TypeError: 'type' object is not subscriptable #23472

flckv · 2023-05-19T14:11:45Z

System Info

** pre-training wav2vec demo**
https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-pretraining/README.md running demo gives error:
File “./run_wav2vec2_pretraining_no_trainer.py", line 783, in
main()
File “./run_wav2vec2_pretraining_no_trainer.py", line 510, in main
vectorized_datasets = raw_datasets.map(
TypeError: 'type' object is not subscriptable

Who can help?

@sanchit-gandhi
@pacman100
@sgugger

Information

The official example scripts
My own modified scripts

Tasks

An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
My own task or dataset (give details below)

Reproduction

just reproducing demo example with provided script and dataset:
https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-pretraining/README.md#demo

Expected behavior

output should be a pre-trained wav2vec model on librispeech dataset

The text was updated successfully, but these errors were encountered:

amyeroberts · 2023-05-19T17:32:40Z

Hi @flckv, thanks for raising this error.

I'm unable to reproduce this error when I run locally on the main branch. Could you share the running environment being used: run transformers-cli env in the terminal and copy-paste the output?

flckv · 2023-05-20T05:06:33Z

Hi @amyeroberts, thanks for the quick reply.

The output of transformers-cli env:

- `transformers` version: 4.26.1
- Platform: Linux-5.4.204-ql-generic-12.0-19-x86_64-with-glibc2.31
- Python version: 3.9.7
- Huggingface_hub version: 0.10.1
- PyTorch version (GPU?): 1.11.0 (True)
- Tensorflow version (GPU?): not installed (NA)
- Flax version (CPU?/GPU?/TPU?): not installed (NA)
- Jax version: not installed
- JaxLib version: not installed
- Using GPU in script?: <fill in>
- Using distributed or parallel set-up in script?: <fill in>

I am running on a cluster with resources:

#SBATCH --job-name=ol            # Job name
#SBATCH --output=/home/flck/output_.%A.txt   # Standard output and error log
#SBATCH --nodes=1                   # Run all processes on a single node    
#SBATCH --ntasks=1                  # Run on a single CPU
#SBATCH --mem=64G                   # Total RAM to be used
#SBATCH --cpus-per-task=8          # Number of CPU cores
#SBATCH --gres=gpu:3                # Number of GPUs (per node)
#SBATCH -p gpu                      # Use the gpu partition
#SBATCH --time=12:00:00             # Specify the time needed for your experiment
#SBATCH --qos=gpu-8                 # To enable the use of up to 8 GPUs

in my .sh file that I run on this cluster has these commands to reproduce the demo :

transformers-cli env
accelerate launch wav2vec/run_wav2vec2_pretraining_no_trainer.py --cache_dir="/dev/shm/" --dataset_name="librispeech_asr" --dataset_config_names clean clean --dataset_split_names validation test --model_name_or_path="patrickvonplaten/wav2vec2-base-v2" --output_dir="./wav2vec2-pretrained-demo" --max_train_steps="20000" --num_warmup_steps="32000" --gradient_accumulation_steps="8" --learning_rate="0.005" --weight_decay="0.01" --max_duration_in_seconds="20.0" --min_duration_in_seconds="2.0" --logging_steps="1" --saving_steps="10000" --per_device_train_batch_size="8" --per_device_eval_batch_size="8" --adam_beta1="0.9" --adam_beta2="0.98" --adam_epsilon="1e-06" --gradient_checkpointing --mask_time_prob="0.65" --mask_time_length="10"
transformers-cli env

Is this what you are asking for?

My guess

I think the error is coming from the fact that the dataset preprocessing (line 473) requires argument "args.audio_column_name" that is not specified in the demo command https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-pretraining/README.md#demo.

