Santacoder finetuned on The-Stack-dedup (GLSL subset) for 1000 steps with a batch size of 2 and full sequence length of 2048. adapted finetuning script found here
Finetuning parameters
python3 train.py --model_path "bigcode/santacoder" \
--dataset_name "bigcode/the-stack-dedup" \
--subset "data/glsl" \
--data_column "content" \
--split "train" \
--seq_length 2048 \
--max_steps 1000 \
--batch_size 2 \
--gradient_accumulation_steps 4 \
--learning_rate 5e-5 \
--num_warmup_steps 100 \
--eval_freq 100 \
--save_freq 100 \
--log_freq 1 \
--output_dir "checkpoint_dir" \
--no_fp16
Main purpose of this model is to explore if finetuning models improves performance on ShaderEval, which reached 0.380 with 300 samples.
License carried over from model, and the finetuning dataset holds the same license.