+ "<SPECIAL_74>",
78
+ "<SPECIAL_75>",
79
+ "<SPECIAL_76>",
80
+ "<SPECIAL_77>",
81
+ "<SPECIAL_78>",
82
+ "<SPECIAL_79>",
83
+ "<SPECIAL_80>",
84
+ "<SPECIAL_81>",
85
+ "<SPECIAL_82>",
86
+ "<SPECIAL_83>",
87
+ "<SPECIAL_84>",
88
+ "<SPECIAL_85>",
89
+ "<SPECIAL_86>",
90
+ "<SPECIAL_87>",
91
+ "<SPECIAL_88>",
92
+ "<SPECIAL_89>",
93
+ "<SPECIAL_90>",
94
+ "<SPECIAL_91>",
95
+ "<SPECIAL_92>",
96
+ "<SPECIAL_93>",
97
+ "<SPECIAL_94>",
98
+ "<SPECIAL_95>",
99
+ "<SPECIAL_96>",
100
+ "<SPECIAL_97>",
101
+ "<SPECIAL_98>",
102
+ "<SPECIAL_99>",
103
+ "<SPECIAL_100>",
104
+ "<SPECIAL_101>",
105
+ "<SPECIAL_102>",
106
+ "<SPECIAL_103>",
107
+ "<SPECIAL_104>",
108
+ "<SPECIAL_105>",
109
+ "<SPECIAL_106>",
110
+ "<SPECIAL_107>",
111
+ "<SPECIAL_108>",
112
+ "<SPECIAL_109>",
113
+ "<SPECIAL_110>",
114
+ "<SPECIAL_111>",
115
+ "<SPECIAL_112>",
116
+ "<SPECIAL_113>",
117
+ "<SPECIAL_114>",
118
+ "<SPECIAL_115>",
119
+ "<SPECIAL_116>",
120
+ "<SPECIAL_117>",
121
+ "<SPECIAL_118>",
122
+ "<SPECIAL_119>",
123
+ "<SPECIAL_120>",
124
+ "<SPECIAL_121>",
125
+ "<SPECIAL_122>",
126
+ "<SPECIAL_123>",
127
+ "<SPECIAL_124>",
128
+ "<SPECIAL_125>",
129
+ "<SPECIAL_126>",
130
+ "<SPECIAL_127>",
131
+ "<SPECIAL_128>",
132
+ "<SPECIAL_129>",
133
+ "<SPECIAL_130>",
134
+ "<SPECIAL_131>",
135
+ "<SPECIAL_132>",
136
+ "<SPECIAL_133>",
137
+ "<SPECIAL_134>",
138
+ "<SPECIAL_135>",
139
+ "<SPECIAL_136>",
140
+ "<SPECIAL_137>",
141
+ "<SPECIAL_138>",
142
+ "<SPECIAL_139>",
143
+ "<SPECIAL_140>",
144
+ "<SPECIAL_141>",
145
+ "<SPECIAL_142>",
146
+ "<SPECIAL_143>",
147
+ "<SPECIAL_144>",
148
+ "<SPECIAL_145>",
149
+ "<SPECIAL_146>",
150
+ "<SPECIAL_147>",
151
+ "<SPECIAL_148>",
152
+ "<SPECIAL_149>",
153
+ "<SPECIAL_150>",
154
+ "<SPECIAL_151>",
155
+ "<SPECIAL_152>",
156
+ "<SPECIAL_153>",
157
+ "<SPECIAL_154>",
158
+ "<SPECIAL_155>",
159
+ "<SPECIAL_156>",
160
+ "<SPECIAL_157>",
161
+ "<SPECIAL_158>",
162
+ "<SPECIAL_159>",
163
+ "<SPECIAL_160>",
164
+ "<SPECIAL_161>",
165
+ "<SPECIAL_162>",
166
+ "<SPECIAL_163>",
167
+ "<SPECIAL_164>",
168
+ "<SPECIAL_165>",
169
+ "<SPECIAL_166>",
170
+ "<SPECIAL_167>",
171
+ "<SPECIAL_168>",
172
+ "<SPECIAL_169>",
173
+ "<SPECIAL_170>",
174
+ "<SPECIAL_171>",
175
+ "<SPECIAL_172>",
176
+ "<SPECIAL_173>",
177
+ "<SPECIAL_174>",
178
+ "<SPECIAL_175>",
179
+ "<SPECIAL_176>",
180
+ "<SPECIAL_177>",
181
+ "<SPECIAL_178>",
182
+ "<SPECIAL_179>",
183
+ "<SPECIAL_180>",
184
+ "<SPECIAL_181>",
185
+ "<SPECIAL_182>",
186
+ "<SPECIAL_183>",
187
+ "<SPECIAL_184>",
188
+ "<SPECIAL_185>",
189
+ "<SPECIAL_186>",
190
+ "<SPECIAL_187>",
191
+ "<SPECIAL_188>",
192
+ "<SPECIAL_189>",
193
+ "<SPECIAL_190>",
194
+ "<SPECIAL_191>",
195
+ "<SPECIAL_192>",
196
+ "<SPECIAL_193>",
197
+ "<SPECIAL_194>",
198
+ "<SPECIAL_195>",
199
+ "<SPECIAL_196>",
200
+ "<SPECIAL_197>",
201
+ "<SPECIAL_198>",
202
+ "<SPECIAL_199>",
203
+ "<SPECIAL_200>",
204
+ "<SPECIAL_201>",
205
+ "<SPECIAL_202>",
206
+ "<SPECIAL_203>",
207
+ "<SPECIAL_204>",
208
+ "<SPECIAL_205>",
209
+ "<SPECIAL_206>",
210
+ "<SPECIAL_207>",
211
+ "<SPECIAL_208>",
212
+ "<SPECIAL_209>",
213
+ "<SPECIAL_210>",
214
+ "<SPECIAL_211>",
215
+ "<SPECIAL_212>",
216
+ "<SPECIAL_213>",
217
+ "<SPECIAL_214>",
218
+ "<SPECIAL_215>",
219
+ "<SPECIAL_216>",
220
+ "<SPECIAL_217>",
221
+ "<SPECIAL_218>",
222
+ "<SPECIAL_219>",
223
+ "<SPECIAL_220>",
224
+ "<SPECIAL_221>",
225
+ "<SPECIAL_222>",
226
+ "<SPECIAL_223>",
227
+ "<SPECIAL_224>",
228
+ "<SPECIAL_225>",
229
+ "<SPECIAL_226>",
230
+ "<SPECIAL_227>",
231
+ "<SPECIAL_228>",
232
+ "<SPECIAL_229>",
233
+ "<SPECIAL_230>",
234
+ "<SPECIAL_231>",
235
+ "<SPECIAL_232>",
236
+ "<SPECIAL_233>",
237
+ "<SPECIAL_234>",
238
+ "<SPECIAL_235>",
239
+ "<SPECIAL_236>",
240
+ "<SPECIAL_237>",
241
+ "<SPECIAL_238>",
242
+ "<SPECIAL_239>",
243
+ "<SPECIAL_240>",
244
+ "<SPECIAL_241>",
245
+ "<SPECIAL_242>",
246
+ "<SPECIAL_243>",
247
+ "<SPECIAL_244>",
248
+ "<SPECIAL_245>",
249
+ "<SPECIAL_246>",
250
+ "<SPECIAL_247>",
251
+ "<SPECIAL_248>",
252
+ "<SPECIAL_249>",
253
+ "<SPECIAL_250>",
254
+ "<SPECIAL_251>",
255
+ "<SPECIAL_252>",
256
+ "<SPECIAL_253>",
257
+ "<SPECIAL_254>",
258
+ "<SPECIAL_255>",
259
+ "<SPECIAL_256>",
260
+ "<SPECIAL_257>",
261
+ "<SPECIAL_258>",
262
+ "<SPECIAL_259>",
263
+ "<SPECIAL_260>",
264
+ "<SPECIAL_261>",
265
+ "<SPECIAL_262>",
266
+ "<SPECIAL_263>",
267
+ "<SPECIAL_264>",
268
+ "<SPECIAL_265>",
269
+ "<SPECIAL_266>",
270
+ "<SPECIAL_267>",
271
+ "<SPECIAL_268>",
272
+ "<SPECIAL_269>",
273
+ "<SPECIAL_270>",
274
+ "<SPECIAL_271>",
275
+ "<SPECIAL_272>",
276
+ "<SPECIAL_273>",
277
+ "<SPECIAL_274>",
278
+ "<SPECIAL_275>",
279
+ "<SPECIAL_276>",
280
+ "<SPECIAL_277>",
281
+ "<SPECIAL_278>",
282
+ "<SPECIAL_279>",
283
+ "<SPECIAL_280>",
284
+ "<SPECIAL_281>",
285
+ "<SPECIAL_282>",
286
+ "<SPECIAL_283>",
287
+ "<SPECIAL_284>",
288
+ "<SPECIAL_285>",
289
+ "<SPECIAL_286>",
290
+ "<SPECIAL_287>",
291
+ "<SPECIAL_288>",
292
+ "<SPECIAL_289>",
293
+ "<SPECIAL_290>",
294
+ "<SPECIAL_291>",
295
+ "<SPECIAL_292>",
296
+ "<SPECIAL_293>",
297
+ "<SPECIAL_294>",
298
+ "<SPECIAL_295>",
299
+ "<SPECIAL_296>",
300
+ "<SPECIAL_297>",
301
+ "<SPECIAL_298>",
302
+ "<SPECIAL_299>",
303
+ "<SPECIAL_300>",
304
+ "<SPECIAL_301>",
305
+ "<SPECIAL_302>",
306
+ "<SPECIAL_303>",
307
+ "<SPECIAL_304>",
308
+ "<SPECIAL_305>",
309
+ "<SPECIAL_306>",
310
+ "<SPECIAL_307>",
311
+ "<SPECIAL_308>",
312
+ "<SPECIAL_309>",
313
+ "<SPECIAL_310>",
314
+ "<SPECIAL_311>",
315
+ "<SPECIAL_312>",
316
+ "<SPECIAL_313>",
317
+ "<SPECIAL_314>",
318
+ "<SPECIAL_315>",
319
+ "<SPECIAL_316>",
320
+ "<SPECIAL_317>",
321
+ "<SPECIAL_318>",
322
+ "<SPECIAL_319>",
323
+ "<SPECIAL_320>",
324
+ "<SPECIAL_321>",
325
+ "<SPECIAL_322>",
326
+ "<SPECIAL_323>",
327
+ "<SPECIAL_324>",
328
+ "<SPECIAL_325>",
329
+ "<SPECIAL_326>",
330
+ "<SPECIAL_327>",
331
+ "<SPECIAL_328>",
332
+ "<SPECIAL_329>",
333
+ "<SPECIAL_330>",
334
+ "<SPECIAL_331>",
335
+ "<SPECIAL_332>",
336
+ "<SPECIAL_333>",
337
+ "<SPECIAL_334>",
338
+ "<SPECIAL_335>",
339
+ "<SPECIAL_336>",
340
+ "<SPECIAL_337>",
341
+ "<SPECIAL_338>",
342
+ "<SPECIAL_339>",
343
+ "<SPECIAL_340>",
344
+ "<SPECIAL_341>",
345
+ "<SPECIAL_342>",
346
+ "<SPECIAL_343>",
347
+ "<SPECIAL_344>",
348
+ "<SPECIAL_345>",
349
+ "<SPECIAL_346>",
350
+ "<SPECIAL_347>",
351
+ "<SPECIAL_348>",
352
+ "<SPECIAL_349>",
353
+ "<SPECIAL_350>",
354
+ "<SPECIAL_351>",
355
+ "<SPECIAL_352>",
356
+ "<SPECIAL_353>",
357
+ "<SPECIAL_354>",
358
+ "<SPECIAL_355>",
359
+ "<SPECIAL_356>",
360
+ "<SPECIAL_357>",
361
+ "<SPECIAL_358>",
362
+ "<SPECIAL_359>",
363
+ "<SPECIAL_360>",
364
+ "<SPECIAL_361>",
365
+ "<SPECIAL_362>",
366
+ "<SPECIAL_363>",
367
+ "<SPECIAL_364>",
368
+ "<SPECIAL_365>",
369
+ "<SPECIAL_366>",
370
+ "<SPECIAL_367>",
371
+ "<SPECIAL_368>",
372
+ "<SPECIAL_369>",
373
+ "<SPECIAL_370>",
374
+ "<SPECIAL_371>",
375
+ "<SPECIAL_372>",
376
+ "<SPECIAL_373>",
377
+ "<SPECIAL_374>",
378
+ "<SPECIAL_375>",
379
+ "<SPECIAL_376>",
380
+ "<SPECIAL_377>",
381
+ "<SPECIAL_378>",
382
+ "<SPECIAL_379>",
383
+ "<SPECIAL_380>",
384
+ "<SPECIAL_381>",
385
+ "<SPECIAL_382>",
386
+ "<SPECIAL_383>",
387
+ "<SPECIAL_384>",
388
+ "<SPECIAL_385>",
389
+ "<SPECIAL_386>",
390
+ "<SPECIAL_387>",
391
+ "<SPECIAL_388>",
392
+ "<SPECIAL_389>",
393
+ "<SPECIAL_390>",
394
+ "<SPECIAL_391>",
395
+ "<SPECIAL_392>",
396
+ "<SPECIAL_393>",
397
+ "<SPECIAL_394>",
398
+ "<SPECIAL_395>",
399
+ "<SPECIAL_396>",
400
+ "<SPECIAL_397>",
401
+ "<SPECIAL_398>",
402
+ "<SPECIAL_399>",
403
+ "<SPECIAL_400>",
404
+ "<SPECIAL_401>",
405
+ "<SPECIAL_402>",
406
+ "<SPECIAL_403>",
407
+ "<SPECIAL_404>",
408
+ "<SPECIAL_405>",
409
+ "<SPECIAL_406>",
410
+ "<SPECIAL_407>",
411
+ "<SPECIAL_408>",
412
+ "<SPECIAL_409>",
413
+ "<SPECIAL_410>",
414
+ "<SPECIAL_411>",
415
+ "<SPECIAL_412>",
416
+ "<SPECIAL_413>",
417
+ "<SPECIAL_414>",
418
+ "<SPECIAL_415>",
419
+ "<SPECIAL_416>",
420
+ "<SPECIAL_417>",
421
+ "<SPECIAL_418>",
422
+ "<SPECIAL_419>",
423
+ "<SPECIAL_420>",
424
+ "<SPECIAL_421>",
425
+ "<SPECIAL_422>",
426
+ "<SPECIAL_423>",
427
+ "<SPECIAL_424>",
428
+ "<SPECIAL_425>",
429
+ "<SPECIAL_426>",
430
+ "<SPECIAL_427>",
431
+ "<SPECIAL_428>",
432
+ "<SPECIAL_429>",
433
+ "<SPECIAL_430>",
434
+ "<SPECIAL_431>",
435
+ "<SPECIAL_432>",
436
+ "<SPECIAL_433>",
437
+ "<SPECIAL_434>",
438
+ "<SPECIAL_435>",
439
+ "<SPECIAL_436>",
440
+ "<SPECIAL_437>",
441
+ "<SPECIAL_438>",
442
+ "<SPECIAL_439>",
443
+ "<SPECIAL_440>",
444
+ "<SPECIAL_441>",
445
+ "<SPECIAL_442>",
446
+ "<SPECIAL_443>",
447
+ "<SPECIAL_444>",
448
+ "<SPECIAL_445>",
449
+ "<SPECIAL_446>",
450
+ "<SPECIAL_447>",
451
+ "<SPECIAL_448>",
452
+ "<SPECIAL_449>",
453
+ "<SPECIAL_450>",
454
+ "<SPECIAL_451>",
455
+ "<SPECIAL_452>",
456
+ "<SPECIAL_453>",
457
+ "<SPECIAL_454>",
458
+ "<SPECIAL_455>",
459
+ "<SPECIAL_456>",
460
+ "<SPECIAL_457>",
461
+ "<SPECIAL_458>",
462
+ "<SPECIAL_459>",
463
+ "<SPECIAL_460>",
464
+ "<SPECIAL_461>",
465
+ "<SPECIAL_462>",
466
+ "<SPECIAL_463>",
467
+ "<SPECIAL_464>",
468
+ "<SPECIAL_465>",
469
+ "<SPECIAL_466>",
470
+ "<SPECIAL_467>",
471
+ "<SPECIAL_468>",
472
+ "<SPECIAL_469>",
473
+ "<SPECIAL_470>",
474
+ "<SPECIAL_471>",
475
+ "<SPECIAL_472>",
476
+ "<SPECIAL_473>",
477
+ "<SPECIAL_474>",
478
+ "<SPECIAL_475>",
479
+ "<SPECIAL_476>",
480
+ "<SPECIAL_477>",
481
+ "<SPECIAL_478>",
482
+ "<SPECIAL_479>",
483
+ "<SPECIAL_480>",
484
+ "<SPECIAL_481>",
485
+ "<SPECIAL_482>",
486
+ "<SPECIAL_483>",
487
+ "<SPECIAL_484>",
488
+ "<SPECIAL_485>",
489
+ "<SPECIAL_486>",
490
+ "<SPECIAL_487>",
491
+ "<SPECIAL_488>",
492
+ "<SPECIAL_489>",
493
+ "<SPECIAL_490>",
494
+ "<SPECIAL_491>",
495
+ "<SPECIAL_492>",
496
+ "<SPECIAL_493>",
497
+ "<SPECIAL_494>",
498
+ "<SPECIAL_495>",
499
+ "<SPECIAL_496>",
500
+ "<SPECIAL_497>",
501
+ "<SPECIAL_498>",
502
+ "<SPECIAL_499>",
503
+ "<SPECIAL_500>",
504
+ "<SPECIAL_501>",
505
+ "<SPECIAL_502>",
506
+ "<SPECIAL_503>",
507
+ "<SPECIAL_504>",
508
+ "<SPECIAL_505>",
509
+ "<SPECIAL_506>",
510
+ "<SPECIAL_507>",
511
+ "<SPECIAL_508>",
512
+ "<SPECIAL_509>",
513
+ "<SPECIAL_510>",
514
+ "<SPECIAL_511>",
515
+ "<SPECIAL_512>",
516
+ "<SPECIAL_513>",
517
+ "<SPECIAL_514>",
518
+ "<SPECIAL_515>",
519
+ "<SPECIAL_516>",
520
+ "<SPECIAL_517>",
521
+ "<SPECIAL_518>",
522
+ "<SPECIAL_519>",
523
+ "<SPECIAL_520>",
524
+ "<SPECIAL_521>",
525
+ "<SPECIAL_522>",
526
+ "<SPECIAL_523>",
527
+ "<SPECIAL_524>",
528
+ "<SPECIAL_525>",
529
+ "<SPECIAL_526>",
530
+ "<SPECIAL_527>",
531
+ "<SPECIAL_528>",
532
+ "<SPECIAL_529>",
533
+ "<SPECIAL_530>",
534
+ "<SPECIAL_531>",
535
+ "<SPECIAL_532>",
536
+ "<SPECIAL_533>",
537
+ "<SPECIAL_534>",
538
+ "<SPECIAL_535>",
539
+ "<SPECIAL_536>",
540
+ "<SPECIAL_537>",
541
+ "<SPECIAL_538>",
542
+ "<SPECIAL_539>",
543
+ "<SPECIAL_540>",
544
+ "<SPECIAL_541>",
545
+ "<SPECIAL_542>",
546
+ "<SPECIAL_543>",
547
+ "<SPECIAL_544>",
548
+ "<SPECIAL_545>",
549
+ "<SPECIAL_546>",
550
+ "<SPECIAL_547>",
551
+ "<SPECIAL_548>",
552
+ "<SPECIAL_549>",
553
+ "<SPECIAL_550>",
554
+ "<SPECIAL_551>",
555
+ "<SPECIAL_552>",
556
+ "<SPECIAL_553>",
557
+ "<SPECIAL_554>",
558
+ "<SPECIAL_555>",
559
+ "<SPECIAL_556>",
560
+ "<SPECIAL_557>",
561
+ "<SPECIAL_558>",
562
+ "<SPECIAL_559>",
563
+ "<SPECIAL_560>",
564
+ "<SPECIAL_561>",
565
+ "<SPECIAL_562>",
566
+ "<SPECIAL_563>",
567
+ "<SPECIAL_564>",
568
+ "<SPECIAL_565>",
569
+ "<SPECIAL_566>",
570
+ "<SPECIAL_567>",
571
+ "<SPECIAL_568>",
572
+ "<SPECIAL_569>",
573
+ "<SPECIAL_570>",
574
+ "<SPECIAL_571>",
575
+ "<SPECIAL_572>",
576
+ "<SPECIAL_573>",
577
+ "<SPECIAL_574>",
578
+ "<SPECIAL_575>",
579
+ "<SPECIAL_576>",
580
+ "<SPECIAL_577>",
581
+ "<SPECIAL_578>",
582
+ "<SPECIAL_579>",
583
+ "<SPECIAL_580>",
584
+ "<SPECIAL_581>",
585
+ "<SPECIAL_582>",
586
+ "<SPECIAL_583>",
587
+ "<SPECIAL_584>",
588
+ "<SPECIAL_585>",
589
+ "<SPECIAL_586>",
590
+ "<SPECIAL_587>",
591
+ "<SPECIAL_588>",
592
+ "<SPECIAL_589>",
593
+ "<SPECIAL_590>",
594
+ "<SPECIAL_591>",
595
+ "<SPECIAL_592>",
596
+ "<SPECIAL_593>",
597
+ "<SPECIAL_594>",
598
+ "<SPECIAL_595>",
599
+ "<SPECIAL_596>",
600
+ "<SPECIAL_597>",
601
+ "<SPECIAL_598>",
602
+ "<SPECIAL_599>",
603
+ "<SPECIAL_600>",
604
+ "<SPECIAL_601>",
605
+ "<SPECIAL_602>",
606
+ "<SPECIAL_603>",
607
+ "<SPECIAL_604>",
608
+ "<SPECIAL_605>",
609
+ "<SPECIAL_606>",
610
+ "<SPECIAL_607>",
611
+ "<SPECIAL_608>",
612
+ "<SPECIAL_609>",
613
+ "<SPECIAL_610>",
614
+ "<SPECIAL_611>",
615
+ "<SPECIAL_612>",
616
+ "<SPECIAL_613>",
617
+ "<SPECIAL_614>",
618
+ "<SPECIAL_615>",
619
+ "<SPECIAL_616>",
620
+ "<SPECIAL_617>",
621
+ "<SPECIAL_618>",
622
+ "<SPECIAL_619>",
623
+ "<SPECIAL_620>",
624
+ "<SPECIAL_621>",
625
+ "<SPECIAL_622>",
626
+ "<SPECIAL_623>",
627
+ "<SPECIAL_624>",
628
+ "<SPECIAL_625>",
629
+ "<SPECIAL_626>",
630
+ "<SPECIAL_627>",
631
+ "<SPECIAL_628>",
632
+ "<SPECIAL_629>",
633
+ "<SPECIAL_630>",
634
+ "<SPECIAL_631>",
635
+ "<SPECIAL_632>",
636
+ "<SPECIAL_633>",
637
+ "<SPECIAL_634>",
638
+ "<SPECIAL_635>",
639
+ "<SPECIAL_636>",
640
+ "<SPECIAL_637>",
641
+ "<SPECIAL_638>",
642
+ "<SPECIAL_639>",
643
+ "<SPECIAL_640>",
644
+ "<SPECIAL_641>",
645
+ "<SPECIAL_642>",
646
+ "<SPECIAL_643>",
647
+ "<SPECIAL_644>",
648
+ "<SPECIAL_645>",
649
+ "<SPECIAL_646>",
650
+ "<SPECIAL_647>",
651
+ "<SPECIAL_648>",
652
+ "<SPECIAL_649>",
653
+ "<SPECIAL_650>",
654
+ "<SPECIAL_651>",
655
+ "<SPECIAL_652>",
656
+ "<SPECIAL_653>",
657
+ "<SPECIAL_654>",
658
+ "<SPECIAL_655>",
659
+ "<SPECIAL_656>",
660
+ "<SPECIAL_657>",
661
+ "<SPECIAL_658>",
662
+ "<SPECIAL_659>",
663
+ "<SPECIAL_660>",
664
+ "<SPECIAL_661>",
665
+ "<SPECIAL_662>",
666
+ "<SPECIAL_663>",
667
+ "<SPECIAL_664>",
668
+ "<SPECIAL_665>",
669
+ "<SPECIAL_666>",
670
+ "<SPECIAL_667>",
671
+ "<SPECIAL_668>",
672
+ "<SPECIAL_669>",
673
+ "<SPECIAL_670>",
674
+ "<SPECIAL_671>",
675
+ "<SPECIAL_672>",
676
+ "<SPECIAL_673>",
677
+ "<SPECIAL_674>",
678
+ "<SPECIAL_675>",
679
+ "<SPECIAL_676>",
680
+ "<SPECIAL_677>",
681
+ "<SPECIAL_678>",
682
+ "<SPECIAL_679>",
683
+ "<SPECIAL_680>",
684
+ "<SPECIAL_681>",
685
+ "<SPECIAL_682>",
686
+ "<SPECIAL_683>",
687
+ "<SPECIAL_684>",
688
+ "<SPECIAL_685>",
689
+ "<SPECIAL_686>",
690
+ "<SPECIAL_687>",
691
+ "<SPECIAL_688>",
692
+ "<SPECIAL_689>",
693
+ "<SPECIAL_690>",
694
+ "<SPECIAL_691>",
695
+ "<SPECIAL_692>",
696
+ "<SPECIAL_693>",
697
+ "<SPECIAL_694>",
698
+ "<SPECIAL_695>",
699
+ "<SPECIAL_696>",
700
+ "<SPECIAL_697>",
701
+ "<SPECIAL_698>",
702
+ "<SPECIAL_699>",
703
+ "<SPECIAL_700>",
704
+ "<SPECIAL_701>",
705
+ "<SPECIAL_702>",
706
+ "<SPECIAL_703>",
707
+ "<SPECIAL_704>",
708
+ "<SPECIAL_705>",
709
+ "<SPECIAL_706>",
710
+ "<SPECIAL_707>",
711
+ "<SPECIAL_708>",
712
+ "<SPECIAL_709>",
713
+ "<SPECIAL_710>",
714
+ "<SPECIAL_711>",
715
+ "<SPECIAL_712>",
716
+ "<SPECIAL_713>",
717
+ "<SPECIAL_714>",
718
+ "<SPECIAL_715>",
719
+ "<SPECIAL_716>",
720
