&`i2= N d dl Z d dlZd dlmZmZmZmZ d dlZd dl m
Z
d dl m Z d dl
mZmZmZmZ d dlmZmZmZmZmZmZ d dlmZ d dlm Z m Z d dl Z d d l m Z n# e!$ r dxZ Z Y nw xY w d d
l"m#Z# d d l$m%Z% n# e!$ r e&xZ%Z#Y nw xY w ej' e( Z) G d d
e Z*dS ) N)DictListOptionalUnion) cloudpickle)DEFAULT_METRIC)UNDEFINED_METRIC_MODEUNDEFINED_SEARCH_SPACEUNRESOLVED_SEARCH_SPACESearcher) CategoricalFloatInteger
LogUniform QuantizedUniform)parse_spec_vars) flatten_dictunflatten_list_dict)AxClient)DataRequiredError) MaxParallelismReachedExceptionc \ e Zd ZdZ ddeeeee f dee dee deee dee dee d ee f fd
Z
d Z dee dee d efd
Z dedee fdZ
ddZd Zedefd ZdefdZdefdZ xZS )AxSearcha& Uses `Ax `_ to optimize hyperparameters.
Ax is a platform for understanding, managing, deploying, and
automating adaptive experiments. Ax provides an easy to use
interface with BoTorch, a flexible, modern library for Bayesian
optimization in PyTorch. More information can be found in https://ax.dev/.
To use this search algorithm, you must install Ax:
.. code-block:: bash
$ pip install ax-platform
Parameters:
space: Parameters in the experiment search space.
Required elements in the dictionaries are: "name" (name of
this parameter, string), "type" (type of the parameter: "range",
"fixed", or "choice", string), "bounds" for range parameters
(list of two values, lower bound first), "values" for choice
parameters (list of values), and "value" for fixed parameters
(single value).
metric: Name of the metric used as objective in this
experiment. This metric must be present in `raw_data` argument
to `log_data`. This metric must also be present in the dict
reported/returned by the Trainable. If None but a mode was passed,
the `ray.tune.result.DEFAULT_METRIC` will be used per default.
mode: One of {min, max}. Determines whether objective is
minimizing or maximizing the metric attribute. Defaults to "max".
points_to_evaluate: Initial parameter suggestions to be run
first. This is for when you already have some good parameters
you want to run first to help the algorithm make better suggestions
for future parameters. Needs to be a list of dicts containing the
configurations.
parameter_constraints: Parameter constraints, such as
"x3 >= x4" or "x3 + x4 >= 2".
outcome_constraints: Outcome constraints of form
"metric_name >= bound", like "m1 <= 3."
ax_client: Optional AxClient instance. If this is set, do
not pass any values to these parameters: `space`, `metric`,
`parameter_constraints`, `outcome_constraints`.
**ax_kwargs: Passed to AxClient instance. Ignored if `AxClient` is not
None.
Tune automatically converts search spaces to Ax's format:
.. code-block:: python
from ray import tune
from ray.tune.search.ax import AxSearch
config = {
"x1": tune.uniform(0.0, 1.0),
"x2": tune.uniform(0.0, 1.0)
}
def easy_objective(config):
for i in range(100):
intermediate_result = config["x1"] + config["x2"] * i
tune.report({"score": intermediate_result})
ax_search = AxSearch()
tuner = tune.Tuner(
easy_objective,
tune_config=tune.TuneConfig(
search_alg=ax_search,
metric="score",
mode="max",
),
param_space=config,
)
tuner.fit()
If you would like to pass the search space manually, the code would
look like this:
.. code-block:: python
from ray import tune
from ray.tune.search.ax import AxSearch
parameters = [
{"name": "x1", "type": "range", "bounds": [0.0, 1.0]},
{"name": "x2", "type": "range", "bounds": [0.0, 1.0]},
]
def easy_objective(config):
for i in range(100):
intermediate_result = config["x1"] + config["x2"] * i
tune.report({"score": intermediate_result})
ax_search = AxSearch(space=parameters, metric="score", mode="max")
tuner = tune.Tuner(
easy_objective,
tune_config=tune.TuneConfig(
search_alg=ax_search,
),
)
tuner.fit()
Nspacemetricmodepoints_to_evaluateparameter_constraintsoutcome_constraints ax_clientc z t
J d |r|dv s
J d t t | || || _ |pi | _ t
|t ri|rgt | \ } }
} |
s| rPt
t j dt | | | }|| _ || _ || _ t% j | | _ g | _ i | _ | j s| j r| d S d S )NzsAx must be installed!
