gpu-dev 0.7.1__tar.gz → 0.7.2__tar.gz

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/CLAUDE.md

@@ -318,7 +318,7 @@ module "us_east_1" {
 
 - **Merge multi-region into single tf state** - HIGH PRIORITY. Kill prod-east1 workspace, refactor into module-per-region in one state. See research notes below. Enables: one `tf apply`, shared AMI (aws_ami_copy), shared Docker (ECR replication already set up), no double builds. Prerequisite for adding west regions.
 - **Add us-west-1 and us-west-2 spot regions** - BLOCKED on single-state refactor. After refactor, adding a region = adding one module block.
-- **Spot UX improvements** - Queue position should be #1 for each type (not cross-type FIFO). Status should show "queued (waiting for capacity)" not just "queued". Interactive picker should show spot GPU counts from east1 not prod.
+- **Spot UX improvements** - Queue position should be #1 for each type (not cross-type FIFO). Status should show "queued (waiting for capacity)" not just "queued". Interactive picker should show spot GPU counts from east1 not prod. NOTE (2026-05-30): spot is now **hidden by default** in `gpu-dev reserve` (interactive picker), `gpu-dev avail`, and watch mode — `cpu-spot` + the us-east-1 spot cluster only appear with `--spot` (reserve/avail flag) or the "⚡ Show spot options" picker entry. Spot was too bloated/half-baked for the default view. CLI-only change (`cli.py` `_show_availability`/`_show_availability_watch`/`avail`/`reserve`, `interactive.py` `select_gpu_type_interactive`).
 - **FQDN for devservers** - Set up proper domain names for development server access
 - **Automated SSH config per reservation** - ✅ DONE - Each reservation now gets `~/.devgpu/<reservation_id>-sshconfig` file, use with `ssh -F ~/.devgpu/<reservation_id>-sshconfig <pod_name>`
 - **Custom Docker image scaffold** - Create Dockerfile with pre-installed packages (Jupyter, etc.)

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gpu-dev
-Version: 0.7.1
+Version: 0.7.2
 Summary: CLI + Python SDK for PyTorch GPU developer server reservations
 Author: PyTorch Team
 Requires-Python: >=3.10

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/cli-tools/gpu-dev-cli/gpu_dev_cli/cli.py

@@ -829,9 +829,9 @@ def reserve(
                 rprint("[red]❌ Could not get GPU availability information[/red]")
                 return
 
-            # Interactive GPU type selection
+            # Interactive GPU type selection (spot hidden unless --spot)
             if gpu_type is None:
-                gpu_type = select_gpu_type_interactive(availability_info)
+                gpu_type = select_gpu_type_interactive(availability_info, show_spot=spot)
                 if gpu_type is None:
                     rprint("[yellow]Reservation cancelled.[/yellow]")
                     return
@@ -3163,8 +3163,11 @@ def _format_gpu_display(gpu_count, gpu_type):
     return f"{gpu_count}x {str(gpu_type).upper()}"
 
 
-def _show_availability() -> None:
-    """Shared function to show GPU availability"""
+def _show_availability(show_spot: bool = False) -> None:
+    """Shared function to show GPU availability.
+
+    Spot SKUs (cpu-spot + the us-east-1 spot cluster) are hidden unless show_spot.
+    """
     try:
         with Live(
             Spinner("dots", text="📡 Checking GPU availability..."), console=console
@@ -3181,7 +3184,7 @@ def _show_availability() -> None:
                 _east1_spot_types = frozenset(Config.ENVIRONMENTS.get("prod-east1", {}).get("spot_types", []))
 
                 def _fetch_east1_spot():
-                    if _env_name != "prod" or not _east1_spot_types:
+                    if not show_spot or _env_name != "prod" or not _east1_spot_types:
                         return {}
                     east1_r = Config.ENVIRONMENTS["prod-east1"]["region"]
                     east1_table = config.session.resource("dynamodb", region_name=east1_r).Table("pytorch-gpu-dev-gpu-availability")
@@ -3247,8 +3250,16 @@ def _show_availability() -> None:
                 "CPU (arm64)": 6,
             }
 
