inp_ids = tokenizer.encode(" I am not happy that he urged me to finish all the hardest tasks in the world", return_tensors="pt")
fig_ids = tokenizer.encode("", add_special_tokens=False, return_tensors="pt")
outs = model.generate(input_ids=inp_ids[:, 1:], fig_ids=fig_ids, forced_bos_token_id=fig_ids.item(), num_beams=5, max_length=60,)
text = tokenizer.decode(outs[0, 2:].tolist(), skip_special_tokens=True, clean_up_tokenization_spaces=False)

AttributeError Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_32792\2188357501.py in
1 inp_ids = tokenizer.encode(" I am not happy that he urged me to finish all the hardest tasks in the world", return_tensors="pt")
2 fig_ids = tokenizer.encode("", add_special_tokens=False, return_tensors="pt")
----> 3 outs = model.generate(input_ids=inp_ids[:, 1:], fig_ids=fig_ids, forced_bos_token_id=fig_ids.item(), num_beams=5, max_length=60,)
4 text = tokenizer.decode(outs[0, 2:].tolist(), skip_special_tokens=True, clean_up_tokenization_spaces=False)

~.conda\envs\transformers\lib\site-packages\torch\autograd\grad_mode.py in decorate_context(*args, **kwargs)
26 def decorate_context(*args, **kwargs):
27 with self.class():
---> 28 return func(*args, **kwargs)
29 return cast(F, decorate_context)
30

~.conda\envs\transformers\lib\site-packages\transformers\generation_utils.py in generate(self, inputs, max_length, min_length, do_sample, early_stopping, num_beams, temperature, penalty_alpha, top_k, top_p, typical_p, repetition_penalty, bad_words_ids, force_words_ids, bos_token_id, pad_token_id, eos_token_id, length_penalty, no_repeat_ngram_size, encoder_no_repeat_ngram_size, num_return_sequences, max_time, max_new_tokens, decoder_start_token_id, use_cache, num_beam_groups, diversity_penalty, prefix_allowed_tokens_fn, logits_processor, renormalize_logits, stopping_criteria, constraints, output_attentions, output_hidden_states, output_scores, return_dict_in_generate, forced_bos_token_id, forced_eos_token_id, remove_invalid_values, synced_gpus, exponential_decay_length_penalty, suppress_tokens, begin_suppress_tokens, forced_decoder_ids, **model_kwargs)
1338 # and added to model_kwargs
1339 model_kwargs = self._prepare_encoder_decoder_kwargs_for_generation(
-> 1340 inputs_tensor, model_kwargs, model_input_name
1341 )
1342

~.conda\envs\transformers\lib\site-packages\transformers\generation_utils.py in _prepare_encoder_decoder_kwargs_for_generation(self, inputs_tensor, model_kwargs, model_input_name)
581 encoder_kwargs["return_dict"] = True
582 encoder_kwargs[model_input_name] = inputs_tensor
--> 583 model_kwargs["encoder_outputs"]: ModelOutput = encoder(**encoder_kwargs)
584
585 return model_kwargs

~.conda\envs\transformers\lib\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 = [], []

~\Desktop\DS_Team_Tasks\NLP\mFLAG\model.py in forward(self, input_ids, attention_mask, fig_ids, head_mask, inputs_embeds, output_attentions, output_hidden_states, return_dict)
228 inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
229
--> 230 embed_pos = self.embed_positions(input_shape)
231
232 hidden_states = inputs_embeds + embed_pos

~.conda\envs\transformers\lib\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 = [], []

~.conda\envs\transformers\lib\site-packages\transformers\models\bart\modeling_bart.py in forward(self, input_ids, past_key_values_length)
132 """`input_ids' shape is expected to be [bsz x seqlen]."""
133
--> 134 bsz, seq_len = input_ids.shape[:2]
135 positions = torch.arange(
136 past_key_values_length, past_key_values_length + seq_len, dtype=torch.long, device=self.weight.device

AttributeError: 'torch.Size' object has no attribute 'shape'

root@DESKTOP-Q3VVK87:/home/cschy/mflag_test# perl multi-bleu.perl test_hyperbole.1 < l2hyperbole.txt
BLEU = 51.47, 68.7/56.4/47.8/40.2 (BP=0.985, ratio=0.985, hyp_len=1332, ref_len=1352)

when I use multi-bleu.perl to calculate bleu, I find the result is lower than your paper given(the same goes for idiom, sarcasm, idiom and simile). l2hyperbole.txt is created by the output of your checkpoint. Did you do anything extra with the output file? Any help would be appreciated. Below is my generated code:

from tokenization_mflag import MFlagTokenizerFast
tokenizer = MFlagTokenizerFast.from_pretrained('laihuiyuan/mFLAG', cache_dir='author')
model = MultiFigurativeGeneration.from_pretrained('laihuiyuan/mFLAG', cache_dir='author')
with open('./data/MultiFG/test_hyperbole.0', 'r') as f, open('./data/outputs/l2hyperbole.txt', 'w') as fw:
    lines = f.readlines()
    for i in lines:
        inp_ids = tokenizer.encode(
            "<literal>" + i.strip(), return_tensors="pt")
        fig_ids = tokenizer.encode("<hyperbole>", add_special_tokens=False, return_tensors="pt")
        outs = model.generate(input_ids=inp_ids[:, 1:], fig_ids=fig_ids, forced_bos_token_id=fig_ids.item(), num_beams=5,
                              max_length=60, )
        text = tokenizer.decode(outs[0, 2:].tolist(), skip_special_tokens=True, clean_up_tokenization_spaces=False)
        print(text)
        fw.write(text + '\n')