metadata

tags:
  - generated_from_trainer
metrics:
  - spearmanr
model-index:
  - name: thermo-predictor-thermo-evotuning-prot_bert
    results: []

thermo-predictor-thermo-evotuning-prot_bert

This model is a fine-tuned version of thundaa/thermo-evotuning-prot_bert on the cradle-bio/tape-thermostability dataset. It achieves the following results on the evaluation set:

Loss: 0.1617
Spearmanr: 0.6914

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 4e-05
train_batch_size: 256
eval_batch_size: 256
seed: 42
gradient_accumulation_steps: 64
total_train_batch_size: 16384
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 1
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Spearmanr
0.4734	0.68	2	0.3146	0.3359
0.4392	1.68	4	0.2936	0.3407
0.4034	2.68	6	0.2633	0.3696
0.3669	3.68	8	0.2437	0.3903
0.3496	4.68	10	0.2377	0.4102
0.3351	5.68	12	0.2285	0.4204
0.3289	6.68	14	0.2267	0.4180
0.3267	7.68	16	0.2258	0.4242
0.3177	8.68	18	0.2206	0.4295
0.3116	9.68	20	0.2150	0.4365
0.3039	10.68	22	0.2115	0.4365
0.2985	11.68	24	0.2062	0.4469
0.2927	12.68	26	0.2045	0.4531
0.2885	13.68	28	0.2005	0.4603
0.2838	14.68	30	0.1987	0.4690
0.2806	15.68	32	0.1975	0.4744
0.2772	16.68	34	0.1970	0.4765
0.2728	17.68	36	0.1939	0.4845
0.2684	18.68	38	0.1931	0.4858
0.2641	19.68	40	0.1925	0.4936
0.2608	20.68	42	0.1905	0.4929
0.2566	21.68	44	0.1886	0.5049
0.2518	22.68	46	0.1875	0.5095
0.2467	23.68	48	0.1869	0.5141
0.2424	24.68	50	0.1859	0.5161
0.2375	25.68	52	0.1850	0.5223
0.2329	26.68	54	0.1851	0.5210
0.2279	27.68	56	0.1850	0.5294
0.2226	28.68	58	0.1837	0.5310

Framework versions

Transformers 4.18.0
Pytorch 1.11.0
Datasets 2.1.0
Tokenizers 0.12.1