You can install AxSearch with the command:
`pip install ax-platform`.minmaxz `mode` must be 'min' or 'max'.)r r r )parcls)axsuperr __init___ax
_ax_kwargs
isinstancedictr loggerwarningr formattypeconvert_search_space_space_parameter_constraints_outcome_constraintscopydeepcopy_points_to_evaluate _parameters_live_trial_mapping_setup_experiment)
selfr r r r r r r! ax_kwargs
resolved_vars domain_vars grid_vars __class__s
p/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/ray/tune/search/ax/ax_search.pyr* zAxSearch.__init__ so
NN *
NN
L>)))+K)))
h && '
#/r eT " " 9u 94CE4J4J 1M;
9i
9+2wDJJOOO 11%88 &;#$7!#'=1C#D#D #% 8 %t{ % " " $ $ $ $ $ % % c 8 | j | j r t | _ | j st d i | j | _ | j j }d}n# t $ r d}Y nw xY w|sk| j st d | j dvrt d | j | j | j | j
| j | j dk nQt | j | j
| j | j | j g r$t d
g d | j j }|j j j rd
nd| _ |j j j j | _ t' |j | _ | j j r t. d d S d S )
NTFzYou have to create an Ax experiment by calling `AxClient.create_experiment()`, or you should pass an Ax search space as the `space` parameter to `AxSearch`, or pass a `param_space` dict to `tune.Tuner()`.r# zmPlease specify the `mode` argument when initializing the `AxSearch` object or pass it to `tune.TuneConfig()`.r% )
parametersobjective_namer r minimizezdIf you create the Ax experiment yourself, do not pass values for these parameters to `AxSearch`: {}.)r r r r r r$ zEDetected sequential enforcement. Be sure to use a ConcurrencyLimiter. )_metric_moder r+ r r,
experiment
ValueErrorr4 create_experimentr5 r6 anyr1 optimization_config objectiverH r namelistrF r: _enforce_sequential_optimizationr/ r0 )r= exphas_experiments rC r< zAxSearch._setup_experiment s < DJ )DLx 3 22$/22DH #(%C !NN # # # "NNN # ) ;
F z // O
H & & ; #|&*&A$($= u,
'
K/-JL
!EEKV F F h! " 5 ? HSUUe
.8?D // 8 4 NN(
s A AAconfigc | j rdS | | }|| _ |r|| _ |r|| _ | dS )NFT)r+ r3 r4 rJ rK r< )r= r r rW specr s rC set_search_propertiesz AxSearch.set_search_properties sb 8 5))&11 "!DL DJ trD trial_idreturnc | j s-t t j | j j d | j r| j s8t t j | j j | j | j | j r8| j
d }| j | \ }n6 | j \ }n# t t f$ r Y d S w xY w|| j |< t! }nJ# t" $ r= t! fdt% D }Y nw xY w|S )Nr )r' r )r' r r r c " i | ] }|| S rI rI ).0krF s rC
z$AxSearch.suggest..8 s EEEaJqMEEErD )r+ RuntimeErrorr
r1 rB __name__rJ rK r r9 pop attach_trialget_next_trialr r r; r AssertionErrorsortedkeys)r= r[ rW trial_indexsuggested_configrF s @rC suggestzAxSearch.suggest s x &- /w
| 4: %, / 4:
# -11!44F&*h&;&;F&C&C #J
*.(*A*A*C*C'
KK24EF
tt
.9 * 2: > >
3EEEE6*//2C2C+D+DEEE s% < C C.-C.<D AEEFc j |r| || | j | dS )zyNotification for the completion of trial.
Data of form key value dictionary of metric names and values.
N)_process_resultr; rd )r= r[ resulterrors rC on_trial_completezAxSearch.on_trial_complete< s?
