Instruction Tuned GPT-NeoXT-20B model on Stanford Alpaca-2 Instruction Tuning dataset (outputs from ChatGPT) (52k data) using Colossal AI
Base Model: togethercomputer/GPT-NeoXT-Chat-Base-20B (not fine-tuned on feedback data)
Training Details :
- Epochs: 5
- Batch Size : 16 instantaneous per device x 1 gradient accumulation steps x 8 gpus = 128
- Max Length : 1024
- Weight Decay : 0
- Learning Rate : 2e-5
- Learning Rate Scheduler Type : Cosine
- Number of warmup steps : 30
- Machine : 8xA100 80GB
Dataset Details :
Dataset : iamplus/Instruction_Tuning
Files :
- stanford_alpaca_it_v2.csv