xrtm-train 0.2.2__tar.gz → 0.2.4__tar.gz

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xrtm-train
-Version: 0.2.2
+Version: 0.2.4
 Summary: The Learning/Optimization layer for XRTM.
 Author-email: XRTM Team <moy@xrtm.org>
 License-Expression: Apache-2.0
@@ -14,8 +14,8 @@ Requires-Dist: click>=8.0.0
 Requires-Dist: rich>=13.0.0
 Requires-Dist: pyyaml>=6.0.0
 Requires-Dist: xrtm-data>=0.2.2
-Requires-Dist: xrtm-eval>=0.2.2
-Requires-Dist: xrtm-forecast>=0.6.2
+Requires-Dist: xrtm-eval>=0.2.3
+Requires-Dist: xrtm-forecast>=0.6.3
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "xrtm-train"
-version = "0.2.2"
+version = "0.2.4"
 description = "The Learning/Optimization layer for XRTM."
 readme = "README.md"
 requires-python = ">=3.11"
@@ -20,8 +20,8 @@ dependencies = [
     "rich>=13.0.0",
     "pyyaml>=6.0.0",
     "xrtm-data>=0.2.2",
-    "xrtm-eval>=0.2.2",
-    "xrtm-forecast>=0.6.2",
+    "xrtm-eval>=0.2.3",
+    "xrtm-forecast>=0.6.3",
 ]
 
 [project.optional-dependencies]
@@ -46,11 +46,6 @@ xrtm-train = "xrtm.train.cli:main"
 package-dir = {"" = "src"}
 packages = {find = {where = ["src"], include = ["xrtm*"], namespaces = true}}
 
-[tool.uv.sources]
-xrtm-data = { workspace = true }
-xrtm-eval = { workspace = true }
-xrtm-forecast = { workspace = true }
-
 [tool.pytest.ini_options]
 testpaths = ["tests"]
 asyncio_mode = "strict"

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/src/xrtm/train/cli/__init__.py

@@ -33,6 +33,7 @@ Example:
 """
 
 import json
+from copy import deepcopy
 from datetime import datetime, timedelta, timezone
 from pathlib import Path
 from typing import Any, Dict, Optional
@@ -86,22 +87,29 @@ DEFAULT_CONFIG = {
 
 def load_config(config_path: Optional[str]) -> Dict[str, Any]:
     r"""Load configuration from YAML file, merged with defaults."""
-    config: Dict[str, Any] = DEFAULT_CONFIG.copy()
+    config: Dict[str, Any] = deepcopy(DEFAULT_CONFIG)
 
     if config_path:
         with open(config_path) as f:
-            user_config = yaml.safe_load(f)
+            user_config = yaml.safe_load(f) or {}
+        if not isinstance(user_config, dict):
+            raise ValueError("Training config must be a YAML mapping")
 
-        # Deep merge
-        for key, value in user_config.items():
-            if isinstance(value, dict) and key in config and isinstance(config[key], dict):
-                config[key].update(value)
-            else:
-                config[key] = value
+        _deep_merge(config, user_config)
 
     return config
 
 
+def _deep_merge(base: Dict[str, Any], overrides: Dict[str, Any]) -> Dict[str, Any]:
+    r"""Recursively merge overrides into base."""
+    for key, value in overrides.items():
+        if isinstance(value, dict) and isinstance(base.get(key), dict):
+            _deep_merge(base[key], value)
+        else:
+            base[key] = value
+    return base
+
+
 # =============================================================================
 # CLI Group
 # =============================================================================
@@ -228,6 +236,7 @@ def prepare(
 @click.option("--epochs", type=int, default=None, help="Number of epochs (overrides config)")
 @click.option("--batch-size", type=int, default=None, help="Batch size (overrides config)")
 @click.option("--dry-run", is_flag=True, help="Validate config without training")
+@click.option("--allow-mock", is_flag=True, help="Run the legacy mock training loop for demos only.")
 def train(
     config: Optional[str],
     data_dir: str,
@@ -238,16 +247,16 @@ def train(
     epochs: Optional[int],
     batch_size: Optional[int],
     dry_run: bool,
+    allow_mock: bool,
 ):
     r"""
-    Train a forecasting model with prior injection.
+    Validate training configuration for prior-injected models.
 
-    Supports full fine-tuning and LoRA. Uses Qwen 3 family by default.
-    Progress and checkpoints are saved automatically.
+    Full model training is not wired into the product path yet. Use --dry-run
+    to validate configuration or --allow-mock for the legacy demo loop.
 
     Example:
-        xrtm-train train -c config.yaml -d samples/ -o models/
-        xrtm-train train -d samples/ -m Qwen/Qwen2.5-1.5B --epochs 3
+        xrtm-train train -c config.yaml -d samples/ --dry-run
     """
     # Load and merge config
     cfg = load_config(config)
@@ -290,6 +299,12 @@ def train(
         console.print("\n[yellow]Dry run complete. No training performed.[/yellow]")
         return
 
+    if not allow_mock:
+        raise click.ClickException(
+            "Real model training is not implemented in this CLI yet. Use --dry-run to validate config, "
+            "or pass --allow-mock to run the legacy demo loop intentionally."
+        )
+
     # Check for GPU dependencies
     try:
         import torch
@@ -393,14 +408,16 @@ def _run_training(cfg: dict, data_dir: str, output: str, resume: Optional[str])
 @click.option("--model", "-m", required=True, type=click.Path(exists=True), help="Model directory")
 @click.option("--data", "-d", "data_path", required=True, type=click.Path(exists=True), help="Test data file")
 @click.option("--output", "-o", type=click.Path(), help="Output file for results (.json)")
-def eval(model: str, data_path: str, output: Optional[str]):
+@click.option("--allow-mock", is_flag=True, help="Emit the legacy mock metrics for demos only.")
+def eval(model: str, data_path: str, output: Optional[str], allow_mock: bool):
     r"""
-    Evaluate a trained model on test data.
+    Validate evaluation inputs for a trained model.
 