1. I tried specifying --audio_column_name= []

I got this error:

 _-- schema metadata --
huggingface: '{"info": {"features": {"id": {"dtype": "string", "_type": "' + 163
to
{'id': Value(dtype='string', id=None), 'audio': Audio(sampling_rate=16000, mono=True, decode=True, id=None), 'duration_ms': Value(dtype='int32', id=None), 'text': Value(dtype='string', id=None), '[]': Audio(sampling_rate=16000, mono=True, decode=True, id=None)}
because column names don't match_

2. I tried specifying --audio_column_name=["audio", "duration_ms", "text"]

error: "run_wav2vec2_pretraining_no_trainer.py: error: unrecognized arguments: duration_ms, text]"

3. I tried specifying --audio_column_name=["audio"], which is the default setting

same issue as in 1.

line 478, in main
    raw_datasets = raw_datasets.cast_column(
raise ValueError(f"Couldn't cast\n{table.schema}\nto\n{features}\nbecause column names don't match")
ValueError: Couldn't cast
id: string
audio: struct<bytes: binary, path: string>
  child 0, bytes: binary
  child 1, path: string
duration_ms: int32
text: string
-- schema metadata --
huggingface: '{"info": {"features": {"id": {"dtype": "string", "_type": "' + 163
to
{'id': Value(dtype='string', id=None), 'audio': Audio(sampling_rate=16000, mono=True, decode=True, id=None), 'duration_ms': Value(dtype='int32', id=None), 'text': Value(dtype='string', id=None), '[audio]': Audio(sampling_rate=16000, mono=True, decode=True, id=None)}
because column names don't match

Any ideas? @amyeroberts @sanchit-gandhi @pacman100 @sgugger

flckv · 2023-05-20T11:11:40Z

here is a more detailed output log content:

The following values were not passed to `accelerate launch` and had defaults used instead:
	`--num_processes` was set to a value of `3`
		More than one GPU was found, enabling multi-GPU training.
		If this was unintended please pass in `--num_processes=1`.
	`--num_machines` was set to a value of `1`
	`--mixed_precision` was set to a value of `'no'`
	`--dynamo_backend` was set to a value of `'no'`
To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
wandb: Currently logged in as: flckv. Use `wandb login --relogin` to force relogin
wandb: wandb version 0.15.3 is available!  To upgrade, please run:
wandb:  $ pip install wandb --upgrade
wandb: Tracking run with wandb version 0.15.2
wandb: Run data is saved locally in /home/flckv/wandb/run-20230519_175326-yuyk0qvn
wandb: Run `wandb offline` to turn off syncing.
wandb: Syncing run rural-morning-6
wandb: ⭐️ View project at https://wandb.ai/flckv/wav2vec2-pretrained-demo
wandb: 🚀 View run at https://wandb.ai/flckv/wav2vec2-pretrained-demo/runs/yuyk0qvn
Downloading and preparing dataset librispeech_asr/clean to /dev/shm/librispeech_asr/clean/2.1.0/cff5df6e7955c80a67f80e27e7e655de71c689e2d2364bece785b972acb37fe7...

Downloading data files:   0%|          | 0/4 [00:00<?, ?it/s]
Downloading data files: 100%|██████████| 4/4 [00:00<00:00, 9828.48it/s]

Extracting data files:   0%|          | 0/4 [00:00<?, ?it/s]
Extracting data files: 100%|██████████| 4/4 [00:00<00:00, 2225.39it/s]


Generating train.100 split: 100%|██████████| 28539/28539 [00:17<00:00, 1834.73 examples/s]
                                                                                          

Generating train.360 split: 100%|██████████| 104014/104014 [01:00<00:00, 1634.61 examples/s]
                                                                                            

Generating validation split: 100%|██████████| 2703/2703 [00:01<00:00, 2341.79 examples/s]
                                                                                         

Generating test split:  93%|█████████▎| 2434/2620 [00:01<00:00, 2338.24 examples/s]
                                                                                   
Dataset librispeech_asr downloaded and prepared to /dev/shm/librispeech_asr/clean/2.1.0/cff5df6e7955c80a67f80e27e7e655de71c689e2d2364bece785b972acb37fe7. Subsequent calls will reuse this data.
Found cached dataset librispeech_asr (/dev/shm/librispeech_asr/clean/2.1.0/cff5df6e7955c80a67f80e27e7e655de71c689e2d2364bece785b972acb37fe7)
Found cached dataset librispeech_asr (/dev/shm/librispeech_asr/clean/2.1.0/cff5df6e7955c80a67f80e27e7e655de71c689e2d2364bece785b972acb37fe7)

