fix(studio/rocm): robustness pass - rocm tag normalisation, Strix routing parity, hardened detection

Robustness pass on top of 76137b2d. Four targeted fixes:

1. install.sh ROCm-tag routing normalisation.
   `rocm7.2.1` would route to https://download.pytorch.org/whl/rocm7.2.1
   which does not exist (PyTorch publishes major.minor URLs only). Same
   for any future patch-level tag. Normalise every rocm{maj.min}* pattern
   to the bare {maj.min} index URL.

2. install.ps1 + studio/setup.ps1 marketing-name fallback.
   The gfx1151 row matched 890M / Strix Halo / HX 37x / HX 38x / AI 9 HX
   but not the actual retail name 'AMD Radeon 8060S Graphics' shipped by
   OEMs (Ryzen AI MAX+ 395). Add '8060S' to the regex.

3. install_python_stack.py Strix + ROCm 7.1 routing parity with install.sh.
   The shell installer reroutes Strix Halo / Point + ROCm 7.1 to
   repo.amd.com/rocm/whl/{gfx}/ (which serves torch 2.11.0+rocm7.13.0
   with the upstream _grouped_mm fix). The Python `studio update` path
   only warned and still installed the broken generic rocm7.1 wheel.
   Mirror the override: detect gfx1151/gfx1150 on ROCm 7.1, route to
   the AMD per-gfx index, honour UNSLOTH_AMD_ROCM_MIRROR override.

4. _detect_windows_gfx_arch amd-smi parsing tightened.
   The amd-smi fallback added in the prior commit used a bare
   `\bgfx[1-9][0-9a-z]{2,3}\b` match against the lowercased stdout,
   which could pick up stray gfx references in warnings / device-name
   strings. Anchor on labelled lines first (Target_Graphics_Version,
   ASIC, Arch, gfx) and fall back to the bare match only when no
   labelled line is present.

Tests: 231 passed, 1 skipped in tests/studio/install/test_rocm_support.py;
sim_5301 23 cases pass (6 new sims for the Strix override + amd-smi parsing).

install.ps1

@@ -1307,7 +1307,7 @@ shell.Run cmd, 0, False
                 $nameArchTable = @(
                     @{ P = "9070 XT|9080";                                        A = "gfx1201" }  # RDNA 4
                     @{ P = "9070|9060";                                            A = "gfx1200" }  # RDNA 4
-                    @{ P = "890M|Strix Halo|HX 37[05]|HX 38[05]|AI 9 HX";        A = "gfx1151" }  # RDNA 3.5 iGPU (Strix Halo)
+                    @{ P = "8060S|890M|Strix Halo|HX 37[05]|HX 38[05]|AI 9 HX";  A = "gfx1151" }  # RDNA 3.5 iGPU (Strix Halo / Radeon 8060S retail)
                     @{ P = "880M|Strix Point|AI 9 36[05]|AI 7 35[05]|AI 5 34[05]"; A = "gfx1150" } # RDNA 3.5 iGPU (Strix Point)
                     @{ P = "RX 7900|RX 7800|RX 7700(?! S)";                       A = "gfx1100" }  # RDNA 3 desktop
                     @{ P = "RX 7600";                                              A = "gfx1102" }  # RDNA 3

install.sh

@@ -1644,9 +1644,17 @@ get_torch_index_url() {
                     echo "$_base/cpu"; return ;;
             esac
             # Supported tags; 6.5+ clips to rocm6.4, 7.3+ caps to rocm7.2.
+            # PyTorch publishes major.minor URLs only (no patch level), so
+            # rocm7.2.1 / rocm6.0.2 / etc. must normalise to rocm7.2 / rocm6.0.
             case "$_rocm_tag" in
-                rocm6.0|rocm6.0.*|rocm6.1|rocm6.1.*|rocm6.2|rocm6.2.*|rocm6.3|rocm6.3.*|rocm6.4|rocm6.4.*|rocm7.0|rocm7.0.*|rocm7.1|rocm7.1.*|rocm7.2|rocm7.2.*)
+                rocm6.0*) echo "$_base/rocm6.0" ;;
-                    echo "$_base/$_rocm_tag" ;;
+                rocm6.1*) echo "$_base/rocm6.1" ;;
+                rocm6.2*) echo "$_base/rocm6.2" ;;
+                rocm6.3*) echo "$_base/rocm6.3" ;;
+                rocm6.4*) echo "$_base/rocm6.4" ;;
+                rocm7.0*) echo "$_base/rocm7.0" ;;
+                rocm7.1*) echo "$_base/rocm7.1" ;;
+                rocm7.2*) echo "$_base/rocm7.2" ;;
                 rocm6.*)
                     # ROCm 6.5+ (no published PyTorch wheels): clip down
                     # to the last supported 6.x wheel set.

