Search results

AttributeError: 'MegatronHFTokenizer' object has no attribute 'additional_special_tokens' — all T5 / BertWordPieceCase nightlies broken

bugSomething isn't workingSomething isn't working

Status: Open.

#4824 In NVIDIA/Megatron-LM;

· balasaajay opened on May 15, 2026

[ROADMAP][2026 Q2] Megatron Core MoE Roadmap

Status: Open.

#4815 In NVIDIA/Megatron-LM;

· Victarry opened on May 15, 2026

Muon QKV split fails for gated-attention QKV projections

bugSomething isn't workingSomething isn't working

community-request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#4731 In NVIDIA/Megatron-LM;

· Moozy23232 opened on May 11, 2026

🐛 CI failure: gpt_grpo_basic_function — BatchEncoding can't be msgpack-serialized after transformers 5.x bump

bugSomething isn't workingSomething isn't working

Status: Open.

#4726 In NVIDIA/Megatron-LM;

· ko3n1g opened on May 11, 2026

🐛 CI failure: gpt_static_inference_tp1_pp1_583m_fp8_cudagraphs — TE 2.15.0 removed FP8GlobalStateManager.set_skip_fp8_weight_update_tensor

bugSomething isn't workingSomething isn't working

Status: Open.

#4707 In NVIDIA/Megatron-LM;

· balasaajay opened on May 8, 2026

Memory leak when log_max_attention_logits is set

bugSomething isn't workingSomething isn't working

community-request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#4699 In NVIDIA/Megatron-LM;

· supershadoe opened on May 8, 2026

Muon CPU offload

community-request

enhancementNew feature or requestNew feature or request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#4691 In NVIDIA/Megatron-LM;

· pengdurice opened on May 8, 2026

Torch FSDP2 crashes in finalize_model_grads because _BaseDataParallel.finish_grad_sync() does not accept force_all_reduce

bugSomething isn't workingSomething isn't working

community-request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#4669 In NVIDIA/Megatron-LM;

· Jayoprell opened on May 7, 2026

[QUESTION]Has MCore v0.17 released?

community-request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#4665 In NVIDIA/Megatron-LM;

· suenphey opened on May 7, 2026

mcore scalar all_reduce fails NCCL FastTrak buffer registration

bugSomething isn't workingSomething isn't working

community-request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#4660 In NVIDIA/Megatron-LM;

· cheonghm opened on May 6, 2026

🐛 CI hang/timeout: hybrid_dynamic_inference_tp1_ep8_nanov3_chunked_prefill exceeds 1h job limit on repeat iteration 3

bugSomething isn't workingSomething isn't working

Status: Open.

#4657 In NVIDIA/Megatron-LM;

· balasaajay opened on May 6, 2026

🐛 CI failure: tests/functional_tests/python_test_utils/test_pretraining_regular_pipeline.py::test_regular_pipeline [mimo_vlm_pretrain_convergence_tp1_pp1_cp1_dp8_seq_packing]

bugSomething isn't workingSomething isn't working

Status: Open.

#4654 In NVIDIA/Megatron-LM;

· balasaajay opened on May 6, 2026