Downloading:   0%|          | 0.00/214 [00:00<?, ?B/s]
Downloading: 100%|██████████| 214/214 [00:00<00:00, 171kB/s]
loading configuration file preprocessor_config.json from cache at /home/flckv/.cache/huggingface/hub/models--patrickvonplaten--wav2vec2-base-v2/snapshots/9371f1849947b4613f451680a8e96d907617ce86/preprocessor_config.json
Feature extractor Wav2Vec2FeatureExtractor {
  "do_normalize": true,
  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
  "feature_size": 1,
  "padding_side": "right",
  "padding_value": 0.0,
  "return_attention_mask": true,
  "sampling_rate": 16000
}


Map:   0%|          | 0/5270 [00:00<?, ? examples/s]
Map:   0%|          | 0/5270 [00:01<?, ? examples/s]
                                                    

> 
> Traceback (most recent call last):
>   File "/home/flckv/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 783, in <module>
>     main()
>   File "/home/flckv/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 510, in main
>     vectorized_datasets = raw_datasets.map(
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/dataset_dict.py", line 852, in map
>     {
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/dataset_dict.py", line 853, in <dictcomp>
>     k: dataset.map(
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 563, in wrapper
>     out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 528, in wrapper
>     out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2953, in map
>     for rank, done, content in Dataset._map_single(**dataset_kwargs):
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 3307, in _map_single
>     example = apply_function_on_filtered_inputs(example, i, offset=offset)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 3210, in apply_function_on_filtered_inputs
>     processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)
>   File "/home/flckv/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 493, in prepare_dataset
>     sample = batch[args.audio_column_name]
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 282, in __getitem__
>     value = self.format(key)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 380, in format
>     return self.formatter.format_column(self.pa_table.select([key]))[0]
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 447, in format_column
>     column = self.python_features_decoder.decode_column(column, pa_table.column_names[0])
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 228, in decode_column
>     return self.features.decode_column(column, column_name) if self.features else column
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/features/features.py", line 1866, in decode_column
>     [decode_nested_example(self[column_name], value) if value is not None else None for value in column]
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/features/features.py", line 1866, in <listcomp>
>     [decode_nested_example(self[column_name], value) if value is not None else None for value in column]
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/features/features.py", line 1308, in decode_nested_example
>     return schema.decode_example(obj, token_per_repo_id=token_per_repo_id)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/features/audio.py", line 164, in decode_example
>     array, sampling_rate = self._decode_non_mp3_file_like(file)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/features/audio.py", line 290, in _decode_non_mp3_file_like
>     array = librosa.to_mono(array)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/lazy_loader/__init__.py", line 77, in __getattr__
>     attr = getattr(submod, name)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/lazy_loader/__init__.py", line 76, in __getattr__
>     submod = importlib.import_module(submod_path)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/importlib/__init__.py", line 127, in import_module
>     return _bootstrap._gcd_import(name[level:], package, level)
>   File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
>   File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
>   File "<frozen importlib._bootstrap>", line 986, in _find_and_load_unlocked
>   File "<frozen importlib._bootstrap>", line 680, in _load_unlocked
>   File "<frozen importlib._bootstrap_external>", line 850, in exec_module
>   File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/librosa/core/audio.py", line 19, in <module>
>     from .convert import frames_to_samples, time_to_samples
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/librosa/core/convert.py", line 7, in <module>
>     from . import notation
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/librosa/core/notation.py", line 8, in <module>
>     from .intervals import INTERVALS
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/librosa/core/intervals.py", line 10, in <module>
>     from numpy.typing import ArrayLike
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/numpy/typing/__init__.py", line 158, in <module>
>     from numpy._typing import (
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/numpy/_typing/__init__.py", line 164, in <module>
>     from ._dtype_like import (
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/numpy/_typing/_dtype_like.py", line 17, in <module>
>     from ._generic_alias import _DType as DType
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/numpy/_typing/_generic_alias.py", line 241, in <module>
>     _DType = np.dtype[ScalarType]
> TypeError: 'type' object is not subscriptable
> wandb: Waiting for W&B process to finish... (failed 1). Press Control-C to abort syncing.
> wandb: 🚀 View run rural-morning-6 at: https://wandb.ai/flckv/wav2vec2-pretrained-demo/runs/yuyk0qvn
> wandb: Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)
> wandb: Find logs at: ./wandb/run-20230519_175326-yuyk0qvn/logs
> WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 3914267 closing signal SIGTERM
> WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 3914268 closing signal SIGTERM
> ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 3914266) of binary: /home/flckv/.conda/envs/vcheckworthy/bin/python
> Traceback (most recent call last):
>   File "/home/flckv/.conda/envs/vcheckworthy/bin/accelerate", line 8, in <module>
>     sys.exit(main())
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/accelerate_cli.py", line 45, in main
>     args.func(args)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/launch.py", line 909, in launch_command
>     multi_gpu_launcher(args)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/launch.py", line 604, in multi_gpu_launcher
>     distrib_run.run(args)
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/torch/distributed/run.py", line 715, in run
>     elastic_launch(
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 131, in __call__
>     return launch_agent(self._config, self._entrypoint, list(args))
>   File "/home/flckv/.conda/envs/vcheckworthy/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 245, in launch_agent
>     raise ChildFailedError(
> torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
> 
> wav2vec/run_wav2vec2_pretraining_no_trainer.py FAILED
> ------------------------------------------------------------
> Failures:
>   <NO_OTHER_FAILURES>
> ------------------------------------------------------------
> Root Cause (first observed failure):
> [0]:
>   time      : 2023-05-19_17:55:10
>   host      : gpu-08
>   rank      : 0 (local_rank: 0)
>   exitcode  : 1 (pid: 3914266)
>   error_file: <N/A>
>   traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
> 
> /var/lib/slurm-llnl/slurmd/job151161/slurm_script: line 42: EOL: command not found
>

