Add Parakeet Hybrid RNNT CTC BPE Model with target language support by ealbasiri · Pull Request #13360 · NVIDIA-NeMo/NeMo

ealbasiri · 2025-04-30T18:16:43Z

Important

The Update branch button must only be pressed in very rare occassions.
An outdated branch is never blocking the merge of a PR.
Please reach out to the automation team before pressing that button.

What does this PR do ?

This PR add support to Parakeet-based AST model for streaming use case
- Architecture: Parakeet-RNNT with target language ID input. The model utilizes a language vector that is a one-hot vector given language ID that is concatenated to ASR embedding from FastConformer and the concatenated vector is fed into RNNT decoder.
- Supports buffered streaming inference
- Support RNNT offline inference
Supported task: ASR and AST
Supported languages:
- English, Arabic, German, Spanish, French
- Any2Any direction supported.
Function Overview
- Target language ID is required while source language ID is not needed. Given target language ID, the model produces transcription if the source language is the same with target language, otherwise it shows translation in target language.
Benefits of this approach:
- One single model to be loaded E2E
- Source language ID input is not required. Utterances from users can be transcribed or translated given target language input; if the target language is the same as the source language, the model performs an ASR task, if the target language is different, the model performs an AST task.
- Word time-stamp generation is applicable. It can be potentially useful for aligned dataset generation for downstream applications.

Collection: [Note which collection this PR will affect]

ASR

Changelog

This PR changes the following:

Add a new Parakeet model design that support target Language ID as an input
Add support for target language ID input in Lhotse dataloader
Add support for the model in offline inference 'transcribe_speech.py'

Usage

This model can be used for ASR/AST tasks as well as word time-stamp generation for downstream application.

For offline inference: 
python examples/asr/transcribe_speech.py model_path=\"/checkpoints/emea_universal_hybrid_loss_01_val_bleu_2220_epoch_104_step_52501.nemo\" dataset_manifest=\"/manifests/fleurs/fleurs_emea_emea_lhotse.dev.ast_update_lang_id_cross_only_unique.json\" output_filename=\"/manifests/evaluations/hybrid_universal_offline_rnnt/fleurs_emea_emea_lhotse.dev.ast_update_lang_id_cross_only_unique_emea_universal_hybrid_loss_01_val_bleu_2220_epoch_104_step_52501.json\" batch_size=32 cuda=0 amp=True decoder_type=rnnt

For training: 
 python examples/asr/ast_hybrid_transducer_ctc/speech_to_text_hybrid_rnnt_ctc_bpe.py  --config-path=/config/ --config-name=fast-conformer_transducer_bpe_tgt_lang_bleu.yaml

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR.
To re-run CI remove and add the label again.
To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

Before your PR is "Ready for review"

Pre checks:

[Y ] Make sure you read and followed Contributor guidelines
[N/A ] Did you write any new necessary tests?
[Y] Did you add or update any necessary documentation?
[N ] Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
- Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

[Y ] New Feature
Bugfix
Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

@nithinraok
Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

Related to # (issue)

nithinraok

Overall:

I think the implementation can be simplified.
Implementation of langid can be generalized for other usecases as well
Unit tests are missing for the current implementation

nithinraok · 2025-05-05T18:34:41Z

examples/asr/ast_hybrid_transducer_ctc/speech_to_text_hybrid_rnnt_ctc_bpe.py

instead of ast, Its better to align this PR in the lines of prompt_transducer so keep in asr_hybrid_transducer_ctc but change name to speech_to_text_hybrid_rnnt_ctc_bpe_prompt.py. Which means this model code need to be generalized for other use cases as well.

nithinraok · 2025-05-05T18:36:35Z

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

+    def output_types(self) -> Optional[Dict[str, NeuralType]]:
+        return {
+            'audio_signal': NeuralType(('B', 'T'), AudioSignal()),
+            'a_sig_length': NeuralType(tuple('B'), LengthsType()),


keep full name-> audio_signal_length

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

nithinraok · 2025-05-05T18:42:09Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+        # Convert to Hydra 1.0 compatible DictConfig
+        cfg = model_utils.convert_model_config_to_dict_config(cfg)
+        cfg = model_utils.maybe_update_config_version(cfg)
+        self._GLOBAL_LANG_MAP = {


move this to end of file and keep in global space. (for better readability)

nithinraok · 2025-05-05T18:50:53Z

examples/asr/ast_hybrid_transducer_ctc/speech_to_text_hybrid_rnnt_ctc_bpe.py

+"""
+# Preparing the Tokenizer for the dataset
+Use the `process_asr_text_tokenizer.py` script under <NEMO_ROOT>/scripts/tokenizers/ in order to prepare the tokenizer.
+


Add a sample manifest line here how it would look for added lang ID support

nithinraok · 2025-05-05T19:30:59Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

example config file is missing. Also add how a sample manifest file would look.

