importpickleimportosimporttimefromitertoolsimportproductimportmatplotlib.pyplotaspltimportnumpyimportpandasfromlightgbmimportLGBMRegressorfromonnxruntimeimportInferenceSession,SessionOptionsfrompsutilimportcpu_countfromsphinx_runpython.runpythonimportrun_cmdfromskl2onnximportto_onnx,update_registered_converterfromskl2onnx.common.shape_calculatorimportcalculate_linear_regressor_output_shapesfromsklearnimportset_configfromsklearn.ensembleimportRandomForestRegressorfromtqdmimporttqdmfromxgboostimportXGBRegressorfromonnxmltools.convert.xgboost.operator_converters.XGBoostimportconvert_xgboostdefskl2onnx_convert_lightgbm(scope,operator,container):fromonnxmltools.convert.lightgbm.operator_converters.LightGbmimport(convert_lightgbm,)options=scope.get_options(operator.raw_operator)operator.split=options.get("split",None)convert_lightgbm(scope,operator,container)update_registered_converter(LGBMRegressor,"LightGbmLGBMRegressor",calculate_linear_regressor_output_shapes,skl2onnx_convert_lightgbm,options={"split":None},)update_registered_converter(XGBRegressor,"XGBoostXGBRegressor",calculate_linear_regressor_output_shapes,convert_xgboost,)# The following instruction reduces the time spent by scikit-learn# to validate the data.set_config(assume_finite=True)

print(f"Number of cores:{cpu_count()}")

Number of cores: 20

try:out,err=run_cmd("lscpu")print(out)exceptExceptionase:print(f"lscpu not available:{e}")

<Popen: returncode: None args: ['lscpu']>

out,err=run_cmd("cat /proc/cpuinfo")print(out)

<Popen: returncode: None args: ['cat', '/proc/cpuinfo']>

defmeasure_inference(fct,X,repeat,max_time=5,quantile=1):"""    Run *repeat* times the same function on data *X*.    :param fct: fonction to run    :param X: data    :param repeat: number of times to run    :param max_time: maximum time to use to measure the inference    :return: number of runs, sum of the time, average, median    """times=[]for_ninrange(repeat):perf=time.perf_counter()fct(X)delta=time.perf_counter()-perftimes.append(delta)iflen(times)<3:continueifmax_timeisnotNoneandsum(times)>=max_time:breaktimes.sort()quantile=0if(len(times)-quantile*2)<3elsequantileifquantile==0:tt=timeselse:tt=times[quantile:-quantile]return(len(times),sum(times),sum(tt)/len(tt),times[len(times)//2])

small=cpu_count()<12ifsmall:N=1000n_features=10n_jobs=[1,cpu_count()//2,cpu_count()]n_ests=[10,20,30]depth=[4,6,8,10]Regressor=RandomForestRegressorelse:N=100000n_features=50n_jobs=[cpu_count(),cpu_count()//2,1]n_ests=[100,200,400]depth=[6,8,10,12,14]Regressor=RandomForestRegressorlegend=f"parallel-nf-{n_features}-"# avoid duplicates on machine with 1 or 2 cores.n_jobs=list(sorted(set(n_jobs),reverse=True))

repeat=7# repeat n times the same inferencequantile=1# exclude extreme timesmax_time=5# maximum number of seconds to spend on one configuration

