Repository Issue Activity (beta)

#8097 Missing input validation could cause unexpected behavior with edge case inputs

#8084 zero.Init silently does not shard (world_size=1) when the process group is uninitialized before from_pretrained -> per-rank full load -> OOM

#8102 [BUG] AutoEP breaks Python 3.9 compatibility -- `TypeError: unsupported operand type(s) for |: 'type' and 'NoneType'`

#7483 [BUG] FlopsProfiler will hit error when sequence parallel enabled

#8123 [Proposal] HeartFlow - Cognitive Engine for DeepSpeed

#8085 zero.Init partitioning of large fused MoE-expert tensors spikes a single GPU (transient full materialization) -> OOM during load even when the sharded model fits

#8074 Add type hints to public API functions

#8101 Consider adding a changelog to track version history

#7961 [BUG][ZeRO-3 / BF16] No public API to reassemble optimizer moments (exp_avg, exp_avg_sq) from sharded checkpoint files post-hoc

#7983 Incorrect variable name: `self.ev_values`

#8104 [Draft] DeepSpeed - Q3 Roadmap

#7942 [BUG][DeepCompile] DeepCompile fails on Qwen1.5-MoE-A2.7B-Chat with `RuntimeError: 'weight' must be 2-D` (LLaMA works in the same environment)

#8092 [REQUEST] Muon → Gefen-Muon

#8098 SuperCompress: Free open-source LLM prompt compression - cut token costs by ~65%

#7918 Fork safety

#7807 [BUG] Cross-partition parameters incorrectly updated when using ZeRO-1/ZeRO-2 with reduce_scatter=true and Muon optimizer

#8093 Exception type too broad in error handling

#4555 How can I extract and log grad norm for individual layers

#8083 Security Advisory: Self-Hosted Runner Risk (Skor: 45/100)

#8072 [BUG] ZeRO-3 + PEFT (LoRA) regression in 0.19.2: TypeError: output tensor must have the same type as input tensor in _allgather_params_coalesced

#8076 [BUG]ds_z3_config error

#8061 [BUG] ZeRO stage 1/2 overlap_comm only waits current stream, but contiguous gradient bucket copies may come from multiple streams

#7964 [REQUEST] Add Trackio as a New Backend for Experiment Monitoring

#7031 Fix - Update DeepSpeed to be PEP517 compliant, update to `pyproject.toml`

#7686 [BUG] Zenflow_stage3 - RuntimeError: narrow() cannot be applied to a 0-dim tensor.