Based on your if cases and everything. Do you really need a new model for this? Why not adjust existing hyb model code? Other than BLEU support I don;t see much use with this new model file. Correct me otherwise.

However if you adjust the code to support for general purpose: like if users want to use this for example add predict token to output it would be great!

As discussed, i think leaving it separate is better for simplicity for now as it's still at developmental stage and potentially, we want to integrate it with Target speaker ASR so we have a model Target speaker ASR/AST. Second reason is that the idea of concatenating target lang ID as a one hot vector to ASR embedding and directly feeding that into the decoder could conceptually apply to other type of decoders such as transformer. So, perhaps it's easier to manipulate if it's separate from current Hybrid Parakeet integration.

nithinraok · 2025-05-05T19:36:26Z

examples/asr/transcribe_speech.py

+    # Special case for EncDecHybridRNNTCTCBPEModelTgtLangID
+    if isinstance(asr_model, EncDecHybridRNNTCTCBPEModelTgtLangID):
+        # Special case for EncDecHybridRNNTCTCBPEModelTgtLangID, where the input manifest is directly passed into the model's transcribe() function
+        filepaths = cfg.dataset_manifest
+        assert (
+            cfg.dataset_manifest is not None
+        ), "dataset_manifest must be provided for EncDecHybridRNNTCTCBPEModelTgtLangID"
+        sorted_manifest_path = None
+    else:
+        filepaths, sorted_manifest_path = prepare_audio_data(cfg)


remove this add support to .transcribe() with a default target_language_id support. See AED models code.

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

nithinraok · 2025-05-05T19:43:31Z

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

+            num_sample_per_mel_frame (int, optional): The number of samples per mel frame. Default is 160.
+            num_mel_frame_per_asr_frame (int, optional): The number of mel frames per ASR frame. Default is 8.


Better to rename these to shift and subsampling factor.

inherited from here

Doesn;t matter, for ASR update accordingly.

what is shift?

https://github.com/NVIDIA/NeMo/blob/92f2defd7686866a470e5e1c51f44b85f31e6326/examples/asr/conf/fastconformer/fast-conformer_transducer_bpe.yaml#L101

nithinraok · 2025-05-05T19:48:07Z

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

+        lang_targets = [
+            torch.transpose(
+                torch.as_tensor(
+                    self.lang_to_target_lang(
+                        c,
+                        self.num_languages,
+                        self.num_sample_per_mel_frame,
+                        self.num_mel_frame_per_asr_frame,
+                    ),
+                    dtype=torch.float32,
+                ),
+                0,
+                1,
+            )
+            for c in cuts
+        ]
+
+        # Create final tensors


why to do this here and not part of training loop? That way you could avoid preparing these matrices and basically avoid this dataset all together.

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

github-actions · 2025-05-30T02:08:25Z

This PR is stale because it has been open for 14 days with no activity. Remove stale label or comment or update or this will be closed in 7 days.

ealbasiri · 2025-05-30T18:13:00Z

Revisions have been integrated. Please review. Thank you!

nithinraok

As mentioned before unit tests are missing
I am pretty sure current script doesn;t run without any errors with config provided, correct them and add a CI-CD run for this script.

nithinraok · 2025-06-02T18:27:06Z

examples/asr/asr_hybrid_transducer_ctc/speech_to_text_hybrid_rnnt_ctc_bpe_prompt.py

+
+
+@hydra_runner(
+    config_path="../conf/conformer/hybrid_transducer_ctc/", config_name="conformer_hybrid_transducer_ctc_bpe"


Shouldn;t the config be: examples/asr/conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml ?

I see its not updated yet. Marking to unresolved

...conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml

examples/asr/transcribe_speech.py

nithinraok · 2025-06-02T18:58:09Z

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py

+            num_sample_per_mel_frame (int, optional): The number of samples per mel frame. Default is 160.
+            num_mel_frame_per_asr_frame (int, optional): The number of mel frames per ASR frame. Default is 8.


