Chat Template

Defend pinning the tokenizer revision alongside the model weights when shipping a fine-tune

PremiumSpot the bug: chat…

Qwen 3.5 SFT data: fill the special tokens that wrap an assistant turn.

Should the system message contribute to the SFT loss?

ShareGPT format: what data shape does it use?

Fill in the SFT JSONL: the three required fields of a typical chat example.

Define supervised fine-tuning (SFT) in one breath.

Roles in a chat training example: pick the three that are standard.

Why does dropping the EOS token from SFT labels produce a model that never stops generating?

Define the role of a chat template when fine-tuning on (system, user, assistant) data

Pick the three first things to check when LoRA training loss sits flat from step 0

Describe the JSON shape of a Claude Opus 4.7 assistant message that calls a tool

You open a HuggingFace model repo and see tokenizer_config.json. What is in it and what is NOT?

A chat template wraps every turn with BOS and EOS tokens. Why does the model need them?

Your GPT-5.5 call bills 30 percent more tokens than the user prompt plus system instruction. Where are the invisible tokens?

A hand built chat prompt performs worse than apply_chat_template. Spot the tokenization issues.

After editing the Jinja chat template, your fine-tuned Llama 3 prints raw special token text. What broke?

Tool calling fine-tune: data shape and the common pitfall

Spot the bug: a Llama-3 fine-tune that skips the end of turn token

Why is chat template integrity the most common 'silent quality killer' in FT?