Modular SpeechLLM implementation for Sept. 2023 submission (SALM) #7634
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
approve the nit change
…#2) approve as no need to review
… to guard the llm input
* Merge heh and zhehuai's initial version of frozen am+llm The previous differences are summarized here: https://docs.google.com/document/d/1zNI4hC6vJtUfcHbrUSPaMuYWRBQdN_36H0P2NiBiuPY/edit This PR includes 1. Finish merging the model, dataset, and config code 2. Previous tests are still enabled and passed (prepare_llm_input, training_step, validation_step) 3. the example training script with LS960 has been run to make sure the training pipeline works The major remaining works are listed here https://docs.google.com/document/d/1o0AM7v4gcTQkPZjE0Vl9TTX4vYnGTrbXEFGWh0UhGlk/edit#bookmark=id.pzvdadt5oxyw --------- Co-authored-by: He Huang (Steve) <[email protected]>
Signed-off-by: zhehuaichen <[email protected]>
* wip Signed-off-by: zhehuaichen <[email protected]> * fix data Signed-off-by: zhehuaichen <[email protected]> * fix consumed_samples Signed-off-by: zhehuaichen <[email protected]> * fix the training restart problem by storing adapter+perception model and init them from the ckpt Signed-off-by: zhehuaichen <[email protected]> * refix state dict Signed-off-by: zhehuaichen <[email protected]> * support wer and inf Signed-off-by: zhehuaichen <[email protected]> * nan guard Signed-off-by: zhehuaichen <[email protected]> * reimpl inf and bug fix Signed-off-by: zhehuaichen <[email protected]> * multi loader Signed-off-by: zhehuaichen <[email protected]> * unfreeze lm Signed-off-by: zhehuaichen <[email protected]> * flag for load am Signed-off-by: zhehuaichen <[email protected]> * tokenizer Signed-off-by: zhehuaichen <[email protected]> * overwrite vocab size Signed-off-by: zhehuaichen <[email protected]> * support bpe dropout Signed-off-by: zhehuaichen <[email protected]> * add tarred datasets Signed-off-by: stevehuang52 <[email protected]> * fix sample_alpha Signed-off-by: stevehuang52 <[email protected]> * fix bpe dropout bugs in the mismatched context in tokenization Signed-off-by: zhehuaichen <[email protected]> * add bleu metric Signed-off-by: stevehuang52 <[email protected]> * update metrics Signed-off-by: stevehuang52 <[email protected]> * support inference and fix a bug in wer calculation Signed-off-by: zhehuaichen <[email protected]> * fix bucketing dataset Signed-off-by: stevehuang52 <[email protected]> * fix bleu implementation Signed-off-by: zhehuaichen <[email protected]> * support question set file per dataset/data loader in preparation for multitask understanding; also fix bleu implementation Signed-off-by: zhehuaichen <[email protected]> * support simple random context for word boosting Signed-off-by: zhehuaichen <[email protected]> * use sacrebleu.corpus_bleu to be consistent with the rest Signed-off-by: zhehuaichen <[email protected]> * make audio_file optional in the data loader Signed-off-by: zhehuaichen <[email protected]> * add a tool to materialize mt and text data Signed-off-by: zhehuaichen <[email protected]> * compatible with tar dataset Signed-off-by: zhehuaichen <[email protected]> * temp fix for metric and speed up materialization Signed-off-by: zhehuaichen <[email protected]> * make num of context configurable Signed-off-by: zhehuaichen <[email protected]> * val_check_interval fix; make manifest dumping consistent with speech models Signed-off-by: zhehuaichen <[email protected]> * random_context_positive_ratio configurable to control precision Signed-off-by: zhehuaichen <[email protected]> * bug fix: freeze_llm flag is not passed to the model cfg Signed-off-by: zhehuaichen <[email protected]> * overwrite tensor_model_parallel_size Signed-off-by: zhehuaichen <[email protected]> * support both stt and ssl models for loading audio encoder Signed-off-by: zhehuaichen <[email protected]> * fix the inference config so as to use sampling; allow inference config update in training Signed-off-by: zhehuaichen <[email protected]> * refactorize and clean up code for preprocessing collections, dataset interface, model inference and rename some classes to be consistent with salm paper. also make sure test passed Signed-off-by: zhehuaichen <[email protected]> * Undo changes in megatron_gpt_peft_models.py and move them to speechllm_models.py; make sure the correctness by test_speechllm_models.py::TestModularizedAudioGPTModel::test_predict_step