Skip to content

FSDP2 Unwrapped Model Still Has DTensor Weights, save_pretrained Fails

After unwrapping an FSDP2 (fully_shard) model the parameters are still DTensors, so save_pretrained or a plain state_dict produces invalid storage or sharded tensors. You must gather a full, unsharded state dict before saving and write only on rank 0.

Quick answer

After unwrapping an FSDP2 (fully_shard) model the parameters are still DTensors, so save_pretrained or a plain state_dict produces invalid storage or sharded tensors.

Distributed Training#accelerate#fsdp2#dtensor#save_pretrained#state-dict#distributed

What this failure is

FSDP2 Unwrapped Model Still Has DTensor Weights. Save_pretrained Fails is a Distributed Training failure seen during ML training runs. After unwrapping an FSDP2 (fully_shard) model the parameters are still DTensors, so save_pretrained or a plain state_dict produces invalid storage or sharded tensors. You must gather a full, unsharded state dict before saving and write only on rank 0. Common tags: Accelerate, Fsdp2, Dtensor, Save_pretrained.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about FSDP2 Unwrapped Model Still Has DTensor Weights, save_pretrained Fails. 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)

FSDP2 keeps parameters as sharded DTensors; simply unwrapping the module does not materialize a full unsharded tensor. Save_pretrained expects plain torch.Tensors, so it chokes on DTensors or saves only the local shard. 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

  • Unwrapped FSDP2 model parameters are DTensors, not plain tensors
  • save_pretrained / get_state_dict raises invalid storage errors
  • Saved checkpoint is sharded or missing weights

Common symptoms and what they mean

SymptomWhy it happens
model weights are torch.distributed.tensor.DTensor after unwrapFSDP2 keeps parameters as sharded DTensors; simply unwrapping the module does not materialize a full unsharded tensor
Invalid storage error when calling get_state_dictsave_pretrained expects plain torch.Tensors, so it chokes on DTensors or saves only the local shard
Checkpoint cannot be reloaded with from_pretrainedFSDP2 keeps parameters as sharded DTensors; simply unwrapping the module does not materialize a full unsharded tensor

Which systems are affected

  • Accelerate FSDP2 (fully_shard) training
  • PyTorch DTensor-based sharding
  • save_pretrained on a sharded model

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: model weights are torch.distributed.tensor.DTensor after unwrap
  • Verified signal present: Invalid storage error when calling get_state_dict
  • Verified signal present: Checkpoint cannot be reloaded with from_pretrained
  • 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

Root cause

  • FSDP2 keeps parameters as sharded DTensors; simply unwrapping the module does not materialize a full unsharded tensor
  • save_pretrained expects plain torch.Tensors, so it chokes on DTensors or saves only the local shard

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.