Doesn;t matter, for ASR update accordingly.

nithinraok · 2025-06-02T19:11:40Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+    @property
+    def input_types(self) -> Optional[Dict[str, NeuralType]]:
+        if hasattr(self.preprocessor, '_sample_rate'):
+            input_signal_eltype = AudioSignal(freq=self.preprocessor._sample_rate)


Thats your branch? I don;t see your point

nithinraok · 2025-06-02T19:14:18Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+                target_lang_id = target_lang_id[:, : encoded.shape[1], :]
+
+            # Concatenate encoded states with language ID
+            concat_enc_states = torch.cat([encoded, target_lang_id], dim=-1)


but we thought the language signal might be less evident at some time steps resulting in wrong language predictions.

Why?

nithinraok · 2025-06-02T19:26:28Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+        return {'loss': loss_value}
+
+    def predict_step(self, batch, batch_idx, dataloader_idx=0):
+        # TODO: add support for CTC decoding


Add support for CTC decoding, its was added for other hyb models as well. See those classes for reference

nithinraok · 2025-06-02T19:28:03Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+        self.bleu.update(
+            predictions=encoded, predictions_lengths=encoded_len, targets=transcript, targets_lengths=transcript_len
+        )
+        bleu_metrics = self.bleu.compute(return_all_metrics=True, prefix="val_")
+        tensorboard_logs.update(
+            {
+                'val_bleu_num': bleu_metrics['val_bleu_num'],
+                'val_bleu_denom': bleu_metrics['val_bleu_denom'],
+                'val_bleu_pred_len': bleu_metrics['val_bleu_pred_len'],
+                'val_bleu_target_len': bleu_metrics['val_bleu_target_len'],
+                'val_bleu': bleu_metrics['val_bleu'],
+            }
+        )
+        self.bleu.reset()


Why to calculate bleu scores when you perform asr task ?

Because the model performs both ASR and AST without specific prompt distinction for the task (ASR/AST)

does that make sense? What is the alternative?

nithinraok · 2025-06-02T19:29:13Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+class EncDecHybridRNNTCTCBPEModelTgtLangID(EncDecHybridRNNTCTCModel, ASRBPEMixin, ASRTranscriptionMixin):
+    """Base class for encoder decoder RNNT-based models with auxiliary CTC decoder/loss and subword tokenization."""
+
+    def __init__(self, cfg: DictConfig, trainer: Trainer = None):


where is .transcribe() method?

nithinraok · 2025-06-02T19:54:36Z

Also rebase with main for latest changes.

...conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml

nithinraok · 2025-06-16T18:06:49Z

...conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml

+trainer:
+  devices: -1 # number of GPUs, -1 would use all available GPUs
+  num_nodes: 1
+  max_epochs: 1000
+  max_steps: -1 # computed at runtime if not set
+  val_check_interval: 1.0 # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
+  accelerator: auto
+  strategy: ddp
+  accumulate_grad_batches: 1
+  gradient_clip_val: 1.0
+  precision: bf16 # 16, 32, or bf16
+  log_every_n_steps: 10  # Interval of logging.
+  enable_progress_bar: True
+  num_sanity_val_steps: 0 # number of steps to perform validation steps for sanity check the validation process before starting the training, setting to 0 disables it
+  check_val_every_n_epoch: 1 # number of evaluations on validation every n epochs
+  sync_batchnorm: true
+  enable_checkpointing: False  # Provided by exp_manager
+  logger: false  # Provided by exp_manager
+  benchmark: false # needs to be false for models with variable-length speech input as it slows down training


Is this correct for lhotse based dataloader? Double check

nemo/collections/asr/data/audio_to_text_lhotse_prompt.py

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_prompt.py

nithinraok · 2025-06-16T18:11:57Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+from nemo.collections.asr.models.hybrid_rnnt_ctc_bpe_models_prompt import EncDecHybridRNNTCTCBPEModelWithPrompt
+
+# Global map of language codes to indices that existed in the old model
+GLOBAL_LANG_MAP = {


is this required now? since you added to cfg

apologies, this whole script should be deleted from the branch. I'll update

nithinraok · 2025-06-16T18:12:32Z

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_tgt_lang.py

+
+
+# For backward compatibility
+class EncDecHybridRNNTCTCBPEModelTgtLangID(EncDecHybridRNNTCTCBPEModelWithPrompt):


Remove this we don;t need anymore no?

nithinraok · 2025-06-16T18:13:09Z

nemo/collections/asr/parts/utils/streaming_utils.py

        return hypothesis


-class BatchedFrameASRTDT(BatchedFrameASRRNNT):


why did you remove this?

