Hello, I have this dataset :
{"version": "0.1.0",
"data":
[
{"id": "hi",
"question": ["hi", "how are you"],
"answers": ["hi!", "how can i help you?"],
"context": ""
},
{"id": "bye",
"question": ["Bye", "good bye", "see you"],
"answers": ["see you later", "have a nice day", "bye", "thanks for visiting"],
"context": ""
},
{"id": "weather",
"question": ["how is the weather", "weather forecast", "weather"],
"answers": ["weather is good", "we have 25 degrees"],
"context": ""
}
]
}
and I am trying to build a question answer bot.
I am using this code:
train = load_dataset('json', data_files='intents.json', field='data', split='train[:80%]')
test = load_dataset('json', data_files='intents.json', field='data', split='train[80%:]')
data = datasets.DatasetDict({"train":train, "test": test})
tokenized = data.map(preprocess_func, batched=True)
#data_collator = DefaultDataCollator()
data_collator = DataCollatorForLanguageModeling(
tokenizer=tokenizer, mlm=True
)
device = "cpu"
model = AutoModelForQuestionAnswering.from_pretrained('bert-base-uncased')
model = model.to(device)
training_args = TrainingArguments(
output_dir="./results",
evaluation_strategy="epoch",
learning_rate=2e-5,
per_device_train_batch_size=2,
num_train_epochs=2,
weight_decay=0.01,
)
trainer = Trainer(
model=model,
args=training_args,
train_dataset=tokenized["train"],
tokenizer=tokenizer,
data_collator=data_collator,
)
trainer.train()
and I am receiving:
BertForQuestionAnswering.forward() got an unexpected keyword argument 'labels'
but I don’t have any labels in the data:
tokenized
DatasetDict({
train: Dataset({
features: ['context', 'id', 'question', 'answers', 'input_ids', 'token_type_ids', 'attention_mask'],
num_rows: 2
})
test: Dataset({
features: ['context', 'id', 'question', 'answers', 'input_ids', 'token_type_ids', 'attention_mask'],
num_rows: 1
})
})
If I use :
DefaultDataCollator()
instead of DataCollatorForLanguageModeling
, I receive:
The model did not return a loss from the inputs, only the following keys: start_logits,end_logits
I am not sure if the preprocess_func
needs more things to do.
Like this:
def preprocess_function(examples):
questions = [q.strip() for q in examples["question"]]
inputs = tokenizer(
questions,
examples["context"],
max_length=512,
truncation="only_second",
return_offsets_mapping=True,
padding="max_length",
)
offset_mapping = inputs.pop("offset_mapping")
answers = examples["answers"]
start_positions = []
end_positions = []
for i, offset in enumerate(offset_mapping):
answer = answers[i]
start_char = answer["answer_start"][0]
end_char = answer["answer_start"][0] + len(answer["text"][0])
sequence_ids = inputs.sequence_ids(i)
# Find the start and end of the context
idx = 0
while sequence_ids[idx] != 1:
idx += 1
context_start = idx
while sequence_ids[idx] == 1:
idx += 1
context_end = idx - 1
# If the answer is not fully inside the context, label it (0, 0)
if offset[context_start][0] > end_char or offset[context_end][1] < start_char:
start_positions.append(0)
end_positions.append(0)
else:
# Otherwise it's the start and end token positions
idx = context_start
while idx <= context_end and offset[idx][0] <= start_char:
idx += 1
start_positions.append(idx - 1)
idx = context_end
while idx >= context_start and offset[idx][1] >= end_char:
idx -= 1
end_positions.append(idx + 1)
inputs["start_positions"] = start_positions
inputs["end_positions"] = end_positions
return inputs