+ "<SPECIAL_717>",
721
+ "<SPECIAL_718>",
722
+ "<SPECIAL_719>",
723
+ "<SPECIAL_720>",
724
+ "<SPECIAL_721>",
725
+ "<SPECIAL_722>",
726
+ "<SPECIAL_723>",
727
+ "<SPECIAL_724>",
728
+ "<SPECIAL_725>",
729
+ "<SPECIAL_726>",
730
+ "<SPECIAL_727>",
731
+ "<SPECIAL_728>",
732
+ "<SPECIAL_729>",
733
+ "<SPECIAL_730>",
734
+ "<SPECIAL_731>",
735
+ "<SPECIAL_732>",
736
+ "<SPECIAL_733>",
737
+ "<SPECIAL_734>",
738
+ "<SPECIAL_735>",
739
+ "<SPECIAL_736>",
740
+ "<SPECIAL_737>",
741
+ "<SPECIAL_738>",
742
+ "<SPECIAL_739>",
743
+ "<SPECIAL_740>",
744
+ "<SPECIAL_741>",
745
+ "<SPECIAL_742>",
746
+ "<SPECIAL_743>",
747
+ "<SPECIAL_744>",
748
+ "<SPECIAL_745>",
749
+ "<SPECIAL_746>",
750
+ "<SPECIAL_747>",
751
+ "<SPECIAL_748>",
752
+ "<SPECIAL_749>",
753
+ "<SPECIAL_750>",
754
+ "<SPECIAL_751>",
755
+ "<SPECIAL_752>",
756
+ "<SPECIAL_753>",
757
+ "<SPECIAL_754>",
758
+ "<SPECIAL_755>",
759
+ "<SPECIAL_756>",
760
+ "<SPECIAL_757>",
761
+ "<SPECIAL_758>",
762
+ "<SPECIAL_759>",
763
+ "<SPECIAL_760>",
764
+ "<SPECIAL_761>",
765
+ "<SPECIAL_762>",
766
+ "<SPECIAL_763>",
767
+ "<SPECIAL_764>",
768
+ "<SPECIAL_765>",
769
+ "<SPECIAL_766>",
770
+ "<SPECIAL_767>",
771
+ "<SPECIAL_768>",
772
+ "<SPECIAL_769>",
773
+ "<SPECIAL_770>",
774
+ "<SPECIAL_771>",
775
+ "<SPECIAL_772>",
776
+ "<SPECIAL_773>",
777
+ "<SPECIAL_774>",
778
+ "<SPECIAL_775>",
779
+ "<SPECIAL_776>",
780
+ "<SPECIAL_777>",
781
+ "<SPECIAL_778>",
782
+ "<SPECIAL_779>",
783
+ "<SPECIAL_780>",
784
+ "<SPECIAL_781>",
785
+ "<SPECIAL_782>",
786
+ "<SPECIAL_783>",
787
+ "<SPECIAL_784>",
788
+ "<SPECIAL_785>",
789
+ "<SPECIAL_786>",
790
+ "<SPECIAL_787>",
791
+ "<SPECIAL_788>",
792
+ "<SPECIAL_789>",
793
+ "<SPECIAL_790>",
794
+ "<SPECIAL_791>",
795
+ "<SPECIAL_792>",
796
+ "<SPECIAL_793>",
797
+ "<SPECIAL_794>",
798
+ "<SPECIAL_795>",
799
+ "<SPECIAL_796>",
800
+ "<SPECIAL_797>",
801
+ "<SPECIAL_798>",
802
+ "<SPECIAL_799>",
803
+ "<SPECIAL_800>",
804
+ "<SPECIAL_801>",
805
+ "<SPECIAL_802>",
806
+ "<SPECIAL_803>",
807
+ "<SPECIAL_804>",
808
+ "<SPECIAL_805>",
809
+ "<SPECIAL_806>",
810
+ "<SPECIAL_807>",
811
+ "<SPECIAL_808>",
812
+ "<SPECIAL_809>",
813
+ "<SPECIAL_810>",
814
+ "<SPECIAL_811>",
815
+ "<SPECIAL_812>",
816
+ "<SPECIAL_813>",
817
+ "<SPECIAL_814>",
818
+ "<SPECIAL_815>",
819
+ "<SPECIAL_816>",
820
+ "<SPECIAL_817>",
821
+ "<SPECIAL_818>",
822
+ "<SPECIAL_819>",
823
+ "<SPECIAL_820>",
824
+ "<SPECIAL_821>",
825
+ "<SPECIAL_822>",
826
+ "<SPECIAL_823>",
827
+ "<SPECIAL_824>",
828
+ "<SPECIAL_825>",
829
+ "<SPECIAL_826>",
830
+ "<SPECIAL_827>",
831
+ "<SPECIAL_828>",
832
+ "<SPECIAL_829>",
833
+ "<SPECIAL_830>",
834
+ "<SPECIAL_831>",
835
+ "<SPECIAL_832>",
836
+ "<SPECIAL_833>",
837
+ "<SPECIAL_834>",
838
+ "<SPECIAL_835>",
839
+ "<SPECIAL_836>",
840
+ "<SPECIAL_837>",
841
+ "<SPECIAL_838>",
842
+ "<SPECIAL_839>",
843
+ "<SPECIAL_840>",
844
+ "<SPECIAL_841>",
845
+ "<SPECIAL_842>",
846
+ "<SPECIAL_843>",
847
+ "<SPECIAL_844>",
848
+ "<SPECIAL_845>",
849
+ "<SPECIAL_846>",
850
+ "<SPECIAL_847>",
851
+ "<SPECIAL_848>",
852
+ "<SPECIAL_849>",
853
+ "<SPECIAL_850>",
854
+ "<SPECIAL_851>",
855
+ "<SPECIAL_852>",
856
+ "<SPECIAL_853>",
857
+ "<SPECIAL_854>",
858
+ "<SPECIAL_855>",
859
+ "<SPECIAL_856>",
860
+ "<SPECIAL_857>",
861
+ "<SPECIAL_858>",
862
+ "<SPECIAL_859>",
863
+ "<SPECIAL_860>",
864
+ "<SPECIAL_861>",
865
+ "<SPECIAL_862>",
866
+ "<SPECIAL_863>",
867
+ "<SPECIAL_864>",
868
+ "<SPECIAL_865>",
869
+ "<SPECIAL_866>",
870
+ "<SPECIAL_867>",
871
+ "<SPECIAL_868>",
872
+ "<SPECIAL_869>",
873
+ "<SPECIAL_870>",
874
+ "<SPECIAL_871>",
875
+ "<SPECIAL_872>",
876
+ "<SPECIAL_873>",
877
+ "<SPECIAL_874>",
878
+ "<SPECIAL_875>",
879
+ "<SPECIAL_876>",
880
+ "<SPECIAL_877>",
881
+ "<SPECIAL_878>",
882
+ "<SPECIAL_879>",
883
+ "<SPECIAL_880>",
884
+ "<SPECIAL_881>",
885
+ "<SPECIAL_882>",
886
+ "<SPECIAL_883>",
887
+ "<SPECIAL_884>",
888
+ "<SPECIAL_885>",
889
+ "<SPECIAL_886>",
890
+ "<SPECIAL_887>",
891
+ "<SPECIAL_888>",
892
+ "<SPECIAL_889>",
893
+ "<SPECIAL_890>",
894
+ "<SPECIAL_891>",
895
+ "<SPECIAL_892>",
896
+ "<SPECIAL_893>",
897
+ "<SPECIAL_894>",
898
+ "<SPECIAL_895>",
899
+ "<SPECIAL_896>",
900
+ "<SPECIAL_897>",
901
+ "<SPECIAL_898>",
902
+ "<SPECIAL_899>",
903
+ "<SPECIAL_900>",
904
+ "<SPECIAL_901>",
905
+ "<SPECIAL_902>",
906
+ "<SPECIAL_903>",
907
+ "<SPECIAL_904>",
908
+ "<SPECIAL_905>",
909
+ "<SPECIAL_906>",
910
+ "<SPECIAL_907>",
911
+ "<SPECIAL_908>",
912
+ "<SPECIAL_909>",
913
+ "<SPECIAL_910>",
914
+ "<SPECIAL_911>",
915
+ "<SPECIAL_912>",
916
+ "<SPECIAL_913>",
917
+ "<SPECIAL_914>",
918
+ "<SPECIAL_915>",
919
+ "<SPECIAL_916>",
920
+ "<SPECIAL_917>",
921
+ "<SPECIAL_918>",
922
+ "<SPECIAL_919>",
923
+ "<SPECIAL_920>",
924
+ "<SPECIAL_921>",
925
+ "<SPECIAL_922>",
926
+ "<SPECIAL_923>",
927
+ "<SPECIAL_924>",
928
+ "<SPECIAL_925>",
929
+ "<SPECIAL_926>",
930
+ "<SPECIAL_927>",
931
+ "<SPECIAL_928>",
932
+ "<SPECIAL_929>",
933
+ "<SPECIAL_930>",
934
+ "<SPECIAL_931>",
935
+ "<SPECIAL_932>",
936
+ "<SPECIAL_933>",
937
+ "<SPECIAL_934>",
938
+ "<SPECIAL_935>",
939
+ "<SPECIAL_936>",
940
+ "<SPECIAL_937>",
941
+ "<SPECIAL_938>",
942
+ "<SPECIAL_939>",
943
+ "<SPECIAL_940>",
944
+ "<SPECIAL_941>",
945
+ "<SPECIAL_942>",
946
+ "<SPECIAL_943>",
947
+ "<SPECIAL_944>",
948
+ "<SPECIAL_945>",
949
+ "<SPECIAL_946>",
950
+ "<SPECIAL_947>",
951
+ "<SPECIAL_948>",
952
+ "<SPECIAL_949>",
953
+ "<SPECIAL_950>",
954
+ "<SPECIAL_951>",
955
+ "<SPECIAL_952>",
956
+ "<SPECIAL_953>",
957
+ "<SPECIAL_954>",
958
+ "<SPECIAL_955>",
959
+ "<SPECIAL_956>",
960
+ "<SPECIAL_957>",
961
+ "<SPECIAL_958>",
962
+ "<SPECIAL_959>",
963
+ "<SPECIAL_960>",
964
+ "<SPECIAL_961>",
965
+ "<SPECIAL_962>",
966
+ "<SPECIAL_963>",
967
+ "<SPECIAL_964>",
968
+ "<SPECIAL_965>",
969
+ "<SPECIAL_966>",
970
+ "<SPECIAL_967>",
971
+ "<SPECIAL_968>",
972
+ "<SPECIAL_969>",
973
+ "<SPECIAL_970>",
974
+ "<SPECIAL_971>",
975
+ "<SPECIAL_972>",
976
+ "<SPECIAL_973>",
977
+ "<SPECIAL_974>",
978
+ "<SPECIAL_975>",
979
+ "<SPECIAL_976>",
980
+ "<SPECIAL_977>",
981
+ "<SPECIAL_978>",
982
+ "<SPECIAL_979>",
983
+ "<SPECIAL_980>",
984
+ "<SPECIAL_981>",
985
+ "<SPECIAL_982>",
986
+ "<SPECIAL_983>",
987
+ "<SPECIAL_984>",
988
+ "<SPECIAL_985>",
989
+ "<SPECIAL_986>",
990
+ "<SPECIAL_987>",
991
+ "<SPECIAL_988>",
992
+ "<SPECIAL_989>",
993
+ "<SPECIAL_990>",
994
+ "<SPECIAL_991>",
995
+ "<SPECIAL_992>",
996
+ "<SPECIAL_993>",
997
+ "<SPECIAL_994>",
998
+ "<SPECIAL_995>",
999
+ "<SPECIAL_996>",
1000
+ "<SPECIAL_997>",
1001
+ "<SPECIAL_998>",
1002
+ "<SPECIAL_999>"
1003
+ ],
1004
+ "bos_token": {
1005
+ "content": "<s>",
1006
+ "lstrip": false,
1007
+ "normalized": false,
1008
+ "rstrip": false,
1009
+ "single_word": false
1010
+ },
1011
+ "eos_token": {
1012
+ "content": "</s>",
1013
+ "lstrip": false,
1014
+ "normalized": false,
1015
+ "rstrip": false,
1016
+ "single_word": false
1017
+ },
1018
+ "pad_token": "</s>",
1019
+ "unk_token": {
1020
+ "content": "<unk>",
1021
+ "lstrip": false,
1022
+ "normalized": false,
1023
+ "rstrip": false,
1024
+ "single_word": false
1025
+ }
1026
+ }
checkpoint-276/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b76085f9923309d873994d444989f7eb6ec074b06f25b58f1e8d7b7741070949
3
+ size 17078037
checkpoint-276/tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-276/trainer_state.json ADDED
@@ -0,0 +1,1973 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.997289972899729,
5
+ "eval_steps": 500,
6
+ "global_step": 276,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0036133694670280035,
13
+ "grad_norm": 0.3959366977214813,
14
+ "learning_rate": 6.25e-06,
15
+ "loss": 0.9323,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.007226738934056007,
20
+ "grad_norm": 0.45551198720932007,
21
+ "learning_rate": 1.25e-05,
22
+ "loss": 1.0507,
23
+ "step": 2
24
+ },
25
+ {
26
+ "epoch": 0.01084010840108401,
27
+ "grad_norm": 0.2823091745376587,
28
+ "learning_rate": 1.8750000000000002e-05,
29
+ "loss": 0.8491,
30
+ "step": 3
31
+ },
32
+ {
33
+ "epoch": 0.014453477868112014,
34
+ "grad_norm": 0.46047303080558777,
35
+ "learning_rate": 2.5e-05,
36
+ "loss": 1.0142,
37
+ "step": 4
38
+ },
39
+ {
40
+ "epoch": 0.018066847335140017,
41
+ "grad_norm": 0.4086349606513977,
42
+ "learning_rate": 3.125e-05,
43
+ "loss": 0.947,
44
+ "step": 5
45
+ },
46
+ {
47
+ "epoch": 0.02168021680216802,
48
+ "grad_norm": 0.457003116607666,
49
+ "learning_rate": 3.7500000000000003e-05,
50
+ "loss": 0.9485,
51
+ "step": 6
52
+ },
53
+ {
54
+ "epoch": 0.025293586269196026,
55
+ "grad_norm": 0.35562458634376526,
56
+ "learning_rate": 4.375e-05,
57
+ "loss": 0.8449,
58
+ "step": 7
59
+ },
60
+ {
61
+ "epoch": 0.028906955736224028,
62
+ "grad_norm": 0.33805516362190247,
63
+ "learning_rate": 5e-05,
64
+ "loss": 0.7379,
65
+ "step": 8
66
+ },
67
+ {
68
+ "epoch": 0.032520325203252036,
69
+ "grad_norm": 0.3412623703479767,
70
+ "learning_rate": 4.9998282347929784e-05,
71
+ "loss": 0.6282,
72
+ "step": 9
73
+ },
74
+ {
75
+ "epoch": 0.036133694670280034,
76
+ "grad_norm": 0.2843680679798126,
77
+ "learning_rate": 4.99931296277454e-05,
78
+ "loss": 0.5503,
79
+ "step": 10
80
+ },
81
+ {
82
+ "epoch": 0.03974706413730804,
83
+ "grad_norm": 0.17628777027130127,
84
+ "learning_rate": 4.998454254749331e-05,
85
+ "loss": 0.512,
86
+ "step": 11
87
+ },
88
+ {
89
+ "epoch": 0.04336043360433604,
90
+ "grad_norm": 0.19055013358592987,
91
+ "learning_rate": 4.997252228714279e-05,
92
+ "loss": 0.5397,
93
+ "step": 12
94
+ },
95
+ {
96
+ "epoch": 0.04697380307136405,
97
+ "grad_norm": 0.08906977623701096,
98
+ "learning_rate": 4.9957070498423854e-05,
99
+ "loss": 0.5458,
100
+ "step": 13
101
+ },
102
+ {
103
+ "epoch": 0.05058717253839205,
104
+ "grad_norm": 0.0917251780629158,
105
+ "learning_rate": 4.993818930460026e-05,
106
+ "loss": 0.5269,
107
+ "step": 14
108
+ },
109
+ {
110
+ "epoch": 0.05420054200542006,
111
+ "grad_norm": 0.0985497236251831,
112
+ "learning_rate": 4.9915881300177725e-05,
113
+ "loss": 0.4135,
114
+ "step": 15
115
+ },
116
+ {
117
+ "epoch": 0.057813911472448055,
118
+ "grad_norm": 0.1111132949590683,
119
+ "learning_rate": 4.9890149550547454e-05,
120
+ "loss": 0.5064,
121
+ "step": 16
122
+ },
123
+ {
124
+ "epoch": 0.06142728093947606,
125
+ "grad_norm": 0.0649256557226181,
126
+ "learning_rate": 4.98609975915649e-05,
127
+ "loss": 0.4804,
128
+ "step": 17
129
+ },
130
+ {
131
+ "epoch": 0.06504065040650407,
132
+ "grad_norm": 0.09687516838312149,
133
+ "learning_rate": 4.982842942906386e-05,
134
+ "loss": 0.3706,
135
+ "step": 18
136
+ },
137
+ {
138
+ "epoch": 0.06865401987353206,
139
+ "grad_norm": 0.14679567515850067,
140
+ "learning_rate": 4.979244953830608e-05,
141
+ "loss": 0.4105,
142
+ "step": 19
143
+ },
144
+ {
145
+ "epoch": 0.07226738934056007,
146
+ "grad_norm": 0.14155593514442444,
147
+ "learning_rate": 4.9753062863366276e-05,
148
+ "loss": 0.4886,
149
+ "step": 20
150
+ },
151
+ {
152
+ "epoch": 0.07588075880758807,
153
+ "grad_norm": 0.14684930443763733,
154
+ "learning_rate": 4.971027481645274e-05,
155
+ "loss": 0.4044,
156
+ "step": 21
157
+ },
158
+ {
159
+ "epoch": 0.07949412827461608,
160
+ "grad_norm": 0.11222010105848312,
161
+ "learning_rate": 4.966409127716367e-05,
162
+ "loss": 0.4361,
163
+ "step": 22
164
+ },
165
+ {
166
+ "epoch": 0.08310749774164408,
167
+ "grad_norm": 0.058118775486946106,
168
+ "learning_rate": 4.96145185916792e-05,
169
+ "loss": 0.4176,
170
+ "step": 23
171
+ },
172
+ {
173
+ "epoch": 0.08672086720867209,
174
+ "grad_norm": 0.06764644384384155,
175
+ "learning_rate": 4.95615635718894e-05,
176
+ "loss": 0.4683,
177
+ "step": 24
178
+ },
179
+ {
180
+ "epoch": 0.09033423667570009,
181
+ "grad_norm": 0.06886276602745056,
182
+ "learning_rate": 4.950523349445824e-05,
183
+ "loss": 0.418,
184
+ "step": 25
185
+ },
186
+ {
187
+ "epoch": 0.0939476061427281,
188
+ "grad_norm": 0.0706636980175972,
189
+ "learning_rate": 4.944553609982363e-05,
190
+ "loss": 0.3967,
191
+ "step": 26
192
+ },
193
+ {
194
+ "epoch": 0.0975609756097561,
195
+ "grad_norm": 0.04914792627096176,
196
+ "learning_rate": 4.938247959113386e-05,
197
+ "loss": 0.4623,
198
+ "step": 27
199
+ },
200
+ {
201
+ "epoch": 0.1011743450767841,
202
+ "grad_norm": 0.05717244744300842,
203
+ "learning_rate": 4.931607263312032e-05,
204
+ "loss": 0.4047,
205
+ "step": 28
206
+ },
207
+ {
208
+ "epoch": 0.10478771454381211,
209
+ "grad_norm": 0.05677526444196701,
210
+ "learning_rate": 4.924632435090696e-05,
211
+ "loss": 0.4251,
212
+ "step": 29
213
+ },
214
+ {
215
+ "epoch": 0.10840108401084012,
216
+ "grad_norm": 0.051282044500112534,
217
+ "learning_rate": 4.917324432875627e-05,
218
+ "loss": 0.4101,
219
+ "step": 30
220
+ },
221
+ {
222
+ "epoch": 0.1120144534778681,
223
+ "grad_norm": 0.05558260530233383,
224
+ "learning_rate": 4.909684260875235e-05,
225
+ "loss": 0.4425,
226
+ "step": 31
227
+ },
228
+ {
229
+ "epoch": 0.11562782294489611,
230
+ "grad_norm": 0.05362090840935707,
231
+ "learning_rate": 4.9017129689421e-05,
232
+ "loss": 0.383,
233
+ "step": 32
234
+ },
235
+ {
236
+ "epoch": 0.11924119241192412,
237
+ "grad_norm": 0.050591859966516495,
238
+ "learning_rate": 4.893411652428712e-05,
239
+ "loss": 0.3988,
240
+ "step": 33
241
+ },
242
+ {
243
+ "epoch": 0.12285456187895212,
244
+ "grad_norm": 0.07354583591222763,
245
+ "learning_rate": 4.8847814520369475e-05,
246
+ "loss": 0.473,
247
+ "step": 34
248
+ },
249
+ {
250
+ "epoch": 0.12646793134598014,
251
+ "grad_norm": 0.07448670268058777,
252
+ "learning_rate": 4.875823553661334e-05,
253
+ "loss": 0.3609,
254
+ "step": 35
255
+ },
256
+ {
257
+ "epoch": 0.13008130081300814,
258
+ "grad_norm": 0.09399361908435822,
259
+ "learning_rate": 4.8665391882260856e-05,
260
+ "loss": 0.3927,
261
+ "step": 36
262
+ },
263
+ {
264
+ "epoch": 0.13369467028003612,
265
+ "grad_norm": 0.061091382056474686,
266
+ "learning_rate": 4.856929631515964e-05,
267
+ "loss": 0.4512,
268
+ "step": 37
269
+ },
270
+ {
271
+ "epoch": 0.13730803974706413,
272
+ "grad_norm": 0.06277038156986237,
273
+ "learning_rate": 4.846996204000967e-05,
274
+ "loss": 0.3961,
275
+ "step": 38
276
+ },
277
+ {
278
+ "epoch": 0.14092140921409213,
279
+ "grad_norm": 0.05277445912361145,
280
+ "learning_rate": 4.8367402706548805e-05,
281
+ "loss": 0.3885,
282
+ "step": 39
283
+ },
284
+ {
285
+ "epoch": 0.14453477868112014,
286
+ "grad_norm": 0.06335710734128952,
287
+ "learning_rate": 4.8261632407677174e-05,
288
+ "loss": 0.4663,
289
+ "step": 40
290
+ },
291
+ {
292
+ "epoch": 0.14814814814814814,
293
+ "grad_norm": 0.05149435997009277,
294
+ "learning_rate": 4.815266567752059e-05,
295
+ "loss": 0.4012,
296
+ "step": 41
297
+ },
298
+ {
299
+ "epoch": 0.15176151761517614,
300
+ "grad_norm": 0.052154790610075,
301
+ "learning_rate": 4.804051748943343e-05,
302
+ "loss": 0.377,
303
+ "step": 42
