fix: install unsloth to /app/pkgs (writable), detect torch version for git extra

training/space_runner.py

@@ -43,25 +43,33 @@ def _run_training():
     # Step 1 — install unsloth on the live GPU
     _training_status = "installing"
     _append_log("Installing unsloth on GPU hardware...")
+
+    import torch
+    tv = torch.__version__.split("+")[0].split(".")   # ['2', '5', '1']
+    torch_tag = f"torch{''.join(tv[:2])}{tv[2] if tv[2] != '0' else ''}"  # torch251
+    _append_log(f"PyTorch {torch.__version__} — will try extras: cu121-ampere-{torch_tag}")
+
+    # Install to /app/pkgs — always writable regardless of HF user UID
+    _PKGS = "/app/pkgs"
+    os.makedirs(_PKGS, exist_ok=True)
+    _pip_base = [sys.executable, "-m", "pip", "install", "--no-cache-dir", "--target", _PKGS]
+
     install_cmds = [
-        # Try PyPI release first (picks up correct CUDA extras automatically)
-        [sys.executable, "-m", "pip", "install", "--upgrade", "--no-cache-dir", "unsloth"],
-        # Fall back to git HEAD with cu121 extra
-        [sys.executable, "-m", "pip", "install", "--upgrade", "--no-cache-dir",
-         "unsloth[cu121-ampere-torch240] @ git+https://github.com/unslothai/unsloth.git"],
+        _pip_base + ["unsloth"],
+        _pip_base + [f"unsloth[cu121-ampere-{torch_tag}] @ git+https://github.com/unslothai/unsloth.git"],
+        _pip_base + ["git+https://github.com/unslothai/unsloth.git"],
     ]
     installed = False
     for cmd in install_cmds:
-        try:
-            result = subprocess.run(cmd, capture_output=True, text=True)
-            if result.returncode == 0:
-                _append_log("unsloth installed successfully.")
-                installed = True
-                break
-            else:
-                _append_log(f"Install attempt failed: {result.stderr[-500:] if result.stderr else 'no output'}")
-        except Exception as e:
-            _append_log(f"Install attempt error: {e}")
+        result = subprocess.run(cmd, capture_output=True, text=True)
+        if result.returncode == 0:
+            if _PKGS not in sys.path:
+                sys.path.insert(0, _PKGS)
+            _append_log("unsloth installed successfully.")
+            installed = True
+            break
+        err = (result.stderr or result.stdout or "")[-600:].strip()
+        _append_log(f"Install attempt failed:\n{err}")
     if not installed:
         _training_status = "failed"
         _append_log("❌ All unsloth install attempts failed. Check logs above.")