alloc 0.0.13__tar.gz → 0.0.14__tar.gz

{alloc-0.0.13 → alloc-0.0.14}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: alloc
-Version: 0.0.13
+Version: 0.0.14
 Summary: Engineer-first training calibration: estimate VRAM fit, profile short runs, and pick GPU configs under real budget constraints.
 Author-email: Alloc Labs <hello@alloclabs.com>
 License-Expression: Apache-2.0

{alloc-0.0.13 → alloc-0.0.14}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "alloc"
-version = "0.0.13"
+version = "0.0.14"
 description = "Engineer-first training calibration: estimate VRAM fit, profile short runs, and pick GPU configs under real budget constraints."
 readme = "README.md"
 license = "Apache-2.0"

{alloc-0.0.13 → alloc-0.0.14}/src/alloc/__init__.py

@@ -9,7 +9,7 @@ _warnings.filterwarnings("ignore", category=FutureWarning, module=r"torch\.cuda"
 _warnings.filterwarnings("ignore", category=DeprecationWarning, module=r"torch\.cuda")
 del _warnings
 
-__version__ = "0.0.13"
+__version__ = "0.0.14"
 
 from alloc.ghost import ghost, GhostReport
 from alloc.callbacks import AllocCallback as HuggingFaceCallback

{alloc-0.0.13 → alloc-0.0.14}/src/alloc/cli.py

@@ -564,8 +564,11 @@ def run(
             else:
                 console.print("[dim]Tip: alloc login --browser to connect your dashboard[/dim]")
 
+    # Propagate non-zero exit code — but NOT when calibrate mode
+    # intentionally killed the process (torchrun exits non-zero on SIGTERM)
     if result.exit_code and result.exit_code != 0:
-        raise typer.Exit(result.exit_code)
+        if result.stop_reason not in ("stable", "timeout"):
+            raise typer.Exit(result.exit_code)
 
 
 @app.command()
@@ -2115,6 +2118,13 @@ def scan(
     """Remote ghost scan via Alloc API — no GPU needed."""
     import httpx
 
+    # When --json, redirect console to stderr so nothing contaminates stdout.
+    from rich.console import Console as _Console
+    if json_output:
+        console = _Console(stderr=True)
+    else:
+        console = _Console()
+
     # Resolve param count from model name or explicit flag
     resolved_param_count = param_count_b or _model_to_params(model)
     if resolved_param_count is None:

{alloc-0.0.13 → alloc-0.0.14}/src/alloc/probe.py

@@ -699,8 +699,8 @@ def probe_command(
         num_gpus_detected=num_gpus_ref[0],
         process_map=process_map_ref[0],
         per_gpu_peak_vram_mb=(
-            [round(per_gpu_peaks_ref[0].get(i, 0), 1) for i in range(num_gpus_ref[0])]
-            if num_gpus_ref[0] >= 1 and per_gpu_peaks_ref[0] else None
+            [round(per_gpu_peaks_ref[0][i], 1) for i in sorted(per_gpu_peaks_ref[0])]
+            if per_gpu_peaks_ref[0] else None
         ),
         detected_interconnect=detected_ic_ref[0],
     )

{alloc-0.0.13 → alloc-0.0.14}/src/alloc.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: alloc
-Version: 0.0.13
+Version: 0.0.14
 Summary: Engineer-first training calibration: estimate VRAM fit, profile short runs, and pick GPU configs under real budget constraints.
 Author-email: Alloc Labs <hello@alloclabs.com>
 License-Expression: Apache-2.0