-    Computes calibration metrics, Brier scores, and generates evaluation report.
+    Full model evaluation is not wired into the product path yet. Use --allow-mock
+    only when intentionally exercising the legacy demo metrics.
 
     Example:
-        xrtm-train eval -m models/final -d samples/test.jsonl
+        xrtm-train eval -m models/final -d samples/test.jsonl --allow-mock
     """
     console.print(Panel(
         f"[bold blue]Evaluating Model[/bold blue]\n"
@@ -423,8 +440,14 @@ def eval(model: str, data_path: str, output: Optional[str]):
     test_samples = _load_jsonl(Path(data_path))
     console.print(f"\nLoaded {len(test_samples)} test samples")
 
+    if not allow_mock:
+        raise click.ClickException(
+            "Real model evaluation is not implemented in this CLI yet. Pass --allow-mock to emit "
+            "the legacy demo metrics intentionally."
+        )
+
     # Mock evaluation
-    console.print("\n[yellow]⚠ Full evaluation pending. Using mock metrics.[/yellow]")
+    console.print("\n[yellow]⚠ Using legacy mock metrics because --allow-mock was provided.[/yellow]")
 
     metrics = {
         "brier_score": 0.15,
@@ -469,7 +492,7 @@ def init_config(output: str, model: str, strategy: str):
     Example:
         xrtm-train init-config -o config.yaml -m Qwen/Qwen2.5-7B
     """
-    config: Dict[str, Any] = DEFAULT_CONFIG.copy()
+    config: Dict[str, Any] = deepcopy(DEFAULT_CONFIG)
     config["model"]["name"] = model  # type: ignore[index]
     config["model"]["strategy"] = strategy  # type: ignore[index]
 

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/src/xrtm/train/version.py

@@ -21,7 +21,7 @@ This module provides the single source of truth for the package version.
 
 __all__ = ["__version__", "__author__", "__contact__", "__license__", "__copyright__"]
 
-__version__ = "0.2.2"
+__version__ = "0.2.4"
 __author__ = "XRTM Team"
 __contact__ = "moy@xrtm.org"
 __license__ = "Apache-2.0"

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/src/xrtm_train.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xrtm-train
-Version: 0.2.2
+Version: 0.2.4
 Summary: The Learning/Optimization layer for XRTM.
 Author-email: XRTM Team <moy@xrtm.org>
 License-Expression: Apache-2.0
@@ -14,8 +14,8 @@ Requires-Dist: click>=8.0.0
 Requires-Dist: rich>=13.0.0
 Requires-Dist: pyyaml>=6.0.0
 Requires-Dist: xrtm-data>=0.2.2
-Requires-Dist: xrtm-eval>=0.2.2
-Requires-Dist: xrtm-forecast>=0.6.2
+Requires-Dist: xrtm-eval>=0.2.3
+Requires-Dist: xrtm-forecast>=0.6.3
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/src/xrtm_train.egg-info/requires.txt

@@ -5,8 +5,8 @@ click>=8.0.0
 rich>=13.0.0
 pyyaml>=6.0.0
 xrtm-data>=0.2.2
-xrtm-eval>=0.2.2
-xrtm-forecast>=0.6.2
+xrtm-eval>=0.2.3
+xrtm-forecast>=0.6.3
 
 [dev]
 pytest>=7.0.0

{xrtm_train-0.2.2 → xrtm_train-0.2.4}/tests/test_cli.py

@@ -18,7 +18,7 @@ from pathlib import Path
 
 from click.testing import CliRunner
 
-from xrtm.train.cli import main
+from xrtm.train.cli import DEFAULT_CONFIG, load_config, main
 
 
 def test_data_prepare_builds_jsonl_samples() -> None:
@@ -102,3 +102,47 @@ def test_data_prepare_can_derive_news_and_priors_from_trade_records() -> None:
         assert rows[0]["current_news"] == "Market probability update: 0.400"
         assert rows[0]["prior_alpha"] == 4.0
         assert rows[0]["target_alpha"] == 6.0
+
+
+def test_load_config_deep_merge_does_not_mutate_defaults() -> None:
+    runner = CliRunner()
+
+    with runner.isolated_filesystem():
+        config_path = Path("config.yaml")
+        config_path.write_text("model:\n  name: custom-model\n")
+
+        config = load_config(str(config_path))
+
+        assert config["model"]["name"] == "custom-model"
+        assert config["model"]["family"] == DEFAULT_CONFIG["model"]["family"]
+        assert DEFAULT_CONFIG["model"]["name"] == "Qwen/Qwen2.5-0.5B"
+
+
+def test_train_command_does_not_run_mock_training_by_default() -> None:
+    runner = CliRunner()
+
+    with runner.isolated_filesystem():
+        samples = Path("samples")
+        samples.mkdir()
+        (samples / "train.jsonl").write_text("{}\n")
+        (samples / "val.jsonl").write_text("{}\n")
+
+        result = runner.invoke(main, ["train", "--data", str(samples)])
+
+        assert result.exit_code != 0
+        assert "Real model training is not implemented" in result.output
+
+
+def test_eval_command_does_not_emit_mock_metrics_by_default() -> None:
+    runner = CliRunner()
+
+    with runner.isolated_filesystem():
+        model = Path("model")
+        model.mkdir()
+        data = Path("test.jsonl")
+        data.write_text("{}\n")
+
+        result = runner.invoke(main, ["eval", "--model", str(model), "--data", str(data)])
+
+        assert result.exit_code != 0
+        assert "Real model evaluation is not implemented" in result.output