sanchit-gandhi · 2023-05-22T15:58:26Z

Hey @flckv! Could you try first updating all your packages to the latest versions?

pip install --upgrade pip
pip install --upgrade soundfile librosa datasets accelerate numpy transformers

The error looks like it's happening when we decode the soundfile (i.e. as we read the soundfile with librosa) - there was recently a big change to how we load audio samples with datasets that might fix this for you huggingface/datasets#5573

flckv · 2023-05-23T09:27:24Z

@sanchit-gandhi Thanks, but now the command is not working:

accelerate launch wav2vec/run_wav2vec2_pretraining_no_trainer.py --cache_dir="/dev/shm/" --dataset_name="librispeech_asr" --dataset_config_names test --dataset_split_names test --model_name_or_path="patrickvonplaten/wav2vec2-base-v2" --output_dir="./wav2vec2-pretrained-demo" --max_train_steps="20000" --num_warmup_steps="32000" --gradient_accumulation_steps="8" --learning_rate="0.005" --weight_decay="0.01" --max_duration_in_seconds="20.0" --min_duration_in_seconds="2.0" --logging_steps="1" --saving_steps="10000" --per_device_train_batch_size="8" --per_device_eval_batch_size="8" --adam_beta1="0.9" --adam_beta2="0.98" --adam_epsilon="1e-06" --gradient_checkpointing --mask_time_prob="0.65" --mask_time_length="10"

ERROR:

Traceback (most recent call last):
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 785, in
main()
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 513, in main
prepare_dataset(raw_datasets["train"]), # loading the audio raise KeyError(f"Column {key} not in the dataset. Current columns in the dataset: {columns}") KeyError: "Column args.audio_column_name not in the dataset. Current columns in the dataset: ['id', 'audio', 'duration_ms', 'text']"
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 493, in prepare_dataset
sample = batch['args.audio_column_name']
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2778, in getitem
return self._getitem(key)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2762, in _getitem
pa_subtable = query_table(self._data, key, indices=self._indices if self._indices is not None else None)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 575, in query_table
_check_valid_column_key(key, table.column_names)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 515, in _check_valid_column_key

raise KeyError(f"Column {key} not in the dataset. Current columns in the dataset: {columns}")
KeyError: "Column args.audio_column_name not in the dataset. Current columns in the dataset: ['id', 'audio', 'duration_ms', 'text']"