304
+ },
305
+ {
306
+ "epoch": 0.15537488708220415,
307
+ "grad_norm": 0.06229854002594948,
308
+ "learning_rate": 4.792520325394111e-05,
309
+ "loss": 0.4677,
310
+ "step": 43
311
+ },
312
+ {
313
+ "epoch": 0.15898825654923215,
314
+ "grad_norm": 0.050992563366889954,
315
+ "learning_rate": 4.780673881662242e-05,
316
+ "loss": 0.4271,
317
+ "step": 44
318
+ },
319
+ {
320
+ "epoch": 0.16260162601626016,
321
+ "grad_norm": 0.057579364627599716,
322
+ "learning_rate": 4.7685140455932267e-05,
323
+ "loss": 0.4096,
324
+ "step": 45
325
+ },
326
+ {
327
+ "epoch": 0.16621499548328816,
328
+ "grad_norm": 0.05966678634285927,
329
+ "learning_rate": 4.756042488096471e-05,
330
+ "loss": 0.4075,
331
+ "step": 46
332
+ },
333
+ {
334
+ "epoch": 0.16982836495031617,
335
+ "grad_norm": 0.055218473076820374,
336
+ "learning_rate": 4.743260922915701e-05,
337
+ "loss": 0.459,
338
+ "step": 47
339
+ },
340
+ {
341
+ "epoch": 0.17344173441734417,
342
+ "grad_norm": 0.05127694830298424,
343
+ "learning_rate": 4.730171106393466e-05,
344
+ "loss": 0.4086,
345
+ "step": 48
346
+ },
347
+ {
348
+ "epoch": 0.17705510388437218,
349
+ "grad_norm": 0.06519781798124313,
350
+ "learning_rate": 4.716774837229804e-05,
351
+ "loss": 0.4418,
352
+ "step": 49
353
+ },
354
+ {
355
+ "epoch": 0.18066847335140018,
356
+ "grad_norm": 0.05895975977182388,
357
+ "learning_rate": 4.7030739562350713e-05,
358
+ "loss": 0.4013,
359
+ "step": 50
360
+ },
361
+ {
362
+ "epoch": 0.1842818428184282,
363
+ "grad_norm": 0.061492159962654114,
364
+ "learning_rate": 4.6890703460769955e-05,
365
+ "loss": 0.3726,
366
+ "step": 51
367
+ },
368
+ {
369
+ "epoch": 0.1878952122854562,
370
+ "grad_norm": 0.05051853135228157,
371
+ "learning_rate": 4.674765931021976e-05,
372
+ "loss": 0.4354,
373
+ "step": 52
374
+ },
375
+ {
376
+ "epoch": 0.1915085817524842,
377
+ "grad_norm": 0.05664265528321266,
378
+ "learning_rate": 4.6601626766706626e-05,
379
+ "loss": 0.4137,
380
+ "step": 53
381
+ },
382
+ {
383
+ "epoch": 0.1951219512195122,
384
+ "grad_norm": 0.06020362302660942,
385
+ "learning_rate": 4.645262589687861e-05,
386
+ "loss": 0.4171,
387
+ "step": 54
388
+ },
389
+ {
390
+ "epoch": 0.1987353206865402,
391
+ "grad_norm": 0.06303560733795166,
392
+ "learning_rate": 4.6300677175267914e-05,
393
+ "loss": 0.3724,
394
+ "step": 55
395
+ },
396
+ {
397
+ "epoch": 0.2023486901535682,
398
+ "grad_norm": 0.06793845444917679,
399
+ "learning_rate": 4.614580148147744e-05,
400
+ "loss": 0.3711,
401
+ "step": 56
402
+ },
403
+ {
404
+ "epoch": 0.20596205962059622,
405
+ "grad_norm": 0.07107391953468323,
406
+ "learning_rate": 4.598802009731167e-05,
407
+ "loss": 0.4428,
408
+ "step": 57
409
+ },
410
+ {
411
+ "epoch": 0.20957542908762422,
412
+ "grad_norm": 0.06567548215389252,
413
+ "learning_rate": 4.582735470385229e-05,
414
+ "loss": 0.3774,
415
+ "step": 58
416
+ },
417
+ {
418
+ "epoch": 0.21318879855465223,
419
+ "grad_norm": 0.05056913569569588,
420
+ "learning_rate": 4.5663827378478975e-05,
421
+ "loss": 0.3584,
422
+ "step": 59
423
+ },
424
+ {
425
+ "epoch": 0.21680216802168023,
426
+ "grad_norm": 0.08128344267606735,
427
+ "learning_rate": 4.5497460591835615e-05,
428
+ "loss": 0.3983,
429
+ "step": 60
430
+ },
431
+ {
432
+ "epoch": 0.2204155374887082,
433
+ "grad_norm": 0.05856931954622269,
434
+ "learning_rate": 4.532827720474268e-05,
435
+ "loss": 0.3486,
436
+ "step": 61
437
+ },
438
+ {
439
+ "epoch": 0.2240289069557362,
440
+ "grad_norm": 0.05503028631210327,
441
+ "learning_rate": 4.515630046505575e-05,
442
+ "loss": 0.3896,
443
+ "step": 62
444
+ },
445
+ {
446
+ "epoch": 0.22764227642276422,
447
+ "grad_norm": 0.047534190118312836,
448
+ "learning_rate": 4.498155400447107e-05,
449
+ "loss": 0.4463,
450
+ "step": 63
451
+ },
452
+ {
453
+ "epoch": 0.23125564588979222,
454
+ "grad_norm": 0.0638430267572403,
455
+ "learning_rate": 4.480406183527823e-05,
456
+ "loss": 0.3977,
457
+ "step": 64
458
+ },
459
+ {
460
+ "epoch": 0.23486901535682023,
461
+ "grad_norm": 0.04974055290222168,
462
+ "learning_rate": 4.462384834706058e-05,
463
+ "loss": 0.3999,
464
+ "step": 65
465
+ },
466
+ {
467
+ "epoch": 0.23848238482384823,
468
+ "grad_norm": 0.06309591233730316,
469
+ "learning_rate": 4.4440938303343804e-05,
470
+ "loss": 0.4275,
471
+ "step": 66
472
+ },
473
+ {
474
+ "epoch": 0.24209575429087624,
475
+ "grad_norm": 0.05192544683814049,
476
+ "learning_rate": 4.425535683819312e-05,
477
+ "loss": 0.4096,
478
+ "step": 67
479
+ },
480
+ {
481
+ "epoch": 0.24570912375790424,
482
+ "grad_norm": 0.057684604078531265,
483
+ "learning_rate": 4.406712945275955e-05,
484
+ "loss": 0.41,
485
+ "step": 68
486
+ },
487
+ {
488
+ "epoch": 0.24932249322493225,
489
+ "grad_norm": 0.0514802448451519,
490
+ "learning_rate": 4.387628201177577e-05,
491
+ "loss": 0.3372,
492
+ "step": 69
493
+ },
494
+ {
495
+ "epoch": 0.2529358626919603,
496
+ "grad_norm": 0.056559968739748,
497
+ "learning_rate": 4.368284074000193e-05,
498
+ "loss": 0.3929,
499
+ "step": 70
500
+ },
501
+ {
502
+ "epoch": 0.2565492321589883,
503
+ "grad_norm": 0.0645717978477478,
504
+ "learning_rate": 4.348683221862212e-05,
505
+ "loss": 0.4353,
506
+ "step": 71
507
+ },
508
+ {
509
+ "epoch": 0.2601626016260163,
510
+ "grad_norm": 0.08638172596693039,
511
+ "learning_rate": 4.328828338159173e-05,
512
+ "loss": 0.3978,
513
+ "step": 72
514
+ },
515
+ {
516
+ "epoch": 0.26377597109304424,
517
+ "grad_norm": 0.05915065109729767,
518
+ "learning_rate": 4.3087221511936434e-05,
519
+ "loss": 0.393,
520
+ "step": 73
521
+ },
522
+ {
523
+ "epoch": 0.26738934056007224,
524
+ "grad_norm": 0.061671093106269836,
525
+ "learning_rate": 4.288367423800319e-05,
526
+ "loss": 0.4187,
527
+ "step": 74
528
+ },
529
+ {
530
+ "epoch": 0.27100271002710025,
531
+ "grad_norm": 0.07420554012060165,
532
+ "learning_rate": 4.267766952966369e-05,
533
+ "loss": 0.3939,
534
+ "step": 75
535
+ },
536
+ {
537
+ "epoch": 0.27461607949412825,
538
+ "grad_norm": 0.07052630186080933,
539
+ "learning_rate": 4.2469235694471043e-05,
540
+ "loss": 0.3435,
541
+ "step": 76
542
+ },
543
+ {
544
+ "epoch": 0.27822944896115626,
545
+ "grad_norm": 0.06885933130979538,
546
+ "learning_rate": 4.225840137376993e-05,
547
+ "loss": 0.4363,
548
+ "step": 77
549
+ },
550
+ {
551
+ "epoch": 0.28184281842818426,
552
+ "grad_norm": 0.05735473707318306,
553
+ "learning_rate": 4.204519553876095e-05,
554
+ "loss": 0.3509,
555
+ "step": 78
556
+ },
557
+ {
558
+ "epoch": 0.28545618789521227,
559
+ "grad_norm": 0.06102309376001358,
560
+ "learning_rate": 4.1829647486519596e-05,
561
+ "loss": 0.3369,
562
+ "step": 79
563
+ },
564
+ {
565
+ "epoch": 0.28906955736224027,
566
+ "grad_norm": 0.06527422368526459,
567
+ "learning_rate": 4.161178683597054e-05,
568
+ "loss": 0.4052,
569
+ "step": 80
570
+ },
571
+ {
572
+ "epoch": 0.2926829268292683,
573
+ "grad_norm": 0.06578138470649719,
574
+ "learning_rate": 4.139164352381758e-05,
575
+ "loss": 0.3586,
576
+ "step": 81
577
+ },
578
+ {
579
+ "epoch": 0.2962962962962963,
580
+ "grad_norm": 0.05465536564588547,
581
+ "learning_rate": 4.116924780042997e-05,
582
+ "loss": 0.3759,
583
+ "step": 82
584
+ },
585
+ {
586
+ "epoch": 0.2999096657633243,
587
+ "grad_norm": 0.08491545915603638,
588
+ "learning_rate": 4.094463022568569e-05,
589
+ "loss": 0.3611,
590
+ "step": 83
591
+ },
592
+ {
593
+ "epoch": 0.3035230352303523,
594
+ "grad_norm": 0.06035340949892998,
595
+ "learning_rate": 4.071782166477213e-05,
596
+ "loss": 0.3537,
597
+ "step": 84
598
+ },
599
+ {
600
+ "epoch": 0.3071364046973803,
601
+ "grad_norm": 0.06220124289393425,
602
+ "learning_rate": 4.0488853283944806e-05,
603
+ "loss": 0.3878,
604
+ "step": 85
605
+ },
606
+ {
607
+ "epoch": 0.3107497741644083,
608
+ "grad_norm": 0.05434149503707886,
609
+ "learning_rate": 4.0257756546244804e-05,
610
+ "loss": 0.3765,
611
+ "step": 86
612
+ },
613
+ {
614
+ "epoch": 0.3143631436314363,
615
+ "grad_norm": 0.06244641914963722,
616
+ "learning_rate": 4.0024563207175316e-05,
617
+ "loss": 0.3668,
618
+ "step": 87
619
+ },
620
+ {
621
+ "epoch": 0.3179765130984643,
622
+ "grad_norm": 0.08008646965026855,
623
+ "learning_rate": 3.978930531033807e-05,
624
+ "loss": 0.3883,
625
+ "step": 88
626
+ },
627
+ {
628
+ "epoch": 0.3215898825654923,
629
+ "grad_norm": 0.06990881264209747,
630
+ "learning_rate": 3.9552015183030136e-05,
631
+ "loss": 0.4611,
632
+ "step": 89
633
+ },
634
+ {
635
+ "epoch": 0.3252032520325203,
636
+ "grad_norm": 0.05660560727119446,
637
+ "learning_rate": 3.93127254318018e-05,
638
+ "loss": 0.3865,
639
+ "step": 90
640
+ },
641
+ {
642
+ "epoch": 0.3288166214995483,
643
+ "grad_norm": 0.05711934715509415,
644
+ "learning_rate": 3.907146893797599e-05,
645
+ "loss": 0.4223,
646
+ "step": 91
647
+ },
648
+ {
649
+ "epoch": 0.3324299909665763,
650
+ "grad_norm": 0.06767363101243973,
651
+ "learning_rate": 3.882827885312999e-05,
652
+ "loss": 0.3481,
653
+ "step": 92
654
+ },
655
+ {
656
+ "epoch": 0.33604336043360433,
657
+ "grad_norm": 0.05866090953350067,
658
+ "learning_rate": 3.858318859454001e-05,
659
+ "loss": 0.4195,
660
+ "step": 93
661
+ },
662
+ {
663
+ "epoch": 0.33965672990063234,
664
+ "grad_norm": 0.05316139757633209,
665
+ "learning_rate": 3.833623184058926e-05,
666
+ "loss": 0.4042,
667
+ "step": 94
668
+ },
669
+ {
670
+ "epoch": 0.34327009936766034,
671
+ "grad_norm": 0.06730002164840698,
672
+ "learning_rate": 3.808744252614012e-05,
673
+ "loss": 0.3717,
674
+ "step": 95
675
+ },
676
+ {
677
+ "epoch": 0.34688346883468835,
678
+ "grad_norm": 0.07342930138111115,
679
+ "learning_rate": 3.783685483787105e-05,
680
+ "loss": 0.4075,
681
+ "step": 96
682
+ },
683
+ {
684
+ "epoch": 0.35049683830171635,
685
+ "grad_norm": 0.07083098590373993,
686
+ "learning_rate": 3.758450320957899e-05,
687
+ "loss": 0.3864,
688
+ "step": 97
689
+ },
690
+ {
691
+ "epoch": 0.35411020776874436,
692
+ "grad_norm": 0.07677371054887772,
693
+ "learning_rate": 3.7330422317447685e-05,
694
+ "loss": 0.393,
695
+ "step": 98
696
+ },
697
+ {
698
+ "epoch": 0.35772357723577236,
699
+ "grad_norm": 0.0808129534125328,
700
+ "learning_rate": 3.707464707528275e-05,
701
+ "loss": 0.3801,
702
+ "step": 99
703
+ },
704
+ {
705
+ "epoch": 0.36133694670280037,
706
+ "grad_norm": 0.06672363728284836,
707
+ "learning_rate": 3.681721262971413e-05,
708
+ "loss": 0.4472,
709
+ "step": 100
710
+ },
711
+ {
712
+ "epoch": 0.36495031616982837,
713
+ "grad_norm": 0.05534950643777847,
714
+ "learning_rate": 3.6558154355366506e-05,
715
+ "loss": 0.3683,
716
+ "step": 101
717
+ },
718
+ {
719
+ "epoch": 0.3685636856368564,
720
+ "grad_norm": 0.06686428934335709,
721
+ "learning_rate": 3.6297507849998344e-05,
722
+ "loss": 0.3455,
723
+ "step": 102
724
+ },
725
+ {
726
+ "epoch": 0.3721770551038844,
727
+ "grad_norm": 0.07248938828706741,
728
+ "learning_rate": 3.6035308929610446e-05,
729
+ "loss": 0.4083,
730
+ "step": 103
731
+ },
732
+ {
733
+ "epoch": 0.3757904245709124,
734
+ "grad_norm": 0.06316327303647995,
735
+ "learning_rate": 3.5771593623524265e-05,
736
+ "loss": 0.3661,
737
+ "step": 104
738
+ },
739
+ {
740
+ "epoch": 0.3794037940379404,
741
+ "grad_norm": 0.08561142534017563,
742
+ "learning_rate": 3.550639816943111e-05,
743
+ "loss": 0.3693,
744
+ "step": 105
745
+ },
746
+ {
747
+ "epoch": 0.3830171635049684,
748
+ "grad_norm": 0.05884739011526108,
749
+ "learning_rate": 3.5239759008412666e-05,
750
+ "loss": 0.4326,
751
+ "step": 106
752
+ },
753
+ {
754
+ "epoch": 0.3866305329719964,
755
+ "grad_norm": 0.06861259788274765,
756
+ "learning_rate": 3.497171277993346e-05,
757
+ "loss": 0.3423,
758
+ "step": 107
759
+ },
760
+ {
761
+ "epoch": 0.3902439024390244,
762
+ "grad_norm": 0.06908590346574783,
763
+ "learning_rate": 3.4702296316806244e-05,
764
+ "loss": 0.4494,
765
+ "step": 108
766
+ },
767
+ {
768
+ "epoch": 0.3938572719060524,
769
+ "grad_norm": 0.07454199343919754,
770
+ "learning_rate": 3.443154664013067e-05,
771
+ "loss": 0.4488,
772
+ "step": 109
773
+ },
774
+ {
775
+ "epoch": 0.3974706413730804,
776
+ "grad_norm": 0.07938794046640396,
777
+ "learning_rate": 3.415950095420616e-05,
778
+ "loss": 0.3938,
779
+ "step": 110
780
+ },
781
+ {
782
+ "epoch": 0.4010840108401084,
783
+ "grad_norm": 0.08505871146917343,
784
+ "learning_rate": 3.3886196641419545e-05,
785
+ "loss": 0.4004,
786
+ "step": 111
787
+ },
788
+ {
789
+ "epoch": 0.4046973803071364,
790
+ "grad_norm": 0.0625777617096901,
791
+ "learning_rate": 3.361167125710832e-05,
792
+ "loss": 0.3863,
793
+ "step": 112
794
+ },
795
+ {
796
+ "epoch": 0.4083107497741644,
797
+ "grad_norm": 0.07772816717624664,
798
+ "learning_rate": 3.333596252440008e-05,
799
+ "loss": 0.3981,
800
+ "step": 113
801
+ },
802
+ {
803
+ "epoch": 0.41192411924119243,
804
+ "grad_norm": 0.06656523048877716,
805
+ "learning_rate": 3.305910832902884e-05,
806
+ "loss": 0.3705,
807
+ "step": 114
808
+ },
809
+ {
810
+ "epoch": 0.41553748870822044,
811
+ "grad_norm": 0.07238256186246872,
812
+ "learning_rate": 3.278114671412917e-05,
813
+ "loss": 0.412,
814
+ "step": 115
815
+ },
816
+ {
817
+ "epoch": 0.41915085817524844,
818
+ "grad_norm": 0.06601731479167938,
819
+ "learning_rate": 3.2502115875008524e-05,
820
+ "loss": 0.3716,
821
+ "step": 116
822
+ },
823
+ {
824
+ "epoch": 0.42276422764227645,
825
+ "grad_norm": 0.0684824138879776,
826
+ "learning_rate": 3.222205415389877e-05,
827
+ "loss": 0.4183,
828
+ "step": 117
829
+ },
830
+ {
831
+ "epoch": 0.42637759710930445,
832
+ "grad_norm": 0.0698830783367157,
833
+ "learning_rate": 3.1941000034687515e-05,
834
+ "loss": 0.3517,
835
+ "step": 118
836
+ },
837
+ {
838
+ "epoch": 0.42999096657633246,
839
+ "grad_norm": 0.05978047475218773,
840
+ "learning_rate": 3.165899213762995e-05,
841
+ "loss": 0.3852,
842
+ "step": 119
843
+ },
844
+ {
845
+ "epoch": 0.43360433604336046,
846
+ "grad_norm": 0.07572682201862335,
847
+ "learning_rate": 3.1376069214041913e-05,
848
+ "loss": 0.4022,
849
+ "step": 120
850
+ },
851
+ {
852
+ "epoch": 0.4372177055103884,
853
+ "grad_norm": 0.07104960829019547,
854
+ "learning_rate": 3.109227014097505e-05,
855
+ "loss": 0.4185,
856
+ "step": 121
857
+ },
858
+ {
859
+ "epoch": 0.4408310749774164,
860
+ "grad_norm": 0.06828156113624573,
861
+ "learning_rate": 3.0807633915874584e-05,
862
+ "loss": 0.4239,
863
+ "step": 122
864
+ },
865
+ {
866
+ "epoch": 0.4444444444444444,
867
+ "grad_norm": 0.057690802961587906,
868
+ "learning_rate": 3.052219965122062e-05,
869
+ "loss": 0.4109,
870
+ "step": 123
871
+ },
872
+ {
873
+ "epoch": 0.4480578139114724,
874
+ "grad_norm": 0.06580954045057297,
875
+ "learning_rate": 3.0236006569153617e-05,
876
+ "loss": 0.359,
877
+ "step": 124
878
+ },
879
+ {
880
+ "epoch": 0.45167118337850043,
881
+ "grad_norm": 0.060349613428115845,
882
+ "learning_rate": 2.9949093996084747e-05,
883
