truncation_strategy is passed twice leading to an error #62
Description
Like this:
In [1]: from transformers import AutoModel, AutoTokenizer
In [2]: from calflops import calculate_flops, calculate_flops_hf
In [3]: model_name = "google/gemma-2b"
In [4]: model = AutoModel.from_pretrained(model_name, cache_dir="hf_cache")
In [5]: tokenizer = AutoTokenizer.from_pretrained(model_name, cache_dir="hf_cache")
In [6]: batch_size, max_seq_length = 1, 128
In [7]: flops, macs, params = calculate_flops(model=model, input_shape=(batch_size, max_seq_length), transformer_tokenizer=tokenizer, print_detailed=True)
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[7], line 1
----> 1 flops, macs, params = calculate_flops(model=model, input_shape=(batch_size, max_seq_length), transformer_tokenizer=tokenizer, print_detailed=True)
File XXX/.venv/lib/python3.13/site-packages/calflops/flops_counter.py:142, in calculate_flops(model, input_shape, transformer_tokenizer, args, kwargs, forward_mode, include_backPropagation, compute_bp_factor, print_results, print_detailed, output_as_string, output_precision, output_unit, ignore_modules)
140 else:
141 assert len(input_shape) == 2, "the format of input_shape must be (batch_size, seq_len) if model is transformers model and auto_generate_transformers_input if True"
--> 142 kwargs = generate_transformer_input(input_shape=input_shape,
143 model_tokenizer=transformer_tokenizer,
144 device=device)
145 else:
146 assert transformer_tokenizer or (len(args) > 0 or len(kwargs) > 0), "input_shape or args or kwargs one of there parameters must specified if auto_generate_input is False"
File XXX/.venv/lib/python3.13/site-packages/calflops/utils.py:115, in generate_transformer_input(model_tokenizer, input_shape, device)
113 inp_seq = ""
114 for _ in range(input_shape[0]):
--> 115 inputs = model_tokenizer.encode_plus(
116 inp_seq,
117 add_special_tokens=True,
118 truncation_strategy='longest_first',
119 )
120 origin_length = len(inputs["input_ids"])
121 padding_length = max_length - origin_length
File XXX/.venv/lib/python3.13/site-packages/transformers/tokenization_utils_base.py:3123, in PreTrainedTokenizerBase.encode_plus(self, text, text_pair, add_special_tokens, padding, truncation, max_length, stride, is_split_into_words, pad_to_multiple_of, padding_side, return_tensors, return_token_type_ids, return_attention_mask, return_overflowing_tokens, return_special_tokens_mask, return_offsets_mapping, return_length, verbose, **kwargs)
3094 """
3095 Tokenize and prepare for the model a sequence or a pair of sequences.
3096
(...) 3111 method).
3112 """
3114 padding_strategy, truncation_strategy, max_length, kwargs = self._get_padding_truncation_strategies(
3115 padding=padding,
3116 truncation=truncation,
(...) 3120 **kwargs,
3121 )
-> 3123 return self._encode_plus(
3124 text=text,
3125 text_pair=text_pair,
3126 add_special_tokens=add_special_tokens,
3127 padding_strategy=padding_strategy,
3128 truncation_strategy=truncation_strategy,
3129 max_length=max_length,
3130 stride=stride,
3131 is_split_into_words=is_split_into_words,
3132 pad_to_multiple_of=pad_to_multiple_of,
3133 padding_side=padding_side,
3134 return_tensors=return_tensors,
3135 return_token_type_ids=return_token_type_ids,
3136 return_attention_mask=return_attention_mask,
3137 return_overflowing_tokens=return_overflowing_tokens,
3138 return_special_tokens_mask=return_special_tokens_mask,
3139 return_offsets_mapping=return_offsets_mapping,
3140 return_length=return_length,
3141 verbose=verbose,
3142 split_special_tokens=kwargs.pop("split_special_tokens", self.split_special_tokens),
3143 **kwargs,
3144 )
TypeError: transformers.tokenization_utils_fast.PreTrainedTokenizerFast._encode_plus() got multiple values for keyword argument 'truncation_strategy'