3 6 2 2 2 $$X.....rD c j | j | }| j gd | j j j j D z }i }|D ][}|| }t
j | st
j | r"| j |d| d S |d f||< \| j
|| d S )Nc & g | ]}|j j S rI )r rR )r_ ocs rC
z,AxSearch._process_result..G s, /
/
/
IN/
/
/
rD z nan/inf metrics reported by )rj reason)rj raw_data) r; rJ r+ rL rP r npisnanisinf
abandon_trialcomplete_trial)r= r[ ro ax_trial_indexmetrics_to_include metric_dictkeyvals rC rn zAxSearch._process_resultE s 1(; "l^ /
/
h)=Q/
/
/
% + +C+Cx}}
&& .D(DD ' #T{K N[ QQQQQrD rY c t | \ }}}|rt d t | d } t | \ }}}d d |D }fd|D }||z S )NzOGrid search parameters cannot be automatically converted to an Ax search space.T)prevent_delimiterc D | }t |t r!t d |j }t |t rSt |t r| d|j |j gdddS t |t r| d|j |j gdddS nt |t rYt |t r| d|j |j dz
gdddS t |t r| d|j |j dz
gdddS n5t |t r t |t r | d |j
d
S t d t! | j t! |j j ) Nz=AxSearch does not support quantization. Dropped quantization.rangefloatT)rR r2 bounds
value_type log_scaleF intchoice)rR r2 valueszLAxSearch does not support parameters of type `{}` with samplers of type `{}`) get_samplerr- r r/ r0 samplerr r lowerupperr r r
categoriesrM r1 r2 rc )r& domainr s rC
resolve_valuez4AxSearch.convert_search_space..resolve_valuef s ((**G'9--
*S "/&%(( $
X gz22 # '#)< ">&-%) 11 # '#)< ">&-%* FG,,
X gz22 # '#)< 1A"B&+%) 11 # '#)< 1A"B&+%* FK00
X gw// X$'VEVWWW 228&LL)4+?+?+H3 3
rD c Z g | ](\ }}d d |D d|d)S )/c 3 4 K | ]}t | V d S Nstrr_ ps rC z;AxSearch.convert_search_space... s( 3 3c!ff 3 3 3 3 3 3rD fixed)rR r2 valuejoin)r_ pathr s rC ru z1AxSearch.convert_search_space.. sQ
c XX 3 3d 3 3 333Ws S S
rD c h g | ].\ }} d d |D | /S )r c 3 4 K | ]}t | V d S r r r s rC r z;AxSearch.convert_search_space... s( "8"8a3q66"8"8"8"8"8"8rD r )r_ r r r s rC ru z1AxSearch.convert_search_space.. sT
f
M#(("8"84"8"8"888& A A
rD )r rM r )rY r? r@ rA fixed_valuesresolved_valuesr s @rC r3 z AxSearch.convert_search_spaceX s 0?0E0E-
{I )
DD9990?0E0E-
{I3 3 3 n
*
+
o--rD checkpoint_pathc | j }t |d 5 }t j || d d d d S # 1 swxY w Y d S )Nwb)__dict__openr dump)r= r save_object
outputFiles rC savez
AxSearch.save s m
/4
(
( 6J [* 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6s
;??c t |d 5 }t j | }d d d n # 1 swxY w Y | j | d S )Nrb)r r loadr update)r= r inputFiler s rC restorezAxSearch.restore s
/4
(
( 6I%*955K 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6
[)))))s 266)NNNNNNN)NF)rc
__module__ __qualname____doc__r r r r r r r* r< rZ rl rq rn staticmethodr3 r r
__classcell__)rB s @rC r r + s c cN 48 $ "3704.2(,.% .%dDJ ./0.%
.% sm .%
%T$Z 0 .% (~
.% &d^.% H%.% .% .% .% .% .%`E E EN
sm
+3C=
BF
# # # # # # J/ / / /R R R& M.4 M. M. M. \M.^6C 6 6 6 6
*s * * * * * * * *rD r )+r7 loggingtypingr r r r numpyrx rayr ray.tune.resultr ray.tune.searchr r
r r ray.tune.search.sampler
r r r r r !ray.tune.search.variant_generatorr ray.tune.utils.utilr r r( ax.service.ax_clientr ImportErrorax.exceptions.corer !ax.exceptions.generation_strategyr Exception getLoggerrc r/ r rI rD rC r s . . . . . . . . . . . . * * * * * * > = = = = = A A A A A A A A
III------- BC444444PPPPPPP C C C9BB"%6%6%6C
8 $ $E* E* E* E* E*x E* E* E* E* E*s$
A A'&A'+ A8 8 BB