-            # Split into categories
-            full_types = {k: v for k, v in availability_info.items() if "mig" not in k}
+            # Split into categories. Hide spot SKUs (e.g. cpu-spot) unless --spot,
+            # but never hide everything if the env is spot-only.
+            def _is_spot(k):
+                return k == "cpu-spot" or k.endswith("-spot")
+            _non_spot_exists = any(not _is_spot(k) for k in availability_info if "mig" not in k)
+            _hide_spot = (not show_spot) and _non_spot_exists
+            full_types = {
+                k: v for k, v in availability_info.items()
+                if "mig" not in k and not (_hide_spot and _is_spot(k))
+            }
             mig_types = {k: v for k, v in availability_info.items() if "mig" in k}
 
             def _sort_by_arch(items):
@@ -3344,8 +3355,12 @@ def _show_availability() -> None:
             rprint("  [green]●[/green]: 1+ full node available - [yellow]●[/yellow]: GPUs available, but no full node - [red]●[/red]: No GPUs available")
 
             # Show usage tip
+            if _hide_spot:
+                rprint(
+                    "\n[dim]💡 Spot instances hidden — pass '--spot' to show (us-east-1, ~70% cheaper, may be preempted)[/dim]"
+                )
             rprint(
-                "\n[dim]💡 Use 'gpu-dev reserve' (interactive) to see all options including MIG slices and spot instances[/dim]"
+                "\n[dim]💡 Use 'gpu-dev reserve' (interactive) to see all options including MIG slices[/dim]"
             )
 
         else:
@@ -3355,10 +3370,13 @@ def _show_availability() -> None:
         rprint(f"[red]❌ Error: {str(e)}[/red]")
 
 
-def _show_availability_watch(interval: int) -> None:
+def _show_availability_watch(interval: int, show_spot: bool = False) -> None:
     _env_name = load_config().user_config.get("environment", "prod")
     _spot_types = frozenset(Config.ENVIRONMENTS.get(_env_name, {}).get("spot_types", []))
 
+    def _is_spot(k):
+        return k == "cpu-spot" or k.endswith("-spot")
+
     """Watch mode for GPU availability with auto-refresh"""
     import time
     from datetime import datetime
@@ -3385,6 +3403,13 @@ def _show_availability_watch(interval: int) -> None:
                     # Get availability data
                     availability_info = reservation_mgr.get_gpu_availability_by_type()
 
+                    # Hide spot SKUs (e.g. cpu-spot) unless --spot, never hide everything.
+                    if availability_info and not show_spot:
+                        if any(not _is_spot(k) for k in availability_info if "mig" not in k):
+                            availability_info = {
+                                k: v for k, v in availability_info.items() if not _is_spot(k)
+                            }
+
                     if availability_info:
                         # GPU architecture mapping (for display)
                         gpu_architectures = {
@@ -4024,8 +4049,14 @@ def help(ctx: click.Context) -> None:
     default=5,
     help="Refresh interval in seconds for watch mode (default: 5)",
 )
+@click.option(
+    "--spot",
+    is_flag=True,
+    default=False,
+    help="Also show spot instances (us-east-1, ~70% cheaper, may be preempted). Hidden by default.",
+)
 @click.pass_context
-def avail(ctx: click.Context, watch: bool, interval: int) -> None:
+def avail(ctx: click.Context, watch: bool, interval: int, spot: bool) -> None:
     """Show GPU availability by type and queue estimates
 
     Displays real-time information about GPU availability for each GPU type.
@@ -4045,9 +4076,9 @@ def avail(ctx: click.Context, watch: bool, interval: int) -> None:
     This helps you choose the right GPU type and understand wait times before reserving.
     """
     if watch:
-        _show_availability_watch(interval)
+        _show_availability_watch(interval, show_spot=spot)
     else:
-        _show_availability()
+        _show_availability(show_spot=spot)
 
 
 @main.command()

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/cli-tools/gpu-dev-cli/gpu_dev_cli/interactive.py