X=numpy.random.randn(N,n_features).astype(numpy.float32)noise=(numpy.random.randn(X.shape[0])/(n_features//5)).astype(numpy.float32)y=X.mean(axis=1)+noisen_train=min(N,N//3)data=[]couples=list(product(n_jobs,depth,n_ests))bar=tqdm(couples)cache_dir="_cache"ifnotos.path.exists(cache_dir):os.mkdir(cache_dir)forn_j,max_depth,n_estimatorsinbar:ifn_j==1andn_estimators>n_ests[0]:# skippingcontinue# parallelizationcache_name=os.path.join(cache_dir,f"nf-{X.shape[1]}-rf-J-{n_j}-E-{n_estimators}-D-{max_depth}.pkl")ifos.path.exists(cache_name):withopen(cache_name,"rb")asf:rf=pickle.load(f)else:bar.set_description(f"J={n_j} E={n_estimators} D={max_depth} train rf")ifn_j==1andissubclass(Regressor,RandomForestRegressor):rf=Regressor(max_depth=max_depth,n_estimators=n_estimators,n_jobs=-1)rf.fit(X[:n_train],y[:n_train])rf.n_jobs=1else:rf=Regressor(max_depth=max_depth,n_estimators=n_estimators,n_jobs=n_j)rf.fit(X[:n_train],y[:n_train])withopen(cache_name,"wb")asf:pickle.dump(rf,f)bar.set_description(f"J={n_j} E={n_estimators} D={max_depth} ISession")so=SessionOptions()so.intra_op_num_threads=n_jcache_name=os.path.join(cache_dir,f"nf-{X.shape[1]}-rf-J-{n_j}-E-{n_estimators}-D-{max_depth}.onnx")ifos.path.exists(cache_name):sess=InferenceSession(cache_name,so,providers=["CPUExecutionProvider"])else:bar.set_description(f"J={n_j} E={n_estimators} D={max_depth} cvt onnx")onx=to_onnx(rf,X[:1])withopen(cache_name,"wb")asf:f.write(onx.SerializeToString())sess=InferenceSession(cache_name,so,providers=["CPUExecutionProvider"])onx_size=os.stat(cache_name).st_size# run once to avoid counting the first runbar.set_description(f"J={n_j} E={n_estimators} D={max_depth} predict1")rf.predict(X)sess.run(None,{"X":X})# fixed dataobs=dict(n_jobs=n_j,max_depth=max_depth,n_estimators=n_estimators,repeat=repeat,max_time=max_time,name=rf.__class__.__name__,n_rows=X.shape[0],n_features=X.shape[1],onnx_size=onx_size,)# baselinebar.set_description(f"J={n_j} E={n_estimators} D={max_depth} predictB")r,t,mean,med=measure_inference(rf.predict,X,repeat=repeat,max_time=max_time)o1=obs.copy()o1.update(dict(avg=mean,med=med,n_runs=r,ttime=t,name="base"))data.append(o1)# onnxruntimebar.set_description(f"J={n_j} E={n_estimators} D={max_depth} predictO")r,t,mean,med=measure_inference(lambdax,sess=sess:sess.run(None,{"X":x}),X,repeat=repeat,max_time=max_time,)o2=obs.copy()o2.update(dict(avg=mean,med=med,n_runs=r,ttime=t,name="ort_"))data.append(o2)

