I'm fairly new to Machine Learning. I've successfully solved errors to do with parameters and model setup.
I'm using this Notebook, where section Apply DocumentClassifier is altered as below.
Jupyter Labs, kernel: conda_mxnet_latest_p37.
Error seems to be more about my laptop's hardware, rather than my code being broken.
Update: I changed batch_size=4, it ran for ages only to crash.
What should be my standard approach to solving this error?
My Code:
with open('filt_gri.txt', 'r') as filehandle:
tags = [current_place.rstrip() for current_place in filehandle.readlines()]
doc_classifier = TransformersDocumentClassifier(model_name_or_path="cross-encoder/nli-distilroberta-base",
task="zero-shot-classification",
labels=tags,
batch_size=4)
# convert to Document using a fieldmap for custom content fields the classification should run on
docs_to_classify = [Document.from_dict(d) for d in docs_sliding_window]
# classify using gpu, batch_size makes sure we do not run out of memory
classified_docs = doc_classifier.predict(docs_to_classify)
# let's see how it looks: there should be a classification result in the meta entry containing labels and scores.
print(classified_docs[0].to_dict())
all_docs = convert_files_to_dicts(dir_path=doc_dir)
preprocessor_sliding_window = PreProcessor(split_overlap=3,
split_length=10,
split_respect_sentence_boundary=False,
split_by='passage')
Error:
INFO - haystack.modeling.utils - Using devices: CUDA
INFO - haystack.modeling.utils - Using devices: CUDA
INFO - haystack.modeling.utils - Number of GPUs: 1
INFO - haystack.modeling.utils - Number of GPUs: 1
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-25-27dfca549a7d> in <module>
14
15 # classify using gpu, batch_size makes sure we do not run out of memory
---> 16 classified_docs = doc_classifier.predict(docs_to_classify)
17
18 # let's see how it looks: there should be a classification result in the meta entry containing labels and scores.
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/haystack/nodes/document_classifier/transformers.py in predict(self, documents)
137 batches = self.get_batches(texts, batch_size=self.batch_size)
138 if self.task == 'zero-shot-classification':
--> 139 batched_predictions = [self.model(batch, candidate_labels=self.labels, truncation=True) for batch in batches]
140 elif self.task == 'text-classification':
141 batched_predictions = [self.model(batch, return_all_scores=self.return_all_scores, truncation=True) for batch in batches]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/haystack/nodes/document_classifier/transformers.py in <listcomp>(.0)
137 batches = self.get_batches(texts, batch_size=self.batch_size)
138 if self.task == 'zero-shot-classification':
--> 139 batched_predictions = [self.model(batch, candidate_labels=self.labels, truncation=True) for batch in batches]
140 elif self.task == 'text-classification':
141 batched_predictions = [self.model(batch, return_all_scores=self.return_all_scores, truncation=True) for batch in batches]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/pipelines/zero_shot_classification.py in __call__(self, sequences, candidate_labels, hypothesis_template, multi_label, **kwargs)
151 sequences = [sequences]
152
--> 153 outputs = super().__call__(sequences, candidate_labels, hypothesis_template)
154 num_sequences = len(sequences)
155 candidate_labels = self._args_parser._parse_labels(candidate_labels)
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/pipelines/base.py in __call__(self, *args, **kwargs)
759 def __call__(self, *args, **kwargs):
760 inputs = self._parse_and_tokenize(*args, **kwargs)
--> 761 return self._forward(inputs)
762
763 def _forward(self, inputs, return_tensors=False):
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/pipelines/base.py in _forward(self, inputs, return_tensors)
780 with torch.no_grad():
781 inputs = self.ensure_tensor_on_device(**inputs)
--> 782 predictions = self.model(**inputs)[0].cpu()
783
784 if return_tensors:
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict)
1162 output_attentions=output_attentions,
1163 output_hidden_states=output_hidden_states,
-> 1164 return_dict=return_dict,
1165 )
1166 sequence_output = outputs[0]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, encoder_hidden_states, encoder_attention_mask, past_key_values, use_cache, output_attentions, output_hidden_states, return_dict)
823 output_attentions=output_attentions,
824 output_hidden_states=output_hidden_states,
--> 825 return_dict=return_dict,
826 )
827 sequence_output = encoder_outputs[0]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_values, use_cache, output_attentions, output_hidden_states, return_dict)
513 encoder_attention_mask,
514 past_key_value,
--> 515 output_attentions,
516 )
517
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions)
398 head_mask,
399 output_attentions=output_attentions,
--> 400 past_key_value=self_attn_past_key_value,
401 )
402 attention_output = self_attention_outputs[0]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions)
328 encoder_attention_mask,
329 past_key_value,
--> 330 output_attentions,
331 )
332 attention_output = self.output(self_outputs[0], hidden_states)
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions)
241 attention_scores = attention_scores relative_position_scores_query relative_position_scores_key
242
--> 243 attention_scores = attention_scores / math.sqrt(self.attention_head_size)
244 if attention_mask is not None:
245 # Apply the attention mask is (precomputed for all layers in RobertaModel forward() function)
RuntimeError: CUDA out of memory. Tried to allocate 3.60 GiB (GPU 0; 14.76 GiB total capacity; 7.33 GiB already allocated; 1.37 GiB free; 12.29 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-25-27dfca549a7d> in <module>
14
15 # classify using gpu, batch_size makes sure we do not run out of memory
---> 16 classified_docs = doc_classifier.predict(docs_to_classify)
17
18 # let's see how it looks: there should be a classification result in the meta entry containing labels and scores.