@@ -50,11 +50,22 @@ def check_interactive_support() -> bool:
     return True
 
 
+def _is_spot_type(gt: str) -> bool:
+    """Spot SKUs hidden from default views: the cpu-spot type + any `*-spot` type."""
+    return gt == "cpu-spot" or gt.endswith("-spot")
+
+
 def select_gpu_type_interactive(
     availability_info: Dict[str, Dict[str, Any]],
     _refresh: bool = False,
+    show_spot: bool = False,
 ) -> Optional[str]:
-    """Interactive GPU type selection with availability table"""
+    """Interactive GPU type selection with availability table.
+
+    Spot SKUs (cpu-spot + the cross-region us-east-1 spot cluster) are hidden by
+    default — pass show_spot=True (CLI `--spot`) or pick the "Show spot options"
+    entry to reveal them.
+    """
     if not check_interactive_support():
         return None
 
@@ -65,11 +76,17 @@ def select_gpu_type_interactive(
         _mgr = ReservationManager(_cfg)
         availability_info = _mgr.get_gpu_availability_by_type() or availability_info
 
+    # Don't hide spot when the whole environment is spot-only (nothing left to show).
+    _non_spot_exists = any(
+        not _is_spot_type(gt) for gt in availability_info if "-mig-" not in gt
+    )
+    _hide_spot = (not show_spot) and _non_spot_exists
+
     # Hide MIG slice SKUs from the top-level selector — reached via the h100 submenu.
     # Direct `--gpu-type h100-mig-1g` still works for non-interactive scripts.
     visible_info = {
         gt: info for gt, info in availability_info.items()
-        if "-mig-" not in gt
+        if "-mig-" not in gt and not (_hide_spot and _is_spot_type(gt))
     }
 
     # Aggregate MIG slice availability per parent type, hinted on the h100/b200 rows.
@@ -86,12 +103,6 @@ def select_gpu_type_interactive(
         )
         return avail, cap
 
-    h100_mig_avail, h100_mig_capacity = _mig_aggregates("h100")
-    b200_mig_avail, b200_mig_capacity = _mig_aggregates("b200")
-    # Backwards-compat aliases for the existing h100 row code below.
-    mig_total_available = h100_mig_avail
-    mig_total_capacity = h100_mig_capacity
-
     # Detect spot types and fetch cross-region spot availability
     from .config import Config, load_config
     _cfg = load_config()
@@ -102,9 +113,10 @@ def select_gpu_type_interactive(
     has_spot_types = len(_spot_types) > 0
 
     # Cross-region: if we're on prod, also fetch prod-east1 spot availability
+    # (skipped entirely when spot is hidden — saves a DynamoDB scan).
     spot_region_info = {}
     spot_region_name = None
-    if _env_name == "prod":
+    if _env_name == "prod" and not _hide_spot:
         east1_env = Config.ENVIRONMENTS.get("prod-east1", {})
         if east1_env:
             spot_region_name = "prod-east1"
@@ -159,7 +171,19 @@ def select_gpu_type_interactive(
             return f"[red]MAINTENANCE[/red]"
         return f"[green]{available}[/green]" if available > 0 else f"[red]{available}[/red]"
 
-    def _build_table(title, items, is_spot=False):
+    def _mig_breakdown(parent):
+        """Compact per-slice availability for a parent, e.g. (['12×1G','4×2G'], 16, 32)."""
+        parts, tot_a, tot_c = [], 0, 0
+        for cgt, ci in sorted((availability_info or {}).items()):
+            if not cgt.startswith(f"{parent}-mig-"):
+                continue
+            a, c = int(ci.get("available", 0)), int(ci.get("total", 0))
+            tot_a += a
+            tot_c += c
+            parts.append(f"{a}×{cgt.rsplit('-', 1)[-1].upper()}")
+        return parts, tot_a, tot_c
+
+    def _build_table(title, items, is_spot=False, with_mig=False):
         console.print(f"\n[cyan]{title}[/cyan]")
         table = Table()
         table.add_column("GPU Type", style="cyan")
@@ -182,16 +206,18 @@ def select_gpu_type_interactive(
                 str(info.get("total", 0)),
                 wait_display,
             )
+            # Compact MIG slice sub-row beneath the parent (h100/b200 today).
+            if with_mig:
+                parts, mig_a, mig_c = _mig_breakdown(gt)
+                if parts:
+                    av = f"[green]{mig_a}[/green]" if mig_a > 0 else f"[red]{mig_a}[/red]"
+                    table.add_row(
+                        f"[dim] └─ MIG {' '.join(parts)}[/dim]",
+                        av, "-", str(mig_c), "[dim]pick parent ↑[/dim]")
         console.print(table)
 
