Weights saved in Google drive of each task -> planning to deploy to HF after approval
https://drive.google.com/drive/folders/1RQkuU-kLtAb1VOuvqNvANv99OR6KFwNC?usp=share_link
Filepaths in lightning ai:
Vulnerability Classification: secure_modern_bert/sentiment_classif/checkpoint_epoch_10.pth
Cross Encoder: secure_modern_bert/embedding_sim_weights/checkpoint_epoch_9.pth
Dual Encoder: secure_modern_bert/modernbert_dual_encoder_mnr/model.safetensors
MLM (base model): secure_modern_bert/final_base_modernsecurebert_pths/checkpoint_epoch_20.pth
Ner: secure_modern_bert/final_ner_original/without_test/checkpoint_epoch_20.pth (Without test of augmented set in)
All possible train data NER: secure_modern_bert/final_ner_original/take_2/checkpoint_epoch_20.pth
dataset.py - Global Dataset for DataLoading script
cont_load.py and primus_load.py - load all datasets from HuggingFaceHub to local disk
All these below scripts got if name == "main"
train.py - Train MLM (SecureBert 2.0) starting from ModernBert
code_mlm_eval.py - Code MLM evaluation for SecureBert 2.0
SecureBert2_mlm_eval.py - Evaluation Verbs.csv and Object_pred.csv for SecureBert 2.0
SecureBert_mlm_eval.py - Evaluation Verbs.csv and Object_pred.csv for SecureBert (Original)
run_modernbert.py - Single Instance Inference for Secure Bert 2.0 MLM
BiEncoder_eval.py - Evaluation for BiEncoder Downstream Task
BiEncoder_train.py - Training BiEncoder model
CrossEncoder_eval.py - Evaluation for Cross Encoder Downstream Task
CrossEncoder_infer.py - Inference for Cross Encoder Downstream Task
CrossEncoder_train.py - Training for Cross Encoder Downstream Task
NER_eval.py - Evaluation for NER Downstream Task
NER_infer.py - Single instance inference for NER Downstream Task
NER_train.py - Training for NER Downstream Task
CodeVuln_eval.py - Code Vulnernability Evaluation
CodeVuln_infer.py - Single Instance Inference for Code Vulnerability
vuln_sentiment_train.py - Train Code Vulnerability downstream task
Weights saved in Google drive of each task -> planning to deploy to HF after approval
https://drive.google.com/drive/folders/1RQkuU-kLtAb1VOuvqNvANv99OR6KFwNC?usp=share_link
Filepaths in lightning ai:
Vulnerability Classification: secure_modern_bert/sentiment_classif/checkpoint_epoch_10.pth
Cross Encoder: secure_modern_bert/embedding_sim_weights/checkpoint_epoch_9.pth
Dual Encoder: secure_modern_bert/modernbert_dual_encoder_mnr/model.safetensors
MLM (base model): secure_modern_bert/final_base_modernsecurebert_pths/checkpoint_epoch_20.pth
Ner: secure_modern_bert/final_ner_original/without_test/checkpoint_epoch_20.pth (Without test of augmented set in)
All possible train data NER: secure_modern_bert/final_ner_original/take_2/checkpoint_epoch_20.pth
dataset.py - Global Dataset for DataLoading script
cont_load.py and primus_load.py - load all datasets from HuggingFaceHub to local disk
All these below scripts got if name == "main"
train.py - Train MLM (SecureBert 2.0) starting from ModernBert
code_mlm_eval.py - Code MLM evaluation for SecureBert 2.0
SecureBert2_mlm_eval.py - Evaluation Verbs.csv and Object_pred.csv for SecureBert 2.0
SecureBert_mlm_eval.py - Evaluation Verbs.csv and Object_pred.csv for SecureBert (Original)
run_modernbert.py - Single Instance Inference for Secure Bert 2.0 MLM
BiEncoder_eval.py - Evaluation for BiEncoder Downstream Task
BiEncoder_train.py - Training BiEncoder model
CrossEncoder_eval.py - Evaluation for Cross Encoder Downstream Task
CrossEncoder_infer.py - Inference for Cross Encoder Downstream Task
CrossEncoder_train.py - Training for Cross Encoder Downstream Task
NER_eval.py - Evaluation for NER Downstream Task
NER_infer.py - Single instance inference for NER Downstream Task
NER_train.py - Training for NER Downstream Task
CodeVuln_eval.py - Code Vulnernability Evaluation
CodeVuln_infer.py - Single Instance Inference for Code Vulnerability
vuln_sentiment_train.py - Train Code Vulnerability downstream task