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/haystack/nodes/document_classifier/transformers.py in predict(self, documents)
137 batches = self.get_batches(texts, batch_size=self.batch_size)
138 if self.task == 'zero-shot-classification':
--> 139 batched_predictions = [self.model(batch, candidate_labels=self.labels, truncation=True) for batch in batches]
140 elif self.task == 'text-classification':
141 batched_predictions = [self.model(batch, return_all_scores=self.return_all_scores, truncation=True) for batch in batches]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/haystack/nodes/document_classifier/transformers.py in <listcomp>(.0)
137 batches = self.get_batches(texts, batch_size=self.batch_size)
138 if self.task == 'zero-shot-classification':
--> 139 batched_predictions = [self.model(batch, candidate_labels=self.labels, truncation=True) for batch in batches]
140 elif self.task == 'text-classification':
141 batched_predictions = [self.model(batch, return_all_scores=self.return_all_scores, truncation=True) for batch in batches]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/pipelines/zero_shot_classification.py in __call__(self, sequences, candidate_labels, hypothesis_template, multi_label, **kwargs)
151 sequences = [sequences]
152
--> 153 outputs = super().__call__(sequences, candidate_labels, hypothesis_template)
154 num_sequences = len(sequences)
155 candidate_labels = self._args_parser._parse_labels(candidate_labels)
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/pipelines/base.py in __call__(self, *args, **kwargs)
759 def __call__(self, *args, **kwargs):
760 inputs = self._parse_and_tokenize(*args, **kwargs)
--> 761 return self._forward(inputs)
762
763 def _forward(self, inputs, return_tensors=False):
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/pipelines/base.py in _forward(self, inputs, return_tensors)
780 with torch.no_grad():
781 inputs = self.ensure_tensor_on_device(**inputs)
--> 782 predictions = self.model(**inputs)[0].cpu()
783
784 if return_tensors:
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict)
1162 output_attentions=output_attentions,
1163 output_hidden_states=output_hidden_states,
-> 1164 return_dict=return_dict,
1165 )
1166 sequence_output = outputs[0]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, encoder_hidden_states, encoder_attention_mask, past_key_values, use_cache, output_attentions, output_hidden_states, return_dict)
823 output_attentions=output_attentions,
824 output_hidden_states=output_hidden_states,
--> 825 return_dict=return_dict,
826 )
827 sequence_output = encoder_outputs[0]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_values, use_cache, output_attentions, output_hidden_states, return_dict)
513 encoder_attention_mask,
514 past_key_value,
--> 515 output_attentions,
516 )
517
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions)
398 head_mask,
399 output_attentions=output_attentions,
--> 400 past_key_value=self_attn_past_key_value,
401 )
402 attention_output = self_attention_outputs[0]
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions)
328 encoder_attention_mask,
329 past_key_value,
--> 330 output_attentions,
331 )
332 attention_output = self.output(self_outputs[0], hidden_states)
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []
~/anaconda3/envs/mxnet_latest_p37/lib/python3.7/site-packages/transformers/models/roberta/modeling_roberta.py in forward(self, hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions)
241 attention_scores = attention_scores relative_position_scores_query relative_position_scores_key
242
--> 243 attention_scores = attention_scores / math.sqrt(self.attention_head_size)
244 if attention_mask is not None:
245 # Apply the attention mask is (precomputed for all layers in RobertaModel forward() function)
RuntimeError: CUDA out of memory. Tried to allocate 3.60 GiB (GPU 0; 14.76 GiB total capacity; 7.33 GiB already allocated; 1.37 GiB free; 12.29 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Original Code:
doc_classifier = TransformersDocumentClassifier(model_name_or_path="cross-encoder/nli-distilroberta-base",
task="zero-shot-classification",
labels=["music", "natural language processing", "history"],
batch_size=16
)
# ----------
# convert to Document using a fieldmap for custom content fields the classification should run on
docs_to_classify = [Document.from_dict(d) for d in docs_sliding_window]
# ----------
# classify using gpu, batch_size makes sure we do not run out of memory
classified_docs = doc_classifier.predict(docs_to_classify)
# ----------
# let's see how it looks: there should be a classification result in the meta entry containing labels and scores.
print(classified_docs[0].to_dict())
Please let me know if there is anything else I should add to post/ clarify.
CodePudding user response：
Reducing the batch_size helped me:
batch_size=2