Skip to content

FSDP Missing Keys / Unexpected Keys

FSDP state_dict missing keys errors prevent checkpoint saving or loading in sharded models. Denpex identifies the source of key mismatches.

Quick answer

FSDP state_dict missing keys errors prevent checkpoint saving or loading in sharded models.

Distributed Training#fsdp#checkpoint#state-dict#model-parallelism#distributed-training#keys

What this failure is

FSDP Missing Keys / Unexpected Keys is a Distributed Training failure seen during ML training runs. FSDP state_dict missing keys errors prevent checkpoint saving or loading in sharded models. Denpex identifies the source of key mismatches. Common tags: Fsdp, Checkpoint, State Dict, Model Parallelism.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about FSDP Missing Keys / Unexpected Keys. 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)

FSDP renames parameters with _fsdp_wrapped_module. prefix that don't exist in the model definition. The checkpoint was saved with a different world size and parameter sharding layout than the current run. FSDP's flatten_params adds a flat_param key that replaces individual weight tensors in the state_dict. Mixed-precision FSDP stores fp32 copies with _orig_mod prefix that don't match the inference model's state_dict. 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

  • FSDP model.save_pretrained() fails with missing keys error
  • Loading sharded checkpoint produces warnings about unexpected or missing keys
  • After resuming training, model weights don't match checkpoint state

Common symptoms and what they mean

SymptomWhy it happens
RuntimeError: Error(s) in loading state_dict for FSDPModel: Missing key(s)FSDP renames parameters with _fsdp_wrapped_module. prefix that don't exist in the model definition
UserWarning: Detected missing keys during FSDP checkpoint loadThe checkpoint was saved with a different world size and parameter sharding layout than the current run
KeyError: 'model.layers.0.self_attn.q_proj.weight' missing from optimizer stateFSDP's flatten_params adds a flat_param key that replaces individual weight tensors in the state_dict
named_parameters() returns different parameter names before and after FSDP wrappingMixed-precision FSDP stores fp32 copies with _orig_mod prefix that don't match the inference model's state_dict

Which systems are affected

  • FSDP training with model parallelism
  • Checkpoint save/load across different GPU counts (world size change)
  • FSDP with mixed-precision training and float32 parameters
  • Checkpoint conversion between FSDP and non-FSDP training

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: Error(s) in loading state_dict for FSDPModel: Missing key(s)
  • Verified signal present: UserWarning: Detected missing keys during FSDP checkpoint load
  • Verified signal present: KeyError: 'model.layers.0.self_attn.q_proj.weight' missing from optimizer state
  • Verified signal present: named_parameters() returns different parameter names before and after FSDP wrapping
  • 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

  • FSDP renames parameters with _fsdp_wrapped_module. prefix that don't exist in the model definition
  • The checkpoint was saved with a different world size and parameter sharding layout than the current run
  • FSDP's flatten_params adds a flat_param key that replaces individual weight tensors in the state_dict
  • Mixed-precision FSDP stores fp32 copies with _orig_mod prefix that don't match the inference model's state_dict

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.