studio/install_python_stack.py

@@ -282,10 +282,21 @@ def _detect_windows_gfx_arch() -> str | None:
                 )
                 if result.returncode != 0:
                     continue
-                text = result.stdout.decode(errors = "replace").lower()
+                text = result.stdout.decode(errors = "replace")
-                m = re.search(r"\bgfx[1-9][0-9a-z]{2,3}\b", text)
+                # Anchor on a labelled gfx line (e.g. "TARGET_GRAPHICS_VERSION: gfx1151"
-                if m:
+                # or "Arch:  gfx1151") to avoid catching stray gfx mentions in
-                    return m.group(0)
+                # warnings or device-name strings. Fall back to a bare token match
+                # only if no labelled line is found.
+                m = re.search(
+                    r"(?im)^\s*(?:target_graphics_version|gfx|arch|asic)\b[^:\r\n]*:\s*(gfx[1-9][0-9a-z]{2,3})\b",
+                    text,
+                )
+                if not m:
+                    m = re.search(r"\bgfx[1-9][0-9a-z]{2,3}\b", text.lower())
+                    if m:
+                        return m.group(0)
+                    continue
+                return m.group(1).lower()
             except Exception:
                 continue
     return None
@@ -610,42 +621,44 @@ def _ensure_rocm_torch() -> None:
 
     rocm_torch_ready = has_hip_torch
 
-    # Strix Halo (gfx1151) segfaults under ROCm 7.1 due to a ROCm driver bug
+    # Strix Halo / Strix Point (gfx1151 / gfx1150) segfault under ROCm 7.1
-    # fixed in ROCm 7.2.  Warn early so users know why training may crash.
+    # in torch._grouped_mm. AMD's per-gfx repo ships torch 2.11.0+rocm7.13.0
+    # with the real fix, so route those hosts there instead of the generic
+    # pytorch.org rocm7.1 wheel. Mirrors install.sh's Strix override.
+    _strix_override_url: "str | None" = None
+    _strix_override_pkgs: "tuple[str, str, str] | None" = None
     if ver < (7, 2):
         gfx_codes = _detect_amd_gfx_codes()
         _strix_gfx = {"gfx1151", "gfx1150"}
-        if _strix_gfx.intersection(gfx_codes):
+        _detected_strix = _strix_gfx.intersection(gfx_codes)
-            _gfx_str = ", ".join(sorted(_strix_gfx.intersection(gfx_codes)))
+        if _detected_strix:
+            _gfx_str = ", ".join(sorted(_detected_strix))
+            _selected_gfx = sorted(_detected_strix)[0]
+            _amd_mirror = (
+                os.environ.get("UNSLOTH_AMD_ROCM_MIRROR")
+                or "https://repo.amd.com/rocm/whl"
+            ).rstrip("/")
+            _strix_override_url = f"{_amd_mirror}/{_selected_gfx}/"
+            _strix_override_pkgs = (
+                "torch>=2.11.0,<2.12.0",
+                "torchvision",
+                "torchaudio",
+            )
             print(
-                f"\n   ⚠️  {_gfx_str} (AMD Strix Halo) detected with ROCm {ver[0]}.{ver[1]}.\n"
+                f"\n   ⚠️  {_gfx_str} (AMD Strix) detected with ROCm {ver[0]}.{ver[1]}.\n"
-                f"   ROCm 7.1 has a known segfault on this GPU when tensors are\n"
+                f"   ROCm 7.1 has a known _grouped_mm segfault on this GPU;\n"
-                f"   moved to the GPU.  Upgrade to ROCm 7.2+ to enable training.\n"
+                f"   routing torch install to AMD's arch-specific index\n"
+                f"   ({_strix_override_url}) which serves torch 2.11.0+rocm7.13.0\n"
+                f"   with the upstream fix.\n"
             )
 
