mikewangwzhl/eeg-to-text

code for AAAI2022 paper "Open Vocabulary Electroencephalography-To-Text Decoding and Zero-shot Sentiment Classification"

eeg-to-text's Issues

Pretrained weights for step 1 not found

I was trying to reproduce the paper and the pretrained weights for step 1 'path_to_step_1_checkpoint.pt' was not in the github. Could you please upload the weights?

Issues with Evaluation Scripts

This issue is related to commit fbaf82d

After running the script: bash ./scripts/eval_decoding.sh , the results came out to be:

corpus BLEU-1 score: 0
corpus BLEU-2 score: 0
corpus BLEU-3 score: 0
corpus BLEU-4 score: 0

{'rouge-1': {'r': 0.0960104371521744, 'p': 0.13671808632706614, 'f': 0.10633835733307583}, 'rouge-2': {'r': 0.011719396402741052, 'p': 0.013988694184239035, 'f': 0.01133032845861094}, 'rouge-l': {'r': 0.09090843088332022, 'p': 0.12862700453138184, 'f': 0.10046980133298505}}

---

Removing the .squeeze and .tolist may have some affect on the results...

I'll be working on this as well @MikeWangWZHL , thanks for acting fast!

ERROR in README?

There is commit dated Jan 19th 2024 but in the README the update date is 19th Jan 2023. Is it a typo?

Issues with eval script

After I run the script: bash ./scripts/eval_decoding.sh, the message is shown below:

TypeError: BartEncoder() got multiple values for keyword argument 'return_dict'
This error seems to indicate that more than one of the same keyword argument is passed when calling the BartAttention module.

---

The detailed output message is as follows:

Traceback (most recent call last):
File "/EEG-To-Text/eval_decoding.py", line 196, in
eval_model(dataloaders, device, tokenizer, criterion, model, output_all_results_path = output_all_results_path)
File "/EEG-To-Text/eval_decoding.py", line 58, in eval_model
predictions=model.generate(input_embeddings_batch, input_masks_batch, input_mask_invert_batch, target_ids_batch,
File "/anaconda3/envs/EEGToText/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/EEG-To-Text/model_decoding.py", line 52, in generate
output=self.pretrained.generate(
File "/anaconda3/envs/EEGToText/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/anaconda3/envs/EEGToText/lib/python3.9/site-packages/transformers/generation_utils.py", line 906, in generate
model_kwargs = self._prepare_encoder_decoder_kwargs_for_generation(input_ids, model_kwargs)
File "/anaconda3/envs/EEGToText/lib/python3.9/site-packages/transformers/generation_utils.py", line 414, in _prepare_encoder_decoder_kwargs_for_generation
model_kwargs["encoder_outputs"]: ModelOutput = encoder(input_ids, return_dict=True, **encoder_kwargs)
TypeError: BartEncoder(
(embed_tokens): Embedding(50265, 1024, padding_idx=1)
(embed_positions): BartLearnedPositionalEmbedding(1026, 1024)
(layers): ModuleList(
(0): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(1): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(2): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(3): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(4): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(5): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(6): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(7): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(8): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(9): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(10): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
(11): BartEncoderLayer(
(self_attn): BartAttention(
(k_proj): Linear(in_features=1024, out_features=1024, bias=True)
(v_proj): Linear(in_features=1024, out_features=1024, bias=True)
(q_proj): Linear(in_features=1024, out_features=1024, bias=True)
(out_proj): Linear(in_features=1024, out_features=1024, bias=True)
)
(self_attn_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
(fc1): Linear(in_features=1024, out_features=4096, bias=True)
(fc2): Linear(in_features=4096, out_features=1024, bias=True)
(final_layer_norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
)
)
(layernorm_embedding): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
) got multiple values for keyword argument 'return_dict'

Question about trainning

Hello! Recently I prepared ZuCo and ZuCo 2.0 dataset, and process them as said in readme. And I start trainning with the default settings. At last,when I access to the decoded text. I find they are basically the same sentence. After that I check the model output. And I find that with various eeg embeddings in, the output are almost the same. And the loss isn't going down. I think I didn't change any code, and I also didn't find any bug in the program. So I wonder if you have meet the same problem, thank you!

Config files are missing

when I run the bash ./scripts/train_decoding.sh, I'm getting this error:
Traceback (most recent call last): File "train_decoding.py", line 229, in <module> with open(f'./config/decoding/{save_name}.json', 'w') as out_config: FileNotFoundError: [Errno 2] No such file or directory: './config/decoding/task1_task2_taskNRv2_finetune_BrainTranslator_skipstep1_b32_20_30_5e-05_5e-07_unique_sent.json'

This original code has a unintentialal but very serious mistake

more details at https://github.com/NeuSpeech/EEG-To-Text

Datset Preprocessing

Hello. As far as I understand, you are storing the data in a pandas dataframe with one column corressponding to EEG signals and the other to text and then converting EEG signals to text, correct? Could you elaborate more on how you've achieved this dataset format so that others can organize the dataset the same way?

EEG to chinese

how to use it to convert the eeg to chinese word？I want to implement such a function that when I think of a word in my mind, such as “中”,it can quickly recognize the word. what should I do?

Anaconda env issue

Hi,
I'm trying to run your project but I failed in creating anaconda environment. I met this issue:

Pip subprocess error:
ERROR: Could not find a version that satisfies the requirement detectron2==0.5+cu111
ERROR: No matching distribution found for detectron2==0.5+cu111
failed

How should I fix this issue?

Concerns about Incorrect Results and a Suggestion for Paper Withdrawal

I don't mean to offend, but the results reported in your paper are incorrect. If, under the correct testing procedure, the results are completely different from what is stated in your paper, I believe it would be best for you to withdraw it to prevent any potential spread of misinformation.