Transformers Latent-Space reasoning Auto Train Once -- a framework meant to utilize a latent-space reasoning model with 'Auto Train Once' pruning ( https://arxiv.org/abs/2403.14729 ) technique to ensure maximum performance with minimum neuron activation/creation/utilization in the LLM.
| __pycache__ | ||
| data | ||
| logs | ||
| processed_data/574992b2b558ca08 | ||
| repositories | ||
| src | ||
| uml_generator.py | ||
| # uml_generator.py | ||
| .env | ||
| .gitignore~ | ||
| .README.swp | ||
| app.log | ||
| errors.log | ||
| LICENSE | ||
| m1.py | ||
| main.py | ||
| README.md | ||
| tlato_v4_1_uml_corrected.png | ||
| uml_generator.py | ||
TLATO
Transformers Latent-Space reasoning Auto Train Once -- a framework meant to utilize a latent-space reasoning model with 'Auto Train Once' pruning ( https://arxiv.org/abs/2403.14729 ) technique to ensure maximum performance with minimum neuron activation/creation/utilization in the LLM.