AttributeError: module 'sacrebleu' has no attribute '__version__'

[dsj96]

env

Thu May  4 11:28:29 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.80.02    Driver Version: 450.80.02    CUDA Version: 11.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla V100-SXM2...  Off  | 00000000:0B:00.0 Off |                    0 |
| N/A   26C    P0    41W / 163W |      0MiB / 32510MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  Tesla V100-SXM2...  Off  | 00000000:89:00.0 Off |                    0 |
| N/A   31C    P0    41W / 163W |      0MiB / 32510MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

transformers                  4.29.0.dev0         /userhome/dsj/transformers
peft                          0.3.0.dev0          /userhome/dsj/peft
evaluate                      0.4.1.dev0          /userhome/dsj/evaluate/src
deepspeed                     0.9.1+7832efbb      /userhome/dsj/deepspeed
sacrebleu                     2.3.1               /userhome/dsj/sacrebleu

run cmd

model=mt0-small
config_file=ds_zero3_cpu.yaml
dataset_name=./data/fairseq/data/wmt10/subset_subset_instruction_train_dev_tst
raw_dataset_name=wmt10
r=8
lora_alpha=32
lora_dropout=0.1
text_column=SRC
label_column=TGT
lr=5e-5
num_epochs=1
batch_size=32
do_test=False
target_max_length=256
GPU=True
max_new_tokens=256
num_beams=1
no_repeat_ngram_size=3

model_name_or_path=./pre_trained_model/$model
output_dir=./trained_checkpoint/model-$model-raw_dataset_name-$raw_dataset_name-r-$r-lora_alpha-$lora_alpha-lora_dropout-$lora_dropout-text_column-$text_column-label_column-$label_column-lr-$lr-num_epochs-$num_epochs-batch_size-$batch_size-target_max_length-$target_max_length-GPU-$GPU-max_new_tokens-$max_new_tokens-num_beams-$num_beams-no_repeat_ngram_size-$no_repeat_ngram_size
echo $output_dir
accelerate launch --config_file=$config_file examples/mt0_peft_lora_ds_zero3_offload.py --model_name_or_path=$model_name_or_path --dataset_name=$dataset_name\
                    --r=$r --lora_alpha=$lora_alpha --lora_dropout=$lora_dropout --text_column=$text_column --label_column=$label_column \
                    --lr=$lr --num_epochs=$num_epochs --batch_size=$batch_size\
                    --target_max_length=$target_max_length --GPU  --output_dir=$output_dir --do_test

script

python script modified from this link

    metric = evaluate.load("sacrebleu")
    eval_result = metric.compute(predictions=eval_preds, references=dataset['validation'][label_column])
    logger.info({"bleu": eval_result["score"]})
    accelerator.print(f"{eval_result=}")

bug repo

when I run the program, I met the following bug repo,

Traceback (most recent call last):
    File "examples/mt0_peft_lora_ds_zero3_offload.py", line 419, in <module>
      main()
    File "examples/mt0_peft_lora_ds_zero3_offload.py", line 347, in main
      metric = evaluate.load("sacrebleu")
    File "/userhome/dsj/evaluate/src/evaluate/loading.py", line 751, in load
      evaluation_instance = evaluation_cls(
    File "/userhome/dsj/evaluate/src/evaluate/module.py", line 191, in __init__
  Traceback (most recent call last):
    File "examples/mt0_peft_lora_ds_zero3_offload.py", line 419, in <module>
      info = self._info()
    File "/root/.cache/huggingface/modules/evaluate_modules/metrics/evaluate-metric--sacrebleu/28676bf65b4f88b276df566e48e603732d0b4afd237603ebdf92acaacf5be99b/sacrebleu.py", line 108, in _info
      if version.parse(scb.__version__) < version.parse("1.4.12"):
  AttributeError: module 'sacrebleu' has no attribute '__version__'
      main()
    File "examples/mt0_peft_lora_ds_zero3_offload.py", line 347, in main
      metric = evaluate.load("sacrebleu")
    File "/userhome/dsj/evaluate/src/evaluate/loading.py", line 751, in load
      evaluation_instance = evaluation_cls(
    File "/userhome/dsj/evaluate/src/evaluate/module.py", line 191, in __init__
      info = self._info()
    File "/root/.cache/huggingface/modules/evaluate_modules/metrics/evaluate-metric--sacrebleu/28676bf65b4f88b276df566e48e603732d0b4afd237603ebdf92acaacf5be99b/sacrebleu.py", line 108, in _info
      if version.parse(scb.__version__) < version.parse("1.4.12"):
  AttributeError: module 'sacrebleu' has no attribute '__version__'

expect

I have tried to change the sacrebleu and evaluate version, delete cache file (in /root/.cache/huggingface/modules/evaluate_modules/metrics), and add parameter evaluate.load("sacrebleu",cache_dir='../huggface_cache') , but it doesn't seem to work. Could you give me some advice?

But it works when I run line by line. This makes me very confused

Python 3.8.10 | packaged by conda-forge | (default, May 11 2021, 07:01:05) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import evaluate
>>> metric = evaluate.load("sacrebleu")