-
Notifications
You must be signed in to change notification settings - Fork 603
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
JSD kernel returns NaN for interior beta when a vocabulary entry underflows under both distributions
Status: Open.#1453 In linkedin/Liger-Kernel;- Status: Open.#1432 In linkedin/Liger-Kernel;
- Status: Open.#1393 In linkedin/Liger-Kernel;
- Status: Open.#1383 In linkedin/Liger-Kernel;
- Status: Open.#1378 In linkedin/Liger-Kernel;
- Status: Open.#1375 In linkedin/Liger-Kernel;
- Status: Open.#1373 In linkedin/Liger-Kernel;
- Status: Open.#1347 In linkedin/Liger-Kernel;
- Status: Open.#1335 In linkedin/Liger-Kernel;
[B300/SM103] Compiled Liger SwiGLU is ~3.1x slower than torch.compile for a Qwen3-shaped BF16 tensor
Status: Open.#1330 In linkedin/Liger-Kernel;- Status: Open.#1327 In linkedin/Liger-Kernel;
- Status: Open.#1323 In linkedin/Liger-Kernel;