task vector optimization checkpoint ready for merging.
trained on MFANN for 12000 steps, however due to a slightly higher training loss, im going to merge this model with the last version and retrain, the goal was to use DARE-TIES to reduce the parameters used per vector, and this model will now be merged with the last model before DARE using TIES alone, and will be subsequently retrained.
- Downloads last month
- 14
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.