+ "loss": 0.3775,
884
+ "step": 125
885
+ },
886
+ {
887
+ "epoch": 0.45528455284552843,
888
+ "grad_norm": 0.07335729151964188,
889
+ "learning_rate": 2.9661501357292033e-05,
890
+ "loss": 0.4043,
891
+ "step": 126
892
+ },
893
+ {
894
+ "epoch": 0.45889792231255644,
895
+ "grad_norm": 0.04954389110207558,
896
+ "learning_rate": 2.9373268171502777e-05,
897
+ "loss": 0.3537,
898
+ "step": 127
899
+ },
900
+ {
901
+ "epoch": 0.46251129177958444,
902
+ "grad_norm": 0.07528957724571228,
903
+ "learning_rate": 2.9084434045463255e-05,
904
+ "loss": 0.467,
905
+ "step": 128
906
+ },
907
+ {
908
+ "epoch": 0.46612466124661245,
909
+ "grad_norm": 0.06106121093034744,
910
+ "learning_rate": 2.8795038668496222e-05,
911
+ "loss": 0.4323,
912
+ "step": 129
913
+ },
914
+ {
915
+ "epoch": 0.46973803071364045,
916
+ "grad_norm": 0.08181653916835785,
917
+ "learning_rate": 2.850512180704715e-05,
918
+ "loss": 0.4208,
919
+ "step": 130
920
+ },
921
+ {
922
+ "epoch": 0.47335140018066846,
923
+ "grad_norm": 0.07354505360126495,
924
+ "learning_rate": 2.821472329921981e-05,
925
+ "loss": 0.3909,
926
+ "step": 131
927
+ },
928
+ {
929
+ "epoch": 0.47696476964769646,
930
+ "grad_norm": 0.09099866449832916,
931
+ "learning_rate": 2.792388304930207e-05,
932
+ "loss": 0.4296,
933
+ "step": 132
934
+ },
935
+ {
936
+ "epoch": 0.48057813911472447,
937
+ "grad_norm": 0.08062151074409485,
938
+ "learning_rate": 2.7632641022282502e-05,
939
+ "loss": 0.4106,
940
+ "step": 133
941
+ },
942
+ {
943
+ "epoch": 0.48419150858175247,
944
+ "grad_norm": 0.09198120981454849,
945
+ "learning_rate": 2.7341037238358774e-05,
946
+ "loss": 0.4064,
947
+ "step": 134
948
+ },
949
+ {
950
+ "epoch": 0.4878048780487805,
951
+ "grad_norm": 0.05343058705329895,
952
+ "learning_rate": 2.704911176743833e-05,
953
+ "loss": 0.404,
954
+ "step": 135
955
+ },
956
+ {
957
+ "epoch": 0.4914182475158085,
958
+ "grad_norm": 0.0657978504896164,
959
+ "learning_rate": 2.6756904723632324e-05,
960
+ "loss": 0.3993,
961
+ "step": 136
962
+ },
963
+ {
964
+ "epoch": 0.4950316169828365,
965
+ "grad_norm": 0.057678401470184326,
966
+ "learning_rate": 2.646445625974347e-05,
967
+ "loss": 0.3804,
968
+ "step": 137
969
+ },
970
+ {
971
+ "epoch": 0.4986449864498645,
972
+ "grad_norm": 0.06898088753223419,
973
+ "learning_rate": 2.6171806561748502e-05,
974
+ "loss": 0.4452,
975
+ "step": 138
976
+ },
977
+ {
978
+ "epoch": 0.5022583559168925,
979
+ "grad_norm": 0.09333262592554092,
980
+ "learning_rate": 2.5878995843276204e-05,
981
+ "loss": 0.3304,
982
+ "step": 139
983
+ },
984
+ {
985
+ "epoch": 0.5058717253839206,
986
+ "grad_norm": 0.06717183440923691,
987
+ "learning_rate": 2.5586064340081516e-05,
988
+ "loss": 0.326,
989
+ "step": 140
990
+ },
991
+ {
992
+ "epoch": 0.5094850948509485,
993
+ "grad_norm": 0.06729979068040848,
994
+ "learning_rate": 2.529305230451666e-05,
995
+ "loss": 0.3934,
996
+ "step": 141
997
+ },
998
+ {
999
+ "epoch": 0.5130984643179766,
1000
+ "grad_norm": 0.09550358355045319,
1001
+ "learning_rate": 2.5e-05,
1002
+ "loss": 0.4733,
1003
+ "step": 142
1004
+ },
1005
+ {
1006
+ "epoch": 0.5167118337850045,
1007
+ "grad_norm": 0.07080523669719696,
1008
+ "learning_rate": 2.4706947695483348e-05,
1009
+ "loss": 0.4039,
1010
+ "step": 143
1011
+ },
1012
+ {
1013
+ "epoch": 0.5203252032520326,
1014
+ "grad_norm": 0.055423106998205185,
1015
+ "learning_rate": 2.441393565991849e-05,
1016
+ "loss": 0.3275,
1017
+ "step": 144
1018
+ },
1019
+ {
1020
+ "epoch": 0.5239385727190605,
1021
+ "grad_norm": 0.06483904272317886,
1022
+ "learning_rate": 2.4121004156723802e-05,
1023
+ "loss": 0.4377,
1024
+ "step": 145
1025
+ },
1026
+ {
1027
+ "epoch": 0.5275519421860885,
1028
+ "grad_norm": 0.06614437699317932,
1029
+ "learning_rate": 2.3828193438251497e-05,
1030
+ "loss": 0.3935,
1031
+ "step": 146
1032
+ },
1033
+ {
1034
+ "epoch": 0.5311653116531165,
1035
+ "grad_norm": 0.08745498955249786,
1036
+ "learning_rate": 2.3535543740256536e-05,
1037
+ "loss": 0.4348,
1038
+ "step": 147
1039
+ },
1040
+ {
1041
+ "epoch": 0.5347786811201445,
1042
+ "grad_norm": 0.07158234715461731,
1043
+ "learning_rate": 2.3243095276367685e-05,
1044
+ "loss": 0.3286,
1045
+ "step": 148
1046
+ },
1047
+ {
1048
+ "epoch": 0.5383920505871725,
1049
+ "grad_norm": 0.06448652595281601,
1050
+ "learning_rate": 2.2950888232561672e-05,
1051
+ "loss": 0.4108,
1052
+ "step": 149
1053
+ },
1054
+ {
1055
+ "epoch": 0.5420054200542005,
1056
+ "grad_norm": 0.07621192187070847,
1057
+ "learning_rate": 2.2658962761641232e-05,
1058
+ "loss": 0.4317,
1059
+ "step": 150
1060
+ },
1061
+ {
1062
+ "epoch": 0.5456187895212286,
1063
+ "grad_norm": 0.07459475100040436,
1064
+ "learning_rate": 2.23673589777175e-05,
1065
+ "loss": 0.3876,
1066
+ "step": 151
1067
+ },
1068
+ {
1069
+ "epoch": 0.5492321589882565,
1070
+ "grad_norm": 0.07355853170156479,
1071
+ "learning_rate": 2.207611695069794e-05,
1072
+ "loss": 0.3506,
1073
+ "step": 152
1074
+ },
1075
+ {
1076
+ "epoch": 0.5528455284552846,
1077
+ "grad_norm": 0.07565652579069138,
1078
+ "learning_rate": 2.17852767007802e-05,
1079
+ "loss": 0.4221,
1080
+ "step": 153
1081
+ },
1082
+ {
1083
+ "epoch": 0.5564588979223125,
1084
+ "grad_norm": 0.07433846592903137,
1085
+ "learning_rate": 2.1494878192952855e-05,
1086
+ "loss": 0.3913,
1087
+ "step": 154
1088
+ },
1089
+ {
1090
+ "epoch": 0.5600722673893406,
1091
+ "grad_norm": 0.07123446464538574,
1092
+ "learning_rate": 2.1204961331503787e-05,
1093
+ "loss": 0.4106,
1094
+ "step": 155
1095
+ },
1096
+ {
1097
+ "epoch": 0.5636856368563685,
1098
+ "grad_norm": 0.0848294198513031,
1099
+ "learning_rate": 2.0915565954536744e-05,
1100
+ "loss": 0.3171,
1101
+ "step": 156
1102
+ },
1103
+ {
1104
+ "epoch": 0.5672990063233966,
1105
+ "grad_norm": 0.06394634395837784,
1106
+ "learning_rate": 2.0626731828497225e-05,
1107
+ "loss": 0.4106,
1108
+ "step": 157
1109
+ },
1110
+ {
1111
+ "epoch": 0.5709123757904245,
1112
+ "grad_norm": 0.06601906567811966,
1113
+ "learning_rate": 2.0338498642707977e-05,
1114
+ "loss": 0.3651,
1115
+ "step": 158
1116
+ },
1117
+ {
1118
+ "epoch": 0.5745257452574526,
1119
+ "grad_norm": 0.0734376311302185,
1120
+ "learning_rate": 2.005090600391526e-05,
1121
+ "loss": 0.3906,
1122
+ "step": 159
1123
+ },
1124
+ {
1125
+ "epoch": 0.5781391147244805,
1126
+ "grad_norm": 0.07122786343097687,
1127
+ "learning_rate": 1.9763993430846395e-05,
1128
+ "loss": 0.4157,
1129
+ "step": 160
1130
+ },
1131
+ {
1132
+ "epoch": 0.5817524841915086,
1133
+ "grad_norm": 0.06590158492326736,
1134
+ "learning_rate": 1.947780034877938e-05,
1135
+ "loss": 0.4267,
1136
+ "step": 161
1137
+ },
1138
+ {
1139
+ "epoch": 0.5853658536585366,
1140
+ "grad_norm": 0.07380690425634384,
1141
+ "learning_rate": 1.9192366084125425e-05,
1142
+ "loss": 0.3748,
1143
+ "step": 162
1144
+ },
1145
+ {
1146
+ "epoch": 0.5889792231255646,
1147
+ "grad_norm": 0.054361093789339066,
1148
+ "learning_rate": 1.890772985902496e-05,
1149
+ "loss": 0.3637,
1150
+ "step": 163
1151
+ },
1152
+ {
1153
+ "epoch": 0.5925925925925926,
1154
+ "grad_norm": 0.06896340101957321,
1155
+ "learning_rate": 1.8623930785958092e-05,
1156
+ "loss": 0.4319,
1157
+ "step": 164
1158
+ },
1159
+ {
1160
+ "epoch": 0.5962059620596206,
1161
+ "grad_norm": 0.08140537887811661,
1162
+ "learning_rate": 1.8341007862370056e-05,
1163
+ "loss": 0.3942,
1164
+ "step": 165
1165
+ },
1166
+ {
1167
+ "epoch": 0.5998193315266486,
1168
+ "grad_norm": 0.07021729648113251,
1169
+ "learning_rate": 1.8058999965312484e-05,
1170
+ "loss": 0.3917,
1171
+ "step": 166
1172
+ },
1173
+ {
1174
+ "epoch": 0.6034327009936766,
1175
+ "grad_norm": 0.06319273263216019,
1176
+ "learning_rate": 1.777794584610124e-05,
1177
+ "loss": 0.3833,
1178
+ "step": 167
1179
+ },
1180
+ {
1181
+ "epoch": 0.6070460704607046,
1182
+ "grad_norm": 0.07088933885097504,
1183
+ "learning_rate": 1.749788412499149e-05,
1184
+ "loss": 0.3326,
1185
+ "step": 168
1186
+ },
1187
+ {
1188
+ "epoch": 0.6106594399277326,
1189
+ "grad_norm": 0.06848324090242386,
1190
+ "learning_rate": 1.721885328587083e-05,
1191
+ "loss": 0.5018,
1192
+ "step": 169
1193
+ },
1194
+ {
1195
+ "epoch": 0.6142728093947606,
1196
+ "grad_norm": 0.07163573056459427,
1197
+ "learning_rate": 1.694089167097116e-05,
1198
+ "loss": 0.3624,
1199
+ "step": 170
1200
+ },
1201
+ {
1202
+ "epoch": 0.6178861788617886,
1203
+ "grad_norm": 0.06683260202407837,
1204
+ "learning_rate": 1.6664037475599923e-05,
1205
+ "loss": 0.4198,
1206
+ "step": 171
1207
+ },
1208
+ {
1209
+ "epoch": 0.6214995483288166,
1210
+ "grad_norm": 0.06273495405912399,
1211
+ "learning_rate": 1.638832874289168e-05,
1212
+ "loss": 0.3388,
1213
+ "step": 172
1214
+ },
1215
+ {
1216
+ "epoch": 0.6251129177958447,
1217
+ "grad_norm": 0.06024303659796715,
1218
+ "learning_rate": 1.611380335858047e-05,
1219
+ "loss": 0.4156,
1220
+ "step": 173
1221
+ },
1222
+ {
1223
+ "epoch": 0.6287262872628726,
1224
+ "grad_norm": 0.08732262253761292,
1225
+ "learning_rate": 1.5840499045793843e-05,
1226
+ "loss": 0.3883,
1227
+ "step": 174
1228
+ },
1229
+ {
1230
+ "epoch": 0.6323396567299007,
1231
+ "grad_norm": 0.06800790876150131,
1232
+ "learning_rate": 1.5568453359869334e-05,
1233
+ "loss": 0.3636,
1234
+ "step": 175
1235
+ },
1236
+ {
1237
+ "epoch": 0.6359530261969286,
1238
+ "grad_norm": 0.08514184504747391,
1239
+ "learning_rate": 1.5297703683193752e-05,
1240
+ "loss": 0.3664,
1241
+ "step": 176
1242
+ },
1243
+ {
1244
+ "epoch": 0.6395663956639567,
1245
+ "grad_norm": 0.0805889442563057,
1246
+ "learning_rate": 1.502828722006655e-05,
1247
+ "loss": 0.3912,
1248
+ "step": 177
1249
+ },
1250
+ {
1251
+ "epoch": 0.6431797651309846,
1252
+ "grad_norm": 0.07321416586637497,
1253
+ "learning_rate": 1.4760240991587337e-05,
1254
+ "loss": 0.4077,
1255
+ "step": 178
1256
+ },
1257
+ {
1258
+ "epoch": 0.6467931345980127,
1259
+ "grad_norm": 0.06993624567985535,
1260
+ "learning_rate": 1.4493601830568887e-05,
1261
+ "loss": 0.3728,
1262
+ "step": 179
1263
+ },
1264
+ {
1265
+ "epoch": 0.6504065040650406,
1266
+ "grad_norm": 0.07736963033676147,
1267
+ "learning_rate": 1.4228406376475742e-05,
1268
+ "loss": 0.3644,
1269
+ "step": 180
1270
+ },
1271
+ {
1272
+ "epoch": 0.6540198735320687,
1273
+ "grad_norm": 0.06840698421001434,
1274
+ "learning_rate": 1.396469107038956e-05,
1275
+ "loss": 0.3936,
1276
+ "step": 181
1277
+ },
1278
+ {
1279
+ "epoch": 0.6576332429990966,
1280
+ "grad_norm": 0.07498890906572342,
1281
+ "learning_rate": 1.3702492150001659e-05,
1282
+ "loss": 0.3948,
1283
+ "step": 182
1284
+ },
1285
+ {
1286
+ "epoch": 0.6612466124661247,
1287
+ "grad_norm": 0.06307978183031082,
1288
+ "learning_rate": 1.34418456446335e-05,
1289
+ "loss": 0.398,
1290
+ "step": 183
1291
+ },
1292
+ {
1293
+ "epoch": 0.6648599819331527,
1294
+ "grad_norm": 0.0843866616487503,
1295
+ "learning_rate": 1.3182787370285865e-05,
1296
+ "loss": 0.3891,
1297
+ "step": 184
1298
+ },
1299
+ {
1300
+ "epoch": 0.6684733514001807,
1301
+ "grad_norm": 0.07880077511072159,
1302
+ "learning_rate": 1.292535292471726e-05,
1303
+ "loss": 0.3812,
1304
+ "step": 185
1305
+ },
1306
+ {
1307
+ "epoch": 0.6720867208672087,
1308
+ "grad_norm": 0.06986968219280243,
1309
+ "learning_rate": 1.2669577682552319e-05,
1310
+ "loss": 0.3851,
1311
+ "step": 186
1312
+ },
1313
+ {
1314
+ "epoch": 0.6757000903342367,
1315
+ "grad_norm": 0.07602784037590027,
1316
+ "learning_rate": 1.2415496790421011e-05,
1317
+ "loss": 0.3956,
1318
+ "step": 187
1319
+ },
1320
+ {
1321
+ "epoch": 0.6793134598012647,
1322
+ "grad_norm": 0.06611546874046326,
1323
+ "learning_rate": 1.2163145162128947e-05,
1324
+ "loss": 0.3629,
1325
+ "step": 188
1326
+ },
1327
+ {
1328
+ "epoch": 0.6829268292682927,
1329
+ "grad_norm": 0.07958898693323135,
1330
+ "learning_rate": 1.1912557473859895e-05,
1331
+ "loss": 0.3647,
1332
+ "step": 189
1333
+ },
1334
+ {
1335
+ "epoch": 0.6865401987353207,
1336
+ "grad_norm": 0.06264237314462662,
1337
+ "learning_rate": 1.1663768159410748e-05,
1338
+ "loss": 0.3797,
1339
+ "step": 190
1340
+ },
1341
+ {
1342
+ "epoch": 0.6901535682023487,
1343
+ "grad_norm": 0.08303744345903397,
1344
+ "learning_rate": 1.1416811405459993e-05,
1345
+ "loss": 0.3754,
1346
+ "step": 191
1347
+ },
1348
+ {
1349
+ "epoch": 0.6937669376693767,
1350
+ "grad_norm": 0.07206673175096512,
1351
+ "learning_rate": 1.1171721146870015e-05,
1352
+ "loss": 0.327,
1353
+ "step": 192
1354
+ },
1355
+ {
1356
+ "epoch": 0.6973803071364046,
1357
+ "grad_norm": 0.06349314749240875,
1358
+ "learning_rate": 1.0928531062024017e-05,
1359
+ "loss": 0.3902,
1360
+ "step": 193
1361
+ },
1362
+ {
1363
+ "epoch": 0.7009936766034327,
1364
+ "grad_norm": 0.07241489738225937,
1365
+ "learning_rate": 1.0687274568198208e-05,
1366
+ "loss": 0.3845,
1367
+ "step": 194
1368
+ },
1369
+ {
1370
+ "epoch": 0.7046070460704607,
1371
+ "grad_norm": 0.06357239931821823,
1372
+ "learning_rate": 1.0447984816969874e-05,
1373
+ "loss": 0.3881,
1374
+ "step": 195
1375
+ },
1376
+ {
1377
+ "epoch": 0.7082204155374887,
1378
+ "grad_norm": 0.06316613405942917,
1379
+ "learning_rate": 1.021069468966194e-05,
1380
+ "loss": 0.4735,
1381
+ "step": 196
1382
+ },
1383
+ {
1384
+ "epoch": 0.7118337850045167,
1385
+ "grad_norm": 0.08076903223991394,
1386
+ "learning_rate": 9.975436792824691e-06,
1387
+ "loss": 0.43,
1388
+ "step": 197
1389
+ },
1390
+ {
1391
+ "epoch": 0.7154471544715447,
1392
+ "grad_norm": 0.0836021676659584,
1393
+ "learning_rate": 9.742243453755202e-06,
1394
+ "loss": 0.3818,
1395
+ "step": 198
1396
+ },
1397
+ {
1398
+ "epoch": 0.7190605239385727,
1399
+ "grad_norm": 0.0713673084974289,
1400
+ "learning_rate": 9.5111467160552e-06,
1401
+ "loss": 0.3846,
1402
+ "step": 199
1403
+ },
1404
+ {
1405
+ "epoch": 0.7226738934056007,
1406
+ "grad_norm": 0.08711904287338257,
1407
+ "learning_rate": 9.282178335227884e-06,
1408
+ "loss": 0.4817,
1409
+ "step": 200
1410
+ },
1411
+ {
1412
+ "epoch": 0.7262872628726287,
1413
+ "grad_norm": 0.05264454334974289,
1414
+ "learning_rate": 9.05536977431431e-06,
1415
+ "loss": 0.3995,
1416
+ "step": 201
1417
+ },
1418
+ {
1419
+ "epoch": 0.7299006323396567,
1420
+ "grad_norm": 0.07466941326856613,
1421
+ "learning_rate": 8.830752199570033e-06,
1422
+ "loss": 0.3718,
1423
+ "step": 202
1424
+ },
1425
+ {
1426
+ "epoch": 0.7335140018066847,
1427
+ "grad_norm": 0.07776648551225662,
1428
+ "learning_rate": 8.608356476182424e-06,
1429
+ "loss": 0.4786,
1430
+ "step": 203
1431
+ },
1432
+ {
1433
+ "epoch": 0.7371273712737128,
1434
+ "grad_norm": 0.06611160188913345,
1435
+ "learning_rate": 8.38821316402946e-06,
1436
+ "loss": 0.3668,
1437
+ "step": 204
1438
+ },
1439
+ {
1440
+ "epoch": 0.7407407407407407,
1441
+ "grad_norm": 0.07174837589263916,
1442
+ "learning_rate": 8.170352513480408e-06,
1443
+ "loss": 0.4016,
1444
+ "step": 205
