syne_tune.optimizer.schedulers.searchers.gp_searcher_utils module
- class syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.MapReward(forward, reverse)[source]
Bases:
object
-
forward:
Callable
[[float
],float
]
-
reverse:
Callable
[[float
],float
]
-
forward:
- syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.map_reward_const_minus_x(const=1.0)[source]
Factory for map_reward argument in GPMultiFidelitySearcher.
- Return type:
- syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.encode_state(state)[source]
- Return type:
Dict
[str
,Any
]
- syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.decode_state(enc_state, hp_ranges)[source]
- Return type:
- syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.decode_state_from_old_encoding(enc_state, hp_ranges)[source]
Decodes
TuningJobState
from encoding done for the old definition ofTuningJobState
. Code maintained for backwards compatibility.Note: Since the old
TuningJobState
did not containtrial_id
, we need to make them up here. We assign these IDs in the ordercandidate_evaluations
,failed_candidates
,pending_candidates
, matching for duplicates.- Parameters:
enc_state (
Dict
[str
,Any
]) –hp_ranges (
HyperparameterRanges
) –
- Return type:
- Returns:
- class syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.ResourceForAcquisitionMap[source]
Bases:
object
In order to use a standard acquisition function (like expected improvement) for multi-fidelity HPO, we need to decide at which
r_acq
we would like to evaluate the AF, w.r.t. the posterior distribution overf(x, r=r_acq)
. This decision can depend on the current state.
- class syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.ResourceForAcquisitionBOHB(threshold, active_metric='target')[source]
Bases:
ResourceForAcquisitionMap
Implements a heuristic proposed in the BOHB paper:
r_acq
is the largestr
such that we have at leastthreshold
observations atr
. If there are less thanthreshold
observations at all levels, the smallest level is returned.
- class syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.ResourceForAcquisitionFirstMilestone[source]
Bases:
ResourceForAcquisitionMap
Here,
r_acq
is the smallest rung level to be attained by a config started from scratch.
- class syne_tune.optimizer.schedulers.searchers.gp_searcher_utils.ResourceForAcquisitionFinal(r_max)[source]
Bases:
ResourceForAcquisitionMap
Here,
r_acq = r_max
is the largest resource level.