Giters
huu4ontocord
/
MDEL
Multi-Domain Expert Learning
Geek Repo:
Geek Repo
Github PK Tool:
Github PK Tool
Stargazers:
67
Watchers:
21
Issues:
29
Forks:
14
huu4ontocord/MDEL Issues
Tokenize the StarCoder dataset
Updated
a year ago
Comments count
1
Get all relevant data for StarCoder into LUMI
Updated
a year ago
Do a small test run
Updated
a year ago
Set up the training configuration
Updated
a year ago
Train 2nd batch of expert models
Updated
a year ago
Investigate Expert Models Having High Perplexity
Updated
a year ago
Comments count
1
Create template for HF dataset config
Updated
a year ago
Evaluate a merged expert model's perplexity
Updated
a year ago
Comments count
3
Integrate with LLM evaluation frameworks
Updated
a year ago
Comments count
3
Train baseline models for evaluation
Updated
a year ago
Comments count
10
Training instruction followers as composable layers and expert layers
Updated
a year ago
Comments count
1
Add script for merging expert models via weight averaging
Updated
a year ago
Comments count
7
Expert merging: c-BTM
Updated
a year ago
Comments count
3
Setup separate environments on Redmond.ai box
Updated
a year ago
Comments count
2
Stabilize Training on Redmond Box
Updated
a year ago
Automatic Training Scripts for All Expert Models
Updated
a year ago
Create minimal example of training on LUMI
Updated
a year ago
Comments count
1
inputs_ids cast to fp16 in deeperspeed bug
Updated
a year ago
Fix HF Hub Upload Error
Closed
a year ago
Report val loss aggregated by data origin
Closed
a year ago
Create minimal example of training on SUMMIT
Updated
a year ago
Dataset generation open issues
Updated
a year ago
Change the dataset mixing script to process files with a wild card
Closed
a year ago
Add code for mixing Pile and Expert data
Closed
a year ago
Add Single Layer HF Trainer
Closed
a year ago
Add genre classifer code to repo
Updated
a year ago
Add pylint
Closed
a year ago
Establish folder structure
Closed
a year ago
Add pre-commit hooks
Closed
a year ago