1445
+ },
1446
+ {
1447
+ "epoch": 0.7443541102077688,
1448
+ "grad_norm": 0.0830477848649025,
1449
+ "learning_rate": 7.954804461239053e-06,
1450
+ "loss": 0.4162,
1451
+ "step": 206
1452
+ },
1453
+ {
1454
+ "epoch": 0.7479674796747967,
1455
+ "grad_norm": 0.08300362527370453,
1456
+ "learning_rate": 7.741598626230079e-06,
1457
+ "loss": 0.3738,
1458
+ "step": 207
1459
+ },
1460
+ {
1461
+ "epoch": 0.7515808491418248,
1462
+ "grad_norm": 0.07526036351919174,
1463
+ "learning_rate": 7.530764305528959e-06,
1464
+ "loss": 0.3576,
1465
+ "step": 208
1466
+ },
1467
+ {
1468
+ "epoch": 0.7551942186088527,
1469
+ "grad_norm": 0.06786955147981644,
1470
+ "learning_rate": 7.3223304703363135e-06,
1471
+ "loss": 0.4152,
1472
+ "step": 209
1473
+ },
1474
+ {
1475
+ "epoch": 0.7588075880758808,
1476
+ "grad_norm": 0.08544765412807465,
1477
+ "learning_rate": 7.116325761996817e-06,
1478
+ "loss": 0.3735,
1479
+ "step": 210
1480
+ },
1481
+ {
1482
+ "epoch": 0.7624209575429087,
1483
+ "grad_norm": 0.06077965721487999,
1484
+ "learning_rate": 6.91277848806356e-06,
1485
+ "loss": 0.3486,
1486
+ "step": 211
1487
+ },
1488
+ {
1489
+ "epoch": 0.7660343270099368,
1490
+ "grad_norm": 0.07332652807235718,
1491
+ "learning_rate": 6.711716618408281e-06,
1492
+ "loss": 0.3734,
1493
+ "step": 212
1494
+ },
1495
+ {
1496
+ "epoch": 0.7696476964769647,
1497
+ "grad_norm": 0.07848729193210602,
1498
+ "learning_rate": 6.513167781377885e-06,
1499
+ "loss": 0.4231,
1500
+ "step": 213
1501
+ },
1502
+ {
1503
+ "epoch": 0.7732610659439928,
1504
+ "grad_norm": 0.07897993177175522,
1505
+ "learning_rate": 6.317159259998073e-06,
1506
+ "loss": 0.3513,
1507
+ "step": 214
1508
+ },
1509
+ {
1510
+ "epoch": 0.7768744354110207,
1511
+ "grad_norm": 0.07235241681337357,
1512
+ "learning_rate": 6.123717988224237e-06,
1513
+ "loss": 0.4069,
1514
+ "step": 215
1515
+ },
1516
+ {
1517
+ "epoch": 0.7804878048780488,
1518
+ "grad_norm": 0.09085345268249512,
1519
+ "learning_rate": 5.932870547240454e-06,
1520
+ "loss": 0.3849,
1521
+ "step": 216
1522
+ },
1523
+ {
1524
+ "epoch": 0.7841011743450768,
1525
+ "grad_norm": 0.07704368233680725,
1526
+ "learning_rate": 5.74464316180689e-06,
1527
+ "loss": 0.4261,
1528
+ "step": 217
1529
+ },
1530
+ {
1531
+ "epoch": 0.7877145438121048,
1532
+ "grad_norm": 0.057720448821783066,
1533
+ "learning_rate": 5.559061696656198e-06,
1534
+ "loss": 0.3711,
1535
+ "step": 218
1536
+ },
1537
+ {
1538
+ "epoch": 0.7913279132791328,
1539
+ "grad_norm": 0.06448069959878922,
1540
+ "learning_rate": 5.37615165293942e-06,
1541
+ "loss": 0.4027,
1542
+ "step": 219
1543
+ },
1544
+ {
1545
+ "epoch": 0.7949412827461608,
1546
+ "grad_norm": 0.08539154380559921,
1547
+ "learning_rate": 5.1959381647217666e-06,
1548
+ "loss": 0.388,
1549
+ "step": 220
1550
+ },
1551
+ {
1552
+ "epoch": 0.7985546522131888,
1553
+ "grad_norm": 0.07000590115785599,
1554
+ "learning_rate": 5.018445995528931e-06,
1555
+ "loss": 0.4122,
1556
+ "step": 221
1557
+ },
1558
+ {
1559
+ "epoch": 0.8021680216802168,
1560
+ "grad_norm": 0.07643178850412369,
1561
+ "learning_rate": 4.843699534944257e-06,
1562
+ "loss": 0.3749,
1563
+ "step": 222
1564
+ },
1565
+ {
1566
+ "epoch": 0.8057813911472448,
1567
+ "grad_norm": 0.06629081815481186,
1568
+ "learning_rate": 4.671722795257327e-06,
1569
+ "loss": 0.3817,
1570
+ "step": 223
1571
+ },
1572
+ {
1573
+ "epoch": 0.8093947606142728,
1574
+ "grad_norm": 0.06171542406082153,
1575
+ "learning_rate": 4.502539408164386e-06,
1576
+ "loss": 0.3474,
1577
+ "step": 224
1578
+ },
1579
+ {
1580
+ "epoch": 0.8130081300813008,
1581
+ "grad_norm": 0.06734922528266907,
1582
+ "learning_rate": 4.336172621521034e-06,
1583
+ "loss": 0.3328,
1584
+ "step": 225
1585
+ },
1586
+ {
1587
+ "epoch": 0.8166214995483289,
1588
+ "grad_norm": 0.09524697810411453,
1589
+ "learning_rate": 4.1726452961477146e-06,
1590
+ "loss": 0.3433,
1591
+ "step": 226
1592
+ },
1593
+ {
1594
+ "epoch": 0.8202348690153568,
1595
+ "grad_norm": 0.06357850879430771,
1596
+ "learning_rate": 4.01197990268834e-06,
1597
+ "loss": 0.3992,
1598
+ "step": 227
1599
+ },
1600
+ {
1601
+ "epoch": 0.8238482384823849,
1602
+ "grad_norm": 0.07560393214225769,
1603
+ "learning_rate": 3.8541985185225645e-06,
1604
+ "loss": 0.3575,
1605
+ "step": 228
1606
+ },
1607
+ {
1608
+ "epoch": 0.8274616079494128,
1609
+ "grad_norm": 0.06906560808420181,
1610
+ "learning_rate": 3.6993228247320877e-06,
1611
+ "loss": 0.3287,
1612
+ "step": 229
1613
+ },
1614
+ {
1615
+ "epoch": 0.8310749774164409,
1616
+ "grad_norm": 0.08411566913127899,
1617
+ "learning_rate": 3.547374103121398e-06,
1618
+ "loss": 0.4115,
1619
+ "step": 230
1620
+ },
1621
+ {
1622
+ "epoch": 0.8346883468834688,
1623
+ "grad_norm": 0.08515972644090652,
1624
+ "learning_rate": 3.398373233293378e-06,
1625
+ "loss": 0.3709,
1626
+ "step": 231
1627
+ },
1628
+ {
1629
+ "epoch": 0.8383017163504969,
1630
+ "grad_norm": 0.06780155003070831,
1631
+ "learning_rate": 3.252340689780245e-06,
1632
+ "loss": 0.3599,
1633
+ "step": 232
1634
+ },
1635
+ {
1636
+ "epoch": 0.8419150858175248,
1637
+ "grad_norm": 0.08019706606864929,
1638
+ "learning_rate": 3.1092965392300417e-06,
1639
+ "loss": 0.3869,
1640
+ "step": 233
1641
+ },
1642
+ {
1643
+ "epoch": 0.8455284552845529,
1644
+ "grad_norm": 0.0702086016535759,
1645
+ "learning_rate": 2.969260437649293e-06,
1646
+ "loss": 0.3846,
1647
+ "step": 234
1648
+ },
1649
+ {
1650
+ "epoch": 0.8491418247515808,
1651
+ "grad_norm": 0.0851154550909996,
1652
+ "learning_rate": 2.8322516277019624e-06,
1653
+ "loss": 0.3434,
1654
+ "step": 235
1655
+ },
1656
+ {
1657
+ "epoch": 0.8527551942186089,
1658
+ "grad_norm": 0.06722518056631088,
1659
+ "learning_rate": 2.6982889360653377e-06,
1660
+ "loss": 0.3349,
1661
+ "step": 236
1662
+ },
1663
+ {
1664
+ "epoch": 0.8563685636856369,
1665
+ "grad_norm": 0.06803542375564575,
1666
+ "learning_rate": 2.5673907708429976e-06,
1667
+ "loss": 0.3526,
1668
+ "step": 237
1669
+ },
1670
+ {
1671
+ "epoch": 0.8599819331526649,
1672
+ "grad_norm": 0.08029063045978546,
1673
+ "learning_rate": 2.4395751190352924e-06,
1674
+ "loss": 0.4286,
1675
+ "step": 238
1676
+ },
1677
+ {
1678
+ "epoch": 0.8635953026196929,
1679
+ "grad_norm": 0.08042778819799423,
1680
+ "learning_rate": 2.3148595440677405e-06,
1681
+ "loss": 0.3739,
1682
+ "step": 239
1683
+ },
1684
+ {
1685
+ "epoch": 0.8672086720867209,
1686
+ "grad_norm": 0.07175204902887344,
1687
+ "learning_rate": 2.1932611833775846e-06,
1688
+ "loss": 0.4156,
1689
+ "step": 240
1690
+ },
1691
+ {
1692
+ "epoch": 0.8708220415537489,
1693
+ "grad_norm": 0.058878783136606216,
1694
+ "learning_rate": 2.074796746058896e-06,
1695
+ "loss": 0.3636,
1696
+ "step": 241
1697
+ },
1698
+ {
1699
+ "epoch": 0.8744354110207768,
1700
+ "grad_norm": 0.08569607883691788,
1701
+ "learning_rate": 1.9594825105665654e-06,
1702
+ "loss": 0.3889,
1703
+ "step": 242
1704
+ },
1705
+ {
1706
+ "epoch": 0.8780487804878049,
1707
+ "grad_norm": 0.07353324443101883,
1708
+ "learning_rate": 1.847334322479413e-06,
1709
+ "loss": 0.4352,
1710
+ "step": 243
1711
+ },
1712
+ {
1713
+ "epoch": 0.8816621499548328,
1714
+ "grad_norm": 0.07135035842657089,
1715
+ "learning_rate": 1.738367592322837e-06,
1716
+ "loss": 0.4265,
1717
+ "step": 244
1718
+ },
1719
+ {
1720
+ "epoch": 0.8852755194218609,
1721
+ "grad_norm": 0.06918162852525711,
1722
+ "learning_rate": 1.6325972934512018e-06,
1723
+ "loss": 0.4295,
1724
+ "step": 245
1725
+ },
1726
+ {
1727
+ "epoch": 0.8888888888888888,
1728
+ "grad_norm": 0.07300789654254913,
1729
+ "learning_rate": 1.5300379599903409e-06,
1730
+ "loss": 0.4226,
1731
+ "step": 246
1732
+ },
1733
+ {
1734
+ "epoch": 0.8925022583559169,
1735
+ "grad_norm": 0.06973148882389069,
1736
+ "learning_rate": 1.4307036848403648e-06,
1737
+ "loss": 0.3368,
1738
+ "step": 247
1739
+ },
1740
+ {
1741
+ "epoch": 0.8961156278229448,
1742
+ "grad_norm": 0.07200148701667786,
1743
+ "learning_rate": 1.3346081177391472e-06,
1744
+ "loss": 0.3924,
1745
+ "step": 248
1746
+ },
1747
+ {
1748
+ "epoch": 0.8997289972899729,
1749
+ "grad_norm": 0.07833510637283325,
1750
+ "learning_rate": 1.2417644633866632e-06,
1751
+ "loss": 0.3274,
1752
+ "step": 249
1753
+ },
1754
+ {
1755
+ "epoch": 0.9033423667570009,
1756
+ "grad_norm": 0.061651114374399185,
1757
+ "learning_rate": 1.1521854796305242e-06,
1758
+ "loss": 0.3705,
1759
+ "step": 250
1760
+ },
1761
+ {
1762
+ "epoch": 0.9069557362240289,
1763
+ "grad_norm": 0.07440148293972015,
1764
+ "learning_rate": 1.0658834757128838e-06,
1765
+ "loss": 0.3715,
1766
+ "step": 251
1767
+ },
1768
+ {
1769
+ "epoch": 0.9105691056910569,
1770
+ "grad_norm": 0.0720466673374176,
1771
+ "learning_rate": 9.828703105789983e-07,
1772
+ "loss": 0.3361,
1773
+ "step": 252
1774
+ },
1775
+ {
1776
+ "epoch": 0.9141824751580849,
1777
+ "grad_norm": 0.08179104328155518,
1778
+ "learning_rate": 9.031573912476554e-07,
1779
+ "loss": 0.3393,
1780
+ "step": 253
1781
+ },
1782
+ {
1783
+ "epoch": 0.9177958446251129,
1784
+ "grad_norm": 0.058865226805210114,
1785
+ "learning_rate": 8.267556712437341e-07,
1786
+ "loss": 0.4249,
1787
+ "step": 254
1788
+ },
1789
+ {
1790
+ "epoch": 0.9214092140921409,
1791
+ "grad_norm": 0.07929901778697968,
1792
+ "learning_rate": 7.536756490930358e-07,
1793
+ "loss": 0.4341,
1794
+ "step": 255
1795
+ },
1796
+ {
1797
+ "epoch": 0.9250225835591689,
1798
+ "grad_norm": 0.07914505153894424,
1799
+ "learning_rate": 6.839273668796747e-07,
1800
+ "loss": 0.3942,
1801
+ "step": 256
1802
+ },
1803
+ {
1804
+ "epoch": 0.928635953026197,
1805
+ "grad_norm": 0.08146975934505463,
1806
+ "learning_rate": 6.175204088661485e-07,
1807
+ "loss": 0.3562,
1808
+ "step": 257
1809
+ },
1810
+ {
1811
+ "epoch": 0.9322493224932249,
1812
+ "grad_norm": 0.08726157248020172,
1813
+ "learning_rate": 5.544639001763718e-07,
1814
+ "loss": 0.4314,
1815
+ "step": 258
1816
+ },
1817
+ {
1818
+ "epoch": 0.935862691960253,
1819
+ "grad_norm": 0.09031800180673599,
1820
+ "learning_rate": 4.947665055417605e-07,
1821
+ "loss": 0.3842,
1822
+ "step": 259
1823
+ },
1824
+ {
1825
+ "epoch": 0.9394760614272809,
1826
+ "grad_norm": 0.0922897681593895,
1827
+ "learning_rate": 4.3843642811059737e-07,
1828
+ "loss": 0.3285,
1829
+ "step": 260
1830
+ },
1831
+ {
1832
+ "epoch": 0.943089430894309,
1833
+ "grad_norm": 0.07188927382230759,
1834
+ "learning_rate": 3.854814083208064e-07,
1835
+ "loss": 0.3839,
1836
+ "step": 261
1837
+ },
1838
+ {
1839
+ "epoch": 0.9467028003613369,
1840
+ "grad_norm": 0.08181816339492798,
1841
+ "learning_rate": 3.3590872283633944e-07,
1842
+ "loss": 0.3651,
1843
+ "step": 262
1844
+ },
1845
+ {
1846
+ "epoch": 0.950316169828365,
1847
+ "grad_norm": 0.0699373111128807,
1848
+ "learning_rate": 2.8972518354725977e-07,
1849
+ "loss": 0.457,
1850
+ "step": 263
1851
+ },
1852
+ {
1853
+ "epoch": 0.9539295392953929,
1854
+ "grad_norm": 0.08292391151189804,
1855
+ "learning_rate": 2.4693713663372644e-07,
1856
+ "loss": 0.4105,
1857
+ "step": 264
1858
+ },
1859
+ {
1860
+ "epoch": 0.957542908762421,
1861
+ "grad_norm": 0.07387669384479523,
1862
+ "learning_rate": 2.0755046169392e-07,
1863
+ "loss": 0.3846,
1864
+ "step": 265
1865
+ },
1866
+ {
1867
+ "epoch": 0.9611562782294489,
1868
+ "grad_norm": 0.08278100937604904,
1869
+ "learning_rate": 1.7157057093614703e-07,
1870
+ "loss": 0.4334,
1871
+ "step": 266
1872
+ },
1873
+ {
1874
+ "epoch": 0.964769647696477,
1875
+ "grad_norm": 0.06216645613312721,
1876
+ "learning_rate": 1.3900240843510993e-07,
1877
+ "loss": 0.4007,
1878
+ "step": 267
1879
+ },
1880
+ {
1881
+ "epoch": 0.9683830171635049,
1882
+ "grad_norm": 0.07292906939983368,
1883
+ "learning_rate": 1.0985044945254764e-07,
1884
+ "loss": 0.4152,
1885
+ "step": 268
1886
+ },
1887
+ {
1888
+ "epoch": 0.971996386630533,
1889
+ "grad_norm": 0.07897216826677322,
1890
+ "learning_rate": 8.411869982228038e-08,
1891
+ "loss": 0.3954,
1892
+ "step": 269
1893
+ },
1894
+ {
1895
+ "epoch": 0.975609756097561,
1896
+ "grad_norm": 0.0776594951748848,
1897
+ "learning_rate": 6.181069539974716e-08,
1898
+ "loss": 0.3449,
1899
+ "step": 270
1900
+ },
1901
+ {
1902
+ "epoch": 0.979223125564589,
1903
+ "grad_norm": 0.07104814052581787,
1904
+ "learning_rate": 4.292950157614717e-08,
1905
+ "loss": 0.3476,
1906
+ "step": 271
1907
+ },
1908
+ {
1909
+ "epoch": 0.982836495031617,
1910
+ "grad_norm": 0.07420724630355835,
1911
+ "learning_rate": 2.7477712857215677e-08,
1912
+ "loss": 0.4095,
1913
+ "step": 272
1914
+ },
1915
+ {
1916
+ "epoch": 0.986449864498645,
1917
+ "grad_norm": 0.06806948781013489,
1918
+ "learning_rate": 1.5457452506698056e-08,
1919
+ "loss": 0.3879,
1920
+ "step": 273
1921
+ },
1922
+ {
1923
+ "epoch": 0.990063233965673,
1924
+ "grad_norm": 0.08909036219120026,
1925
+ "learning_rate": 6.870372254602631e-09,
1926
+ "loss": 0.3327,
1927
+ "step": 274
1928
+ },
1929
+ {
1930
+ "epoch": 0.993676603432701,
1931
+ "grad_norm": 0.07509468495845795,
1932
+ "learning_rate": 1.7176520702238964e-09,
1933
+ "loss": 0.4033,
1934
+ "step": 275
1935
+ },
1936
+ {
1937
+ "epoch": 0.997289972899729,
1938
+ "grad_norm": 0.06269805878400803,
1939
+ "learning_rate": 0.0,
1940
+ "loss": 0.4076,
1941
+ "step": 276
1942
+ },
1943
+ {
1944
+ "epoch": 0.997289972899729,
1945
+ "eval_loss": 0.35787180066108704,
1946
+ "eval_runtime": 515.6409,
1947
+ "eval_samples_per_second": 1.422,
1948
+ "eval_steps_per_second": 0.357,
1949
+ "step": 276
1950
+ }
1951
+ ],
1952
+ "logging_steps": 1,
1953
+ "max_steps": 276,
1954
+ "num_input_tokens_seen": 0,
1955
+ "num_train_epochs": 1,
1956
+ "save_steps": 500,
1957
+ "stateful_callbacks": {
1958
+ "TrainerControl": {
1959
+ "args": {
1960
+ "should_epoch_stop": false,
1961
+ "should_evaluate": false,
1962
+ "should_log": false,
1963
+ "should_save": true,
1964
+ "should_training_stop": true
1965
+ },
1966
+ "attributes": {}
1967
+ }
1968
+ },
1969
+ "total_flos": 1.247726843172225e+18,
1970
+ "train_batch_size": 1,
1971
+ "trial_name": null,
1972
+ "trial_params": null
1973
+ }
checkpoint-276/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0711f8a2740f3c9b234d40aa556791e27049ad82916726bec6fd8213bee302a
3
+ size 8312
checkpoint-276/zero_to_fp32.py ADDED
@@ -0,0 +1,760 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example:
14
+ # python zero_to_fp32.py . output_dir/
15
+ # or
16
+ # python zero_to_fp32.py . output_dir/ --safe_serialization
17
+
18
+ import argparse
19
+ import torch
20
+ import glob
21
+ import math
22
+ import os
23
+ import re
24
+ import gc
25
+ import json
26
+ import numpy as np
27
+ from tqdm import tqdm
28
+ from collections import OrderedDict
29
+ from dataclasses import dataclass
30
+
31
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
32
+ # DeepSpeed data structures it has to be available in the current python environment.