Signed-off-by: zhehuaichen <[email protected]> * update default inference config and test golden value accordingly Signed-off-by: zhehuaichen <[email protected]> * integration test and minor fix Signed-off-by: zhehuaichen <[email protected]> * nit bug fix on manifest_filepath introduced by code cleanup Signed-off-by: zhehuaichen <[email protected]> * update workspace/ files; consider moving to examples later Signed-off-by: zhehuaichen <[email protected]> * further remove unnecessary stuff in the inference implementation Signed-off-by: zhehuaichen <[email protected]> * revert the update in default end_string to be compatible with legacy models Signed-off-by: zhehuaichen <[email protected]> --------- Signed-off-by: zhehuaichen <[email protected]> Signed-off-by: stevehuang52 <[email protected]> Co-authored-by: stevehuang52 <[email protected]> Co-authored-by: He Huang (Steve) <[email protected]>
… speechllm stuff under nemo/collections/multimodal/speechllm Signed-off-by: zhehuaichen <[email protected]>
zhehuaichen
requested review from
titu1994,
blisc,
nithinraok and
krishnacpuvvada
|
@zhehuaichen why merging to https://github.com/NVIDIA/NeMo/tree/modular_speechllm ? instead of main? |
@nithinraok Because the project is still under development, and we want to put the branch under NVIDIA/NeMo to facilitate collaboration with more people. The scripts under |
stevehuang52
reviewed
Oct 8, 2023
stevehuang52
requested changes
Oct 8, 2023
…s since the main branch has LLaMA support Signed-off-by: zhehuaichen <[email protected]>
|
Thanks for the review. In this PR we want to move our experimental branch from the private repository to NVIDIA/NeMo to facilitate collaboration with more people while the project is still under development. Later we will send another PR to merge to the main branch. We will move the scripts under workspace/ to examples at the time. |
zhehuaichen
added a commit
that referenced
this pull request
Oct 13, 2023
) * add initial impl of ModularizedSpeechGPTModel and integration test * fix typo in the test name (#1) approve the nit change * clean a initial version of example config; make sure it works by test (#2) approve as no need to review * add the test for training_step and fix the code correspondingly (test passed now) (#3) * add test for validation_step (#4) * mv audio and text emb concat to prepare_llm_input so as to write test to guard the llm input * Merge heh and zhehuai's initial version of frozen am+llm (#5) * Merge heh and zhehuai's initial version of frozen am+llm The previous differences are summarized here: https://docs.google.com/document/d/1zNI4hC6vJtUfcHbrUSPaMuYWRBQdN_36H0P2NiBiuPY/edit This PR includes 1. Finish merging the model, dataset, and config code 2. Previous tests are still enabled and passed (prepare_llm_input, training_step, validation_step) 3. the example training script with LS960 has been run to make sure the training pipeline works The major remaining works are listed here https://docs.google.com/document/d/1o0AM7v4gcTQkPZjE0Vl9TTX4vYnGTrbXEFGWh0UhGlk/edit#bookmark=id.pzvdadt5oxyw --------- Co-authored-by: He Huang (Steve) <[email protected]> * fix a nit init bug broke test (#6) Signed-off-by: zhehuaichen <[email protected]> * Clean up implementation for SALM paper and sync to NEMO v1.20.0 (#18) * wip Signed-off-by: zhehuaichen <[email protected]> * fix data Signed-off-by: zhehuaichen <[email protected]> * fix consumed_samples Signed-off-by: zhehuaichen <[email protected]> * fix the training restart problem by storing adapter+perception model and init them from the ckpt Signed-off-by: zhehuaichen <[email protected]> * refix state dict Signed-off-by: zhehuaichen <[email protected]> * support wer and inf Signed-off-by: zhehuaichen <[email protected]> * nan guard Signed-off-by: zhehuaichen <[email protected]> * reimpl inf and bug fix Signed-off-by: zhehuaichen <[email protected]> * multi loader Signed-off-by: zhehuaichen <[email protected]> * unfreeze lm Signed-off-by: zhehuaichen <[email protected]> * flag for load am Signed-off-by: zhehuaichen <[email protected]> * tokenizer Signed-off-by: zhehuaichen <[email protected]> * overwrite vocab size Signed-off-by: zhehuaichen <[email protected]> * support bpe dropout Signed-off-by: zhehuaichen <[email protected]> * add tarred datasets Signed-off-by: stevehuang52 <[email protected]> * fix sample_alpha Signed-off-by: stevehuang52 <[email protected]> * fix bpe dropout bugs in the