Reference model logits shape does not match policy logits shape vocab mismatch

The reference model and the policy model have different vocabulary sizes, so their logits cannot be compared. DPO scores the policy against the reference token by token, a vocab difference of even a few entries makes every comparison meaningless. This entry explains how to confirm the cause, apply the fix, and separate it from adjacent rlhf failures.

The literal signature is "Reference model logits shape does not match policy logits shape vocab mismatch". It is a distributed training failure associated with TRL, OpenRLHF, Ray, and actor-critic training. The line identifies the failing operation or subsystem, while the surrounding evidence decides whether it is the initiating fault or a downstream symptom.

The reference model and the policy model have different vocabulary sizes, so their logits cannot be compared. DPO scores the policy against the reference token by token, a vocab difference of even a few entries makes every comparison meaningless. The failure becomes visible at this call site because the operation first requires the missing resource, valid state, healthy peer, or correct result. Earlier log lines and a known-good control carry more causal value than the final wrapper exception.

• The workload stops or loses forward progress after emitting "Reference model logits shape does not match policy logits shape vocab mismatch".
• A retry on the same configuration reproduces the failure because the causal state has not changed.
• The outer framework exception can hide the rank, node, allocation, or dependency that failed first.
• Increasing timeouts or reducing workload size can suppress the symptom without correcting the cause.

• TRL, OpenRLHF, Ray, and actor-critic training
• production-shaped multi-accelerator workloads
• containerized and bare-metal deployments of the same stack

Apply the following checklist to a small reproduction: each box below is a positive signal that you are looking at this exact failure rather than a sibling in the same taxonomy.

• ✓Find the first occurrence of "Reference model logits shape does not match policy logits shape vocab mismatch" and preserve at least 100 lines before it.
• ✓Identify which rank, node, device, or process emitted the first related warning.
• ✓this happens when the policy was fine-tuned with resize_token_embeddings() after adding tokens, while the reference was loaded from the original checkpoint. The policy is right; the reference is stale.
• ✓Repeat the same input after the targeted change and require the signature to disappear.
• ✓Resume from job start only after the control passes.

Reference model logits shape does not match policy logits shape vocab mismatch

Timestamps and exact values vary across runs, but the pattern. An info-level start, an early WARN, an ERROR carrying the symptom. Is the actual fingerprint you should alert on. The Denpex platform flags this combination automatically.

check both vocabulary sizes, model.config.vocab_size on each. The difference is usually a handful of added special tokens (chat template, padding, tool tokens). This changes the condition in the causal diagnosis instead of hiding the outer exception. The repeated control proves ownership before recovery from job start.

# Preserve evidence before restarting
rg -n -i 'error|exception|timeout|failed' <log-file>
nvidia-smi
python -m torch.utils.collect_env

# Find the exact signature in the complete log
rg -n -F -- "Reference model logits shape does not match policy logits shape vocab mismatch" <log-file>

Adapt the snippet to your framework. The same pattern holds for PyTorch Lightning, Hugging Face Trainer, DeepSpeed, Megatron-LM, and vLLM training wrappers. Where the wrapper exposes a config flag (for examplelr_scheduler_type in Trainer), prefer the flag over the imperative API to keep the schedule declarative and reproducible.

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.

• The reference model and the policy model have different vocabulary sizes, so their logits cannot be compared. DPO scores the policy against the reference token by token, a vocab difference of even a few entries makes every comparison meaningless.
• The decisive evidence is the first log line that precedes "Reference model logits shape does not match policy logits shape vocab mismatch" and differs from a healthy run.
• A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.