33
+ from deepspeed.utils import logger
34
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
35
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
36
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
37
+
38
+
39
+ @dataclass
40
+ class zero_model_state:
41
+ buffers: dict()
42
+ param_shapes: dict()
43
+ shared_params: list
44
+ ds_version: int
45
+ frozen_param_shapes: dict()
46
+ frozen_param_fragments: dict()
47
+
48
+
49
+ debug = 0
50
+
51
+ # load to cpu
52
+ device = torch.device('cpu')
53
+
54
+
55
+ def atoi(text):
56
+ return int(text) if text.isdigit() else text
57
+
58
+
59
+ def natural_keys(text):
60
+ '''
61
+ alist.sort(key=natural_keys) sorts in human order
62
+ http://nedbatchelder.com/blog/200712/human_sorting.html
63
+ (See Toothy's implementation in the comments)
64
+ '''
65
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
66
+
67
+
68
+ def get_model_state_file(checkpoint_dir, zero_stage):
69
+ if not os.path.isdir(checkpoint_dir):
70
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
71
+
72
+ # there should be only one file
73
+ if zero_stage <= 2:
74
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
75
+ elif zero_stage == 3:
76
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
77
+
78
+ if not os.path.exists(file):
79
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
80
+
81
+ return file
82
+
83
+
84
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
85
+ # XXX: need to test that this simple glob rule works for multi-node setup too
86
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
87
+
88
+ if len(ckpt_files) == 0:
89
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
90
+
91
+ return ckpt_files
92
+
93
+
94
+ def get_optim_files(checkpoint_dir):
95
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
96
+
97
+
98
+ def get_model_state_files(checkpoint_dir):
99
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
100
+
101
+
102
+ def parse_model_states(files):
103
+ zero_model_states = []
104
+ for file in files:
105
+ state_dict = torch.load(file, map_location=device, weights_only=False)
106
+
107
+ if BUFFER_NAMES not in state_dict:
108
+ raise ValueError(f"{file} is not a model state checkpoint")
109
+ buffer_names = state_dict[BUFFER_NAMES]
110
+ if debug:
111
+ print("Found buffers:", buffer_names)
112
+
113
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
114
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
115
+ param_shapes = state_dict[PARAM_SHAPES]
116
+
117
+ # collect parameters that are included in param_shapes
118
+ param_names = []
119
+ for s in param_shapes:
120
+ for name in s.keys():
121
+ param_names.append(name)
122
+
123
+ # update with frozen parameters
124
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
125
+ if frozen_param_shapes is not None:
126
+ if debug:
127
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
128
+ param_names += list(frozen_param_shapes.keys())
129
+
130
+ # handle shared params
131
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
132
+
133
+ ds_version = state_dict.get(DS_VERSION, None)
134
+
135
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
136
+
137
+ z_model_state = zero_model_state(buffers=buffers,
138
+ param_shapes=param_shapes,
139
+ shared_params=shared_params,
140
+ ds_version=ds_version,
141
+ frozen_param_shapes=frozen_param_shapes,
142
+ frozen_param_fragments=frozen_param_fragments)
143
+ zero_model_states.append(z_model_state)
144
+
145
+ return zero_model_states
146
+
147
+
148
+ def parse_optim_states(files, ds_checkpoint_dir):
149
+ total_files = len(files)
150
+ state_dicts = []
151
+ for f in tqdm(files, desc='Loading checkpoint shards'):
152
+ state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
153
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
154
+ # and also handle the case where it was already removed by another helper script
155
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
156
+ state_dicts.append(state_dict)
157
+
158
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
159
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
160
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
161
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
162
+
163
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
164
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
165
+ # use the max of the partition_count to get the dp world_size.
166
+
167
+ if type(world_size) is list:
168
+ world_size = max(world_size)
169
+
170
+ if world_size != total_files:
171
+ raise ValueError(
172
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
173
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
174
+ )
175
+
176
+ # the groups are named differently in each stage
177
+ if zero_stage <= 2:
178
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
179
+ elif zero_stage == 3:
180
+ fp32_groups_key = FP32_FLAT_GROUPS
181
+ else:
182
+ raise ValueError(f"unknown zero stage {zero_stage}")
183
+
184
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
185
+ return zero_stage, world_size, fp32_flat_groups
186
+
187
+
188
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
189
+ """
190
+ Returns fp32 state_dict reconstructed from ds checkpoint
191
+
192
+ Args:
193
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
194
+
195
+ """
196
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
197
+
198
+ optim_files = get_optim_files(ds_checkpoint_dir)
199
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
200
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
201
+
202
+ model_files = get_model_state_files(ds_checkpoint_dir)
203
+
204
+ zero_model_states = parse_model_states(model_files)
205
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
206
+
207
+ if zero_stage <= 2:
208
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
209
+ exclude_frozen_parameters)
210
+ elif zero_stage == 3:
211
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
212
+ exclude_frozen_parameters)
213
+
214
+
215
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
216
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
217
+ return
218
+
219
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
220
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
221
+
222
+ if debug:
223
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
224
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
225
+
226
+ wanted_params = len(frozen_param_shapes)
227
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
228
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
229
+ print(f'Frozen params: Have {avail_numel} numels to process.')
230
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
231
+
232
+ total_params = 0
233
+ total_numel = 0
234
+ for name, shape in frozen_param_shapes.items():
235
+ total_params += 1
236
+ unpartitioned_numel = shape.numel()
237
+ total_numel += unpartitioned_numel
238
+
239
+ state_dict[name] = frozen_param_fragments[name]
240
+
241
+ if debug:
242
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
243
+
244
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
245
+
246
+
247
+ def _has_callable(obj, fn):
248
+ attr = getattr(obj, fn, None)
249
+ return callable(attr)
250
+
251
+
252
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
253
+ param_shapes = zero_model_states[0].param_shapes
254
+
255
+ # Reconstruction protocol:
256
+ #
257
+ # XXX: document this
258
+
259
+ if debug:
260
+ for i in range(world_size):
261
+ for j in range(len(fp32_flat_groups[0])):
262
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
263
+
264
+ # XXX: memory usage doubles here (zero2)
265
+ num_param_groups = len(fp32_flat_groups[0])
266
+ merged_single_partition_of_fp32_groups = []
267
+ for i in range(num_param_groups):
268
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
269
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
270
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
271
+ avail_numel = sum(
272
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
273
+
274
+ if debug:
275
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
276
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
277
+ # not asserting if there is a mismatch due to possible padding
278
+ print(f"Have {avail_numel} numels to process.")
279
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
280
+
281
+ # params
282
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
283
+ # out-of-core computing solution
284
+ total_numel = 0
285
+ total_params = 0
286
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
287
+ offset = 0
288
+ avail_numel = full_single_fp32_vector.numel()
289
+ for name, shape in shapes.items():
290
+
291
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
292
+ total_numel += unpartitioned_numel
293
+ total_params += 1
294
+
295
+ if debug:
296
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
297
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
298
+ offset += unpartitioned_numel
299
+
300
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
301
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
302
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
303
+ # live optimizer object, so we are checking that the numbers are within the right range
304
+ align_to = 2 * world_size
305
+
306
+ def zero2_align(x):
307
+ return align_to * math.ceil(x / align_to)
308
+
309
+ if debug:
310
+ print(f"original offset={offset}, avail_numel={avail_numel}")
311
+
312
+ offset = zero2_align(offset)
313
+ avail_numel = zero2_align(avail_numel)
314
+
315
+ if debug:
316
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
317
+
318
+ # Sanity check
319
+ if offset != avail_numel:
320
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
321
+
322
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
323
+
324
+
325
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
326
+ exclude_frozen_parameters):
327
+ state_dict = OrderedDict()
328
+
329
+ # buffers
330
+ buffers = zero_model_states[0].buffers
331
+ state_dict.update(buffers)
332
+ if debug:
333
+ print(f"added {len(buffers)} buffers")
334
+
335
+ if not exclude_frozen_parameters:
336
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
337
+
338
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
339
+
340
+ # recover shared parameters
341
+ for pair in zero_model_states[0].shared_params:
342
+ if pair[1] in state_dict:
343
+ state_dict[pair[0]] = state_dict[pair[1]]
344
+
345
+ return state_dict
346
+
347
+
348
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
349
+ remainder = unpartitioned_numel % world_size
350
+ padding_numel = (world_size - remainder) if remainder else 0
351
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
352
+ return partitioned_numel, padding_numel
353
+
354
+
355
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
356
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
357
+ return
358
+
359
+ if debug:
360
+ for i in range(world_size):
361
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
362
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
363
+
364
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
365
+ wanted_params = len(frozen_param_shapes)
366
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
367
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
368
+ print(f'Frozen params: Have {avail_numel} numels to process.')
369
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
370
+
371
+ total_params = 0
372
+ total_numel = 0
373
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
374
+ total_params += 1
375
+ unpartitioned_numel = shape.numel()
376
+ total_numel += unpartitioned_numel
377
+
378
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
379
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
380
+
381
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
382
+
383
+ if debug:
384
+ print(
385
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
386
+ )
387
+
388
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
389
+
390
+
391
+ class GatheredTensor:
392
+ """
393
+ A pseudo tensor that collects partitioned weights.
394
+ It is more memory efficient when there are multiple groups.
395
+ """
396
+
397
+ def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
398
+ self.flat_groups = flat_groups
399
+ self.flat_groups_offset = flat_groups_offset
400
+ self.offset = offset
401
+ self.partitioned_numel = partitioned_numel
402
+ self.shape = shape
403
+ self.dtype = self.flat_groups[0][0].dtype
404
+
405
+ def contiguous(self):
406
+ """
407
+ Merge partitioned weights from flat_groups into a single tensor.
408
+ """
409
+ end_idx = self.offset + self.partitioned_numel
410
+ world_size = len(self.flat_groups)
411
+ pad_flat_param_chunks = []
412
+
413
+ for rank_i in range(world_size):
414
+ # for each rank, we need to collect weights from related group/groups
415
+ flat_groups_at_rank_i = self.flat_groups[rank_i]
416
+ start_group_id = None
417
+ end_group_id = None
418
+ for group_id in range(len(self.flat_groups_offset)):
419
+ if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
420
+ start_group_id = group_id
421
+ if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
422
+ end_group_id = group_id
423
+ break
424
+ # collect weights from related group/groups
425
+ for group_id in range(start_group_id, end_group_id + 1):
426
+ flat_tensor = flat_groups_at_rank_i[group_id]
427
+ start_offset = self.offset - self.flat_groups_offset[group_id]
428
+ end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
429
+ pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
430
+
431
+ # collect weights from all ranks
432
+ pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
433
+ param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
434
+ return param
435
+
436
+
437
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
438
+ param_shapes = zero_model_states[0].param_shapes
439
+ avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
440
+
441
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
442
+ # param, re-consolidating each param, while dealing with padding if any
443
+
444
+ # merge list of dicts, preserving order
445
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
446
+
447
+ if debug:
448
+ for i in range(world_size):
449
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
450
+
451
+ wanted_params = len(param_shapes)
452
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
453
+ # not asserting if there is a mismatch due to possible padding
454
+ avail_numel = fp32_flat_groups[0].numel() * world_size
455
+ print(f"Trainable params: Have {avail_numel} numels to process.")
456
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
457
+
458
+ # params
459
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
460
+ # out-of-core computing solution
461
+ offset = 0
462
+ total_numel = 0
463
+ total_params = 0
464
+ flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
465
+ for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
466
+ unpartitioned_numel = shape.numel()
467
+ total_numel += unpartitioned_numel
468
+ total_params += 1
469
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
470
+
471
+ if debug:
472
+ print(
473
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
474
+ )
475
+
476
+ # memory efficient tensor
477
+ tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
478
+ state_dict[name] = tensor
479
+ offset += partitioned_numel
480
+
481
+ offset *= world_size
482
+
483
+ # Sanity check
484
+ if offset != avail_numel:
485
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
486
+
487
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
488
+
489
+
490
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
491
+ exclude_frozen_parameters):
492
+ state_dict = OrderedDict()
493
+
494
+ # buffers
495
+ buffers = zero_model_states[0].buffers
496
+ state_dict.update(buffers)
497
+ if debug:
498
+ print(f"added {len(buffers)} buffers")
499
+
500
+ if not exclude_frozen_parameters:
501
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
502
+
503
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
504
+
505
+ # recover shared parameters
506
+ for pair in zero_model_states[0].shared_params:
507
+ if pair[1] in state_dict:
508
+ state_dict[pair[0]] = state_dict[pair[1]]
509
+
510
+ return state_dict
511
+
512
+
513
+ def to_torch_tensor(state_dict, return_empty_tensor=False):
514
+ """
515
+ Convert state_dict of GatheredTensor to torch tensor
516
+ """
517
+ torch_state_dict = {}
518
+ converted_tensors = {}
519
+ for name, tensor in state_dict.items():
520
+ tensor_id = id(tensor)
521
+ if tensor_id in converted_tensors: # shared tensors
522
+ shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
523
+ torch_state_dict[name] = shared_tensor
524
+ else:
525
+ converted_tensors[tensor_id] = name
526
+ if return_empty_tensor:
527
+ torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
528
+ else:
529
+ torch_state_dict[name] = tensor.contiguous()
530
+ return torch_state_dict
531
+
532
+
533
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
534
+ tag=None,
535
+ exclude_frozen_parameters=False,
536
+ lazy_mode=False):
537
+ """
538
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
539
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
540
+ via a model hub.
541
+
542
+ Args:
543
+ - ``checkpoint_dir``: path to the desired checkpoint folder
544
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
545
+ - ``exclude_frozen_parameters``: exclude frozen parameters
546
+ - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
547
+ Convert the pesduo tensor to torch tensor by ``.contiguous()``
548
+
549
+ Returns:
550
+ - pytorch ``state_dict``
551
+
552
+ A typical usage might be ::
553
+
554
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
555
+ # do the training and checkpoint saving
556
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
557
+ model = model.cpu() # move to cpu
558
+ model.load_state_dict(state_dict)
559
+ # submit to model hub or save the model to share with others
560
+
561
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
562
+ application. i.e. you will need to re-initialize the deepspeed engine, since
563
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
564
+
565
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
566
+
567
+ Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
568
+ You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
569
+ the checkpoint. Or you can load state_dict in lazy mode ::
570
+
571
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
572
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
573
+ for name, lazy_tensor in state_dict.item():
574
+ tensor = lazy_tensor.contiguous() # to cpu
575
+ print(name, tensor)
576
+ # del tensor to release memory if it no longer in use
577
+ """
578
+ if tag is None:
579
+ latest_path = os.path.join(checkpoint_dir, 'latest')
580
+ if os.path.isfile(latest_path):
581
+ with open(latest_path, 'r') as fd:
582
+ tag = fd.read().strip()
583
+ else:
584
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
585
+
586
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
587
+
588
+ if not os.path.isdir(ds_checkpoint_dir):
589
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
590
+
591
+ state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
592
+ if lazy_mode:
593
+ return state_dict
594
+ else:
595
+ return to_torch_tensor(state_dict)
596
+
597
+
598
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
599
+ output_dir,
600
+ max_shard_size="5GB",
601
+ safe_serialization=False,
602
+ tag=None,
603
+ exclude_frozen_parameters=False):
604
+ """
605
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
606
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
607
+
608
+ Args:
609
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
610
+ - ``output_dir``: directory to the pytorch fp32 state_dict output files
611
+ - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
612
+ - ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
613
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
614
+ - ``exclude_frozen_parameters``: exclude frozen parameters
615
+ """
616
+
617
+ # Dependency pre-check
618
+ if safe_serialization:
619
+ try:
620
+ from safetensors.torch import save_file
621
+ except ImportError:
622
+ print('If you want to use `safe_serialization`, please `pip install safetensors`')
623
+ raise
624
+ if max_shard_size is not None:
625
+ try:
626
+ from huggingface_hub import split_torch_state_dict_into_shards
627
+ except ImportError:
628
+ print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
629
+ raise
630
+
631
+ # Convert zero checkpoint to state_dict
632
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
633
+ tag,
634
+ exclude_frozen_parameters,
635
+ lazy_mode=True)
636
+
637
+ # Shard the model if it is too big.
638
+ weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
639
+ if max_shard_size is not None:
640
+ filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
641
+ # an memory-efficient approach for sharding
642
+ empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
643
+ state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
644
+ filename_pattern=filename_pattern,
645
+ max_shard_size=max_shard_size)
646
+ else:
647
+ from collections import namedtuple
648
+ StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
649
+ state_dict_split = StateDictSplit(is_sharded=False,
650
+ filename_to_tensors={weights_name: list(state_dict.keys())})
651
+
652
+ # Save the model by shard
653
+ os.makedirs(output_dir, exist_ok=True)
654
+ filename_to_tensors = state_dict_split.filename_to_tensors.items()
655
+ for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
656
+ shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
657
+ shard_state_dict = to_torch_tensor(shard_state_dict)
658
+ output_path = os.path.join(output_dir, shard_file)
659
+ if safe_serialization:
660
+ save_file(shard_state_dict, output_path, metadata={"format": "pt"})
661
+ else:
662
+ torch.save(shard_state_dict, output_path)
663
+ # release the memory of current shard
664
+ for tensor_name in list(shard_state_dict.keys()):
665
+ del state_dict[tensor_name]
666
+ del shard_state_dict[tensor_name]
667
+ del shard_state_dict
668
+ gc.collect()
669
+
670
+ # Save index if sharded
671
+ if state_dict_split.is_sharded:
672
+ index = {
673
+ "metadata": state_dict_split.metadata,
674
+ "weight_map": state_dict_split.tensor_to_filename,
675
+ }
676
+ save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
677
+ save_index_file = os.path.join(output_dir, save_index_file)
678
+ with open(save_index_file, "w", encoding="utf-8") as f:
679
+ content = json.dumps(index, indent=2, sort_keys=True) + "\n"
680
+ f.write(content)
681
+
682
+
683
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
684
+ """
685
+ 1. Put the provided model to cpu
686
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
687
+ 3. Load it into the provided model
688
+
689
+ Args:
690
+ - ``model``: the model object to update
691
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
692
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
693
+
694
+ Returns:
695
+ - ``model`: modified model
696
+
697
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
698
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
699
+ conveniently placed for you in the checkpoint folder.
700
+
701
+ A typical usage might be ::
702
+
703
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
704
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
705
+ # submit to model hub or save the model to share with others
706
+
707
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
708
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
709
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
710
+
711
+ """
712
+ logger.info(f"Extracting fp32 weights")
713
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
714
+
715
+ logger.info(f"Overwriting model with fp32 weights")
716
+ model = model.cpu()
717
+ model.load_state_dict(state_dict, strict=False)
718
+
719
+ return model
720
+
721
+
722
+ if __name__ == "__main__":
723
+ parser = argparse.ArgumentParser()
724
+ parser.add_argument("checkpoint_dir",
725
+ type=str,
726
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
727
+ parser.add_argument("output_dir",
728
+ type=str,
729
+ help="directory to the pytorch fp32 state_dict output files"
730
+ "(e.g. path/checkpoint-12-output/)")
731
+ parser.add_argument(
732
+ "--max_shard_size",
733
+ type=str,
734
+ default="5GB",
735
+ help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
736
+ "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
737
+ "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
738
+ "without CPU OOM issues.")
