-
Notifications
You must be signed in to change notification settings - Fork 4.5k
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
MiMo LLM-only Nemotron provider rejects absent RADIO bridge metadata
bugSomething isn't workingSomething isn't workingStatus: Open.#7540 In NVIDIA/Megatron-LM;- Status: Open.#7524 In NVIDIA/Megatron-LM;
Avoid re-arming unrelated warnings before distributed initialization
waiting-on-customerWaiting on the original author to respondWaiting on the original author to respondStatus: Open.#7515 In NVIDIA/Megatron-LM;[Inference] Expert tensor parallelism is rejected for inference-optimized MoE and has no regression coverage
waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#7484 In NVIDIA/Megatron-LM;current_max_attn_logitsbreaks CUDA Graph replay because its Python-level assignment is not re-executedbugSomething isn't workingSomething isn't workingwaiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#7475 In NVIDIA/Megatron-LM;[Bug] Vocab-parallel label smoothing uses the local vocabulary size and the local mean log-probability
bugSomething isn't workingSomething isn't workingwaiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#7464 In NVIDIA/Megatron-LM;[Bug] 256K DSA training with Megatron fails while 32K succeeds (dsa_kernel_backend="cudnn", experimental_attention_variant="dsa")
bugSomething isn't workingSomething isn't workingwaiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#7463 In NVIDIA/Megatron-LM;- Status: Open.#7460 In NVIDIA/Megatron-LM;
Silent numerical-correctness bugs in Megatron Core parallelism composition
bugSomething isn't workingSomething isn't workingStatus: Open.#7452 In NVIDIA/Megatron-LM;[BUG] Strict loading rejects checkpoints with idle rerun state
waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#7428 In NVIDIA/Megatron-LM;[BUG] MFSDP v2 with HybridEP: intermittent bad gradients in eager training
bugSomething isn't workingSomething isn't workingStatus: Open.[Bug] Fine-grained activation offloading: the inline reload in
tensor_popraces its own D2H, corrupting the warmup iterationbugSomething isn't workingSomething isn't workingwaiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#7398 In NVIDIA/Megatron-LM;