-    # Section 1: Full GPUs & CPUs
-    _build_table("━━━ Full GPUs & CPUs ━━━", full_gpus)
-
-    # Section 2: MIG Slices
-    if mig_gpus:
-        console.print("[dim]  Sliced GPUs — isolated fractions of a physical GPU. Perfect for smaller jobs[/dim]")
-        console.print("[dim]  that don\'t need full performance or VRAM.[/dim]")
-        _build_table("━━━ 🔬 MIG Slices ━━━", mig_gpus)
+    # Section 1: Full GPUs & CPUs (MIG slices shown as a compact sub-row per parent)
+    _build_table("━━━ Full GPUs & CPUs ━━━", full_gpus, with_mig=True)
 
     # Section 3: Spot Instances (cross-region) — custom table with per-node + price
     if spot_gpus:
@@ -249,10 +275,10 @@ def select_gpu_type_interactive(
             label = f"{status_indicator} {gt.upper()} ({avail}/{total} available)"
             if ql > 0:
                 label += f" - {ql} in queue"
-            if gt == "h100" and mig_total_capacity > 0:
-                label += f" — also {mig_total_available}/{mig_total_capacity} MIG slices"
-            elif gt == "b200" and b200_mig_capacity > 0:
-                label += f" — also {b200_mig_avail}/{b200_mig_capacity} MIG slices"
+            # Any parent with MIG children gets a slice hint (h100/b200 today).
+            mig_avail, mig_cap = _mig_aggregates(gt)
+            if mig_cap > 0:
+                label += f" — also {mig_avail}/{mig_cap} MIG slices"
             choices.append(questionary.Choice(title=label, value=gt))
 
     if mig_gpus:
@@ -296,11 +322,15 @@ def select_gpu_type_interactive(
             choices.append(questionary.Choice(title=label, value=f"spot:{gt}"))
 
     choices.append(questionary.Separator("───"))
+    if _hide_spot:
+        choices.append(questionary.Choice(
+            title="⚡ Show spot options (us-east-1, ~70% cheaper, may be preempted)",
+            value="_show_spot"))
     choices.append(questionary.Choice(title="🔄 Refresh availability", value="_refresh"))
 
     console.print()
 
-    # Interactive selection — loop on refresh
+    # Interactive selection — loop on refresh / spot toggle
     while True:
         try:
             answer = questionary.select(
@@ -309,7 +339,10 @@ def select_gpu_type_interactive(
 
             if answer == "_refresh":
                 console.print("[dim]Refreshing...[/dim]")
-                return select_gpu_type_interactive(availability_info, _refresh=True)
+                return select_gpu_type_interactive(
+                    availability_info, _refresh=True, show_spot=show_spot)
+            if answer == "_show_spot":
+                return select_gpu_type_interactive(availability_info, show_spot=True)
             return answer
         except (KeyboardInterrupt, EOFError):
             console.print("\n[yellow]Selection cancelled.[/yellow]")

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/gpu_dev.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gpu-dev
-Version: 0.7.1
+Version: 0.7.2
 Summary: CLI + Python SDK for PyTorch GPU developer server reservations
 Author: PyTorch Team
 Requires-Python: >=3.10

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gpu-dev"
-version = "0.7.1"
+version = "0.7.2"
 description = "CLI + Python SDK for PyTorch GPU developer server reservations"
 authors = [{name = "PyTorch Team"}]
 readme = "cli-tools/gpu-dev-cli/README.md"

{gpu_dev-0.7.1 → gpu_dev-0.7.2}/sdk/python/src/gpu_dev/__init__.py

@@ -63,4 +63,4 @@ try:
 
     __version__ = _pkg_version("gpu-dev")
 except Exception:
-    __version__ = "0.7.1"
+    __version__ = "0.7.2"