nithinraok

Add .transcribe() tests
Add timestamp check test
CI-CD run test

...conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml

nithinraok · 2025-06-17T14:11:58Z

...conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml

+  devices: -1 # number of GPUs, -1 would use all available GPUs
+  num_nodes: 1
+  max_epochs: -1
+  max_steps: 10000
+  val_check_interval: 1.0 # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
+  accelerator: auto
+  strategy: ddp
+  accumulate_grad_batches: 1
+  gradient_clip_val: 1.0
+  precision: bf16 # 16, 32, or bf16
+  log_every_n_steps: 10  # Interval of logging.
+  enable_progress_bar: True
+  num_sanity_val_steps: 0 # number of steps to perform validation steps for sanity check the validation process before starting the training, setting to 0 disables it
+  check_val_every_n_epoch: 1 # number of evaluations on validation every n epochs
+  sync_batchnorm: true
+  enable_checkpointing: False  # Provided by exp_manager
+  logger: false  # Provided by exp_manager
+  benchmark: false # needs to be false for models with variable-length speech input as it slows down training
+  use_distributed_sampler: false


lhotse parameters are not updated here yet, this will error out/gets stuck

nithinraok · 2025-06-17T14:14:44Z

tests/collections/asr/test_asr_hybrid_rnnt_ctc_model_bpe_prompt.py

+        not NUMBA_RNNT_LOSS_AVAILABLE,
+        reason='RNNTLoss has not been compiled with appropriate numba version.',
+    )
+    @pytest.mark.with_downloads()


why to run with_downloads(). which tests are based on Test Folder?

nemo/collections/asr/models/hybrid_rnnt_ctc_bpe_models_prompt.py

tests/collections/asr/test_asr_hybrid_rnnt_ctc_model_bpe_prompt.py

Signed-off-by: Enas Albasiri <[email protected]>

Signed-off-by: ealbasiri <[email protected]> Signed-off-by: Enas Albasiri <[email protected]>

Signed-off-by: Enas Albasiri <[email protected]>

…ang-apr30

Signed-off-by: Enas Albasiri <[email protected]>

nithinraok

Overall looks fine now.
Few things:

Update integration test as mentioned
Add documentation

nithinraok · 2025-07-21T17:16:19Z

...conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe_prompt.yaml

@@ -0,0 +1,372 @@
+# The model would have two decoders: RNNT (Transducer) and CTC


Add a manifest example line here on how it should look for training this model

nithinraok · 2025-07-21T17:26:23Z

tests/functional_tests/L0_Unit_Tests_GPU_ASR_Hybrid_Prompt.sh

+python -c "from nemo.collections.asr.models import EncDecHybridRNNTCTCBPEModelWithPrompt" && \
+NEMO_NUMBA_MINVER=0.53 CUDA_VISIBLE_DEVICES=0 \
+coverage run -a --data-file=/workspace/.coverage --source=/workspace/ \
+-m pytest tests/collections/asr/test_asr_hybrid_rnnt_ctc_model_bpe_prompt.py \
+-v


This is not for running tests, this is for integration see example and run for few steps only: https://github.com/ealbasiri/NeMo/blob/hybrid-parakeet-tgt-lang-apr30/tests/functional_tests/ASR_dev_run_Speech_to_Text.sh

github-actions · 2025-08-07T02:12:33Z

This PR is stale because it has been open for 14 days with no activity. Remove stale label or comment or update or this will be closed in 7 days.

github-actions · 2025-08-15T02:12:19Z

This PR was closed because it has been inactive for 7 days since being marked as stale.

github-actions bot added the ASR label Apr 30, 2025

nithinraok requested changes May 5, 2025

View reviewed changes

jmayank1511 reviewed May 15, 2025

View reviewed changes

nemo/collections/asr/data/audio_to_text_lhotse_target_language.py Outdated Show resolved Hide resolved

github-actions bot added stale common labels May 30, 2025

github-actions bot removed the stale label May 31, 2025

nithinraok requested changes Jun 2, 2025

View reviewed changes

github-actions bot removed the common label Jun 2, 2025

ealbasiri force-pushed the hybrid-parakeet-tgt-lang-apr30 branch from df8c0fb to 967eb94 Compare June 4, 2025 19:00

github-actions bot added the common label Jun 4, 2025

ealbasiri force-pushed the hybrid-parakeet-tgt-lang-apr30 branch from 967eb94 to 2afbd35 Compare June 16, 2025 16:43

nithinraok requested changes Jun 16, 2025

View reviewed changes

nithinraok requested changes Jun 17, 2025

View reviewed changes

ealbasiri force-pushed the hybrid-parakeet-tgt-lang-apr30 branch from f9d5379 to a59043a Compare June 30, 2025 16:38

