Santacoder finetuned on
Shadertoys-fine for 1000 steps with a batch size of 2 and full sequence length of 2048.
adapted finetuning script found
here
1python3 train.py --model_path "bigcode/santacoder" \
2--dataset_name "Vipitis/Shadertoys-fine" \
3--data_column "code" \
4--split "train" \
5--seq_length 2048 \
6--max_steps 1000 \
7--batch_size 2 \
8--gradient_accumulation_steps 4 \
9--learning_rate 5e-5 \
10--num_warmup_steps 100 \
11--eval_freq 100 \
12--save_freq 100 \
13--log_freq 1 \
14--output_dir "checkpoint_dir" \
15--no_fp16
16
17
Main purpose of this model is to explore if finetuning models improves performance on
ShaderEval, which reached 0.567 with 300 samples and 0.59749 on all samples.
While the train/test split is held out, there is a lot of data contamination. The model results can't be trusted for this simple benchmark.
Better tasks for the benchmark will be developed and tested against these models.
License carried over from model, however training data has an undefied license. Check details in
Shadertoys.