  0%|          | 0/45 [00:00<?, ?it/s]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=100 D=6 ISession:   0%|          | 0/45 [00:00<?, ?it/s]J=20 E=100 D=6 predict1:   0%|          | 0/45 [00:00<?, ?it/s]J=20 E=100 D=6 predictB:   0%|          | 0/45 [00:00<?, ?it/s]J=20 E=100 D=6 predictO:   0%|          | 0/45 [00:01<?, ?it/s]J=20 E=100 D=6 predictO:   2%|▏         | 1/45 [00:03<02:35,  3.54s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=200 D=6 ISession:   2%|▏         | 1/45 [00:03<02:35,  3.54s/it]J=20 E=200 D=6 predict1:   2%|▏         | 1/45 [00:03<02:35,  3.54s/it]J=20 E=200 D=6 predictB:   2%|▏         | 1/45 [00:04<02:35,  3.54s/it]J=20 E=200 D=6 predictO:   2%|▏         | 1/45 [00:06<02:35,  3.54s/it]J=20 E=200 D=6 predictO:   4%|▍         | 2/45 [00:09<03:35,  5.00s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=400 D=6 ISession:   4%|▍         | 2/45 [00:09<03:35,  5.00s/it]J=20 E=400 D=6 predict1:   4%|▍         | 2/45 [00:09<03:35,  5.00s/it]J=20 E=400 D=6 predictB:   4%|▍         | 2/45 [00:10<03:35,  5.00s/it]J=20 E=400 D=6 predictO:   4%|▍         | 2/45 [00:13<03:35,  5.00s/it]J=20 E=400 D=6 predictO:   7%|▋         | 3/45 [00:18<04:40,  6.68s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=100 D=8 ISession:   7%|▋         | 3/45 [00:18<04:40,  6.68s/it]J=20 E=100 D=8 predict1:   7%|▋         | 3/45 [00:18<04:40,  6.68s/it]J=20 E=100 D=8 predictB:   7%|▋         | 3/45 [00:18<04:40,  6.68s/it]J=20 E=100 D=8 predictO:   7%|▋         | 3/45 [00:19<04:40,  6.68s/it]J=20 E=100 D=8 predictO:   9%|▉         | 4/45 [00:21<03:37,  5.30s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=200 D=8 ISession:   9%|▉         | 4/45 [00:21<03:37,  5.30s/it]J=20 E=200 D=8 predict1:   9%|▉         | 4/45 [00:21<03:37,  5.30s/it]J=20 E=200 D=8 predictB:   9%|▉         | 4/45 [00:22<03:37,  5.30s/it]J=20 E=200 D=8 predictO:   9%|▉         | 4/45 [00:24<03:37,  5.30s/it]J=20 E=200 D=8 predictO:  11%|█         | 5/45 [00:26<03:34,  5.37s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=400 D=8 ISession:  11%|█         | 5/45 [00:26<03:34,  5.37s/it]J=20 E=400 D=8 predict1:  11%|█         | 5/45 [00:27<03:34,  5.37s/it]J=20 E=400 D=8 predictB:  11%|█         | 5/45 [00:27<03:34,  5.37s/it]J=20 E=400 D=8 predictO:  11%|█         | 5/45 [00:30<03:34,  5.37s/it]J=20 E=400 D=8 predictO:  13%|█▎        | 6/45 [00:35<04:05,  6.30s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=100 D=10 ISession:  13%|█▎        | 6/45 [00:35<04:05,  6.30s/it]J=20 E=100 D=10 predict1:  13%|█▎        | 6/45 [00:35<04:05,  6.30s/it]J=20 E=100 D=10 predictB:  13%|█▎        | 6/45 [00:35<04:05,  6.30s/it]J=20 E=100 D=10 predictO:  13%|█▎        | 6/45 [00:36<04:05,  6.30s/it]J=20 E=100 D=10 predictO:  16%|█▌        | 7/45 [00:37<03:17,  5.21s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=200 D=10 ISession:  16%|█▌        | 7/45 [00:38<03:17,  5.21s/it]J=20 E=200 D=10 predict1:  16%|█▌        | 7/45 [00:38<03:17,  5.21s/it]J=20 E=200 D=10 predictB:  16%|█▌        | 7/45 [00:38<03:17,  5.21s/it]J=20 E=200 D=10 predictO:  16%|█▌        | 7/45 [00:40<03:17,  5.21s/it]J=20 E=200 D=10 predictO:  18%|█▊        | 8/45 [00:42<03:05,  