pymc-devs · fonnesbeck · Feb 4, 2018 · Feb 1, 2018 · Feb 1, 2018 · Feb 1, 2018
diff --git a/RELEASE-NOTES.md b/RELEASE-NOTES.md
@@ -16,6 +16,7 @@
 ### Deprecations
 
 - DIC and BPIC calculations have been removed
+- `njobs` and `nchains` deprecated in favor of `cores` and `chains` for `sample`
 
 ## PyMC 3.3 (January 9, 2018)
 

diff --git a/benchmarks/benchmarks/benchmarks.py b/benchmarks/benchmarks/benchmarks.py
@@ -115,11 +115,11 @@ def time_drug_evaluation(self):
             pm.Deterministic('difference of stds', group1_std - group2_std)
             pm.Deterministic(
                 'effect size', diff_of_means / np.sqrt((group1_std**2 + group2_std**2) / 2))
-            pm.sample(20000, njobs=4, chains=4)
+            pm.sample(20000, cores=4, chains=4)
 
     def time_glm_hierarchical(self):
         with glm_hierarchical_model():
-            pm.sample(draws=20000, njobs=4, chains=4)
+            pm.sample(draws=20000, cores=4, chains=4)
 
 
 class NUTSInitSuite(object):
@@ -141,7 +141,7 @@ def track_glm_hierarchical_ess(self, init):
         with glm_hierarchical_model():
             start, step = pm.init_nuts(init=init, chains=self.chains, progressbar=False, random_seed=123)
             t0 = time.time()
-            trace = pm.sample(draws=self.draws, step=step, njobs=4, chains=self.chains,
+            trace = pm.sample(draws=self.draws, step=step, cores=4, chains=self.chains,
                               start=start, random_seed=100)
             tot = time.time() - t0
         ess = pm.effective_n(trace, ('mu_a',))['mu_a']
@@ -154,7 +154,7 @@ def track_marginal_mixture_model_ess(self, init):
                                    progressbar=False, random_seed=123)
             start = [{k: v for k, v in start.items()} for _ in range(self.chains)]
             t0 = time.time()
-            trace = pm.sample(draws=self.draws, step=step, njobs=4, chains=self.chains,
+            trace = pm.sample(draws=self.draws, step=step, cores=4, chains=self.chains,
                               start=start, random_seed=100)
             tot = time.time() - t0
         ess = pm.effective_n(trace, ('mu',))['mu'].min()  # worst case
@@ -178,7 +178,7 @@ def track_glm_hierarchical_ess(self, step):
             if step is not None:
                 step = step()
             t0 = time.time()
-            trace = pm.sample(draws=self.draws, step=step, njobs=4, chains=4,
+            trace = pm.sample(draws=self.draws, step=step, cores=4, chains=4,
                               random_seed=100)
             tot = time.time() - t0
         ess = pm.effective_n(trace, ('mu_a',))['mu_a']

diff --git a/docs/source/notebooks/AR.ipynb b/docs/source/notebooks/AR.ipynb
@@ -152,7 +152,7 @@
     "with pm.Model() as ar1:\n",
     "    beta = pm.Normal('beta', mu=0, sd=tau)\n",
     "    data = pm.AR('y', beta, sd=1.0, observed=y)\n",
-    "    trace = pm.sample(1000, njobs=4)\n",
+    "    trace = pm.sample(1000, cores=4)\n",
     "    \n",
     "pm.traceplot(trace);"
    ]
@@ -278,7 +278,7 @@
     "with pm.Model() as ar2:\n",
     "    beta = pm.Normal('beta', mu=0, sd=tau, shape=2)\n",
     "    data = pm.AR('y', beta, sd=1.0, observed=y)\n",
-    "    trace = pm.sample(1000, njobs=4)\n",
+    "    trace = pm.sample(1000, cores=4)\n",
     "    \n",
     "pm.traceplot(trace);"
    ]
@@ -340,7 +340,7 @@
     "    beta = pm.Normal('beta', mu=0, sd=tau)\n",
     "    beta2 = pm.Uniform('beta2')\n",
     "    data = pm.AR('y', [beta, beta2], sd=1.0, observed=y)\n",
-    "    trace = pm.sample(1000, njobs=4)\n",
+    "    trace = pm.sample(1000, cores=4)\n",
     "    \n",
     "pm.traceplot(trace);"
    ]

diff --git a/docs/source/notebooks/BEST.ipynb b/docs/source/notebooks/BEST.ipynb
@@ -260,7 +260,7 @@
    ],
    "source": [
     "with model:\n",
-    "    trace = pm.sample(2000, njobs=2)"
+    "    trace = pm.sample(2000, cores=2)"
    ]
   },
   {

diff --git a/docs/source/notebooks/GLM-linear.ipynb b/docs/source/notebooks/GLM-linear.ipynb
@@ -198,7 +198,7 @@
     "                        sd=sigma, observed=y)\n",
     "    \n",
     "    # Inference!\n",
-    "    trace = sample(3000, njobs=2) # draw 3000 posterior samples using NUTS sampling"
+    "    trace = sample(3000, cores=2) # draw 3000 posterior samples using NUTS sampling"
    ]
   },
   {
@@ -234,7 +234,7 @@
     "    # specify glm and pass in data. The resulting linear model, its likelihood and \n",
     "    # and all its parameters are automatically added to our model.\n",
     "    glm.GLM.from_formula('y ~ x', data)\n",
-    "    trace = sample(3000, njobs=2) # draw 3000 posterior samples using NUTS sampling"
+    "    trace = sample(3000, cores=2) # draw 3000 posterior samples using NUTS sampling"
    ]
   },
   {

diff --git a/docs/source/notebooks/GLM-negative-binomial-regression.ipynb b/docs/source/notebooks/GLM-negative-binomial-regression.ipynb
@@ -452,7 +452,7 @@
     "    # C = pm.approx_hessian(start)\n",
     "    # trace = pm.sample(4000, step=pm.NUTS(scaling=C))\n",
     "    \n",
-    "    trace = pm.sample(2000, njobs=2)"
+    "    trace = pm.sample(2000, cores=2)"
    ]
   },
   {

diff --git a/docs/source/notebooks/GLM-poisson-regression.ipynb b/docs/source/notebooks/GLM-poisson-regression.ipynb
@@ -640,7 +640,7 @@
    ],
    "source": [
     "with mdl_fish:\n",
-    "    trc_fish = pm.sample(2000, tune=1000, njobs=4)[1000:]"
+    "    trc_fish = pm.sample(2000, tune=1000, cores=4)[1000:]"
    ]
   },
   {

diff --git a/docs/source/notebooks/GLM-robust.ipynb b/docs/source/notebooks/GLM-robust.ipynb
@@ -142,7 +142,7 @@
    "source": [
     "with pm.Model() as model:\n",
     "    pm.glm.GLM.from_formula('y ~ x', data)\n",
-    "    trace = pm.sample(2000, njobs=2)"
+    "    trace = pm.sample(2000, cores=2)"
    ]
   },
   {
@@ -262,7 +262,7 @@
     "with pm.Model() as model_robust:\n",
     "    family = pm.glm.families.StudentT()\n",
     "    pm.glm.GLM.from_formula('y ~ x', data, family=family)\n",
-    "    trace_robust = pm.sample(2000, njobs=2)\n",
+    "    trace_robust = pm.sample(2000, cores=2)\n",
     "\n",
     "plt.figure(figsize=(7, 5))\n",
     "plt.plot(x_out, y_out, 'x')\n",

diff --git a/docs/source/notebooks/GLM-rolling-regression.ipynb b/docs/source/notebooks/GLM-rolling-regression.ipynb
@@ -327,7 +327,7 @@
    ],
    "source": [
     "with model_randomwalk:\n",
-    "    trace_rw = pm.sample(tune=2000, njobs=4, samples=200, \n",
+    "    trace_rw = pm.sample(tune=2000, cores=4, samples=200, \n",
     "                         nuts_kwargs=dict(target_accept=.9))"
    ]
   },

diff --git a/docs/source/notebooks/GLM.ipynb b/docs/source/notebooks/GLM.ipynb
@@ -88,7 +88,7 @@
     "    lm = glm.LinearComponent.from_formula('y ~ x', data)\n",
     "    sigma = Uniform('sigma', 0, 20)\n",
     "    y_obs = Normal('y_obs', mu=lm.y_est, sd=sigma, observed=y)\n",
-    "    trace = sample(2000, njobs=2)\n",
+    "    trace = sample(2000, cores=2)\n",
     "\n",
     "plt.figure(figsize=(5, 5))\n",
     "plt.plot(x, y, 'x')\n",
@@ -135,7 +135,7 @@
    "source": [
     "with Model() as model:\n",
     "    GLM.from_formula('y ~ x', data)\n",
-    "    trace = sample(2000, njobs=2)\n",
+    "    trace = sample(2000, cores=2)\n",
     "\n",
     "plt.figure(figsize=(5, 5))\n",
     "plt.plot(x, y, 'x')\n",
@@ -194,7 +194,7 @@
    "source": [
     "with Model() as model:\n",
     "    GLM.from_formula('y ~ x', data_outlier)\n",
-    "    trace = sample(2000, njobs=2)\n",
+    "    trace = sample(2000, cores=2)\n",
     "\n",
     "plt.figure(figsize=(5, 5))\n",
     "plt.plot(x_out, y_out, 'x')\n",
@@ -244,7 +244,7 @@
     "                                   priors={'nu': 1.5,\n",
     "                                           'lam': Uniform.dist(0, 20)})\n",
     "    GLM.from_formula('y ~ x', data_outlier, family=family)    \n",
-    "    trace = sample(2000, njobs=2)\n",
+    "    trace = sample(2000, cores=2)\n",
     "\n",
     "plt.figure(figsize=(5, 5))\n",
     "plt.plot(x_out, y_out, 'x')\n",
@@ -301,7 +301,7 @@
     "          'prcnt_take': Normal.dist(mu=grp_mean, sd=grp_sd)\n",
     "    }\n",
     "    GLM.from_formula('sat_t ~ spend + stu_tea_rat + salary + prcnt_take', sat_data, priors=priors)\n",
-    "    trace_sat = sample(2000, njobs=2)"
+    "    trace_sat = sample(2000, cores=2)"
    ]
   },
   {
@@ -358,7 +358,7 @@
     "    intercept = Normal.dist(mu=sat_data.sat_t.mean(), sd=sat_data.sat_t.std())\n",
     "    GLM.from_formula('sat_t ~ spend + stu_tea_rat + salary + prcnt_take', sat_data,\n",
     "        priors={'Intercept': intercept, 'Regressor': slope})\n",
-    "    trace_sat = sample(2000, njobs=2)"
+    "    trace_sat = sample(2000, cores=2)"
    ]
   },
   {
@@ -417,7 +417,7 @@
     "    GLM.from_formula('sat_t ~ spend + stu_tea_rat + salary + prcnt_take', sat_data,\n",
     "                priors={'Intercept': intercept, 'Regressor': slope})\n",
     "\n",
-    "    trace_sat = sample(2000, njobs=2)"
+    "    trace_sat = sample(2000, cores=2)"
    ]
   },
   {
@@ -599,7 +599,7 @@
    "source": [
     "with Model() as model_htwt:\n",
     "    GLM.from_formula('male ~ height + weight', htwt_data, family=glm.families.Binomial())\n",
-    "    trace_htwt = sample(2000, njobs=2)"
+    "    trace_htwt = sample(2000, cores=2)"
    ]
   },
   {
@@ -747,7 +747,7 @@
     "    GLM.from_formula('male ~ height + weight', htwt_data, family=glm.families.Binomial(),\n",
     "                    priors=priors)\n",
     "    \n",
-    "    trace_lasso = sample(500, njobs=2)\n",
+    "    trace_lasso = sample(500, cores=2)\n",
     "    \n",
     "trace_df = trace_to_dataframe(trace_lasso)\n",
     "scatter_matrix(trace_df, figsize=(8, 8));\n",

diff --git a/docs/source/notebooks/LKJ.ipynb b/docs/source/notebooks/LKJ.ipynb
@@ -288,7 +288,7 @@
    ],
    "source": [
     "with model:\n",
-    "    trace = pm.sample(random_seed=SEED, njobs=4)"
+    "    trace = pm.sample(random_seed=SEED, cores=4)"
    ]
   },
   {

diff --git a/docs/source/notebooks/MvGaussianRandomWalk_demo.ipynb b/docs/source/notebooks/MvGaussianRandomWalk_demo.ipynb
@@ -139,7 +139,7 @@
     "\n",
     "        sd = pm.Uniform('sd', 0, 1)\n",
     "        likelihood = pm.Normal('y', mu=regression, sd=sd, observed=y_t)\n",
-    "        trace = pm.sample(n_samples, njobs=4)\n",
+    "        trace = pm.sample(n_samples, cores=4)\n",
     "\n",
     "    return trace, y_scaler, t_scaler, t_section"
    ]

diff --git a/docs/source/notebooks/PyMC3_tips_and_heuristic.ipynb b/docs/source/notebooks/PyMC3_tips_and_heuristic.ipynb
@@ -484,7 +484,7 @@
     "    # Proportion sptial variance\n",
     "    alpha = pm.Deterministic('alpha', sd_c/(sd_h+sd_c))\n",
     "\n",
-    "    trace1 = pm.sample(3e3, njobs=2, tune=1000, nuts_kwargs={'max_treedepth': 15})"
+    "    trace1 = pm.sample(3e3, cores=2, tune=1000, nuts_kwargs={'max_treedepth': 15})"
    ]
   },
   {
@@ -702,7 +702,7 @@
     "    # Proportion sptial variance\n",
     "    alpha = pm.Deterministic('alpha', sd_c/(sd_h+sd_c))\n",
     "\n",
-    "    trace2 = pm.sample(3e3, njobs=2, tune=1000, nuts_kwargs={'max_treedepth': 15})"
+    "    trace2 = pm.sample(3e3, cores=2, tune=1000, nuts_kwargs={'max_treedepth': 15})"
    ]
   },
   {
@@ -856,7 +856,7 @@
     "    # Likelihood\n",
     "    Yi = pm.Poisson('Yi', mu=mu.ravel(), observed=O)\n",
     "\n",
-    "    trace3 = pm.sample(3e3, njobs=2, tune=1000)"
+    "    trace3 = pm.sample(3e3, cores=2, tune=1000)"
    ]
   },
   {
@@ -1104,7 +1104,7 @@
     "    # Likelihood\n",
     "    Yi = pm.Poisson('Yi', mu=mu.ravel(), observed=O)\n",
     "\n",
-    "    trace4 = pm.sample(3e3, njobs=2, tune=1000)"
+    "    trace4 = pm.sample(3e3, cores=2, tune=1000)"
    ]
   },
   {

diff --git a/docs/source/notebooks/api_quickstart.ipynb b/docs/source/notebooks/api_quickstart.ipynb
@@ -814,7 +814,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "You can also run multiple chains in parallel using the `njobs` kwarg:"
+    "You can also run multiple chains in parallel using the `cores` kwarg:"
    ]
   },
   {
@@ -837,7 +837,7 @@
     "    mu = pm.Normal('mu', mu=0, sd=1)\n",
     "    obs = pm.Normal('obs', mu=mu, sd=1, observed=np.random.randn(100))\n",
     "    \n",
-    "    trace = pm.sample(njobs=4)"
+    "    trace = pm.sample(cores=4)"
    ]
   },
   {
@@ -1009,7 +1009,7 @@
     "    \n",
     "    step1 = pm.Metropolis(vars=[mu])\n",
     "    step2 = pm.Slice(vars=[sd])\n",
-    "    trace = pm.sample(10000, step=[step1, step2], njobs=4)"
+    "    trace = pm.sample(10000, step=[step1, step2], cores=4)"
    ]
   },
   {
@@ -1159,7 +1159,7 @@
    "source": [
     "with pm.Model() as model:\n",
     "    x = pm.Normal('x', mu=0, sd=1, shape=100)    \n",
-    "    trace = pm.sample(njobs=4)\n",
+    "    trace = pm.sample(cores=4)\n",
     "    \n",
     "pm.energyplot(trace);"
    ]

diff --git a/docs/source/notebooks/hierarchical_partial_pooling.ipynb b/docs/source/notebooks/hierarchical_partial_pooling.ipynb
@@ -170,7 +170,7 @@
    ],
    "source": [
     "with baseball_model:\n",
-    "    trace = pm.sample(2000, tune=1000, nchains=2,\n",
+    "    trace = pm.sample(2000, tune=1000, chains=2,\n",
     "                      nuts_kwargs={'target_accept': 0.95})"
    ]
   },

diff --git a/docs/source/notebooks/normalizing_flows_overview.ipynb b/docs/source/notebooks/normalizing_flows_overview.ipynb
@@ -438,7 +438,7 @@
     "pm.set_tt_rng(42)\n",
     "np.random.seed(42)\n",
     "with pot1m:\n",
-    "    trace = pm.sample(1000, init='auto', njobs=2, start=[dict(pot1=np.array([-2, 0])),\n",
+    "    trace = pm.sample(1000, init='auto', cores=2, start=[dict(pot1=np.array([-2, 0])),\n",
     "                                                         dict(pot1=np.array([2, 0]))])"
    ]
   },
@@ -975,7 +975,7 @@
    ],
    "source": [
     "with pot_m:\n",
-    "    traceNUTS = pm.sample(3000, tune=1000, target_accept=0.9, njobs=2)"
+    "    traceNUTS = pm.sample(3000, tune=1000, target_accept=0.9, cores=2)"
    ]
   },
   {

diff --git a/docs/source/notebooks/probabilistic_matrix_factorization.ipynb b/docs/source/notebooks/probabilistic_matrix_factorization.ipynb
@@ -810,17 +810,17 @@
     "    basename = 'pmf-mcmc-d%d' % self.dim\n",
     "    return os.path.join(DATA_DIR, basename)\n",
     "\n",
-    "def _draw_samples(self, nsamples=1000, njobs=2):\n",
+    "def _draw_samples(self, nsamples=1000, cores=2):\n",
     "    # First make sure the trace_dir does not already exist.\n",
     "    if os.path.isdir(self.trace_dir):\n",
     "        shutil.rmtree(self.trace_dir)\n",
     "\n",
     "    with self.model:\n",
-    "        logging.info('drawing %d samples using %d jobs' % (nsamples, njobs))\n",
+    "        logging.info('drawing %d samples using %d jobs' % (nsamples, cores))\n",
     "        backend = pm.backends.Text(self.trace_dir)\n",
     "        logging.info('backing up trace to directory: %s' % self.trace_dir)\n",
     "        self.trace = pm.sample(draws=nsamples, init='advi',\n",
-    "                               n_init=150000, njobs=njobs, trace=backend)\n",
+    "                               n_init=150000, cores=cores, trace=backend)\n",
     "        \n",
     "def _load_trace(self):\n",
     "    with self.model:\n",
@@ -837,7 +837,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "We could define some kind of default trace property like we did for the MAP, but that would mean using possibly nonsensical values for `nsamples` and `njobs`. Better to leave it as a non-optional call to `draw_samples`. Finally, we'll need a function to make predictions using our inferred values for $U$ and $V$. For user $i$ and joke $j$, a prediction is generated by drawing from $\\mathcal{N}(U_i V_j^T, \\alpha)$. To generate predictions from the sampler, we generate an $R$ matrix for each $U$ and $V$ sampled, then we combine these by averaging over the $K$ samples.\n",
+    "We could define some kind of default trace property like we did for the MAP, but that would mean using possibly nonsensical values for `nsamples` and `cores`. Better to leave it as a non-optional call to `draw_samples`. Finally, we'll need a function to make predictions using our inferred values for $U$ and $V$. For user $i$ and joke $j$, a prediction is generated by drawing from $\\mathcal{N}(U_i V_j^T, \\alpha)$. To generate predictions from the sampler, we generate an $R$ matrix for each $U$ and $V$ sampled, then we combine these by averaging over the $K$ samples.\n",
     "\n",
     "\\begin{equation}\n",
     "P(R_{ij}^* \\given R, \\alpha, \\alpha_U, \\alpha_V) \\approx\n",

diff --git a/docs/source/notebooks/rugby_analytics.ipynb b/docs/source/notebooks/rugby_analytics.ipynb
@@ -815,7 +815,7 @@
    ],
    "source": [
     "with model:\n",
-    "    trace = pm.sample(1000, tune=1000, njobs=3)\n",
+    "    trace = pm.sample(1000, tune=1000, cores=3)\n",
     "    pm.traceplot(trace)"
    ]
   },