739
+ parser.add_argument(
740
+ "--safe_serialization",
741
+ default=False,
742
+ action='store_true',
743
+ help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
744
+ parser.add_argument("-t",
745
+ "--tag",
746
+ type=str,
747
+ default=None,
748
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
749
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
750
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
751
+ args = parser.parse_args()
752
+
753
+ debug = args.debug
754
+
755
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
756
+ args.output_dir,
757
+ max_shard_size=args.max_shard_size,
758
+ safe_serialization=args.safe_serialization,
759
+ tag=args.tag,
760
+ exclude_frozen_parameters=args.exclude_frozen_parameters)
config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": true,
3
+ "_name_or_path": "/kaggle/input/mistral-small-24b/transformers/mistral-small-24b-instruct-2501/2",
4
+ "architectures": [
5
+ "MistralForCausalLM"
6
+ ],
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "eos_token_id": 2,
10
+ "head_dim": 128,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 5120,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 32768,
15
+ "max_position_embeddings": 32768,
16
+ "model_type": "mistral",
17
+ "num_attention_heads": 32,
18
+ "num_hidden_layers": 40,
19
+ "num_key_value_heads": 8,
20
+ "quantization_config": {
21
+ "_load_in_4bit": true,
22
+ "_load_in_8bit": false,
23
+ "bnb_4bit_compute_dtype": "bfloat16",
24
+ "bnb_4bit_quant_storage": "bfloat16",
25
+ "bnb_4bit_quant_type": "nf4",
26
+ "bnb_4bit_use_double_quant": true,
27
+ "llm_int8_enable_fp32_cpu_offload": false,
28
+ "llm_int8_has_fp16_weight": false,
29
+ "llm_int8_skip_modules": null,
30
+ "llm_int8_threshold": 6.0,
31
+ "load_in_4bit": true,
32
+ "load_in_8bit": false,
33
+ "quant_method": "bitsandbytes"
34
+ },
35
+ "rms_norm_eps": 1e-05,
36
+ "rope_theta": 100000000.0,
37
+ "sliding_window": null,
38
+ "tie_word_embeddings": false,
39
+ "torch_dtype": "bfloat16",
40
+ "transformers_version": "4.48.3",
41
+ "use_cache": false,
42
+ "vocab_size": 131072
43
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,1026 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<unk>",
4
+ "<s>",
5
+ "</s>",
6
+ "[INST]",
7
+ "[/INST]",
8
+ "[AVAILABLE_TOOLS]",
9
+ "[/AVAILABLE_TOOLS]",
10
+ "[TOOL_RESULTS]",
11
+ "[/TOOL_RESULTS]",
12
+ "[TOOL_CALLS]",
13
+ "[IMG]",
14
+ "<pad>",
15
+ "[IMG_BREAK]",
16
+ "[IMG_END]",
17
+ "[PREFIX]",
18
+ "[MIDDLE]",
19
+ "[SUFFIX]",
20
+ "[SYSTEM_PROMPT]",
21
+ "[/SYSTEM_PROMPT]",
22
+ "[TOOL_CONTENT]",
23
+ "<SPECIAL_20>",
24
+ "<SPECIAL_21>",
25
+ "<SPECIAL_22>",
26
+ "<SPECIAL_23>",
27
+ "<SPECIAL_24>",
28
+ "<SPECIAL_25>",
29
+ "<SPECIAL_26>",
30
+ "<SPECIAL_27>",
31
+ "<SPECIAL_28>",
32
+ "<SPECIAL_29>",
33
+ "<SPECIAL_30>",
34
+ "<SPECIAL_31>",
35
+ "<SPECIAL_32>",
36
+ "<SPECIAL_33>",
37
+ "<SPECIAL_34>",
38
+ "<SPECIAL_35>",
39
+ "<SPECIAL_36>",
40
+ "<SPECIAL_37>",
41
+ "<SPECIAL_38>",
42
+ "<SPECIAL_39>",
43
+ "<SPECIAL_40>",
44
+ "<SPECIAL_41>",
45
+ "<SPECIAL_42>",
46
+ "<SPECIAL_43>",
47
+ "<SPECIAL_44>",
48
+ "<SPECIAL_45>",
49
+ "<SPECIAL_46>",
50
+ "<SPECIAL_47>",
51
+ "<SPECIAL_48>",
52
+ "<SPECIAL_49>",
53
+ "<SPECIAL_50>",
54
+ "<SPECIAL_51>",
55
+ "<SPECIAL_52>",
56
+ "<SPECIAL_53>",
57
+ "<SPECIAL_54>",
58
+ "<SPECIAL_55>",
59
+ "<SPECIAL_56>",
60
+ "<SPECIAL_57>",
61
+ "<SPECIAL_58>",
62
+ "<SPECIAL_59>",
63
+ "<SPECIAL_60>",
64
+ "<SPECIAL_61>",
65
+ "<SPECIAL_62>",
66
+ "<SPECIAL_63>",
67
+ "<SPECIAL_64>",
68
+ "<SPECIAL_65>",
69
+ "<SPECIAL_66>",
70
+ "<SPECIAL_67>",
71
+ "<SPECIAL_68>",
72
+ "<SPECIAL_69>",
73
+ "<SPECIAL_70>",
74
+ "<SPECIAL_71>",
75
+ "<SPECIAL_72>",
76
+ "<SPECIAL_73>",
77
+ "<SPECIAL_74>",
78
+ "<SPECIAL_75>",
79
+ "<SPECIAL_76>",
80
+ "<SPECIAL_77>",
81
+ "<SPECIAL_78>",
82
+ "<SPECIAL_79>",
83
+ "<SPECIAL_80>",
84
+ "<SPECIAL_81>",
85
+ "<SPECIAL_82>",
86
+ "<SPECIAL_83>",
87
+ "<SPECIAL_84>",
88
+ "<SPECIAL_85>",
89
+ "<SPECIAL_86>",
90
+ "<SPECIAL_87>",
91
+ "<SPECIAL_88>",
92
+ "<SPECIAL_89>",
93
+ "<SPECIAL_90>",
94
+ "<SPECIAL_91>",
95
+ "<SPECIAL_92>",
96
+ "<SPECIAL_93>",
97
+ "<SPECIAL_94>",
98
+ "<SPECIAL_95>",
99
+ "<SPECIAL_96>",
100
+ "<SPECIAL_97>",
101
+ "<SPECIAL_98>",
102
+ "<SPECIAL_99>",
103
+ "<SPECIAL_100>",
104
+ "<SPECIAL_101>",
105
+ "<SPECIAL_102>",
106
+ "<SPECIAL_103>",
107
+ "<SPECIAL_104>",
108
+ "<SPECIAL_105>",
109
+ "<SPECIAL_106>",
110
+ "<SPECIAL_107>",
111
+ "<SPECIAL_108>",
112
+ "<SPECIAL_109>",
113
+ "<SPECIAL_110>",
114
+ "<SPECIAL_111>",
115
+ "<SPECIAL_112>",
116
+ "<SPECIAL_113>",
117
+ "<SPECIAL_114>",
118
+ "<SPECIAL_115>",
119
+ "<SPECIAL_116>",
120
+ "<SPECIAL_117>",
121
+ "<SPECIAL_118>",
122
+ "<SPECIAL_119>",
123
+ "<SPECIAL_120>",
124
+ "<SPECIAL_121>",
125
+ "<SPECIAL_122>",
126
+ "<SPECIAL_123>",
127
+ "<SPECIAL_124>",
128
+ "<SPECIAL_125>",
129
+ "<SPECIAL_126>",
130
+ "<SPECIAL_127>",
131
+ "<SPECIAL_128>",
132
+ "<SPECIAL_129>",
133
+ "<SPECIAL_130>",
134
+ "<SPECIAL_131>",
135
+ "<SPECIAL_132>",
136
+ "<SPECIAL_133>",
137
+ "<SPECIAL_134>",
138
+ "<SPECIAL_135>",
139
+ "<SPECIAL_136>",
140
+ "<SPECIAL_137>",
141
+ "<SPECIAL_138>",
142
+ "<SPECIAL_139>",
143
+ "<SPECIAL_140>",
144
+ "<SPECIAL_141>",
145
+ "<SPECIAL_142>",
146
+ "<SPECIAL_143>",
147
+ "<SPECIAL_144>",
148
+ "<SPECIAL_145>",
149
+ "<SPECIAL_146>",
150
+ "<SPECIAL_147>",
151
+ "<SPECIAL_148>",
152
+ "<SPECIAL_149>",
153
+ "<SPECIAL_150>",
154
+ "<SPECIAL_151>",
155
+ "<SPECIAL_152>",
156
+ "<SPECIAL_153>",
157
+ "<SPECIAL_154>",
158
+ "<SPECIAL_155>",
159
+ "<SPECIAL_156>",
160
+ "<SPECIAL_157>",
161
+ "<SPECIAL_158>",
162
+ "<SPECIAL_159>",
163
+ "<SPECIAL_160>",
164
+ "<SPECIAL_161>",
165
+ "<SPECIAL_162>",
166
+ "<SPECIAL_163>",
167
+ "<SPECIAL_164>",
168
+ "<SPECIAL_165>",
169
+ "<SPECIAL_166>",
170
+ "<SPECIAL_167>",
171
+ "<SPECIAL_168>",
172
+ "<SPECIAL_169>",
173
+ "<SPECIAL_170>",
174
+ "<SPECIAL_171>",
175
+ "<SPECIAL_172>",
176
+ "<SPECIAL_173>",
177
+ "<SPECIAL_174>",
178
+ "<SPECIAL_175>",
179
+ "<SPECIAL_176>",
180
+ "<SPECIAL_177>",
181
+ "<SPECIAL_178>",
182
+ "<SPECIAL_179>",
183
+ "<SPECIAL_180>",
184
+ "<SPECIAL_181>",
185
+ "<SPECIAL_182>",
186
+ "<SPECIAL_183>",
187
+ "<SPECIAL_184>",
188
+ "<SPECIAL_185>",
189
+ "<SPECIAL_186>",
190
+ "<SPECIAL_187>",
191
+ "<SPECIAL_188>",
192
+ "<SPECIAL_189>",
193
+ "<SPECIAL_190>",
194
+ "<SPECIAL_191>",
195
+ "<SPECIAL_192>",
196
+ "<SPECIAL_193>",
197
+ "<SPECIAL_194>",
198
+ "<SPECIAL_195>",
199
+ "<SPECIAL_196>",
200
+ "<SPECIAL_197>",
201
+ "<SPECIAL_198>",
202
+ "<SPECIAL_199>",
203
+ "<SPECIAL_200>",
204
+ "<SPECIAL_201>",
205
+ "<SPECIAL_202>",
206
+ "<SPECIAL_203>",
207
+ "<SPECIAL_204>",
208
+ "<SPECIAL_205>",
209
+ "<SPECIAL_206>",
210
+ "<SPECIAL_207>",
211
+ "<SPECIAL_208>",
212
+ "<SPECIAL_209>",
213
+ "<SPECIAL_210>",
214
+ "<SPECIAL_211>",
215
+ "<SPECIAL_212>",
216
+ "<SPECIAL_213>",
217
+ "<SPECIAL_214>",
218
+ "<SPECIAL_215>",
219
+ "<SPECIAL_216>",
220
+ "<SPECIAL_217>",
221
+ "<SPECIAL_218>",
222
+ "<SPECIAL_219>",
223
+ "<SPECIAL_220>",
224
+ "<SPECIAL_221>",
225
+ "<SPECIAL_222>",
226
+ "<SPECIAL_223>",
227
+ "<SPECIAL_224>",
228
+ "<SPECIAL_225>",
229
+ "<SPECIAL_226>",
230
+ "<SPECIAL_227>",
231
+ "<SPECIAL_228>",
232
+ "<SPECIAL_229>",
233
+ "<SPECIAL_230>",
234
+ "<SPECIAL_231>",
235
+ "<SPECIAL_232>",
236
+ "<SPECIAL_233>",
237
+ "<SPECIAL_234>",
238
+ "<SPECIAL_235>",
239
+ "<SPECIAL_236>",
240
+ "<SPECIAL_237>",
241
+ "<SPECIAL_238>",
242
+ "<SPECIAL_239>",
243
+ "<SPECIAL_240>",
244
+ "<SPECIAL_241>",
245
+ "<SPECIAL_242>",
246
+ "<SPECIAL_243>",
247
+ "<SPECIAL_244>",
248
+ "<SPECIAL_245>",
249
+ "<SPECIAL_246>",
250
+ "<SPECIAL_247>",
251
+ "<SPECIAL_248>",
252
+ "<SPECIAL_249>",
253
+ "<SPECIAL_250>",
254
+ "<SPECIAL_251>",
255
+ "<SPECIAL_252>",
256
+ "<SPECIAL_253>",
257
+ "<SPECIAL_254>",
258
+ "<SPECIAL_255>",
259
+ "<SPECIAL_256>",
260
+ "<SPECIAL_257>",
261
+ "<SPECIAL_258>",
262
+ "<SPECIAL_259>",
263
+ "<SPECIAL_260>",
264
+ "<SPECIAL_261>",
265
+ "<SPECIAL_262>",
266
+ "<SPECIAL_263>",
267
+ "<SPECIAL_264>",
268
+ "<SPECIAL_265>",
269
+ "<SPECIAL_266>",
270
+ "<SPECIAL_267>",
271
+ "<SPECIAL_268>",
272
+ "<SPECIAL_269>",
273
+ "<SPECIAL_270>",
274
+ "<SPECIAL_271>",
275
+ "<SPECIAL_272>",
276
+ "<SPECIAL_273>",
277
+ "<SPECIAL_274>",
278
+ "<SPECIAL_275>",
279
+ "<SPECIAL_276>",
280
+ "<SPECIAL_277>",
281
+ "<SPECIAL_278>",
282
+ "<SPECIAL_279>",
283
+ "<SPECIAL_280>",
284
+ "<SPECIAL_281>",
285
+ "<SPECIAL_282>",
286
+ "<SPECIAL_283>",
287
+ "<SPECIAL_284>",
288
+ "<SPECIAL_285>",
289
+ "<SPECIAL_286>",
290
+ "<SPECIAL_287>",
291
+ "<SPECIAL_288>",
292
+ "<SPECIAL_289>",
293
+ "<SPECIAL_290>",
294
+ "<SPECIAL_291>",
295
+ "<SPECIAL_292>",
296
+ "<SPECIAL_293>",
297
+ "<SPECIAL_294>",
298
+ "<SPECIAL_295>",
299
+ "<SPECIAL_296>",
300
+ "<SPECIAL_297>",
301
+ "<SPECIAL_298>",
302
+ "<SPECIAL_299>",
303
+ "<SPECIAL_300>",
304
+ "<SPECIAL_301>",
305
+ "<SPECIAL_302>",
306
+ "<SPECIAL_303>",
307
+ "<SPECIAL_304>",
308
+ "<SPECIAL_305>",
309
+ "<SPECIAL_306>",
310
+ "<SPECIAL_307>",
311
+ "<SPECIAL_308>",
312
+ "<SPECIAL_309>",
313
+ "<SPECIAL_310>",
314
+ "<SPECIAL_311>",
315
+ "<SPECIAL_312>",
316
+ "<SPECIAL_313>",
317
+ "<SPECIAL_314>",
318
+ "<SPECIAL_315>",
319
+ "<SPECIAL_316>",
320
+ "<SPECIAL_317>",
321
+ "<SPECIAL_318>",
322
+ "<SPECIAL_319>",
323
+ "<SPECIAL_320>",
324
+ "<SPECIAL_321>",
325
+ "<SPECIAL_322>",
326
+ "<SPECIAL_323>",
327
+ "<SPECIAL_324>",
328
+ "<SPECIAL_325>",
329
+ "<SPECIAL_326>",
330
+ "<SPECIAL_327>",
331
+ "<SPECIAL_328>",
332
+ "<SPECIAL_329>",
333
+ "<SPECIAL_330>",
334
+ "<SPECIAL_331>",
335
+ "<SPECIAL_332>",
336
+ "<SPECIAL_333>",
337
+ "<SPECIAL_334>",
338
+ "<SPECIAL_335>",
339
+ "<SPECIAL_336>",
340
+ "<SPECIAL_337>",
341
+ "<SPECIAL_338>",
342
+ "<SPECIAL_339>",
343
+ "<SPECIAL_340>",
344
+ "<SPECIAL_341>",
345
+ "<SPECIAL_342>",
346
+ "<SPECIAL_343>",
347
+ "<SPECIAL_344>",
348
+ "<SPECIAL_345>",
349
+ "<SPECIAL_346>",
350
+ "<SPECIAL_347>",
351
+ "<SPECIAL_348>",
352
+ "<SPECIAL_349>",
353
+ "<SPECIAL_350>",
354
+ "<SPECIAL_351>",
355
+ "<SPECIAL_352>",
356
+ "<SPECIAL_353>",
357
+ "<SPECIAL_354>",
358
+ "<SPECIAL_355>",
359
+ "<SPECIAL_356>",
360
+ "<SPECIAL_357>",
361
+ "<SPECIAL_358>",
362
+ "<SPECIAL_359>",
363
+ "<SPECIAL_360>",
364
+ "<SPECIAL_361>",
365
+ "<SPECIAL_362>",
366
+ "<SPECIAL_363>",
367
+ "<SPECIAL_364>",
368
+ "<SPECIAL_365>",
369
+ "<SPECIAL_366>",
370
+ "<SPECIAL_367>",
371
+ "<SPECIAL_368>",
372
+ "<SPECIAL_369>",
373
+ "<SPECIAL_370>",
374
+ "<SPECIAL_371>",
375
+ "<SPECIAL_372>",
376
+ "<SPECIAL_373>",
377
+ "<SPECIAL_374>",
378
+ "<SPECIAL_375>",
379
+ "<SPECIAL_376>",
380
+ "<SPECIAL_377>",
381
+ "<SPECIAL_378>",
382
+ "<SPECIAL_379>",
383
+ "<SPECIAL_380>",
384
+ "<SPECIAL_381>",
385
+ "<SPECIAL_382>",
386
+ "<SPECIAL_383>",
387
+ "<SPECIAL_384>",
388
+ "<SPECIAL_385>",
389
+ "<SPECIAL_386>",
390
+ "<SPECIAL_387>",
391
+ "<SPECIAL_388>",
392
+ "<SPECIAL_389>",
393
+ "<SPECIAL_390>",
394
+ "<SPECIAL_391>",
395
+ "<SPECIAL_392>",
396
+ "<SPECIAL_393>",
397
+ "<SPECIAL_394>",
398
+ "<SPECIAL_395>",
399
+ "<SPECIAL_396>",
400
+ "<SPECIAL_397>",
401
+ "<SPECIAL_398>",
402
+ "<SPECIAL_399>",
403
+ "<SPECIAL_400>",
404
+ "<SPECIAL_401>",
405
+ "<SPECIAL_402>",
406
+ "<SPECIAL_403>",
407
+ "<SPECIAL_404>",
408
+ "<SPECIAL_405>",
409
+ "<SPECIAL_406>",
410
+ "<SPECIAL_407>",
411
+ "<SPECIAL_408>",
412
+ "<SPECIAL_409>",
413
+ "<SPECIAL_410>",
414
+ "<SPECIAL_411>",
415
+ "<SPECIAL_412>",
416
+ "<SPECIAL_413>",
417
+ "<SPECIAL_414>",
418
+ "<SPECIAL_415>",
419
+ "<SPECIAL_416>",
420
+ "<SPECIAL_417>",
421
+ "<SPECIAL_418>",
422
+ "<SPECIAL_419>",
423
+ "<SPECIAL_420>",
424
+ "<SPECIAL_421>",
425
+ "<SPECIAL_422>",
426
+ "<SPECIAL_423>",
427
+ "<SPECIAL_424>",
428
+ "<SPECIAL_425>",
429
+ "<SPECIAL_426>",
430
+ "<SPECIAL_427>",
431
+ "<SPECIAL_428>",
432
+ "<SPECIAL_429>",
433
+ "<SPECIAL_430>",
434
+ "<SPECIAL_431>",
435
+ "<SPECIAL_432>",
436
+ "<SPECIAL_433>",
437
+ "<SPECIAL_434>",
438
+ "<SPECIAL_435>",
439
+ "<SPECIAL_436>",
440
+ "<SPECIAL_437>",
441
+ "<SPECIAL_438>",
442
+ "<SPECIAL_439>",
443
+ "<SPECIAL_440>",
444
+ "<SPECIAL_441>",
445
+ "<SPECIAL_442>",
446
+ "<SPECIAL_443>",
447
+ "<SPECIAL_444>",
448
+ "<SPECIAL_445>",
449
+ "<SPECIAL_446>",
450
+ "<SPECIAL_447>",
451
+ "<SPECIAL_448>",
452
+ "<SPECIAL_449>",
453
+ "<SPECIAL_450>",
454
+ "<SPECIAL_451>",
455
+ "<SPECIAL_452>",
456
+ "<SPECIAL_453>",
457
+ "<SPECIAL_454>",
458
+ "<SPECIAL_455>",
459
+ "<SPECIAL_456>",
460
+ "<SPECIAL_457>",
461
+ "<SPECIAL_458>",
462
+ "<SPECIAL_459>",
463
+ "<SPECIAL_460>",
464
+ "<SPECIAL_461>",
465
+ "<SPECIAL_462>",
466
+ "<SPECIAL_463>",
467
+ "<SPECIAL_464>",
468
+ "<SPECIAL_465>",
469
+ "<SPECIAL_466>",
470
+ "<SPECIAL_467>",
471
+ "<SPECIAL_468>",
472
+ "<SPECIAL_469>",
473
+ "<SPECIAL_470>",
474
+ "<SPECIAL_471>",
475
+ "<SPECIAL_472>",
