I am performing a sweep but one of the of the value has not been explored even once after 30+ runs. The entire config is given below but the main part is the pool_type parameter.
pool_type:
values:
- cls
- mean
- sum
Only the values mean and sum have been used so far. Another photo attached shows that (I have grouped them by pool type and only two show up)
early_terminate:
min_iter: 3
type: hyperband
method: bayes
metric:
goal: maximize
name: precision
name: kinase_VEGFR2_sweep_full
parameters:
batch_size:
values:
- 16
- 32
- 64
drop_prob:
distribution: uniform
max: 0.5
min: 0.1
epochs:
value: 100
hidden_size1:
values:
- 128
- 256
- 512
hidden_size2:
values:
- 32
- 64
- 128
- 256
loss:
values:
- bce
lr:
distribution: uniform
max: 0.001
min: 1e-05
model:
values:
- thaonguyen217/farm_molecular_representation
- ibm/MoLFormer-XL-both-10pct
- seyonec/PubChem10M_SMILES_BPE_450k
- DeepChem/ChemBERTa-10M-MLM
- DeepChem/ChemBERTa-77M-MLM
- DeepChem/ChemBERTa-10M-MTR
- DeepChem/ChemBERTa-77M-MTR
pool_type:
values:
- cls
- mean
- sum
weight_decay:
distribution: uniform
max: 0.01
min: 1e-07