Instruction Tuned GPT-NeoXT-20B model on Instruction Tuning dataset as listed below (~560k data) using Colossal AI

Base Model: togethercomputer/GPT-NeoXT-Chat-Base-20B (GPT-NeoXT-Chat-Base-20B-v0.16 - fine-tuned on feedback data)

Training Details :

• Epochs: 2
• 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 : 240
• Machine : 8xA100 80GB

Dataset Details :

Dataset : iamplus/Instruction_Tuning

Files :

• stanford_alpaca_it_v2.csv
• ColossalChat.csv
• unified_chip2.csv
• iamai_summarization_v1.csv
• iamai_v1.csv