Skip to content

vllm.entrypoints.openai.tool_parsers.longcat_tool_parser

LongcatFlashToolParser

Bases: Hermes2ProToolParser

Source code in vllm/entrypoints/openai/tool_parsers/longcat_tool_parser.py 
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
@ToolParserManager.register_module("longcat")
class LongcatFlashToolParser(Hermes2ProToolParser):

    def __init__(self, tokenizer: AnyTokenizer):
        super().__init__(tokenizer)

        self.tool_call_start_token: str = "<longcat_tool_call>"
        self.tool_call_end_token: str = "</longcat_tool_call>"

        self.tool_call_regex = re.compile(
            r"<longcat_tool_call>(.*?)</longcat_tool_call>|<longcat_tool_call>(.*)",
            re.DOTALL)

        self.tool_call_start_token_ids = self.model_tokenizer.encode(
            self.tool_call_start_token, add_special_tokens=False)
        self.tool_call_end_token_ids = self.model_tokenizer.encode(
            self.tool_call_end_token, add_special_tokens=False)

        self.tool_call_start_token_array = [
            self.model_tokenizer.decode([token_id])
            for token_id in self.tool_call_start_token_ids
        ]

        self.tool_call_end_token_array = [
            self.model_tokenizer.decode([token_id])
            for token_id in self.tool_call_end_token_ids
        ]

tool_call_end_token instance-attribute 

tool_call_end_token: str = '</longcat_tool_call>'

tool_call_end_token_array instance-attribute 

tool_call_end_token_array = [
    (decode([token_id]))
    for token_id in (tool_call_end_token_ids)
]

tool_call_end_token_ids instance-attribute 

tool_call_end_token_ids = encode(
    tool_call_end_token, add_special_tokens=False
)

tool_call_regex instance-attribute 

tool_call_regex = compile(
    "<longcat_tool_call>(.*?)</longcat_tool_call>|<longcat_tool_call>(.*)",
    DOTALL,
)

tool_call_start_token instance-attribute 

tool_call_start_token: str = '<longcat_tool_call>'

tool_call_start_token_array instance-attribute 

tool_call_start_token_array = [
    (decode([token_id]))
    for token_id in (tool_call_start_token_ids)
]

tool_call_start_token_ids instance-attribute 

tool_call_start_token_ids = encode(
    tool_call_start_token, add_special_tokens=False
)

__init__ 

__init__(tokenizer: AnyTokenizer)
Source code in vllm/entrypoints/openai/tool_parsers/longcat_tool_parser.py 
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
def __init__(self, tokenizer: AnyTokenizer):
    super().__init__(tokenizer)

    self.tool_call_start_token: str = "<longcat_tool_call>"
    self.tool_call_end_token: str = "</longcat_tool_call>"

    self.tool_call_regex = re.compile(
        r"<longcat_tool_call>(.*?)</longcat_tool_call>|<longcat_tool_call>(.*)",
        re.DOTALL)

    self.tool_call_start_token_ids = self.model_tokenizer.encode(
        self.tool_call_start_token, add_special_tokens=False)
    self.tool_call_end_token_ids = self.model_tokenizer.encode(
        self.tool_call_end_token, add_special_tokens=False)

    self.tool_call_start_token_array = [
        self.model_tokenizer.decode([token_id])
        for token_id in self.tool_call_start_token_ids
    ]

    self.tool_call_end_token_array = [
        self.model_tokenizer.decode([token_id])
        for token_id in self.tool_call_end_token_ids
    ]