Skip to content

DeepSpeed 0.14.x Regression: 'Expected all tensors to be on the same device' (ZeRO-3 / Adam Offload)

After upgrading to DeepSpeed >0.14.0 (e.g. 0.14.2), ZeRO-3 / optimizer-offload training crashes with 'Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!'. A regression in the 0.14.x ZeRO-3 path left some tensors on CPU; upgrading past the fix (or downgrading) resolves it.

Quick answer

After upgrading to DeepSpeed >0.

Distributed Training#deepspeed#zero-3#optimizer-offload#device-mismatch#regression#distributed

What this failure is

DeepSpeed 0.14.x Regression: 'Expected all tensors to be on the same device' (ZeRO-3 / Adam Offload) is a Distributed Training failure seen during ML training runs. After upgrading to DeepSpeed >0.14.0 (e.g. 0.14.2), ZeRO-3 / optimizer-offload training crashes with 'Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!'. A regression in the 0.14.x ZeRO-3 path left some tensors on CPU; upgrading past the fix (or downgrading) resolves it. Common tags: Deepspeed, Zero 3, Optimizer Offload, Device Mismatch.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about DeepSpeed 0.14.x Regression: 'Expected all tensors to be on the same device' (ZeRO-3 / Adam Offload). Paste your own crash log or traceback below and get the real root cause for YOUR run, not this generic entry. No account, no card. Logs are masked at ingress and never saved to account history.

training_logs.txt
No log to hand? Try one:

3 free diagnoses/day

Want 14 days on the Scale plan?

Request an evaluation code. A verified workplace organization activates up to 50 diagnoses a day, alerts, history, and follow-up questions. No credit card or automatic subscription.

Evaluate one incident

Why it happens (the mechanism)

A regression introduced in DeepSpeed 0.14.x ZeRO-3/optimizer-offload code (a single line in stage3.py) left some optimizer/parameter tensors on CPU while siblings moved to CUDA. The mismatched devices then collide during an elementwise/optimizer op; the line was later reverted/fixed on master. Taken together, these mechanisms explain why the failure is reproducible, why it tends to surface on specific workloads or scales, and why generic mitigation attempts often fall short without addressing the underlying cause.

What you'll observe

  • Training that worked on 0.13.x crashes after upgrading DeepSpeed to 0.14.x
  • Device-mismatch RuntimeError during the optimizer step or ZeRO-3 resume
  • Adam offload / ZeRO-3 fp16/bf16 affected

Common symptoms and what they mean

SymptomWhy it happens
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!A regression introduced in DeepSpeed 0.14.x ZeRO-3/optimizer-offload code (a single line in stage3.py) left some optimizer/parameter tensors on CPU while siblings moved to CUDA
Traceback through deepspeed/runtime/zero/stage3.pyThe mismatched devices then collide during an elementwise/optimizer op; the line was later reverted/fixed on master
Only on DeepSpeed >0.14.0A regression introduced in DeepSpeed 0.14.x ZeRO-3/optimizer-offload code (a single line in stage3.py) left some optimizer/parameter tensors on CPU while siblings moved to CUDA

Which systems are affected

  • DeepSpeed 0.14.x with ZeRO-3 and/or optimizer offload
  • HF Trainer resume tests with ds_optim/ds_scheduler
  • Mixed CPU-offload + CUDA optimizer state

How to confirm this is the problem

Use this checklist to test the hypothesis against a small reproduction. No single line proves the root cause, so preserve the preceding events and compare one variable at a time.

  • Reproduce the failure from a clean checkpoint/seed: the symptom must appear without warm-up state from a previous run.
  • Verified signal present: RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
  • Verified signal present: Traceback through deepspeed/runtime/zero/stage3.py
  • Verified signal present: Only on DeepSpeed >0.14.0
  • A targeted fix from the "How to fix it" section eliminates or substantially reduces the symptom within one validation pass.

The fix and the prevention pattern

The root cause is on this page and stays free. A free account adds the exact remediation steps, saved history, and the fix on every entry in the encyclopedia.

Sign up free. Unlock the full analysis

No credit card. Daily allowance follows verified trust tier. Instant access.

Diagnose this failure in VS Code

Select the traceback or open the failed terminal, then run Denpex locally to see the initiating rank, collateral failures, exact fix, and verification command without uploading the log.

Install the free VS Code extension

DeepSpeed errors in context

DeepSpeed changes when parameters, gradients and optimizer state are created, partitioned, gathered and offloaded. The hub separates ZeRO, memory, checkpoint and pipeline failures by lifecycle phase.

Compare every deepspeed error side by side

Root cause

  • A regression introduced in DeepSpeed 0.14.x ZeRO-3/optimizer-offload code (a single line in stage3.py) left some optimizer/parameter tensors on CPU while siblings moved to CUDA
  • The mismatched devices then collide during an elementwise/optimizer op; the line was later reverted/fixed on master

The fix and how to prevent it

Evaluate Denpex on your own logs

Request a Scale evaluation code. A verified workplace organization activates 14 days with up to 50 diagnoses a day. Every account keeps its current diagnosis allowance and gets a verification path. No card or automatic subscription.

We send a single-use code tied to that address. Static provider and TLD rules do not reject valid addresses. Account trust determines the benefit after signup.

Don't just read the fix, diagnose your run

The encyclopedia tells you what went wrong. Denpex tells you what went wrong in YOUR training run. With your logs, your config, and your stack.