Spaces:
Sleeping
Sleeping
Upload 2 files
Browse files- config.yaml +33 -0
- mapping.json +52 -0
config.yaml
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
base_model_config:
|
| 2 |
+
base_model: mbert-uncased
|
| 3 |
+
apply_lora_flag: True
|
| 4 |
+
freeze_transformer_flag: False
|
| 5 |
+
num_unique_labels: 20
|
| 6 |
+
|
| 7 |
+
training_config:
|
| 8 |
+
# directory
|
| 9 |
+
output_dir: ../model_checkpoints/lora
|
| 10 |
+
logging_dir: ../logs/lora
|
| 11 |
+
|
| 12 |
+
# LoRA config
|
| 13 |
+
lora_rank: 8
|
| 14 |
+
lora_alpha: 16
|
| 15 |
+
target_modules: ["query", "key", "value"]
|
| 16 |
+
lora_dropout: 0.1
|
| 17 |
+
|
| 18 |
+
# training config
|
| 19 |
+
learning_rate: 0.00005
|
| 20 |
+
eval_steps: 500
|
| 21 |
+
save_steps: 500
|
| 22 |
+
logging_steps: 100
|
| 23 |
+
per_device_train_batch_size: 32
|
| 24 |
+
per_device_eval_batch_size: 32
|
| 25 |
+
num_train_epochs: 20
|
| 26 |
+
weight_decay: 0.01
|
| 27 |
+
early_stopping_patience: 5
|
| 28 |
+
early_stopping_threshold: 0.01
|
| 29 |
+
|
| 30 |
+
inference_config:
|
| 31 |
+
# directory
|
| 32 |
+
model_path: ../model_checkpoints/lora
|
| 33 |
+
saving_dir: ../results/lora/test_split.csv
|
mapping.json
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"models": {
|
| 3 |
+
"mbert-cased": "google-bert/bert-base-multilingual-cased",
|
| 4 |
+
"mbert-uncased": "google-bert/bert-base-multilingual-uncased"
|
| 5 |
+
},
|
| 6 |
+
|
| 7 |
+
"id2label": {
|
| 8 |
+
"0": "ar",
|
| 9 |
+
"1": "bg",
|
| 10 |
+
"2": "de",
|
| 11 |
+
"3": "el",
|
| 12 |
+
"4": "en",
|
| 13 |
+
"5": "es",
|
| 14 |
+
"6": "fr",
|
| 15 |
+
"7": "hi",
|
| 16 |
+
"8": "it",
|
| 17 |
+
"9": "ja",
|
| 18 |
+
"10": "nl",
|
| 19 |
+
"11": "pl",
|
| 20 |
+
"12": "pt",
|
| 21 |
+
"13": "ru",
|
| 22 |
+
"14": "sw",
|
| 23 |
+
"15": "th",
|
| 24 |
+
"16": "tr",
|
| 25 |
+
"17": "ur",
|
| 26 |
+
"18": "vi",
|
| 27 |
+
"19": "zh"
|
| 28 |
+
},
|
| 29 |
+
|
| 30 |
+
"language_codes":{
|
| 31 |
+
"ar": "arabic",
|
| 32 |
+
"bg": "bulgarian",
|
| 33 |
+
"de": "german",
|
| 34 |
+
"el": "modern greek",
|
| 35 |
+
"en": "english",
|
| 36 |
+
"es": "spanish",
|
| 37 |
+
"fr": "french",
|
| 38 |
+
"hi": "hindi",
|
| 39 |
+
"it": "italian",
|
| 40 |
+
"ja": "japanese",
|
| 41 |
+
"nl": "dutch",
|
| 42 |
+
"pl": "polish",
|
| 43 |
+
"pt": "portuguese",
|
| 44 |
+
"ru": "russian",
|
| 45 |
+
"sw": "swahili",
|
| 46 |
+
"th": "thai",
|
| 47 |
+
"tr": "turkish",
|
| 48 |
+
"ur": "urdu",
|
| 49 |
+
"vi": "vietnamese",
|
| 50 |
+
"zh": "chinese"
|
| 51 |
+
}
|
| 52 |
+
}
|