microsoft/KBLaM

View on GitHub

Official Implementation of "KBLaM: Knowledge Base augmented Language Model"

Stars: 1.5k
Forks: 125
Open beginner issues: 0
Indexed issues: 33
Avg merge: 22d 8m
Merged PRs (30d): 1

Dominant language: Jupyter Notebook
License: MIT
Last GitHub push: Sep 14, 2026
Latest indexed: Sep 19, 2026
Contributing guide: No contributing guide
Code of conduct: Code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

33 open issues indexed

Huge discrepancy between training PRED and generation inference results on Llama-3-8B.Bad performance for generation modes. Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

microsoft/KBLaM#96 · Opened Feb 6, 2026
About KB_size during training Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#94 · Opened Oct 7, 2025
Available pre-trained linear adapter's weights Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

microsoft/KBLaM#93 · Opened Oct 6, 2025
Issue with Embedding shape Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

microsoft/KBLaM#92 · Opened Oct 3, 2025
Future model support (Qwen3 dens/moe) Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/KBLaM#90 · 2 comments · Opened Aug 29, 2025
Working with newer Transformer versions Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/KBLaM#89 · 6 comments · Opened Aug 27, 2025
Guidance for Integrating KBLaM with Food/Nutrition Knowledge Base Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/KBLaM#82 · 7 comments · Opened Jul 29, 2025
Where is the sep_query_head stored? Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

microsoft/KBLaM#81 · 1 comment · Opened Jul 23, 2025
ImportError: cannot import name 'LlamaDynamicNTKScalingRotaryEmbedding' from 'transformers.models.llama.modeling_llama' Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

microsoft/KBLaM#80 · 1 comment · Opened Jul 23, 2025
In experiments/train.py，training labels problem Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#79 · 2 comments · Opened Jul 21, 2025
Has anyone had a good trained model Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

microsoft/KBLaM#75 · 8 comments · Opened Jul 8, 2025
Bug: encoder.pt and config not loaded correct during resume Open

Difficulty 2/5 1-3 hours Newbie friendliness 45/100

microsoft/KBLaM#72 · 1 comment · 3 reactions · Opened Jun 26, 2025
Readme train.py command doesn't create saved output once training is complete Open

Difficulty 2/5 1-3 hours Newbie friendliness 45/100

microsoft/KBLaM#71 · 8 comments · 2 reactions · Opened Jun 24, 2025
eval kb generation mode does not work well Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#70 · Opened Jun 23, 2025
Questions about the Rouge metrics Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#66 · 10 comments · Opened Jun 11, 2025
Bad performance for three generation modes. Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#64 · 20 comments · Opened May 28, 2025
Question about train.py Open

Difficulty 3/5 1-2 days Newbie friendliness 15/100

microsoft/KBLaM#62 · 5 comments · Opened May 26, 2025
Question about eval parameter. Open

Difficulty 2/5 1-3 hours Newbie friendliness 25/100

microsoft/KBLaM#61 · 5 comments · Opened May 22, 2025
Training Effects Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/KBLaM#60 · Opened May 21, 2025
There is a formula missing an 'l' in the paper Open

Difficulty 1/5 Under an hour Newbie friendliness 45/100

microsoft/KBLaM#57 · 8 comments · 1 reaction · Opened May 19, 2025
Bug in `train_utils.py` Open

Difficulty 2/5 1-3 hours Newbie friendliness 62/100

microsoft/KBLaM#56 · 11 comments · Opened May 16, 2025
Has anyone trained with Phi models before? How was the effectiveness? Open

Difficulty 5/5 Over a week Newbie friendliness 10/100

microsoft/KBLaM#55 · 1 comment · Opened May 16, 2025
Question about dataset Open

Difficulty 2/5 1-3 hours Newbie friendliness 25/100

microsoft/KBLaM#52 · 2 comments · Opened May 16, 2025
Questions concerning train.py and its parameters Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#51 · 1 comment · Opened May 1, 2025
Generating KB embeddings does not work if AZURE OPENAI API KEY is set Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

microsoft/KBLaM#49 · Opened Apr 24, 2025
Who can provide the train prameters of erron dataset？ Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

microsoft/KBLaM#47 · 1 comment · Opened Apr 17, 2025
Can I train with 3*4090 GPU？ Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/KBLaM#46 · 1 comment · Opened Apr 17, 2025
Who can provide the synthetic npy file Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

microsoft/KBLaM#43 · 3 comments · Opened Apr 14, 2025
After running train nothing is happening. Stuck on Running on Cuda. Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

microsoft/KBLaM#41 · Opened Apr 10, 2025
Error in __init_rope of KblamLlamaAttention Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

microsoft/KBLaM#37 · 2 comments · Opened Apr 3, 2025
cannot import name 'LlamaDynamicNTKScalingRotaryEmbedding' Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

microsoft/KBLaM#35 · 6 comments · 1 reaction · Opened Mar 31, 2025
Python3.12 - on linux discovery issues Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

microsoft/KBLaM#29 · 24 comments · Opened Mar 19, 2025
Release artifacts (models, dataset) on Hugging Face Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

microsoft/KBLaM#25 · 8 comments · 14 reactions · Opened Feb 17, 2025

microsoft/KBLaM

Get new issues in your inbox