Upload 5 files
Browse files- model_config.yaml +26 -0
- preprocessor/tokenizer.json +0 -0
- preprocessor/tokenizer_config.yaml +36 -0
- train/dataset_config.yaml +16 -0
- train/train_config.yaml +11 -0
model_config.yaml
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: roberta_text_classification
|
| 2 |
+
config_type: model
|
| 3 |
+
task: text_classification
|
| 4 |
+
num_labels: 3
|
| 5 |
+
id2label:
|
| 6 |
+
0: negative
|
| 7 |
+
1: positive
|
| 8 |
+
2: neutral
|
| 9 |
+
attention_probs_dropout_prob: 0.1
|
| 10 |
+
bos_token_id: 0
|
| 11 |
+
eos_token_id: 2
|
| 12 |
+
gradient_checkpointing: false
|
| 13 |
+
hidden_act: gelu
|
| 14 |
+
hidden_dropout_prob: 0.1
|
| 15 |
+
hidden_size: 768
|
| 16 |
+
initializer_range: 0.02
|
| 17 |
+
intermediate_size: 3072
|
| 18 |
+
layer_norm_eps: 1.0e-12
|
| 19 |
+
max_position_embeddings: 514
|
| 20 |
+
num_attention_heads: 12
|
| 21 |
+
num_hidden_layers: 12
|
| 22 |
+
pad_token_id: 1
|
| 23 |
+
position_embedding_type: absolute
|
| 24 |
+
type_vocab_size: 1
|
| 25 |
+
use_cache: true
|
| 26 |
+
vocab_size: 42000
|
preprocessor/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
preprocessor/tokenizer_config.yaml
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: bpe_tokenizer
|
| 2 |
+
config_type: preprocessor
|
| 3 |
+
pretrained_path: hezarai/roberta-base-fa
|
| 4 |
+
max_length: 512
|
| 5 |
+
truncation_strategy: longest_first
|
| 6 |
+
truncation_direction: right
|
| 7 |
+
stride: 0
|
| 8 |
+
padding_strategy: longest
|
| 9 |
+
padding_direction: right
|
| 10 |
+
pad_to_multiple_of: 0
|
| 11 |
+
pad_token_id: 0
|
| 12 |
+
pad_token: <pad>
|
| 13 |
+
pad_token_type_id: 0
|
| 14 |
+
unk_token: <unk>
|
| 15 |
+
special_tokens:
|
| 16 |
+
- <s>
|
| 17 |
+
- <pad>
|
| 18 |
+
- </s>
|
| 19 |
+
- <unk>
|
| 20 |
+
- <mask>
|
| 21 |
+
- <|endoftext|>
|
| 22 |
+
- <|startoftext|>
|
| 23 |
+
- <nl>
|
| 24 |
+
- <hs>
|
| 25 |
+
- <sep>
|
| 26 |
+
- <cls>
|
| 27 |
+
continuing_subword_prefix: ''
|
| 28 |
+
end_of_word_suffix: ''
|
| 29 |
+
fuse_unk: false
|
| 30 |
+
train_config:
|
| 31 |
+
name: bpe_tokenizer
|
| 32 |
+
config_type: preprocessor
|
| 33 |
+
vocab_size: 30000
|
| 34 |
+
min_frequency: 2
|
| 35 |
+
limit_alphabet: 1000
|
| 36 |
+
show_progress: true
|
train/dataset_config.yaml
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: text_classification
|
| 2 |
+
config_type: dataset
|
| 3 |
+
task: text_classification
|
| 4 |
+
path: hezarai/sentiment_digikala_snappfood
|
| 5 |
+
tokenizer_path: hezarai/roberta-base-fa
|
| 6 |
+
label_field: label
|
| 7 |
+
text_field: text
|
| 8 |
+
id2label:
|
| 9 |
+
0: negative
|
| 10 |
+
1: positive
|
| 11 |
+
2: neutral
|
| 12 |
+
label2id:
|
| 13 |
+
negative: 0
|
| 14 |
+
positive: 1
|
| 15 |
+
neutral: 2
|
| 16 |
+
num_labels: 3
|
train/train_config.yaml
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: roberta_text_classification
|
| 2 |
+
config_type: train
|
| 3 |
+
device: cuda
|
| 4 |
+
init_weights_from: hezarai/roberta-base-fa
|
| 5 |
+
seed: 42
|
| 6 |
+
batch_size: 8
|
| 7 |
+
metrics:
|
| 8 |
+
f1:
|
| 9 |
+
task: multiclass
|
| 10 |
+
num_train_epochs: 5
|
| 11 |
+
checkpoints_dir: checkpoints/
|