Fix sarix sigma_pooling='shared' bug and update dependencies

This commit fixes a failing test for SARIX models with shared sigma pooling
and updates the sarix dependency to resolve an upstream bug.

Problem:
--------
The test `test_sarix_shared_sigma_pooling_multiple_batches` was failing with
a reshape error when using `sigma_pooling='shared'` with multiple locations.
The error occurred in the sarix library at line 392 where it incorrectly used
`theta` instead of `sigma` when reshaping arrays:
  TypeError: cannot reshape array of shape (100, 5, 6) into shape (100, 1, 1)

Root Cause:
-----------
The installed sarix package (v0.0.1 from elray1/sarix) contained a bug where
the variable name was wrong in the sigma pooling code block. The reichlab/sarix
repository had already fixed this bug in v0.2.0.

Fixes:
------
1. Updated sarix dependency from elray1/sarix to reichlab/sarix (v0.2.0)
2. Updated requires-python from >=3.9 to >=3.11 (required by newer sarix)
3. Regenerated uv.lock with updated dependencies
4. Regenerated requirements.txt and requirements-dev.txt
5. Added explicit string conversion for output_type_id in CSV output
6. Fixed test assertion to handle pandas type inference for output_type_id

Test Results:
-------------
✅ test_sarix - PASSED
✅ test_sarix_shared_sigma_pooling_multiple_batches - PASSED (was failing)
✅ test_drop_level_feats - PASSED

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: nickreich

pyproject.toml

@@ -3,7 +3,7 @@ name = "idmodels"
 description = "A Python module for modeling infectious disease."
 license = {text = "MIT License"}
 readme = "README.md"
-requires-python = ">=3.9"
+requires-python = ">=3.11"
 classifiers = [
     "Programming Language :: Python :: 3",
     "License :: OSI Approved :: MIT License",

requirements/requirements-dev.txt

@@ -20,10 +20,6 @@ botocore==1.35.36
     # via aiobotocore
 cfgv==3.4.0
     # via pre-commit
-colorama==0.4.6
-    # via
-    #   pytest
-    #   tqdm
 contourpy==1.3.0
     # via matplotlib
 coverage==7.6.4
@@ -42,19 +38,19 @@ frozenlist==1.5.0
     #   aiosignal
 fsspec==2024.10.0
     # via s3fs
-iddata @ git+https://github.com/reichlab/iddata@3ad0ac0dc6d7f14488628a49bfb5228ca0643e1b
+iddata @ git+https://github.com/reichlab/iddata@c28849b2a02ab84e2f82876f16fee2ac60814877
     # via idmodels (pyproject.toml)
 identify==2.6.1
     # via pre-commit
 idna==3.10
     # via yarl
 iniconfig==2.0.0
     # via pytest
-jax==0.4.35
+jax==0.8.0
     # via
     #   numpyro
     #   sarix
-jaxlib==0.4.35
+jaxlib==0.8.0
     # via
     #   jax
     #   numpyro
@@ -100,7 +96,7 @@ numpy==2.1.3
     #   scikit-learn
     #   scipy
     #   timeseriesutils
-numpyro==0.15.3
+numpyro==0.19.0
     # via sarix
 opt-einsum==3.4.0
     # via jax
@@ -112,7 +108,6 @@ pandas==2.2.3
     # via
     #   idmodels (pyproject.toml)
     #   iddata
-    #   sarix
     #   timeseriesutils
 pillow==11.0.0
     # via matplotlib
@@ -147,7 +142,7 @@ ruff==0.7.2
     # via idmodels (pyproject.toml)
 s3fs==2024.10.0
     # via iddata
-sarix @ git+https://github.com/elray1/sarix@1c8995942d49afbb66637f0dc0662e1248606af4
+sarix @ git+https://github.com/reichlab/sarix@cedaebed367d9fc00e2b34cdfc6db09eef6a99a3
     # via idmodels (pyproject.toml)
 scikit-learn==1.5.2
     # via idmodels (pyproject.toml)

requirements/requirements.txt

@@ -18,8 +18,6 @@ attrs==24.2.0
     #   pymmwr
 botocore==1.35.36
     # via aiobotocore
-colorama==0.4.6
-    # via tqdm
 contourpy==1.3.0
     # via matplotlib
 cycler==0.12.1
@@ -32,15 +30,15 @@ frozenlist==1.5.0
     #   aiosignal
 fsspec==2024.10.0
     # via s3fs
-iddata @ git+https://github.com/reichlab/iddata@3ad0ac0dc6d7f14488628a49bfb5228ca0643e1b
+iddata @ git+https://github.com/reichlab/iddata@c28849b2a02ab84e2f82876f16fee2ac60814877
     # via idmodels (pyproject.toml)
 idna==3.10
     # via yarl
-jax==0.4.35
+jax==0.8.0
     # via
     #   numpyro
     #   sarix
-jaxlib==0.4.35
+jaxlib==0.8.0
     # via
     #   jax
     #   numpyro
@@ -84,7 +82,7 @@ numpy==2.1.3
     #   scikit-learn
     #   scipy
     #   timeseriesutils
-numpyro==0.15.3
+numpyro==0.19.0
     # via sarix
 opt-einsum==3.4.0
     # via jax
@@ -94,7 +92,6 @@ pandas==2.2.3
     # via
     #   idmodels (pyproject.toml)
     #   iddata
-    #   sarix
     #   timeseriesutils
 pillow==11.0.0
     # via matplotlib
@@ -117,7 +114,7 @@ rich==13.9.4
     # via iddata
 s3fs==2024.10.0
     # via iddata
-sarix @ git+https://github.com/elray1/sarix@1c8995942d49afbb66637f0dc0662e1248606af4
+sarix @ git+https://github.com/reichlab/sarix@cedaebed367d9fc00e2b34cdfc6db09eef6a99a3
     # via idmodels (pyproject.toml)
 scikit-learn==1.5.2
     # via idmodels (pyproject.toml)

src/idmodels/sarix.py

@@ -106,6 +106,8 @@ def run(self, run_config):
             run_config=run_config,
             model_config=self.model_config
         )
+        # Ensure output_type_id is string to avoid pandas inferring it as float when reading
+        preds_df["output_type_id"] = preds_df["output_type_id"].astype(str)
         preds_df.to_csv(save_path, index=False)
 
 

tests/integration/test_sarix.py

@@ -136,7 +136,8 @@ def test_sarix_shared_sigma_pooling_multiple_batches(tmp_path):
         "Output should contain predictions for all input locations"
     assert all(actual_df["output_type"] == "quantile"), \
         "All outputs should be quantiles"
-    assert set(actual_df["output_type_id"].unique()) == set(run_config.q_labels), \
+    # Convert output_type_id to string for comparison since pandas may infer numeric types
+    assert set(actual_df["output_type_id"].astype(str).unique()) == set(run_config.q_labels), \
         "Output should contain all specified quantile levels"
     assert actual_df["value"].notna().all(), \
         "All predictions should be non-null"