Hugging Face's logo

Spaces:
poccio
/
ExtEnD
Build error

App Files Community
ExtEnD
File size: 4,937 Bytes 
67a2b13
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
 
callbacks:
  callbacks: []
data:
  datamodule:
    _target_: classy.data.data_modules.ClassyDataModule
    task: ${task}
    dataset_path: data/aida
    train_dataset:
      _target_: classy.data.dataset.hf.classification.HFQADataset.from_file
      transformer_model: ${transformer_model}
      additional_special_tokens: ${model.additional_special_tokens}
      min_length: 5
      max_length: 500
      tokens_per_batch: 2000
      max_batch_size: 10
      section_size: 10000
      prebatch: true
      materialize: false
      for_inference: false
    validation_dataset:
      _target_: classy.data.dataset.hf.classification.HFQADataset.from_file
      transformer_model: ${transformer_model}
      additional_special_tokens: ${model.additional_special_tokens}
      min_length: 5
      max_length: 500
      tokens_per_batch: 2000
      max_batch_size: 10
      section_size: 10000
      prebatch: true
      materialize: true
      for_inference: true
    validation_split_size: 0.1
    test_split_size: 0.1
    max_nontrain_split_size: 10000
    shuffle_dataset: true
device:
  gpus:
  - 0
  precision: 32
  amp_level: O0
model:
  _target_: classy.pl_modules.hf.classification.HFQAPLModule
  transformer_model: ${transformer_model}
  additional_special_tokens: []
  optim_conf:
    _target_: classy.optim.factories.RAdamFactory
    lr: 1.0e-05
    weight_decay: 0.01
    no_decay_params:
    - bias
    - LayerNorm.weight
prediction:
  dataset:
    _target_: classy.data.dataset.hf.classification.HFQADataset.from_samples
    transformer_model: ${transformer_model}
    additional_special_tokens: ${model.additional_special_tokens}
    min_length: -1
    max_length: -1
    tokens_per_batch: 800
    max_batch_size: -1
    section_size: 10000
    prebatch: true
    materialize: false
    for_inference: true
training:
  seed: 12
  pl_trainer:
    _target_: pytorch_lightning.Trainer
    accumulate_grad_batches: 4
    gradient_clip_val: 10.0
    val_check_interval: 1.0
    max_steps: 1000000
  early_stopping_callback:
    _target_: pytorch_lightning.callbacks.EarlyStopping
    monitor: ${callbacks_monitor}
    mode: ${callbacks_mode}
    patience: 25
  model_checkpoint_callback:
    _target_: classy.pl_callbacks.best_checkpoint.ModelCheckpointWithBest
    monitor: ${callbacks_monitor}
    mode: ${callbacks_mode}
    verbose: true
    save_top_k: 3
    dirpath: checkpoints
    save_last: true
  resume_from: null
logging:
  wandb:
    use_wandb: true
    project_name: esc-ed
    experiment_name: aida-longformer-large-*sep-gam-cand-shuffle
    anonymous: null
    run_id: null
task: qa
project_name: classy
exp_name: esc-aida-longformer-large-gam-cand-shuffle
exp_folder: ./experiments/${exp_name}
transformer_model: bert-base-cased
callbacks_monitor: val_accuracy
callbacks_mode: max
profiles:
  supported_tasks:
  - qa
  - sentence-pair
  - sequence
  - token
  - generation
  transformer_model: allenai/longformer-large-4096
  candidates_separator: '*'
  training:
    pl_trainer:
      accumulate_grad_batches: 8
      val_check_interval: 2048
      max_steps: 100000
  model:
    _target_: src.esc_ed_module.ESCModule
    additional_special_tokens: []
    transformer_model: ${transformer_model}
    attention_window: 64
    modify_global_attention: true
    optim_conf:
      _target_: classy.optim.factories.RAdamFactory
      lr: 1.0e-05
      weight_decay: 0.01
      no_decay_params:
      - bias
      - LayerNorm.weight
  data:
    datamodule:
      train_dataset:
        _target_: src.data.esc_ed_dataset.ESCEDDataset.from_file
        transformer_model: ${transformer_model}
        additional_special_tokens: ${model.additional_special_tokens}
        candidates_separator: ${candidates_separator}
        shuffle_candidates_prob: 0.0
        min_length: 0
        max_length: 1024
        tokens_per_batch: 1024
        max_batch_size: 10
        section_size: 20000
        prebatch: true
        materialize: false
        for_inference: false
      validation_dataset:
        _target_: src.data.esc_ed_dataset.ESCEDDataset.from_file
        transformer_model: ${transformer_model}
        additional_special_tokens: ${model.additional_special_tokens}
        candidates_separator: ${candidates_separator}
        min_length: 0
        max_length: 1024
        tokens_per_batch: 2048
        max_batch_size: 10
        section_size: 10000
        prebatch: true
        materialize: true
        for_inference: false
      shuffle_dataset: true
  prediction:
    dataset:
      _target_: src.data.esc_ed_dataset.ESCEDDataset.from_samples
      transformer_model: ${transformer_model}
      additional_special_tokens: ${model.additional_special_tokens}
      candidates_separator: ${candidates_separator}
      min_length: -1
      max_length: -1
      tokens_per_batch: 2048
      max_batch_size: -1
      section_size: 10000
      prebatch: true
      materialize: false
      for_inference: true