mismatched context in tokenization Signed-off-by: zhehuaichen <[email protected]> * add bleu metric Signed-off-by: stevehuang52 <[email protected]> * update metrics Signed-off-by: stevehuang52 <[email protected]> * support inference and fix a bug in wer calculation Signed-off-by: zhehuaichen <[email protected]> * fix bucketing dataset Signed-off-by: stevehuang52 <[email protected]> * fix bleu implementation Signed-off-by: zhehuaichen <[email protected]> * support question set file per dataset/data loader in preparation for multitask understanding; also fix bleu implementation Signed-off-by: zhehuaichen <[email protected]> * support simple random context for word boosting Signed-off-by: zhehuaichen <[email protected]> * use sacrebleu.corpus_bleu to be consistent with the rest Signed-off-by: zhehuaichen <[email protected]> * make audio_file optional in the data loader Signed-off-by: zhehuaichen <[email protected]> * add a tool to materialize mt and text data Signed-off-by: zhehuaichen <[email protected]> * compatible with tar dataset Signed-off-by: zhehuaichen <[email protected]> * temp fix for metric and speed up materialization Signed-off-by: zhehuaichen <[email protected]> * make num of context configurable Signed-off-by: zhehuaichen <[email protected]> * val_check_interval fix; make manifest dumping consistent with speech models Signed-off-by: zhehuaichen <[email protected]> * random_context_positive_ratio configurable to control precision Signed-off-by: zhehuaichen <[email protected]> * bug fix: freeze_llm flag is not passed to the model cfg Signed-off-by: zhehuaichen <[email protected]> * overwrite tensor_model_parallel_size Signed-off-by: zhehuaichen <[email protected]> * support both stt and ssl models for loading audio encoder Signed-off-by: zhehuaichen <[email protected]> * fix the inference config so as to use sampling; allow inference config update in training Signed-off-by: zhehuaichen <[email protected]> * refactorize and clean up code for preprocessing collections, dataset interface, model inference and rename some classes to be consistent with salm paper. also make sure test passed Signed-off-by: zhehuaichen <[email protected]> * Undo changes in megatron_gpt_peft_models.py and move them to speechllm_models.py; make sure the correctness by test_speechllm_models.py::TestModularizedAudioGPTModel::test_predict_step Signed-off-by: zhehuaichen <[email protected]> * update default inference config and test golden value accordingly Signed-off-by: zhehuaichen <[email protected]> * integration test and minor fix Signed-off-by: zhehuaichen <[email protected]> * nit bug fix on manifest_filepath introduced by code cleanup Signed-off-by: zhehuaichen <[email protected]> * update workspace/ files; consider moving to examples later Signed-off-by: zhehuaichen <[email protected]> * further remove unnecessary stuff in the inference implementation Signed-off-by: zhehuaichen <[email protected]> * revert the update in default end_string to be compatible with legacy models Signed-off-by: zhehuaichen <[email protected]> --------- Signed-off-by: zhehuaichen <[email protected]> Signed-off-by: stevehuang52 <[email protected]> Co-authored-by: stevehuang52 <[email protected]> Co-authored-by: He Huang (Steve) <[email protected]> * rename 'ModularizedAudioGPTModel' to 'ModularAudioGPTLoRAModel'; move speechllm stuff under nemo/collections/multimodal/speechllm Signed-off-by: zhehuaichen <[email protected]> * update copyright; remove workspace/scripts and workspace/tools folders since the main branch has LLaMA support Signed-off-by: zhehuaichen <[email protected]> --------- Signed-off-by: zhehuaichen <[email protected]> Signed-off-by: stevehuang52 <[email protected]> Co-authored-by: Zhehuai Chen <[email protected]> Co-authored-by: He Huang (Steve) <[email protected]> Co-authored-by: stevehuang52 <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do ?
Include Modular SpeechLLM implementation for Sept. 2023 submission (SALM). The implementation integrates a pretrained FC and a frozen LLM, and can do ASR, AST, and in-context learning for keyword boosting.
The current implementation is based off of r1.20.0 release.
Collection: multimodal
Usage
refer to
workspace/run_sft_audio_lm.pyBefore your PR is "Ready for review"
Pre checks:
PR Type:
If you haven't finished some of the above items you can still open "Draft" PR.
Who can review?
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.