|
--- |
|
language: |
|
- bn |
|
metrics: |
|
- code_eval |
|
- accuracy |
|
- bertscore |
|
--- |
|
|
|
from transformers import AutoTokenizer, AutoModelForSequenceClassification |
|
irony_name = "raquiba/sarcasm-detection-BanglaSARC" |
|
tokenizer_irony = AutoTokenizer.from_pretrained(irony_name) |
|
model_irony = AutoModelForSequenceClassification.from_pretrained(irony_name) |
|
irony_pipeline = pipeline("sentiment-analysis", model=model_irony, tokenizer=tokenizer_irony, device=0,max_length=512, padding=True, truncation=True) |
|
|
|
#Model Evaluation |
|
tokenizer = AutoTokenizer.from_pretrained(irony_name) |
|
df_train, df_test = tokenized_data(df_eval) |
|
model_irony = AutoModelForSequenceClassification.from_pretrained(irony_name, num_labels=2, ignore_mismatched_sizes=True).to(device) |
|
training_args = TrainingArguments("test-trainer-banglaBERT", {'reprocess_input_data': True}, evaluation_strategy="epoch") |
|
trainer_irony.train() |