Add `{% generation %} to support tokenizer.apply_chat_template return_assistant_tokens_mask
#25
by
zeromquan
- opened
model_name_or_path = "Qwen/Qwen3-Coder-30B-A3B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(
model_name_or_path,
trust_remote_code=True,
use_fast=True,
)
example = {
"messages": [
{"role": "user", "content": "Hello!"},
{"role": "assistant", "content": "Hi! How can I help you? "},
],
}
processed = tokenizer .apply_chat_template(
example["messages"],
return_dict=True,
return_assistant_tokens_mask=True,
tools=example.get("tools"),
)
assert 1 in processed["assistant_masks"]