[Bugfix][Tokenizer] Replace bare asserts in the DeepSeek V4 encoder (#53747)
Signed-off-by: mhuzaifa3 <mhuzaifa3@outlook.com>
This commit is contained in:
@@ -403,3 +403,27 @@ def test_deepseek_v4_matches_reference_golden_fixtures(case_id, kwargs):
|
|||||||
|
|
||||||
expected = (FIXTURES_DIR / f"test_output_{case_id}.txt").read_text()
|
expected = (FIXTURES_DIR / f"test_output_{case_id}.txt").read_text()
|
||||||
assert prompt == expected
|
assert prompt == expected
|
||||||
|
|
||||||
|
|
||||||
|
def test_deepseek_v4_rejects_empty_developer_content():
|
||||||
|
with pytest.raises(ValueError):
|
||||||
|
_tokenizer().apply_chat_template(
|
||||||
|
[{"role": "developer", "content": ""}],
|
||||||
|
tokenize=False,
|
||||||
|
enable_thinking=True,
|
||||||
|
reasoning_effort="low",
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.parametrize(
|
||||||
|
"kwargs",
|
||||||
|
[
|
||||||
|
{"thinking_mode": "bogus"},
|
||||||
|
{"thinking_mode": "thinking", "reasoning_effort": "turbo"},
|
||||||
|
],
|
||||||
|
)
|
||||||
|
def test_deepseek_v4_encode_messages_rejects_invalid_arguments(kwargs):
|
||||||
|
from vllm.tokenizers.deepseek_v4_encoding import encode_messages
|
||||||
|
|
||||||
|
with pytest.raises(ValueError):
|
||||||
|
encode_messages([{"role": "user", "content": "Hello"}], **kwargs)
|
||||||
|
|||||||
@@ -225,8 +225,12 @@ def render_message(
|
|||||||
Returns:
|
Returns:
|
||||||
Encoded string for this message.
|
Encoded string for this message.
|
||||||
"""
|
"""
|
||||||
assert 0 <= index < len(messages)
|
if not (0 <= index < len(messages)):
|
||||||
assert thinking_mode in ["chat", "thinking"], f"Invalid thinking_mode `{thinking_mode}`"
|
raise ValueError(
|
||||||
|
f"Index {index} out of range for messages list of length {len(messages)}"
|
||||||
|
)
|
||||||
|
if thinking_mode not in ["chat", "thinking"]:
|
||||||
|
raise ValueError(f"Invalid thinking_mode `{thinking_mode}`")
|
||||||
|
|
||||||
prompt = ""
|
prompt = ""
|
||||||
msg = messages[index]
|
msg = messages[index]
|
||||||
@@ -248,10 +252,11 @@ def render_message(
|
|||||||
tool_calls = tool_calls_from_openai_format(tool_calls)
|
tool_calls = tool_calls_from_openai_format(tool_calls)
|
||||||
|
|
||||||
reasoning_effort = reasoning_effort or DEFAULT_REASONING_EFFORT
|
reasoning_effort = reasoning_effort or DEFAULT_REASONING_EFFORT
|
||||||
assert reasoning_effort in REASONING_EFFORT_PROMPTS, (
|
if reasoning_effort not in REASONING_EFFORT_PROMPTS:
|
||||||
f"Invalid reasoning effort: {reasoning_effort}, expected one of "
|
raise ValueError(
|
||||||
f"{list(REASONING_EFFORT_PROMPTS)}"
|
f"Invalid reasoning effort: {reasoning_effort}, expected one of "
|
||||||
)
|
f"{list(REASONING_EFFORT_PROMPTS)}"
|
||||||
|
)
|
||||||
if index == 0 and thinking_mode == "thinking":
|
if index == 0 and thinking_mode == "thinking":
|
||||||
prompt += REASONING_EFFORT_PROMPTS[reasoning_effort]
|
prompt += REASONING_EFFORT_PROMPTS[reasoning_effort]
|
||||||
|
|
||||||
@@ -263,7 +268,8 @@ def render_message(
|
|||||||
prompt += "\n\n" + response_format_template.format(schema=to_json(response_format))
|
prompt += "\n\n" + response_format_template.format(schema=to_json(response_format))
|
||||||
|
|
||||||
elif role == "developer":
|
elif role == "developer":
|
||||||
assert content, f"Invalid message for role `{role}`: {msg}"
|
if not content:
|
||||||
|
raise ValueError(f"Invalid message for role `{role}`: {msg}")
|
||||||
|
|
||||||
content_developer = USER_SP_TOKEN
|
content_developer = USER_SP_TOKEN
|
||||||
content_developer += content
|
content_developer += content
|
||||||
@@ -362,7 +368,10 @@ def render_message(
|
|||||||
task = messages[index].get("task")
|
task = messages[index].get("task")
|
||||||
if task is not None:
|
if task is not None:
|
||||||
# Task special token for internal classification tasks
|
# Task special token for internal classification tasks
|
||||||
assert task in VALID_TASKS, f"Invalid task: '{task}'. Valid tasks are: {list(VALID_TASKS)}"
|
if task not in VALID_TASKS:
|
||||||
|
raise ValueError(
|
||||||
|
f"Invalid task: '{task}'. Valid tasks are: {list(VALID_TASKS)}"
|
||||||
|
)
|
||||||
task_sp_token = DS_TASK_SP_TOKENS[task]
|
task_sp_token = DS_TASK_SP_TOKENS[task]
|
||||||
|
|
||||||
if task != "action":
|
if task != "action":
|
||||||
|
|||||||
Reference in New Issue
Block a user