5.01s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=400 D=10 ISession:  18%|█▊        | 8/45 [00:42<03:05,  5.01s/it]J=20 E=400 D=10 predict1:  18%|█▊        | 8/45 [00:43<03:05,  5.01s/it]J=20 E=400 D=10 predictB:  18%|█▊        | 8/45 [00:44<03:05,  5.01s/it]J=20 E=400 D=10 predictO:  18%|█▊        | 8/45 [00:48<03:05,  5.01s/it]J=20 E=400 D=10 predictO:  20%|██        | 9/45 [00:52<04:00,  6.69s/it]~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(~/vv/this312/lib/python3.12/site-packages/sklearn/base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestRegressor from version 1.6.dev0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations  warnings.warn(J=20 E=100 D=12 ISession:  20%|██        | 9/45 [00:52<04:00,  6.69s/it]J=20 E=100 D=12 predict1:  20%|██        | 9/45 [00:53<04:00,  6.69s/it]J=20 E=100 D=12 predictB:  20%|██        | 9/45 [00:53<04:00,  6.69s/it]J=20 E=100 D=12 predictO:  20%|██        | 9/45 [00:54<04:00,  6.69s/it]J=20 E=100 D=12 predictO:  22%|██▏       | 10/45 [00:57<03:33,  6.09s/it]J=20 E=200 D=12 ISession:  22%|██▏       | 10/45 [00:57<03:33,  6.09s/it]J=20 E=200 D=12 predict1:  22%|██▏       | 10/45 [00:59<03:33,  6.09s/it]J=20 E=200 D=12 predictB:  22%|██▏       | 10/45 [01:00<03:33,  6.09s/it]J=20 E=200 D=12 predictO:  22%|██▏       | 10/45 [01:03<03:33,  6.09s/it]J=20 E=200 D=12 predictO:  24%|██▍       | 11/45 [01:08<04:20,  7.67s/it]J=20 E=400 D=12 ISession:  24%|██▍       | 11/45 [01:09<04:20,  7.67s/it]J=20 E=400 D=12 predict1:  24%|██▍       | 11/45 [01:11<04:20,  7.67s/it]J=20 E=400 D=12 predictB:  24%|██▍       | 11/45 [01:14<04:20,  7.67s/it]J=20 E=400 D=12 predictO:  24%|██▍       | 11/45 [01:18<04:20,  7.67s/it]J=20 E=400 D=12 predictO:  27%|██▋       | 12/45 [01:24<05:28,  9.95s/it]J=20 E=100 D=14 ISession:  27%|██▋       | 12/45 [01:24<05:28,  9.95s/it]J=20 E=100 D=14 predict1:  27%|██▋       | 12/45 [01:25<05:28,  9.95s/it]J=20 E=100 D=14 predictB:  27%|██▋       | 12/45 [01:26<05:28,  9.95s/it]J=20 E=100 D=14 predictO:  27%|██▋       | 12/45 [01:27<05:28,  9.95s/it]J=20 E=100 D=14 predictO:  29%|██▉       | 13/45 [01:31<04:53,  9.18s/it]J=20 E=200 D=14 ISession:  29%|██▉       | 13/45 [01:31<04:53,  9.18s/it]J=20 E=200 D=14 predict1:  29%|██▉       | 13/45 [01:33<04:53,  9.18s/it]J=20 E=200 D=14 predictB:  29%|██▉       | 13/45 [01:35<04:53,  9.18s/it]J=20 E=200 D=14 predictO:  29%|██▉       | 13/45 [01:38<04:53,  9.18s/it]J=20 E=200 D=14 predictO:  31%|███       | 14/45 [01:43<05:09,  9.97s/it]J=20 E=400 D=14 ISession:  31%|███       | 14/45 [01:43<05:09,  9.97s/it]J=20 E=400 D=14 predict1:  31%|███       | 14/45 [01:48<05:09,  9.97s/it]J=20 E=400 D=14 predictB:  31%|███       | 14/45 [01:51<05:09,  9.97s/it]J=20 E=400 D=14 predictO:  31%|███       | 14/45 [01:56<05:09,  9.97s/it]J=20 E=400 D=14 predictO:  33%|███▎      | 15/45 [02:04<06:38, 13.28s/it]J=10 E=100 D=6 ISession:  33%|███▎      | 15/45 [02:04<06:38, 13.28s/it]J=10 E=100 D=6 predict1:  33%|███▎      | 15/45 [02:04<06:38, 13.28s/it]J=10 E=100 D=6 predictB:  33%|███▎      | 15/45 [02:04<06:38, 13.28s/it]J=10 E=100 D=6 predictO:  33%|███▎      | 15/45 [02:05<06:38, 13.28s/it]J=10 E=100 D=6 predictO:  36%|███▌      | 16/45 [02:06<04:45,  9.85s/it]J=10 E=200 D=6 ISession:  36%|███▌      | 16/45 [02:06<04:45,  9.85s/it]J=10 E=200 D=6 predict1:  36%|███▌      | 16/45 [02:06<04:45,  9.85s/it]J=10 E=200 D=6 predictB:  36%|███▌      | 16/45 [02:06<04:45,  9.85s/it]J=10 E=200 D=6 predictO:  36%|███▌      | 16/45 [02:08<04:45,  9.85s/it]J=10 E=200 D=6 predictO:  38%|███▊      | 17/45 [02:09<03:41,  7.91s/it]J=10 E=400 D=6 train rf:  38%|███▊      | 17/45 [02:09<03:41,  7.91s/it]J=10 E=400 D=6 ISession:  38%|███▊      | 17/45 [03:06<03:41,  7.91s/it]J=10 E=400 D=6 cvt onnx:  38%|███▊      | 17/45 [03:06<03:41,  7.91s/it]J=10 E=400 D=6 predict1:  38%|███▊      | 17/45 [03:06<03:41,  7.91s/it]J=10 E=400 D=6 predictB:  38%|███▊      | 17/45 [03:07<03:41,  7.91s/it]J=10 E=400 D=6 predictO:  38%|███▊      | 17/45 [03:10<03:41,  7.91s/it]J=10 E=400 D=6 predictO:  40%|████      | 18/45 [03:13<11:04, 24.61s/it]J=10 E=100 D=8 train rf:  40%|████      | 18/45 [03:13<11:04, 24.61s/it]J=10 E=100 D=8 ISession:  40%|████      | 18/45 [03:31<11:04, 24.61s/it]J=10 E=100 D=8 cvt onnx:  40%|████      | 18/45 [03:31<11:04, 24.61s/it]J=10 E=100 D=8 predict1:  40%|████      | 18/45 [03:31<11:04, 24.61s/it]J=10 E=100 D=8 predictB:  40%|████      | 18/45 [03:32<11:04, 24.61s/it]J=10 E=100 D=8 predictO:  40%|████      | 18/45 [03:32<11:04, 24.61s/it]J=10 E=100 D=8 predictO:  42%|████▏     | 19/45 [03:33<10:08, 23.40s/it]J=10 E=200 D=8 train rf:  42%|████▏     | 19/45 [03:33<10:08, 23.40s/it]J=10 E=200 D=8 ISession:  42%|████▏     | 19/45 [04:15<10:08, 23.40s/it]J=10 E=200 D=8 cvt onnx:  42%|████▏     | 19/45 [04:15<10:08, 23.40s/it]J=10 E=200 D=8 predict1:  42%|████▏     | 19/45 [04:16<10:08, 23.40s/it]J=10 E=200 D=8 predictB:  42%|████▏     | 19/45 [04:16<10:08, 23.40s/it]J=10 E=200 D=8 predictO:  42%|████▏     | 19/45 [04:18<10:08, 23.40s/it]J=10 E=200 D=8 predictO:  44%|████▍     | 20/45 [04:19<12:35, 30.20s/it]J=10 E=400 D=8 train rf:  44%|████▍     | 20/45 [04:19<12:35, 30.20s/it]J=10 E=400 D=8 ISession:  44%|████▍     | 20/45 [05:32<12:35, 30.20s/it]J=10 E=400 D=8 cvt onnx:  44%|████▍     | 20/45 [05:32<12:35, 30.20s/it]J=10 E=400 D=8 predict1:  44%|████▍     | 20/45 [05:33<12:35, 30.20s/it]J=10 E=400 D=8 predictB:  44%|████▍     | 20/45 [05:35<12:35, 30.20s/it]J=10 E=400 D=8 predictO:  44%|████▍     | 20/45 [05:38<12:35, 30.20s/it]J=10 E=400 D=8 predictO:  47%|████▋     | 21/45 [05:41<18:18, 45.76s/it]J=10 E=100 D=10 train rf:  47%|████▋     | 21/45 [05:41<18:18, 45.76s/it]J=10 E=100 D=10 ISession:  47%|████▋     | 21/45 [06:04<18:18, 45.76s/it]J=10 E=100 D=10 cvt onnx:  47%|████▋     | 21/45 [06:04<18:18, 45.76s/it]J=10 E=100 D=10 predict1:  47%|████▋     | 21/45 [06:05<18:18, 45.76s/it]J=10 E=100 D=10 predictB:  47%|████▋     | 21/45 [06:05<18:18, 45.76s/it]J=10 E=100 D=10 predictO:  47%|████▋     | 21/45 [06:06<18:18, 45.76s/it]J=10 E=100 D=10 predictO:  49%|████▉     | 22/45 [06:07<15:14, 39.78s/it]J=10 E=200 D=10 train rf:  49%|████▉     | 22/45 [06:07<15:14, 39.78s/it]J=10 E=200 D=10 ISession:  49%|████▉     | 22/45 [06:50<15:14, 39.78s/it]J=10 E=200 D=10 cvt onnx:  49%|████▉     | 22/45 [06:50<15:14, 39.78s/it]J=10 E=200 D=10 predict1:  49%|████▉     | 22/45 [06:52<15:14, 39.78s/it]J=10 E=200 D=10 predictB:  49%|████▉     | 22/45 [06:53<15:14, 39.78s/it]J=10 E=200 D=10 predictO:  49%|████▉     | 22/45 [06:55<15:14, 39.78s/it]J=10 E=200 D=10 predictO:  51%|█████     | 23/45 [06:57<15:40, 42.75s/it]J=10 E=400 D=10 train rf:  51%|█████     | 23/45 [06:57<15:40, 42.75s/it]J=10 E=400 D=10 ISession:  51%|█████     | 23/45 [08:25<15:40, 42.75s/it]J=10 E=400 D=10 cvt onnx:  51%|█████     | 23/45 [08:25<15:40, 42.75s/it]J=10 E=400 D=10 predict1:  51%|█████     | 23/45 [08:31<15:40, 42.75s/it]J=10 E=400 D=10 predictB:  51%|█████     | 23/45 [08:32<15:40, 42.75s/it]J=10 E=400 D=10 predictO:  51%|█████     | 23/45 [08:35<15:40, 42.75s/it]J=10 E=400 D=10 predictO:  53%|█████▎    | 24/45 [08:40<21:17, 60.83s/it]J=10 E=100 D=12 train rf:  53%|█████▎    | 24/45 [08:40<21:17, 60.83s/it]J=10 E=100 D=12 ISession:  53%|█████▎    | 24/45 [09:06<21:17, 60.83s/it]J=10 E=100 D=12 cvt onnx:  53%|█████▎    | 24/45 [09:06<21:17, 60.83s/it]J=10 E=100 D=12 predict1:  53%|█████▎    | 24/45 [09:11<21:17, 60.83s/it]J=10 E=100 D=12 predictB:  53%|█████▎    | 24/45 [09:11<21:17, 60.83s/it]J=10 E=100 D=12 predictO:  53%|█████▎    | 24/45 [09:13<21:17, 60.83s/it]J=10 E=100 D=12 predictO:  56%|█████▌    | 25/45 [09:14<17:37, 52.90s/it]J=10 E=200 D=12 train rf:  56%|█████▌    | 25/45 [09:14<17:37, 52.90s/it]J=10 E=200 D=12 ISession:  56%|█████▌    | 25/45 [10:07<17:37, 52.90s/it]J=10 E=200 D=12 cvt onnx:  56%|█████▌    | 25/45 [10:07<17:37, 52.90s/it]J=10 E=200 D=12 predict1:  56%|█████▌    | 25/45 [10:16<17:37, 52.90s/it]J=10 E=200 D=12 predictB:  56%|█████▌    | 25/45 [10:17<17:37, 52.90s/it]J=10 E=200 D=12 predictO:  56%|█████▌    | 25/45 [10:19<17:37, 52.90s/it]J=10 E=200 D=12 predictO:  58%|█████▊    | 26/45 [10:23<18:14, 57.63s/it]J=10 E=400 D=12 train rf:  58%|█████▊    | 26/45 [10:23<18:14, 57.63s/it]J=10 E=400 D=12 ISession:  58%|█████▊    | 26/45 [12:02<18:14, 57.63s/it]J=10 E=400 D=12 cvt onnx:  58%|█████▊    | 26/45 [12:02<18:14, 57.63s/it]J=10 E=400 D=12 predict1:  58%|█████▊    | 26/45 [12:19<18:14, 57.63s/it]J=10 E=400 D=12 predictB:  58%|█████▊    | 26/45 [12:21<18:14, 57.63s/it]J=10 E=400 D=12 predictO:  58%|█████▊    | 26/45 [12:26<18:14, 57.63s/it]J=10 E=400 D=12 predictO:  60%|██████    | 27/45 [12:31<23:37, 78.77s/it]J=10 E=100 D=14 train rf:  60%|██████    | 27/45 [12:31<23:37, 78.77s/it]J=10 E=100 D=14 ISession:  60%|██████    | 27/45 [12:58<23:37, 78.77s/it]J=10 E=100 D=14 cvt onnx:  60%|██████    | 27/45 [12:58<23:37, 78.77s/it]J=10 E=100 D=14 predict1:  60%|██████    | 27/45 [13:08<23:37, 78.77s/it]J=10 E=100 D=14 predictB:  60%|██████    | 27/45 [13:09<23:37, 78.77s/it]J=10 E=100 D=14 predictO:  60%|██████    | 27/45 [13:11<23:37, 78.77s/it]J=10 E=100 D=14 predictO:  62%|██████▏   | 28/45 [13:13<19:11, 67.75s/it]J=10 E=200 D=14 train rf:  62%|██████▏   | 28/45 [13:13<19:11, 67.75s/it]J=10 E=200 D=14 ISession:  62%|██████▏   | 28/45 [14:08<19:11, 67.75s/it]J=10 E=200 D=14 cvt onnx:  62%|██████▏   | 28/45 [14:08<19:11, 67.75s/it]J=10 E=200 D=14 predict1:  62%|██████▏   | 28/45 [14:28<19:11, 67.75s/it]J=10 E=200 D=14 predictB:  62%|██████▏   | 28/45 [14:29<19:11, 67.75s/it]J=10 E=200 D=14 predictO:  62%|██████▏   | 28/45 [14:33<19:11, 67.75s/it]J=10 E=200 D=14 predictO:  64%|██████▍   | 29/45 [14:38<19:27, 72.97s/it]J=10 E=400 D=14 train rf:  64%|██████▍   | 29/45 [14:38<19:27, 72.97s/it]J=10 E=400 D=14 ISession:  64%|██████▍   | 29/45 [16:27<19:27, 72.97s/it]J=10 E=400 D=14 cvt onnx:  64%|██████▍   | 29/45 [16:27<19:27, 72.97s/it]J=10 E=400 D=14 predict1:  64%|██████▍   | 29/45 [17:07<19:27, 72.97s/it]J=10 E=400 D=14 predictB:  64%|██████▍   | 29/45 [17:10<19:27, 72.97s/it]J=10 E=400 D=14 predictO:  64%|██████▍   | 29/45 [17:16<19:27, 72.97s/it]J=10 E=400 D=14 predictO:  67%|██████▋   | 30/45 [17:22<25:04, 100.31s/it]J=1 E=100 D=6 train rf:  67%|██████▋   | 30/45 [17:22<25:04, 100.31s/it]J=1 E=100 D=6 ISession:  67%|██████▋   | 30/45 [17:32<25:04, 100.31s/it]J=1 E=100 D=6 cvt onnx:  67%|██████▋   | 30/45 [17:32<25:04, 100.31s/it]J=1 E=100 D=6 predict1:  67%|██████▋   | 30/45 [17:32<25:04, 100.31s/it]J=1 E=100 D=6 predictB:  67%|██████▋   | 30/45 [17:33<25:04, 100.31s/it]J=1 E=100 D=6 predictO:  67%|██████▋   | 30/45 [17:39<25:04, 100.31s/it]J=1 E=100 D=6 predictO:  69%|██████▉   | 31/45 [17:42<17:44, 76.04s/it]J=1 E=100 D=8 train rf:  69%|██████▉   | 31/45 [17:42<17:44, 76.04s/it]J=1 E=100 D=8 ISession:  69%|██████▉   | 31/45 [17:55<17:44, 76.04s/it]J=1 E=100 D=8 cvt onnx:  69%|██████▉   | 31/45 [17:55<17:44, 76.04s/it]J=1 E=100 D=8 predict1:  69%|██████▉   | 31/45 [17:56<17:44, 76.04s/it]J=1 E=100 D=8 predictB:  69%|██████▉   | 31/45 [17:58<17:44, 76.04s/it]J=1 E=100 D=8 predictO:  69%|██████▉   | 31/45 [18:03<17:44, 76.04s/it]J=1 E=100 D=8 predictO:  76%|███████▌  | 34/45 [18:07<06:59, 38.13s/it]J=1 E=100 D=10 train rf:  76%|███████▌  | 34/45 [18:07<06:59, 38.13s/it]J=1 E=100 D=10 ISession:  76%|███████▌  | 34/45 [18:22<06:59, 38.13s/it]J=1 E=100 D=10 cvt onnx:  76%|███████▌  | 34/45 [18:22<06:59, 38.13s/it]J=1 E=100 D=10 predict1:  76%|███████▌  | 34/45 [18:24<06:59, 38.13s/it]J=1 E=100 D=10 predictB:  76%|███████▌  | 34/45 [18:26<06:59, 38.13s/it]J=1 E=100 D=10 predictO:  76%|███████▌  | 34/45 [18:31<06:59, 38.13s/it]J=1 E=100 D=10 predictO:  82%|████████▏ | 37/45 [18:37<03:24, 25.53s/it]J=1 E=100 D=12 train rf:  82%|████████▏ | 37/45 [18:37<03:24, 25.53s/it]J=1 E=100 D=12 ISession:  82%|████████▏ | 37/45 [18:54<03:24, 25.53s/it]J=1 E=100 D=12 cvt onnx:  82%|████████▏ | 37/45 [18:54<03:24, 25.53s/it]J=1 E=100 D=12 predict1:  82%|████████▏ | 37/45 [18:58<03:24, 25.53s/it]J=1 E=100 D=12 predictB:  82%|████████▏ | 37/45 [19:02<03:24, 25.53s/it]J=1 E=100 D=12 predictO:  82%|████████▏ | 37/45 [19:07<03:24, 25.53s/it]J=1 E=100 D=12 predictO:  89%|████████▉ | 40/45 [19:12<01:40, 20.15s/it]J=1 E=100 D=14 train rf:  89%|████████▉ | 40/45 [19:12<01:40, 20.15s/it]J=1 E=100 D=14 ISession:  89%|████████▉ | 40/45 [19:31<01:40, 20.15s/it]J=1 E=100 D=14 cvt onnx:  89%|████████▉ | 40/45 [19:31<01:40, 20.15s/it]J=1 E=100 D=14 predict1:  89%|████████▉ | 40/45 [19:40<01:40, 20.15s/it]J=1 E=100 D=14 predictB:  89%|████████▉ | 40/45 [19:46<01:40, 20.15s/it]J=1 E=100 D=14 predictO:  89%|████████▉ | 40/45 [19:53<01:40, 20.15s/it]J=1 E=100 D=14 predictO:  96%|█████████▌| 43/45 [20:04<00:38, 19.18s/it]J=1 E=100 D=14 predictO: 100%|██████████| 45/45 [20:04<00:00, 26.78s/it]

name=os.path.join(cache_dir,"plot_beanchmark_rf")print(f"Saving data into{name!r}")df=pandas.DataFrame(data)df2=df.copy()df2["legend"]=legenddf2.to_csv(f"{name}-{legend}.csv",index=False)

Saving data into '_cache/plot_beanchmark_rf'

df

70 rows × 13 columns

n_rows=len(n_jobs)n_cols=len(n_ests)fig,axes=plt.subplots(n_rows,n_cols,figsize=(4*n_cols,4*n_rows))fig.suptitle(f"{rf.__class__.__name__}\nX.shape={X.shape}")forn_j,n_estimatorsintqdm(product(n_jobs,n_ests)):i=n_jobs.index(n_j)j=n_ests.index(n_estimators)ax=axes[i,j]subdf=df[(df.n_estimators==n_estimators)&(df.n_jobs==n_j)]ifsubdf.shape[0]==0:continuepiv=subdf.pivot(index="max_depth",columns="name",values=["avg","med"])piv.plot(ax=ax,title=f"jobs={n_j}, trees={n_estimators}")ax.set_ylabel(f"n_jobs={n_j}",fontsize="small")ax.set_xlabel("max_depth",fontsize="small")# ratioax2=ax.twinx()piv1=subdf.pivot(index="max_depth",columns="name",values="avg")piv1["speedup"]=piv1.base/piv1.ort_ax2.plot(piv1.index,piv1.speedup,"b--",label="speedup avg")piv1=subdf.pivot(index="max_depth",columns="name",values="med")piv1["speedup"]=piv1.base/piv1.ort_ax2.plot(piv1.index,piv1.speedup,"y--",label="speedup med")ax2.legend(fontsize="x-small")# 1ax2.plot(piv1.index,[1for_inpiv1.index],"k--",label="no speedup")foriinrange(axes.shape[0]):forjinrange(axes.shape[1]):axes[i,j].legend(fontsize="small")fig.tight_layout()fig.savefig(f"{name}-{legend}.png")# plt.show()

0it [00:00, ?it/s]3it [00:00, 25.48it/s]9it [00:00, 43.67it/s]~/github/onnx-array-api/_doc/examples/plot_benchmark_rf.py:307: UserWarning: No artists with labels found to put in legend.  Note that artists whose label start with an underscore are ignored when legend() is called with no argument.  axes[i, j].legend(fontsize="small")

DownloadJupyternotebook:plot_benchmark_rf.ipynb

DownloadPythonsourcecode:plot_benchmark_rf.py

Downloadzipped:plot_benchmark_rf.zip