babyai_v1
Fine-tuned model using LoRA on Affine validator datasets.
Training Details
- Base Model: ./models/Affine-ofdt-k4
- Training Method: LoRA (merged)
- LoRA Rank: 4
- LoRA Alpha: 4
- Learning Rate: 1e-06
- Epochs: 1
- Final Loss: 0.41884476563026163
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("./checkpoints/babyai_v1")
tokenizer = AutoTokenizer.from_pretrained("./checkpoints/babyai_v1")
- Downloads last month
- 167