I am running bertxl summarization on a GPU Ubuntu instance in AWS ec2 and it runs fine on my windows machine using CPU. But throws an error running there. Any assistance would be helpful. I am ready to give any other extra information if this was not sufficient.
File "/home/ubuntu/neus/cron_summary_desc.py", line 282, in store_summaries
sum_df = extract_summaries(df, min_desc_length)
File "/home/ubuntu/neus/cron_summary_desc.py", line 344, in extract_summaries
mini_sum_xl = ''.join(model_xl(para, num_sentences = sent_count))
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/summary_processor.py", line 234, in call
return self.run(body, ratio, min_length, max_length,
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/summaryprocessor.py", line 202, in run
sentences, = self.cluster_runner(sentences, ratio, algorithm, use_first, num_sentences)
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/summary_processor.py", line 108, in cluster_runner
hidden = self.model(sentences)
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/transformer_embeddings/bert_embedding.py", line 173, in call
return self.create_matrix(content, hidden, reduce_option, hidden_concat)
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/transformer_embeddings/bert_embedding.py", line 151, in create_matrix
return np.asarray([
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/transformer_embeddings/bert_embedding.py", line 152, in
np.squeeze(self.extract_embeddings(
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/summarizer/transformer_embeddings/bert_embedding.py", line 108, in extract_embeddings
pooled, hidden_states = self.model(tokens_tensor)[-2:]
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(*input, *kwargs)
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/transformers/models/xlnet/modeling_xlnet.py", line 1180, in forward
word_emb_k = self.word_embedding(input_ids)
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(input, **kwargs)
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/torch/nn/modules/sparse.py", line 158, in forward
return F.embedding(
File "/home/ubuntu/neus/neusenv/lib/python3.8/site-packages/torch/nn/functional.py", line 2183, in embedding
return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)
RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int; but got torch.cuda.FloatTensor instead (while checking arguments for embedding)
I am running bertxl summarization on a GPU Ubuntu instance in AWS ec2 and it runs fine on my windows machine using CPU. But throws an error running there. Any assistance would be helpful. I am ready to give any other extra information if this was not sufficient.