github-actions bot removed the common label Jun 30, 2025

nithinraok added the Run CICD label Jun 30, 2025

github-advanced-security bot found potential problems Jun 30, 2025

View reviewed changes

ko3n1g added Run CICD and removed Run CICD labels Jul 1, 2025

ealbasiri force-pushed the hybrid-parakeet-tgt-lang-apr30 branch from 2047fb7 to c6b3215 Compare July 1, 2025 15:16

ko3n1g added Run CICD and removed Run CICD labels Jul 1, 2025

chtruong814 added skip-linting Run CICD and removed Run CICD labels Jul 2, 2025

Enas Albasiri and others added 11 commits July 2, 2025 12:27

add straming iferacne support for RNNT with target lang ID support

1f0f843

Signed-off-by: Enas Albasiri <[email protected]>

update streaming_utils-- rebase

e9e7e68

Signed-off-by: Enas Albasiri <[email protected]>

modifed Parakeet with target lang to Parakeet with prompt

31c258e

Signed-off-by: Enas Albasiri <[email protected]>

added unit tests and modifed files to reflect revisions

4c40f47

Signed-off-by: Enas Albasiri <[email protected]>

added transcribe function to the model and test for it

be25af8

Signed-off-by: Enas Albasiri <[email protected]>

added CI-CD run test and timestamps test

4b399fa

Signed-off-by: Enas Albasiri <[email protected]>

Apply isort and black reformatting

49aaa0f

Signed-off-by: ealbasiri <[email protected]> Signed-off-by: Enas Albasiri <[email protected]>

fix CodeQL failing tests

5d63849

Signed-off-by: Enas Albasiri <[email protected]>

Fix empty f-string issue in audio_to_text_lhotse_prompt

85619f0

Signed-off-by: Enas Albasiri <[email protected]>

keep transcription.py without changes

c03f185

Signed-off-by: Enas Albasiri <[email protected]>

keep transcribe_speech no change

b9bd2b4

Signed-off-by: Enas Albasiri <[email protected]>

ealbasiri force-pushed the hybrid-parakeet-tgt-lang-apr30 branch from caee120 to b9bd2b4 Compare July 3, 2025 14:36

ko3n1g added Run CICD and removed Run CICD labels Jul 3, 2025

ko3n1g temporarily deployed to test July 3, 2025 15:12 — with GitHub Actions Inactive

add more robus to coda graph in model forward and forward unit test

3afbb93

Signed-off-by: Enas Albasiri <[email protected]>

ko3n1g added Run CICD and removed Run CICD labels Jul 7, 2025

ko3n1g temporarily deployed to test July 8, 2025 17:14 — with GitHub Actions Inactive

ealbasiri added 2 commits July 11, 2025 09:46

Merge remote-tracking branch 'origin/main' into hybrid-parakeet-tgt-l…

bfb22e4

…ang-apr30

fixed failing ci test

13b9086

Signed-off-by: Enas Albasiri <[email protected]>

ko3n1g added Run CICD and removed Run CICD labels Jul 11, 2025

ko3n1g temporarily deployed to test July 11, 2025 22:04 — with GitHub Actions Inactive

nithinraok requested changes Jul 21, 2025

View reviewed changes

github-actions bot added the stale label Aug 7, 2025

github-actions bot closed this Aug 15, 2025

ealbasiri mentioned this pull request Aug 22, 2025

Add Parakeet Hybrid RNNT CTC BPE Model with Prompt support #14561

Merged

6 tasks

		num_sample_per_mel_frame (int, optional): The number of samples per mel frame. Default is 160.
		num_mel_frame_per_asr_frame (int, optional): The number of mel frames per ASR frame. Default is 8.



		@hydra_runner(
		config_path="../conf/conformer/hybrid_transducer_ctc/", config_name="conformer_hybrid_transducer_ctc_bpe"



		# For backward compatibility
		class EncDecHybridRNNTCTCBPEModelTgtLangID(EncDecHybridRNNTCTCBPEModelWithPrompt):

		return hypothesis


		class BatchedFrameASRTDT(BatchedFrameASRRNNT):

		@@ -0,0 +1,372 @@
		# The model would have two decoders: RNNT (Transducer) and CTC

Conversation

ealbasiri commented Apr 30, 2025 • edited by nithinraok Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do ?

Changelog

Usage

GitHub Actions CI

Before your PR is "Ready for review"

Who can review?

Additional Information

Uh oh!

nithinraok left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented May 30, 2025

Uh oh!

ealbasiri commented May 30, 2025

Uh oh!

nithinraok left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nithinraok commented Jun 2, 2025

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ealbasiri commented Apr 30, 2025 •

edited by nithinraok

Loading