     if not has_hip_torch:
-        # Select best matching wheel tag (newest ROCm version <= installed)
+        if _strix_override_url is not None and _strix_override_pkgs is not None:
-        tag = next(
+            index_url = _strix_override_url
-            (
+            _torch_pkg, _vision_pkg, _audio_pkg = _strix_override_pkgs
-                t
+            print(f"   Strix ROCm 7.1 override -- installing torch from {index_url}")
-                for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
-                if ver >= (maj, mn)
-            ),
-            None,
-        )
-        if tag is None:
-            print(
-                f"   No PyTorch wheel for ROCm {ver[0]}.{ver[1]} -- "
-                f"skipping torch reinstall"
-            )
-        else:
-            index_url = f"{_PYTORCH_WHL_BASE}/{tag}"
-            print(f"   ROCm {ver[0]}.{ver[1]} -- installing torch from {index_url}")
-            _torch_pkg, _vision_pkg, _audio_pkg = _ROCM_TORCH_PKG_SPECS.get(
-                tag, _ROCM_TORCH_PKG_SPECS["_default"]
-            )
             pip_install(
-                f"ROCm torch ({tag})",
+                "ROCm torch (Strix arch-specific)",
                 "--force-reinstall",
                 "--no-cache-dir",
                 _torch_pkg,
@@ -656,6 +669,39 @@ def _ensure_rocm_torch() -> None:
                 constrain = False,
             )
             rocm_torch_ready = True
+        else:
+            # Select best matching wheel tag (newest ROCm version <= installed)
+            tag = next(
+                (
+                    t
+                    for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
+                    if ver >= (maj, mn)
+                ),
+                None,
+            )
+            if tag is None:
+                print(
+                    f"   No PyTorch wheel for ROCm {ver[0]}.{ver[1]} -- "
+                    f"skipping torch reinstall"
+                )
+            else:
+                index_url = f"{_PYTORCH_WHL_BASE}/{tag}"
+                print(f"   ROCm {ver[0]}.{ver[1]} -- installing torch from {index_url}")
+                _torch_pkg, _vision_pkg, _audio_pkg = _ROCM_TORCH_PKG_SPECS.get(
+                    tag, _ROCM_TORCH_PKG_SPECS["_default"]
+                )
+                pip_install(
+                    f"ROCm torch ({tag})",
+                    "--force-reinstall",
+                    "--no-cache-dir",
+                    _torch_pkg,
+                    _vision_pkg,
+                    _audio_pkg,
+                    "--index-url",
+                    index_url,
+                    constrain = False,
+                )
+                rocm_torch_ready = True
 
     # Install bitsandbytes only when torch links against ROCm. Prefers the
     # continuous-release_main wheel (bnb PR #1887 4-bit GEMV fix) and falls

studio/setup.ps1

@@ -792,7 +792,7 @@ if (-not $HasNvidiaSmi) {
             $nameArchTable = @(
                 @{ P = "9070 XT|9080";                                        A = "gfx1201" }  # RDNA 4
                 @{ P = "9070|9060";                                            A = "gfx1200" }  # RDNA 4
-                @{ P = "890M|Strix Halo|HX 37[05]|HX 38[05]|AI 9 HX";        A = "gfx1151" }  # RDNA 3.5 iGPU (Strix Halo)
+                @{ P = "8060S|890M|Strix Halo|HX 37[05]|HX 38[05]|AI 9 HX";  A = "gfx1151" }  # RDNA 3.5 iGPU (Strix Halo / Radeon 8060S retail)
                 @{ P = "880M|Strix Point|AI 9 36[05]|AI 7 35[05]|AI 5 34[05]"; A = "gfx1150" } # RDNA 3.5 iGPU (Strix Point)
                 @{ P = "RX 7900|RX 7800|RX 7700(?! S)";                       A = "gfx1100" }  # RDNA 3 desktop
                 @{ P = "RX 7600";                                              A = "gfx1102" }  # RDNA 3