476
+ "<SPECIAL_473>",
477
+ "<SPECIAL_474>",
478
+ "<SPECIAL_475>",
479
+ "<SPECIAL_476>",
480
+ "<SPECIAL_477>",
481
+ "<SPECIAL_478>",
482
+ "<SPECIAL_479>",
483
+ "<SPECIAL_480>",
484
+ "<SPECIAL_481>",
485
+ "<SPECIAL_482>",
486
+ "<SPECIAL_483>",
487
+ "<SPECIAL_484>",
488
+ "<SPECIAL_485>",
489
+ "<SPECIAL_486>",
490
+ "<SPECIAL_487>",
491
+ "<SPECIAL_488>",
492
+ "<SPECIAL_489>",
493
+ "<SPECIAL_490>",
494
+ "<SPECIAL_491>",
495
+ "<SPECIAL_492>",
496
+ "<SPECIAL_493>",
497
+ "<SPECIAL_494>",
498
+ "<SPECIAL_495>",
499
+ "<SPECIAL_496>",
500
+ "<SPECIAL_497>",
501
+ "<SPECIAL_498>",
502
+ "<SPECIAL_499>",
503
+ "<SPECIAL_500>",
504
+ "<SPECIAL_501>",
505
+ "<SPECIAL_502>",
506
+ "<SPECIAL_503>",
507
+ "<SPECIAL_504>",
508
+ "<SPECIAL_505>",
509
+ "<SPECIAL_506>",
510
+ "<SPECIAL_507>",
511
+ "<SPECIAL_508>",
512
+ "<SPECIAL_509>",
513
+ "<SPECIAL_510>",
514
+ "<SPECIAL_511>",
515
+ "<SPECIAL_512>",
516
+ "<SPECIAL_513>",
517
+ "<SPECIAL_514>",
518
+ "<SPECIAL_515>",
519
+ "<SPECIAL_516>",
520
+ "<SPECIAL_517>",
521
+ "<SPECIAL_518>",
522
+ "<SPECIAL_519>",
523
+ "<SPECIAL_520>",
524
+ "<SPECIAL_521>",
525
+ "<SPECIAL_522>",
526
+ "<SPECIAL_523>",
527
+ "<SPECIAL_524>",
528
+ "<SPECIAL_525>",
529
+ "<SPECIAL_526>",
530
+ "<SPECIAL_527>",
531
+ "<SPECIAL_528>",
532
+ "<SPECIAL_529>",
533
+ "<SPECIAL_530>",
534
+ "<SPECIAL_531>",
535
+ "<SPECIAL_532>",
536
+ "<SPECIAL_533>",
537
+ "<SPECIAL_534>",
538
+ "<SPECIAL_535>",
539
+ "<SPECIAL_536>",
540
+ "<SPECIAL_537>",
541
+ "<SPECIAL_538>",
542
+ "<SPECIAL_539>",
543
+ "<SPECIAL_540>",
544
+ "<SPECIAL_541>",
545
+ "<SPECIAL_542>",
546
+ "<SPECIAL_543>",
547
+ "<SPECIAL_544>",
548
+ "<SPECIAL_545>",
549
+ "<SPECIAL_546>",
550
+ "<SPECIAL_547>",
551
+ "<SPECIAL_548>",
552
+ "<SPECIAL_549>",
553
+ "<SPECIAL_550>",
554
+ "<SPECIAL_551>",
555
+ "<SPECIAL_552>",
556
+ "<SPECIAL_553>",
557
+ "<SPECIAL_554>",
558
+ "<SPECIAL_555>",
559
+ "<SPECIAL_556>",
560
+ "<SPECIAL_557>",
561
+ "<SPECIAL_558>",
562
+ "<SPECIAL_559>",
563
+ "<SPECIAL_560>",
564
+ "<SPECIAL_561>",
565
+ "<SPECIAL_562>",
566
+ "<SPECIAL_563>",
567
+ "<SPECIAL_564>",
568
+ "<SPECIAL_565>",
569
+ "<SPECIAL_566>",
570
+ "<SPECIAL_567>",
571
+ "<SPECIAL_568>",
572
+ "<SPECIAL_569>",
573
+ "<SPECIAL_570>",
574
+ "<SPECIAL_571>",
575
+ "<SPECIAL_572>",
576
+ "<SPECIAL_573>",
577
+ "<SPECIAL_574>",
578
+ "<SPECIAL_575>",
579
+ "<SPECIAL_576>",
580
+ "<SPECIAL_577>",
581
+ "<SPECIAL_578>",
582
+ "<SPECIAL_579>",
583
+ "<SPECIAL_580>",
584
+ "<SPECIAL_581>",
585
+ "<SPECIAL_582>",
586
+ "<SPECIAL_583>",
587
+ "<SPECIAL_584>",
588
+ "<SPECIAL_585>",
589
+ "<SPECIAL_586>",
590
+ "<SPECIAL_587>",
591
+ "<SPECIAL_588>",
592
+ "<SPECIAL_589>",
593
+ "<SPECIAL_590>",
594
+ "<SPECIAL_591>",
595
+ "<SPECIAL_592>",
596
+ "<SPECIAL_593>",
597
+ "<SPECIAL_594>",
598
+ "<SPECIAL_595>",
599
+ "<SPECIAL_596>",
600
+ "<SPECIAL_597>",
601
+ "<SPECIAL_598>",
602
+ "<SPECIAL_599>",
603
+ "<SPECIAL_600>",
604
+ "<SPECIAL_601>",
605
+ "<SPECIAL_602>",
606
+ "<SPECIAL_603>",
607
+ "<SPECIAL_604>",
608
+ "<SPECIAL_605>",
609
+ "<SPECIAL_606>",
610
+ "<SPECIAL_607>",
611
+ "<SPECIAL_608>",
612
+ "<SPECIAL_609>",
613
+ "<SPECIAL_610>",
614
+ "<SPECIAL_611>",
615
+ "<SPECIAL_612>",
616
+ "<SPECIAL_613>",
617
+ "<SPECIAL_614>",
618
+ "<SPECIAL_615>",
619
+ "<SPECIAL_616>",
620
+ "<SPECIAL_617>",
621
+ "<SPECIAL_618>",
622
+ "<SPECIAL_619>",
623
+ "<SPECIAL_620>",
624
+ "<SPECIAL_621>",
625
+ "<SPECIAL_622>",
626
+ "<SPECIAL_623>",
627
+ "<SPECIAL_624>",
628
+ "<SPECIAL_625>",
629
+ "<SPECIAL_626>",
630
+ "<SPECIAL_627>",
631
+ "<SPECIAL_628>",
632
+ "<SPECIAL_629>",
633
+ "<SPECIAL_630>",
634
+ "<SPECIAL_631>",
635
+ "<SPECIAL_632>",
636
+ "<SPECIAL_633>",
637
+ "<SPECIAL_634>",
638
+ "<SPECIAL_635>",
639
+ "<SPECIAL_636>",
640
+ "<SPECIAL_637>",
641
+ "<SPECIAL_638>",
642
+ "<SPECIAL_639>",
643
+ "<SPECIAL_640>",
644
+ "<SPECIAL_641>",
645
+ "<SPECIAL_642>",
646
+ "<SPECIAL_643>",
647
+ "<SPECIAL_644>",
648
+ "<SPECIAL_645>",
649
+ "<SPECIAL_646>",
650
+ "<SPECIAL_647>",
651
+ "<SPECIAL_648>",
652
+ "<SPECIAL_649>",
653
+ "<SPECIAL_650>",
654
+ "<SPECIAL_651>",
655
+ "<SPECIAL_652>",
656
+ "<SPECIAL_653>",
657
+ "<SPECIAL_654>",
658
+ "<SPECIAL_655>",
659
+ "<SPECIAL_656>",
660
+ "<SPECIAL_657>",
661
+ "<SPECIAL_658>",
662
+ "<SPECIAL_659>",
663
+ "<SPECIAL_660>",
664
+ "<SPECIAL_661>",
665
+ "<SPECIAL_662>",
666
+ "<SPECIAL_663>",
667
+ "<SPECIAL_664>",
668
+ "<SPECIAL_665>",
669
+ "<SPECIAL_666>",
670
+ "<SPECIAL_667>",
671
+ "<SPECIAL_668>",
672
+ "<SPECIAL_669>",
673
+ "<SPECIAL_670>",
674
+ "<SPECIAL_671>",
675
+ "<SPECIAL_672>",
676
+ "<SPECIAL_673>",
677
+ "<SPECIAL_674>",
678
+ "<SPECIAL_675>",
679
+ "<SPECIAL_676>",
680
+ "<SPECIAL_677>",
681
+ "<SPECIAL_678>",
682
+ "<SPECIAL_679>",
683
+ "<SPECIAL_680>",
684
+ "<SPECIAL_681>",
685
+ "<SPECIAL_682>",
686
+ "<SPECIAL_683>",
687
+ "<SPECIAL_684>",
688
+ "<SPECIAL_685>",
689
+ "<SPECIAL_686>",
690
+ "<SPECIAL_687>",
691
+ "<SPECIAL_688>",
692
+ "<SPECIAL_689>",
693
+ "<SPECIAL_690>",
694
+ "<SPECIAL_691>",
695
+ "<SPECIAL_692>",
696
+ "<SPECIAL_693>",
697
+ "<SPECIAL_694>",
698
+ "<SPECIAL_695>",
699
+ "<SPECIAL_696>",
700
+ "<SPECIAL_697>",
701
+ "<SPECIAL_698>",
702
+ "<SPECIAL_699>",
703
+ "<SPECIAL_700>",
704
+ "<SPECIAL_701>",
705
+ "<SPECIAL_702>",
706
+ "<SPECIAL_703>",
707
+ "<SPECIAL_704>",
708
+ "<SPECIAL_705>",
709
+ "<SPECIAL_706>",
710
+ "<SPECIAL_707>",
711
+ "<SPECIAL_708>",
712
+ "<SPECIAL_709>",
713
+ "<SPECIAL_710>",
714
+ "<SPECIAL_711>",
715
+ "<SPECIAL_712>",
716
+ "<SPECIAL_713>",
717
+ "<SPECIAL_714>",
718
+ "<SPECIAL_715>",
719
+ "<SPECIAL_716>",
720
+ "<SPECIAL_717>",
721
+ "<SPECIAL_718>",
722
+ "<SPECIAL_719>",
723
+ "<SPECIAL_720>",
724
+ "<SPECIAL_721>",
725
+ "<SPECIAL_722>",
726
+ "<SPECIAL_723>",
727
+ "<SPECIAL_724>",
728
+ "<SPECIAL_725>",
729
+ "<SPECIAL_726>",
730
+ "<SPECIAL_727>",
731
+ "<SPECIAL_728>",
732
+ "<SPECIAL_729>",
733
+ "<SPECIAL_730>",
734
+ "<SPECIAL_731>",
735
+ "<SPECIAL_732>",
736
+ "<SPECIAL_733>",
737
+ "<SPECIAL_734>",
738
+ "<SPECIAL_735>",
739
+ "<SPECIAL_736>",
740
+ "<SPECIAL_737>",
741
+ "<SPECIAL_738>",
742
+ "<SPECIAL_739>",
743
+ "<SPECIAL_740>",
744
+ "<SPECIAL_741>",
745
+ "<SPECIAL_742>",
746
+ "<SPECIAL_743>",
747
+ "<SPECIAL_744>",
748
+ "<SPECIAL_745>",
749
+ "<SPECIAL_746>",
750
+ "<SPECIAL_747>",
751
+ "<SPECIAL_748>",
752
+ "<SPECIAL_749>",
753
+ "<SPECIAL_750>",
754
+ "<SPECIAL_751>",
755
+ "<SPECIAL_752>",
756
+ "<SPECIAL_753>",
757
+ "<SPECIAL_754>",
758
+ "<SPECIAL_755>",
759
+ "<SPECIAL_756>",
760
+ "<SPECIAL_757>",
761
+ "<SPECIAL_758>",
762
+ "<SPECIAL_759>",
763
+ "<SPECIAL_760>",
764
+ "<SPECIAL_761>",
765
+ "<SPECIAL_762>",
766
+ "<SPECIAL_763>",
767
+ "<SPECIAL_764>",
768
+ "<SPECIAL_765>",
769
+ "<SPECIAL_766>",
770
+ "<SPECIAL_767>",
771
+ "<SPECIAL_768>",
772
+ "<SPECIAL_769>",
773
+ "<SPECIAL_770>",
774
+ "<SPECIAL_771>",
775
+ "<SPECIAL_772>",
776
+ "<SPECIAL_773>",
777
+ "<SPECIAL_774>",
778
+ "<SPECIAL_775>",
779
+ "<SPECIAL_776>",
780
+ "<SPECIAL_777>",
781
+ "<SPECIAL_778>",
782
+ "<SPECIAL_779>",
783
+ "<SPECIAL_780>",
784
+ "<SPECIAL_781>",
785
+ "<SPECIAL_782>",
786
+ "<SPECIAL_783>",
787
+ "<SPECIAL_784>",
788
+ "<SPECIAL_785>",
789
+ "<SPECIAL_786>",
790
+ "<SPECIAL_787>",
791
+ "<SPECIAL_788>",
792
+ "<SPECIAL_789>",
793
+ "<SPECIAL_790>",
794
+ "<SPECIAL_791>",
795
+ "<SPECIAL_792>",
796
+ "<SPECIAL_793>",
797
+ "<SPECIAL_794>",
798
+ "<SPECIAL_795>",
799
+ "<SPECIAL_796>",
800
+ "<SPECIAL_797>",
801
+ "<SPECIAL_798>",
802
+ "<SPECIAL_799>",
803
+ "<SPECIAL_800>",
804
+ "<SPECIAL_801>",
805
+ "<SPECIAL_802>",
806
+ "<SPECIAL_803>",
807
+ "<SPECIAL_804>",
808
+ "<SPECIAL_805>",
809
+ "<SPECIAL_806>",
810
+ "<SPECIAL_807>",
811
+ "<SPECIAL_808>",
812
+ "<SPECIAL_809>",
813
+ "<SPECIAL_810>",
814
+ "<SPECIAL_811>",
815
+ "<SPECIAL_812>",
816
+ "<SPECIAL_813>",
817
+ "<SPECIAL_814>",
818
+ "<SPECIAL_815>",
819
+ "<SPECIAL_816>",
820
+ "<SPECIAL_817>",
821
+ "<SPECIAL_818>",
822
+ "<SPECIAL_819>",
823
+ "<SPECIAL_820>",
824
+ "<SPECIAL_821>",
825
+ "<SPECIAL_822>",
826
+ "<SPECIAL_823>",
827
+ "<SPECIAL_824>",
828
+ "<SPECIAL_825>",
829
+ "<SPECIAL_826>",
830
+ "<SPECIAL_827>",
831
+ "<SPECIAL_828>",
832
+ "<SPECIAL_829>",
833
+ "<SPECIAL_830>",
834
+ "<SPECIAL_831>",
835
+ "<SPECIAL_832>",
836
+ "<SPECIAL_833>",
837
+ "<SPECIAL_834>",
838
+ "<SPECIAL_835>",
839
+ "<SPECIAL_836>",
840
+ "<SPECIAL_837>",
841
+ "<SPECIAL_838>",
842
+ "<SPECIAL_839>",
843
+ "<SPECIAL_840>",
844
+ "<SPECIAL_841>",
845
+ "<SPECIAL_842>",
846
+ "<SPECIAL_843>",
847
+ "<SPECIAL_844>",
848
+ "<SPECIAL_845>",
849
+ "<SPECIAL_846>",
850
+ "<SPECIAL_847>",
851
+ "<SPECIAL_848>",
852
+ "<SPECIAL_849>",
853
+ "<SPECIAL_850>",
854
+ "<SPECIAL_851>",
855
+ "<SPECIAL_852>",
856
+ "<SPECIAL_853>",
857
+ "<SPECIAL_854>",
858
+ "<SPECIAL_855>",
859
+ "<SPECIAL_856>",
860
+ "<SPECIAL_857>",
861
+ "<SPECIAL_858>",
862
+ "<SPECIAL_859>",
863
+ "<SPECIAL_860>",
864
+ "<SPECIAL_861>",
865
+ "<SPECIAL_862>",
866
+ "<SPECIAL_863>",
867
+ "<SPECIAL_864>",
868
+ "<SPECIAL_865>",
869
+ "<SPECIAL_866>",
870
+ "<SPECIAL_867>",
871
+ "<SPECIAL_868>",
872
+ "<SPECIAL_869>",
873
+ "<SPECIAL_870>",
874
+ "<SPECIAL_871>",
875
+ "<SPECIAL_872>",
876
+ "<SPECIAL_873>",
877
+ "<SPECIAL_874>",
878
+ "<SPECIAL_875>",
879
+ "<SPECIAL_876>",
880
+ "<SPECIAL_877>",
881
+ "<SPECIAL_878>",
882
+ "<SPECIAL_879>",
883
+ "<SPECIAL_880>",
884
+ "<SPECIAL_881>",
885
+ "<SPECIAL_882>",
886
+ "<SPECIAL_883>",
887
+ "<SPECIAL_884>",
888
+ "<SPECIAL_885>",
889
+ "<SPECIAL_886>",
890
+ "<SPECIAL_887>",
891
+ "<SPECIAL_888>",
892
+ "<SPECIAL_889>",
893
+ "<SPECIAL_890>",
894
+ "<SPECIAL_891>",
895
+ "<SPECIAL_892>",
896
+ "<SPECIAL_893>",
897
+ "<SPECIAL_894>",
898
+ "<SPECIAL_895>",
899
+ "<SPECIAL_896>",
900
+ "<SPECIAL_897>",
901
+ "<SPECIAL_898>",
902
+ "<SPECIAL_899>",
903
+ "<SPECIAL_900>",
904
+ "<SPECIAL_901>",
905
+ "<SPECIAL_902>",
906
+ "<SPECIAL_903>",
907
+ "<SPECIAL_904>",
908
+ "<SPECIAL_905>",
909
+ "<SPECIAL_906>",
910
+ "<SPECIAL_907>",
911
+ "<SPECIAL_908>",
912
+ "<SPECIAL_909>",
913
+ "<SPECIAL_910>",
914
+ "<SPECIAL_911>",
915
+ "<SPECIAL_912>",
916
+ "<SPECIAL_913>",
917
+ "<SPECIAL_914>",
918
+ "<SPECIAL_915>",
919
+ "<SPECIAL_916>",
920
+ "<SPECIAL_917>",
921
+ "<SPECIAL_918>",
922
+ "<SPECIAL_919>",
923
+ "<SPECIAL_920>",
924
+ "<SPECIAL_921>",
925
+ "<SPECIAL_922>",
926
+ "<SPECIAL_923>",
927
+ "<SPECIAL_924>",
928
+ "<SPECIAL_925>",
929
+ "<SPECIAL_926>",
930
+ "<SPECIAL_927>",
931
+ "<SPECIAL_928>",
932
+ "<SPECIAL_929>",
933
+ "<SPECIAL_930>",
934
+ "<SPECIAL_931>",
935
+ "<SPECIAL_932>",
936
+ "<SPECIAL_933>",
937
+ "<SPECIAL_934>",
938
+ "<SPECIAL_935>",
939
+ "<SPECIAL_936>",
940
+ "<SPECIAL_937>",
941
+ "<SPECIAL_938>",
942
+ "<SPECIAL_939>",
943
+ "<SPECIAL_940>",
944
+ "<SPECIAL_941>",
945
+ "<SPECIAL_942>",
946
+ "<SPECIAL_943>",
947
+ "<SPECIAL_944>",
948
+ "<SPECIAL_945>",
949
+ "<SPECIAL_946>",
950
+ "<SPECIAL_947>",
951
+ "<SPECIAL_948>",
952
+ "<SPECIAL_949>",
953
+ "<SPECIAL_950>",
954
+ "<SPECIAL_951>",
955
+ "<SPECIAL_952>",
956
+ "<SPECIAL_953>",
957
+ "<SPECIAL_954>",
958
+ "<SPECIAL_955>",
959
+ "<SPECIAL_956>",
960
+ "<SPECIAL_957>",
961
+ "<SPECIAL_958>",
962
+ "<SPECIAL_959>",
963
+ "<SPECIAL_960>",
964
+ "<SPECIAL_961>",
965
+ "<SPECIAL_962>",
966
+ "<SPECIAL_963>",
967
+ "<SPECIAL_964>",
968
+ "<SPECIAL_965>",
969
+ "<SPECIAL_966>",
970
+ "<SPECIAL_967>",
971
+ "<SPECIAL_968>",
972
+ "<SPECIAL_969>",
973
+ "<SPECIAL_970>",
974
+ "<SPECIAL_971>",
975
+ "<SPECIAL_972>",
976
+ "<SPECIAL_973>",
977
+ "<SPECIAL_974>",
978
+ "<SPECIAL_975>",
979
+ "<SPECIAL_976>",
980
+ "<SPECIAL_977>",
981
+ "<SPECIAL_978>",
982
+ "<SPECIAL_979>",
983
+ "<SPECIAL_980>",
984
+ "<SPECIAL_981>",
985
+ "<SPECIAL_982>",
986
+ "<SPECIAL_983>",
987
+ "<SPECIAL_984>",
988
+ "<SPECIAL_985>",
989
+ "<SPECIAL_986>",
990
+ "<SPECIAL_987>",
991
+ "<SPECIAL_988>",
992
+ "<SPECIAL_989>",
993
+ "<SPECIAL_990>",
994
+ "<SPECIAL_991>",
995
+ "<SPECIAL_992>",
996
+ "<SPECIAL_993>",
997
+ "<SPECIAL_994>",
998
+ "<SPECIAL_995>",
999
+ "<SPECIAL_996>",
1000
+ "<SPECIAL_997>",
1001
+ "<SPECIAL_998>",
1002
+ "<SPECIAL_999>"
1003
+ ],
1004
+ "bos_token": {
1005
+ "content": "<s>",
1006
+ "lstrip": false,
1007
+ "normalized": false,
1008
+ "rstrip": false,
1009
+ "single_word": false
1010
+ },
1011
+ "eos_token": {
1012
+ "content": "</s>",
1013
+ "lstrip": false,
1014
+ "normalized": false,
1015
+ "rstrip": false,
1016
+ "single_word": false
1017
+ },
1018
+ "pad_token": "</s>",
1019
+ "unk_token": {
1020
+ "content": "<unk>",
1021
+ "lstrip": false,
1022
+ "normalized": false,
1023
+ "rstrip": false,
1024
+ "single_word": false
1025
+ }
1026
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b76085f9923309d873994d444989f7eb6ec074b06f25b58f1e8d7b7741070949
3
+ size 17078037
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff