@pikaa-ai/pikaa 0.3.1 → 0.3.4

package/skills/pymatgen/scripts/symmetry_sensitivity_report.py

@@ -1,212 +0,0 @@
-#!/usr/bin/env python3
-"""Report space-group sensitivity across explicit symmetry tolerances."""
-
-from __future__ import annotations
-
-import argparse
-import math
-from typing import Any
-
-from _common import (
-    CliError,
-    DEFAULT_MAX_INPUT_BYTES,
-    DEFAULT_MAX_OUTPUT_BYTES,
-    DEFAULT_MAX_SITES,
-    checked_output_file,
-    emit_json,
-    load_structure,
-    positive_int,
-    write_json_new,
-)
-
-
-def parse_tolerances(value: str, label: str) -> list[float]:
-    """Parse a unique comma-separated finite positive float list."""
-    pieces = [piece.strip() for piece in value.split(",") if piece.strip()]
-    if not pieces:
-        raise CliError(f"{label} must contain at least one value")
-    result: list[float] = []
-    for piece in pieces:
-        try:
-            number = float(piece)
-        except ValueError as exc:
-            raise CliError(f"{label} contains a non-number: {piece!r}") from exc
-        if not math.isfinite(number) or number <= 0:
-            raise CliError(f"{label} values must be finite and positive")
-        if number not in result:
-            result.append(number)
-    if len(result) > 10:
-        raise CliError(f"{label} may contain at most 10 unique values")
-    return result
-
-
-def analyze_grid(
-    structure: Any,
-    *,
-    symprec_values: list[float],
-    angle_values: list[float],
-) -> dict[str, Any]:
-    """Evaluate a bounded Cartesian product of symmetry tolerances."""
-    combinations = len(symprec_values) * len(angle_values)
-    if combinations > 25:
-        raise CliError("symmetry tolerance grid may contain at most 25 combinations")
-    from pymatgen.symmetry.analyzer import SpacegroupAnalyzer
-
-    rows: list[dict[str, Any]] = []
-    assignments: set[tuple[str, int]] = set()
-    failures = 0
-    for symprec in symprec_values:
-        for angle in angle_values:
-            try:
-                analyzer = SpacegroupAnalyzer(
-                    structure,
-                    symprec=symprec,
-                    angle_tolerance=angle,
-                )
-                symbol = analyzer.get_space_group_symbol()
-                number = int(analyzer.get_space_group_number())
-                assignments.add((symbol, number))
-                rows.append(
-                    {
-                        "symprec_angstrom": symprec,
-                        "angle_tolerance_degrees": angle,
-                        "space_group_symbol": symbol,
-                        "space_group_number": number,
-                        "crystal_system": str(analyzer.get_crystal_system()),
-                        "point_group_symbol": analyzer.get_point_group_symbol(),
-                        "symmetry_operations": len(
-                            analyzer.get_symmetry_operations()
-                        ),
-                        "equivalent_site_groups": len(
-                            analyzer.get_symmetrized_structure().equivalent_indices
-                        ),
-                    }
-                )
-            except (RuntimeError, TypeError, ValueError) as exc:
-                failures += 1
-                rows.append(
-                    {
-                        "symprec_angstrom": symprec,
-                        "angle_tolerance_degrees": angle,
-                        "error": f"{type(exc).__name__}: {exc}"[:500],
-                    }
-                )
-    return {
-        "backend": "spglib through pymatgen",
-        "grid": rows,
-        "combinations": combinations,
-        "failures": failures,
-        "distinct_assignments": [
-            {"space_group_symbol": symbol, "space_group_number": number}
-            for symbol, number in sorted(assignments, key=lambda item: item[1])
-        ],
-        "tolerance_sensitive": len(assignments) > 1 or failures > 0,
-        "interpretation": (
-            "A tolerance-sensitive assignment must be reported with its exact "
-            "symprec and angle_tolerance; it is not a unique structure invariant."
-        ),
-    }
-
-
-def build_parser() -> argparse.ArgumentParser:
-    parser = argparse.ArgumentParser(
-        description=(
-            "Compare space-group assignments across a bounded tolerance grid. "
-            "The structure is not standardized or written."
-        )
-    )
-    parser.add_argument("structure_file")
-    parser.add_argument("--structure-index", type=int, default=0)
-    parser.add_argument(
-        "--symprec",
-        default="0.001,0.01,0.1",
-        help="Comma-separated distance tolerances in angstrom",
-    )
-    parser.add_argument(
-        "--angle-tolerance",
-        default="1,5",
-        help="Comma-separated angle tolerances in degrees",
-    )
-    parser.add_argument(
-        "--allow-disordered",
-        action="store_true",
-        help="Acknowledge that symmetry assignment for disorder may be misleading",
-    )
-    parser.add_argument("--output", help="New JSON output; default stdout")
-    parser.add_argument(
-        "--max-input-bytes",
-        type=positive_int,
-        default=DEFAULT_MAX_INPUT_BYTES,
-    )
-    parser.add_argument("--max-sites", type=positive_int, default=DEFAULT_MAX_SITES)
-    parser.add_argument(
-        "--max-output-bytes",
-        type=positive_int,
-        default=DEFAULT_MAX_OUTPUT_BYTES,
-    )
-    return parser
-
-
-def main() -> int:
-    args = build_parser().parse_args()
-    try:
-        if args.structure_index < 0:
-            raise CliError("--structure-index must be non-negative")
-        symprec_values = parse_tolerances(args.symprec, "--symprec")
-        angle_values = parse_tolerances(
-            args.angle_tolerance, "--angle-tolerance"
-        )
-        structure, input_path, parse_report = load_structure(
-            args.structure_file,
-            structure_index=args.structure_index,
-            max_bytes=args.max_input_bytes,
-            max_sites=args.max_sites,
-        )
-        if not structure.is_ordered and not args.allow_disordered:
-            raise CliError(
-                "disordered structure requires --allow-disordered after reviewing "
-                "occupancies and the chosen symmetry model"
-            )
-        report = {
-            "ok": True,
-            "analysis": "symmetry_tolerance_sensitivity",
-            "input": parse_report,
-            "structure": {
-                "formula": structure.composition.reduced_formula,
-                "sites": len(structure),
-                "ordered": bool(structure.is_ordered),
-                "periodic_boundary_conditions": [
-                    bool(value) for value in structure.lattice.pbc
-                ],
-            },
-            "symmetry": analyze_grid(
-                structure,
-                symprec_values=symprec_values,
-                angle_values=angle_values,
-            ),
-            "disorder_acknowledged": bool(args.allow_disordered),
-            "structure_modified": False,
-        }
-        if args.output:
-            output = checked_output_file(args.output, input_paths=(input_path,))
-            write_json_new(output, report, max_bytes=args.max_output_bytes)
-            emit_json(
-                {
-                    "ok": True,
-                    "output": output.name,
-                    "overwrote_existing": False,
-                    "tolerance_sensitive": report["symmetry"][
-                        "tolerance_sensitive"
-                    ],
-                }
-            )
-        else:
-            emit_json(report)
-        return 0
-    except (CliError, ImportError, OSError, RuntimeError, TypeError, ValueError) as exc:
-        emit_json({"ok": False, "error": f"{type(exc).__name__}: {exc}"[:1000]})
-        return 2
-
-
-if __name__ == "__main__":
-    raise SystemExit(main())

package/skills/pymc/assets/hierarchical_model_template.py

@@ -1,332 +0,0 @@
-"""
-PyMC Hierarchical/Multilevel Model Template
-
-This template provides a complete workflow for Bayesian hierarchical models,
-useful for grouped/nested data (e.g., students within schools, patients within hospitals).
-
-Customize the sections marked with # TODO
-"""
-
-import pymc as pm
-import arviz as az
-import numpy as np
-import pandas as pd
-import matplotlib.pyplot as plt
-
-# =============================================================================
-# 1. DATA PREPARATION
-# =============================================================================
-
-# TODO: Load your data with group structure
-# Example:
-# df = pd.read_csv('data.csv')
-# groups = df['group_id'].values
-# X = df['predictor'].values
-# y = df['outcome'].values
-
-# For demonstration: Generate hierarchical data
-np.random.seed(42)
-n_groups = 10
-n_per_group = 20
-n_obs = n_groups * n_per_group
-
-# True hierarchical structure
-true_mu_alpha = 5.0
-true_sigma_alpha = 2.0
-true_mu_beta = 1.5
-true_sigma_beta = 0.5
-true_sigma = 1.0
-
-group_alphas = np.random.normal(true_mu_alpha, true_sigma_alpha, n_groups)
-group_betas = np.random.normal(true_mu_beta, true_sigma_beta, n_groups)
-
-# Generate data
-groups = np.repeat(np.arange(n_groups), n_per_group)
-X = np.random.randn(n_obs)
-y = group_alphas[groups] + group_betas[groups] * X + np.random.randn(n_obs) * true_sigma
-
-# TODO: Customize group names
-group_names = [f'Group_{i}' for i in range(n_groups)]
-
-# =============================================================================
-# 2. BUILD HIERARCHICAL MODEL
-# =============================================================================
-
-print("Building hierarchical model...")
-
-coords = {
-    'groups': group_names,
-    'obs': np.arange(n_obs)
-}
-
-with pm.Model(coords=coords) as hierarchical_model:
-    # Data containers (for later predictions)
-    X_data = pm.Data('X_data', X)
-    groups_data = pm.Data('groups_data', groups)
-
-    # Hyperpriors (population-level parameters)
-    # TODO: Adjust hyperpriors based on your domain knowledge
-    mu_alpha = pm.Normal('mu_alpha', mu=0, sigma=10)
-    sigma_alpha = pm.HalfNormal('sigma_alpha', sigma=5)
-
-    mu_beta = pm.Normal('mu_beta', mu=0, sigma=10)
-    sigma_beta = pm.HalfNormal('sigma_beta', sigma=5)
-
-    # Group-level parameters (non-centered parameterization)
-    # Non-centered parameterization improves sampling efficiency
-    alpha_offset = pm.Normal('alpha_offset', mu=0, sigma=1, dims='groups')
-    alpha = pm.Deterministic('alpha', mu_alpha + sigma_alpha * alpha_offset, dims='groups')
-
-    beta_offset = pm.Normal('beta_offset', mu=0, sigma=1, dims='groups')
-    beta = pm.Deterministic('beta', mu_beta + sigma_beta * beta_offset, dims='groups')
-
-    # Observation-level model
-    mu = alpha[groups_data] + beta[groups_data] * X_data
-
-    # Observation noise
-    sigma = pm.HalfNormal('sigma', sigma=5)
-
-    # Likelihood; tie shape to X_data so prediction data can have a new row count
-    y_obs = pm.Normal('y_obs', mu=mu, sigma=sigma, observed=y, shape=X_data.shape[0], dims='obs')
-
-print("Model built successfully!")
-print(f"Groups: {n_groups}")
-print(f"Observations: {n_obs}")
-
-# =============================================================================
-# 3. PRIOR PREDICTIVE CHECK
-# =============================================================================
-
-print("\nRunning prior predictive check...")
-with hierarchical_model:
-    prior_pred = pm.sample_prior_predictive(draws=500, random_seed=42)
-
-# Visualize prior predictions
-fig, ax = plt.subplots(figsize=(10, 6))
-az.plot_ppc(prior_pred, group='prior', num_pp_samples=100, ax=ax)
-ax.set_title('Prior Predictive Check')
-plt.tight_layout()
-plt.savefig('hierarchical_prior_check.png', dpi=300, bbox_inches='tight')
-print("Prior predictive check saved to 'hierarchical_prior_check.png'")
-
-# =============================================================================
-# 4. FIT MODEL
-# =============================================================================
-
-print("\nFitting hierarchical model...")
-print("(This may take a few minutes due to model complexity)")
-
-with hierarchical_model:
-    # MCMC sampling with higher target_accept for hierarchical models
-    idata = pm.sample(
-        draws=2000,
-        tune=2000,  # More tuning for hierarchical models
-        chains=4,
-        target_accept=0.95,  # Higher for better convergence
-        random_seed=42,
-        idata_kwargs={'log_likelihood': True}
-    )
-
-print("Sampling complete!")
-
-# =============================================================================
-# 5. CHECK DIAGNOSTICS
-# =============================================================================
-
-print("\n" + "="*60)
-print("DIAGNOSTICS")
-print("="*60)
-
-# Summary for key parameters
-summary = az.summary(
-    idata,
-    var_names=['mu_alpha', 'sigma_alpha', 'mu_beta', 'sigma_beta', 'sigma', 'alpha', 'beta']
-)
-print("\nParameter Summary:")
-print(summary)
-
-# Check convergence
-bad_rhat = summary[summary['r_hat'] > 1.01]
-if len(bad_rhat) > 0:
-    print(f"\n⚠️  WARNING: {len(bad_rhat)} parameters with R-hat > 1.01")
-    print(bad_rhat[['r_hat']])
-else:
-    print("\n✓ All R-hat values < 1.01 (good convergence)")
-
-# Check effective sample size
-low_ess = summary[summary['ess_bulk'] < 400]
-if len(low_ess) > 0:
-    print(f"\n⚠️  WARNING: {len(low_ess)} parameters with ESS < 400")
-    print(low_ess[['ess_bulk']].head(10))
-else:
-    print("\n✓ All ESS values > 400 (sufficient samples)")
-
-# Check divergences
-divergences = idata.sample_stats.diverging.sum().item()
-if divergences > 0:
-    print(f"\n⚠️  WARNING: {divergences} divergent transitions")
-    print("   This is common in hierarchical models - non-centered parameterization already applied")
-    print("   Consider even higher target_accept or stronger hyperpriors")
-else:
-    print("\n✓ No divergences")
-
-# Trace plots for hyperparameters
-az.plot_trace_dist(
-    idata,
-    var_names=['mu_alpha', 'sigma_alpha', 'mu_beta', 'sigma_beta', 'sigma'],
-)
-plt.tight_layout()
-plt.savefig('hierarchical_trace_plots.png', dpi=300, bbox_inches='tight')
-print("\nTrace plots saved to 'hierarchical_trace_plots.png'")
-
-# =============================================================================
-# 6. POSTERIOR PREDICTIVE CHECK
-# =============================================================================
-
-print("\nRunning posterior predictive check...")
-with hierarchical_model:
-    pm.sample_posterior_predictive(idata, extend_inferencedata=True, random_seed=42)
-
-# Visualize fit
-fig, ax = plt.subplots(figsize=(10, 6))
-az.plot_ppc(idata, num_pp_samples=100, ax=ax)
-ax.set_title('Posterior Predictive Check')
-plt.tight_layout()
-plt.savefig('hierarchical_posterior_check.png', dpi=300, bbox_inches='tight')
-print("Posterior predictive check saved to 'hierarchical_posterior_check.png'")
-
-# =============================================================================
-# 7. ANALYZE HIERARCHICAL STRUCTURE
-# =============================================================================
-
-print("\n" + "="*60)
-print("POPULATION-LEVEL (HYPERPARAMETER) ESTIMATES")
-print("="*60)
-
-# Population-level estimates
-hyper_summary = summary.loc[['mu_alpha', 'sigma_alpha', 'mu_beta', 'sigma_beta', 'sigma']]
-print(hyper_summary[['mean', 'sd', 'hdi_3%', 'hdi_97%']])
-
-# Forest plot for group-level parameters
-fig, axes = plt.subplots(1, 2, figsize=(14, 8))
-
-# Group intercepts
-az.plot_forest(idata, var_names=['alpha'], combined=True, ax=axes[0])
-axes[0].set_title('Group-Level Intercepts (α)')
-axes[0].set_yticklabels(group_names)
-axes[0].axvline(idata.posterior['mu_alpha'].mean().item(), color='red', linestyle='--', label='Population mean')
-axes[0].legend()
-
-# Group slopes
-az.plot_forest(idata, var_names=['beta'], combined=True, ax=axes[1])
-axes[1].set_title('Group-Level Slopes (β)')
-axes[1].set_yticklabels(group_names)
-axes[1].axvline(idata.posterior['mu_beta'].mean().item(), color='red', linestyle='--', label='Population mean')
-axes[1].legend()
-
-plt.tight_layout()
-plt.savefig('group_level_estimates.png', dpi=300, bbox_inches='tight')
-print("\nGroup-level estimates saved to 'group_level_estimates.png'")
-
-# Shrinkage visualization
-fig, axes = plt.subplots(1, 2, figsize=(12, 5))
-
-# Intercepts
-alpha_samples = idata.posterior['alpha'].values.reshape(-1, n_groups)
-alpha_means = alpha_samples.mean(axis=0)
-mu_alpha_mean = idata.posterior['mu_alpha'].mean().item()
-
-axes[0].scatter(range(n_groups), alpha_means, alpha=0.6)
-axes[0].axhline(mu_alpha_mean, color='red', linestyle='--', label='Population mean')
-axes[0].set_xlabel('Group')
-axes[0].set_ylabel('Intercept')
-axes[0].set_title('Group Intercepts (showing shrinkage to population mean)')
-axes[0].legend()
-
-# Slopes
-beta_samples = idata.posterior['beta'].values.reshape(-1, n_groups)
-beta_means = beta_samples.mean(axis=0)
-mu_beta_mean = idata.posterior['mu_beta'].mean().item()
-
-axes[1].scatter(range(n_groups), beta_means, alpha=0.6)
-axes[1].axhline(mu_beta_mean, color='red', linestyle='--', label='Population mean')
-axes[1].set_xlabel('Group')
-axes[1].set_ylabel('Slope')
-axes[1].set_title('Group Slopes (showing shrinkage to population mean)')
-axes[1].legend()
-
-plt.tight_layout()
-plt.savefig('shrinkage_plot.png', dpi=300, bbox_inches='tight')
-print("Shrinkage plot saved to 'shrinkage_plot.png'")
-
-# =============================================================================
-# 8. PREDICTIONS FOR NEW DATA
-# =============================================================================
-
-# TODO: Specify new data
-# For existing groups:
-# new_X = np.array([...])
-# new_groups = np.array([0, 1, 2, ...])  # Existing group indices
-
-# For a new group (predict using population-level parameters):
-# Just use mu_alpha and mu_beta
-
-print("\n" + "="*60)
-print("PREDICTIONS FOR NEW DATA")
-print("="*60)
-
-# Example: Predict for existing groups
-new_X = np.array([-2, -1, 0, 1, 2])
-new_groups = np.array([0, 2, 4, 6, 8])  # Select some groups
-
-with hierarchical_model:
-    pm.set_data({'X_data': new_X, 'groups_data': new_groups}, coords={'obs': np.arange(len(new_X))})
-
-    post_pred = pm.sample_posterior_predictive(
-        idata,
-        var_names=['y_obs'],
-        predictions=True,
-        random_seed=42
-    )
-
-y_pred_samples = post_pred.predictions['y_obs']
-y_pred_mean = y_pred_samples.mean(dim=['chain', 'draw']).values
-y_pred_hdi = az.hdi(y_pred_samples, hdi_prob=0.95).values
-
-print(f"Predictions for existing groups:")
-print(f"{'Group':<10} {'X':<10} {'Mean':<15} {'95% HDI Lower':<15} {'95% HDI Upper':<15}")
-print("-"*65)
-for i, g in enumerate(new_groups):
-    print(f"{group_names[g]:<10} {new_X[i]:<10.2f} {y_pred_mean[i]:<15.3f} {y_pred_hdi[i, 0]:<15.3f} {y_pred_hdi[i, 1]:<15.3f}")
-
-# Predict for a new group (using population parameters)
-print(f"\nPrediction for a NEW group (using population-level parameters):")
-new_X_newgroup = np.array([0.0])
-
-# Manually compute using population parameters
-mu_alpha_samples = idata.posterior['mu_alpha'].values.flatten()
-mu_beta_samples = idata.posterior['mu_beta'].values.flatten()
-sigma_samples = idata.posterior['sigma'].values.flatten()
-
-# Predicted mean for new group
-y_pred_newgroup = mu_alpha_samples + mu_beta_samples * new_X_newgroup[0]
-y_pred_mean_newgroup = y_pred_newgroup.mean()
-y_pred_hdi_newgroup = az.hdi(y_pred_newgroup, hdi_prob=0.95)
-
-print(f"X = {new_X_newgroup[0]:.2f}")
-print(f"Predicted mean: {y_pred_mean_newgroup:.3f}")
-print(f"95% HDI: [{y_pred_hdi_newgroup[0]:.3f}, {y_pred_hdi_newgroup[1]:.3f}]")
-
-# =============================================================================
-# 9. SAVE RESULTS
-# =============================================================================
-
-idata.to_netcdf('hierarchical_model_results.nc')
-print("\nResults saved to 'hierarchical_model_results.nc'")
-
-summary.to_csv('hierarchical_model_summary.csv')
-print("Summary saved to 'hierarchical_model_summary.csv'")
-
-print("\n" + "="*60)
-print("ANALYSIS COMPLETE")
-print("="*60)