Traceback (most recent call last):
File "/home/flck/.conda/envs/vcheckworthy/bin/accelerate", line 8, in
sys.exit(main())
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/accelerate_cli.py", line 45, in main
args.func(args)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/launch.py", line 918, in launch_command
simple_launcher(args)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/launch.py", line 580, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['/home/flck/.conda/envs/vcheckworthy/bin/python', 'wav2vec/run_wav2vec2_pretraining_no_trainer.py', '--cache_dir=/dev/shm/', '--dataset_name=librispeech_asr', '--dataset_config_names', 'test', '--dataset_split_names', 'test', '--model_name_or_path=patrickvonplaten/wav2vec2-base-v2', '--output_dir=./wav2vec2-pretrained-demo', '--max_train_steps=20000', '--num_warmup_steps=32000', '--gradient_accumulation_steps=8', '--learning_rate=0.005', '--weight_decay=0.01', '--max_duration_in_seconds=20.0', '--min_duration_in_seconds=2.0', '--logging_steps=1', '--saving_steps=10000', '--per_device_train_batch_size=8', '--per_device_eval_batch_size=8', '--adam_beta1=0.9', '--adam_beta2=0.98', '--adam_epsilon=1e-06', '--gradient_checkpointing', '--mask_time_prob=0.65', '--mask_time_length=10']' returned non-zero exit status 1.

/var/lib/slurm-llnl/slurmd/job153086/slurm_script: line 45: EOL: command not found

WHEN I specify this in the command args

accelerate launch wav2vec/run_wav2vec2_pretraining_no_trainer.py --cache_dir="/dev/shm/" --dataset_name="librispeech_asr" --dataset_config_names test --dataset_split_names test --model_name_or_path="patrickvonplaten/wav2vec2-base-v2" --output_dir="./wav2vec2-pretrained-demo"--audio_column_name=["id", "audio", "duration_ms", "text"]--max_train_steps="20000" --num_warmup_steps="32000" --gradient_accumulation_steps="8" --learning_rate="0.005" --weight_decay="0.01" --max_duration_in_seconds="20.0" --min_duration_in_seconds="2.0" --logging_steps="1" --saving_steps="10000" --per_device_train_batch_size="8" --per_device_eval_batch_size="8" --adam_beta1="0.9" --adam_beta2="0.98" --adam_epsilon="1e-06" --gradient_checkpointing --mask_time_prob="0.65" --mask_time_length="10"

then the error is:

run_wav2vec2_pretraining_no_trainer.py: error: unrecognized arguments: audio, duration_ms, text]

WHEN I only add "id":
accelerate launch wav2vec/run_wav2vec2_pretraining_no_trainer.py --cache_dir="/dev/shm/" --dataset_name="librispeech_asr" --dataset_config_names test --dataset_split_names test --model_name_or_path="patrickvonplaten/wav2vec2-base-v2" --output_dir="./wav2vec2-pretrained-demo"--audio_column_name=["id"]--max_train_steps="20000" --num_warmup_steps="32000" --gradient_accumulation_steps="8" --learning_rate="0.005" --weight_decay="0.01" --max_duration_in_seconds="20.0" --min_duration_in_seconds="2.0" --logging_steps="1" --saving_steps="10000" --per_device_train_batch_size="8" --per_device_eval_batch_size="8" --adam_beta1="0.9" --adam_beta2="0.98" --adam_epsilon="1e-06" --gradient_checkpointing --mask_time_prob="0.65" --mask_time_length="10"

Traceback (most recent call last):
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 785, in
main()
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 478, in main
raw_datasets = raw_datasets.cast_column(
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/dataset_dict.py", line 310, in cast_column
return DatasetDict({k: dataset.cast_column(column=column, feature=feature) for k, dataset in self.items()})
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/dataset_dict.py", line 310, in
return DatasetDict({k: dataset.cast_column(column=column, feature=feature) for k, dataset in self.items()})
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/fingerprint.py", line 511, in wrapper
out = func(dataset, *args, **kwargs)
File "/home/flck.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2082, in cast_column
dataset._data = dataset._data.cast(dataset.features.arrow_schema)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/table.py", line 1152, in cast
return MemoryMappedTable(table_cast(self.table, *args, **kwargs), self.path, replays)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/table.py", line 2290, in table_cast
return cast_table_to_schema(table, schema)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/table.py", line 2248, in cast_table_to_schema
raise ValueError(f"Couldn't cast\n{table.schema}\nto\n{features}\nbecause column names don't match")

ValueError: Couldn't cast
id: string
audio: struct<bytes: binary, path: string>
  child 0, bytes: binary
  child 1, path: string
duration_ms: int32
text: string

