Circuit-Forge/buddymon
2
0
Fork 0
Code Issues 2 Pull requests Projects Releases 1 Packages Wiki Activity Actions

feat: expand MemoryLeech patterns + add CudaCrash bug_monster

Browse source 
MemoryLeech now catches: malloc failures, std::bad_alloc, Java OOM,
GC overhead limit, JavaScript heap OOM, OOMKilled, oom-killer,
macOS malloc region failures.

CudaCrash is a new uncommon bug_monster (strength 65, 130 XP) for
GPU/VRAM OOM: torch.cuda.OutOfMemoryError, CUDA error: out of memory,
cuDNN/CUBLAS allocation failures, device-side assert triggered.
This commit is contained in:
pyr0ball 2026-04-02 22:33:52 -07:00
parent 0c311b099b
commit 55747068e1
1 changed files with 37 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

38
lib/catalog.json
View file

@ -276,7 +276,15 @@
"heap.*exhausted", "heap.*exhausted",
"memory leak", "memory leak",
"Cannot allocate memory", "Cannot allocate memory",
"Killed.*memory" "Killed.*memory",
"malloc.*failed",
"std::bad_alloc",
"java\\.lang\\.OutOfMemoryError",
"GC overhead limit exceeded",
"JavaScript heap out of memory",
"OOMKilled",
"oom-killer",
"malloc: can't allocate region"
], ],
"weaken_actions": [ "weaken_actions": [
{"action": "write_failing_test", "strength_reduction": 20}, {"action": "write_failing_test", "strength_reduction": 20},
@ -284,6 +292,34 @@
{"action": "add_documenting_comment", "strength_reduction": 10} {"action": "add_documenting_comment", "strength_reduction": 10}
], ],
"flavor": "It was already there when you opened the task manager." "flavor": "It was already there when you opened the task manager."
},
"CudaCrash": {
"id": "CudaCrash",
"display": "⚡ CudaCrash",
"type": "bug_monster",
"rarity": "uncommon",
"base_strength": 65,
"xp_reward": 130,
"catchable": true,
"defeatable": true,
"description": "Lives in VRAM. Detonates the moment your batch size is one too many. Doesn't negotiate.",
"error_patterns": [
"CUDA out of memory",
"torch\\.cuda\\.OutOfMemoryError",
"CUDA error: out of memory",
"RuntimeError.*CUDA.*memory",
"cuDNN.*insufficient memory",
"CUBLAS_STATUS_ALLOC_FAILED",
"out of GPU memory",
"VRAM.*exhausted",
"device-side assert triggered"
],
"weaken_actions": [
{"action": "isolate_reproduction", "strength_reduction": 30},
{"action": "write_failing_test", "strength_reduction": 20},
{"action": "add_documenting_comment", "strength_reduction": 10}
],
"flavor": "Your model fit in VRAM yesterday. You added one layer."
} }
}, },