-- schema metadata --
huggingface: '{"info": {"features": {"id": {"dtype": "string", "_type": "' + 163
to
{'id': Value(dtype='string', id=None), 'audio': Audio(sampling_rate=16000, mono=True, decode=True, id=None), 'duration_ms': Value(dtype='int32', id=None), 'text': Value(dtype='string', id=None), '[id]': Audio(sampling_rate=16000, mono=True, decode=True, id=None)}
because column names don't match
wandb: Waiting for W&B process to finish... (failed 1). Press Control-C to abort syncing.
wandb: - 0.028 MB of 0.028 MB uploaded (0.000 MB deduped)
wandb: \ 0.028 MB of 0.032 MB uploaded (0.000 MB deduped)
wandb: | 0.035 MB of 0.035 MB uploaded (0.000 MB deduped)
wandb: 🚀 View run helpful-voice-29 at: https://wandb.ai/flck/wav2vec2-pretrained-demo/runs/tnmnebg6
wandb: Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)
wandb: Find logs at: ./wandb/run-20230523_133136-tnmnebg6/logs
Traceback (most recent call last):
File "/home/flck/.conda/envs/vcheckworthy/bin/accelerate", line 8, in
sys.exit(main())
File "/home/flck.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/accelerate_cli.py", line 45, in main
args.func(args)
File "/homeflck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/launch.py", line 918, in launch_command
simple_launcher(args)
File "/home/flck.conda/envs/vcheckworthy/lib/python3.9/site-packages/accelerate/commands/launch.py", line 580, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['/home/flck/.conda/envs/vcheckworthy/bin/python', 'wav2vec/run_wav2vec2_pretraining_no_trainer.py', '--cache_dir=/dev/shm/', '--dataset_name=librispeech_asr', '--dataset_config_names', 'test', '--dataset_split_names', 'test', '--model_name_or_path=patrickvonplaten/wav2vec2-base-v2', '--output_dir=./wav2vec2-pretrained-demo', '--audio_column_name=[id]', '--max_train_steps=20000', '--num_warmup_steps=32000', '--gradient_accumulation_steps=8', '--learning_rate=0.005', '--weight_decay=0.01', '--max_duration_in_seconds=20.0', '--min_duration_in_seconds=2.0', '--logging_steps=1', '--saving_steps=10000', '--per_device_train_batch_size=8', '--per_device_eval_batch_size=8', '--adam_beta1=0.9', '--adam_beta2=0.98', '--adam_epsilon=1e-06', '--gradient_checkpointing', '--mask_time_prob=0.65', '--mask_time_length=10']' returned non-zero exit status 1.

Copy-and-paste the text below in your GitHub issue and FILL OUT the two last points.

transformers version: 4.29.2
Platform: Linux-5.4.204-ql-generic-12.0-19-x86_64-with-glibc2.31
Python version: 3.9.7
Huggingface_hub version: 0.14.1
Safetensors version: not installed
PyTorch version (GPU?): 1.11.0 (False)
Tensorflow version (GPU?): not installed (NA)
Flax version (CPU?/GPU?/TPU?): not installed (NA)
Jax version: not installed
JaxLib version: not installed
Using GPU in script?:
Using distributed or parallel set-up in script?:

/var/lib/slurm-llnl/slurmd/job153092/slurm_script: line 50: EOL: command not found
/var/lib/slurm-llnl/slurmd/job153092/slurm_script: line 53: /home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py: Permission denied

sanchit-gandhi · 2023-05-23T16:45:59Z

Hey @flckv - great! Glad updating to the latest packages fixed the previous error. Can you try setting:

--audio_column_name="audio"

Here we just need to pick-out the correct column name for the audio inputs (which in this case is "audio")

flckv · 2023-05-24T04:21:56Z

hey @sanchit-gandhi yes it is great! the column name is still not interpreted :

I added what you said:

accelerate launch wav2vec/run_wav2vec2_pretraining_no_trainer.py --cache_dir="/dev/shm/" --dataset_name="librispeech_asr" --dataset_config_names test --dataset_split_names test --model_name_or_path="patrickvonplaten/wav2vec2-base-v2" --output_dir="./wav2vec2-pretrained-demo" --audio_column_name="audio" --max_train_steps="20000" --num_warmup_steps="32000" --gradient_accumulation_steps="8" --learning_rate="0.005" --weight_decay="0.01" --max_duration_in_seconds="20.0" --min_duration_in_seconds="2.0" --logging_steps="1" --saving_steps="10000" --per_device_train_batch_size="8" --per_device_eval_batch_size="8" --adam_beta1="0.9" --adam_beta2="0.98" --adam_epsilon="1e-06" --gradient_checkpointing --mask_time_prob="0.65" --mask_time_length="10"

tried also:

--audio_column_name='audio'
--audio_column_name=['audio']
--audio_column_name=["audio"]

BUT :

Traceback (most recent call last):
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 785, in
main()
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 513, in main
prepare_dataset(raw_datasets["train"]), raise KeyError(f"Column {key} not in the dataset. Current columns in the dataset: {columns}") KeyError: "Column args.audio_column_name not in the dataset. Current columns in the dataset: ['id', 'audio', 'duration_ms', 'text']"
File "/home/flck/wav2vec/run_wav2vec2_pretraining_no_trainer.py", line 493, in prepare_dataset
sample = batch['args.audio_column_name']
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2778, in getitem
return self._getitem(key)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2762, in _getitem
pa_subtable = query_table(self._data, key, indices=self._indices if self._indices is not None else None)
File "/home/flcks/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 575, in query_table
_check_valid_column_key(key, table.column_names)
File "/home/flck/.conda/envs/vcheckworthy/lib/python3.9/site-packages/datasets/formatting/formatting.py", line 515, in _check_valid_column_key
raise KeyError(f"Column {key} not in the dataset. Current columns in the dataset: {columns}")
KeyError: "Column args.audio_column_name not in the dataset. Current columns in the dataset: ['id', 'audio', 'duration_ms', 'text']"

sanchit-gandhi · 2023-05-24T15:36:34Z

Can you double check you haven't changed the parser args for audio_column_name?

transformers/examples/pytorch/speech-pretraining/run_wav2vec2_pretraining_no_trainer.py

Lines 112 to 117 in 50a56be

    
           parser.add_argument( 
        
               "--audio_column_name", 
        
               type=str, 
        
               default="audio", 
        
               help="Column in the dataset that contains speech file path. Defaults to 'audio'", 
        
           )

I can't see the check that is erroring out for you on the example script. Your error is occurring on line 513. If I check line 513 in the example, I get something completely different to the audio column name check:

transformers/examples/pytorch/speech-pretraining/run_wav2vec2_pretraining_no_trainer.py

Line 513 in 3d7baef

lambda x: x > min_length,

Could you make sure you are using the latest version of the script? You can just copy it from main.

flckv · 2023-05-31T05:26:44Z

@sanchit-gandhi thanks, you were right. It works now.

flckv closed this as completed May 31, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TypeError: 'type' object is not subscriptable #23472

TypeError: 'type' object is not subscriptable #23472

flckv commented May 19, 2023

amyeroberts commented May 19, 2023

flckv commented May 20, 2023 •

edited

Loading

flckv commented May 20, 2023 •

edited by sanchit-gandhi

Loading

sanchit-gandhi commented May 22, 2023 •

edited

Loading

flckv commented May 23, 2023 •

edited

Loading

sanchit-gandhi commented May 23, 2023

flckv commented May 24, 2023 •

edited

Loading

sanchit-gandhi commented May 24, 2023

flckv commented May 31, 2023

TypeError: 'type' object is not subscriptable #23472

TypeError: 'type' object is not subscriptable #23472

Comments

flckv commented May 19, 2023

System Info

Who can help?

Information

Tasks

Reproduction

Expected behavior

amyeroberts commented May 19, 2023

flckv commented May 20, 2023 • edited Loading

The output of transformers-cli env:

My guess

1. I tried specifying --audio_column_name= []

2. I tried specifying --audio_column_name=["audio", "duration_ms", "text"]

3. I tried specifying --audio_column_name=["audio"], which is the default setting

flckv commented May 20, 2023 • edited by sanchit-gandhi Loading

sanchit-gandhi commented May 22, 2023 • edited Loading

flckv commented May 23, 2023 • edited Loading

sanchit-gandhi commented May 23, 2023

flckv commented May 24, 2023 • edited Loading

sanchit-gandhi commented May 24, 2023

flckv commented May 31, 2023

flckv commented May 20, 2023 •

edited

Loading

flckv commented May 20, 2023 •

edited by sanchit-gandhi

Loading

sanchit-gandhi commented May 22, 2023 •

edited

Loading

flckv commented May 23, 2023 •

edited

Loading

flckv commented May 24, 2023 •

edited

Loading