Add files using upload-large-folder tool

deepseek/lib/python3.10/site-packages/lmformatenforcer/__init__.py

@@ -0,0 +1,23 @@
+__all__ = ['CharacterLevelParser', 
+           'CharacterLevelParserConfig',
+           'StringParser', 
+           'RegexParser', 
+           'UnionParser',
+           'SequenceParser',
+           'JsonSchemaParser',
+           'TokenEnforcer',
+           'TokenEnforcerTokenizerData',
+           'LMFormatEnforcerException',
+           'FormatEnforcerAnalyzer',]
+
+from .characterlevelparser import CharacterLevelParser, CharacterLevelParserConfig, StringParser, UnionParser, SequenceParser
+from .regexparser import RegexParser
+from .jsonschemaparser import JsonSchemaParser
+from .tokenenforcer import TokenEnforcer, TokenEnforcerTokenizerData
+from .exceptions import LMFormatEnforcerException
+try:
+    from .analyzer import FormatEnforcerAnalyzer
+except ImportError as e:
+    import logging
+    logging.warning(e)
+    FormatEnforcerAnalyzer = None

deepseek/lib/python3.10/site-packages/lmformatenforcer/consts.py

@@ -0,0 +1,20 @@
+COMPLETE_ALPHABET = "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!@#$%^&*()_+-=[]{};:,./<>? `'\""
+DEFAULT_MAX_CONSECUTIVE_WHITESPACES = 12
+DEFAULT_FORCE_JSON_FIELD_ORDER = False
+DEFAULT_MAX_JSON_ARRAY_LENGTH = 20
+WHITESPACE_CHARACTERS = " \t\n\r"
+BACKSLASH = "\\"
+BACKSLASH_ESCAPING_CHARACTERS = '"\\/bfnrt'  # Characters allowed after an escaping backslash, except unicode
+BACKSLACH_UNICODE_ESCAPE = "u"
+
+CONFIG_ENV_VAR_MAX_CONSECUTIVE_WHITESPACES = 'LMFE_MAX_CONSECUTIVE_WHITESPACES'
+"""Environment variable for externally controlling how many consective whitespaces the 
+JsonSchemaParser will allow. Default: 12"""
+
+CONFIG_ENV_VAR_STRICT_JSON_FIELD_ORDER = 'LMFE_STRICT_JSON_FIELD_ORDER'
+"""Environment variable for externally controlling whether the JsonSchemaParser will force 
+fields to appear in the order of the 'required' field in the schema. Default: false"""
+
+CONFIG_ENV_VAR_MAX_JSON_ARRAY_LENGTH = 'LMFE_MAX_JSON_ARRAY_LENGTH'
+"""Environment variable for externally controlling what is the maximal JSON array length,
+if not specified by the schema. Default: 20"""

deepseek/lib/python3.10/site-packages/lmformatenforcer/exceptions.py

@@ -0,0 +1,3 @@
+class LMFormatEnforcerException(Exception):
+    """Base class for exceptions in this module."""
+    pass

deepseek/lib/python3.10/site-packages/lmformatenforcer/integrations/haystackv2.py

@@ -0,0 +1,74 @@
+try:
+    from haystack import component
+    from canals import Component
+except ImportError:
+    raise ImportError('haystack is not installed. Please install it with "pip install farm-haystack" or "pip install haystack-ai')
+
+import enum
+from typing import Any, Callable, Dict, List, Optional
+from lmformatenforcer import CharacterLevelParser
+
+
+class _ModelType(enum.Enum):
+    HUGGINGFACE = 'HuggingFaceLocalGenerator'
+    # VLLM = 'vLLMLocalInvocationLayer' TODO: Add this when vLLM has Haystack V2 support
+
+@component
+class LMFormatEnforcerLocalGenerator:
+    """A generator component for Haystack V2 API that activates the LMFormatEnforcer on the generated text. 
+    It wraps a local generator, and should be added to the pipeline instead of it"""
+    def __init__(self, model_component: Component, character_level_parser: Optional[CharacterLevelParser] = None):
+        """Initialize the generator component
+        :param model_component: A local generator component to wrap
+        :param character_level_parser: A CharacterLevelParser that will be used to enforce the format of the generated"""
+        self.model_component = model_component
+        self.character_level_parser = character_level_parser
+        self._model_type = self._resolve_model_type()
+        self.token_enforcer_fn: Optional[Callable] = None
+
+    @component.output_types(replies=List[str])
+    def run(self, prompt: str, generation_kwargs: Optional[Dict[str, Any]] = None):
+        try:
+            self._inject_enforcer_into_model()
+            kwargs = {}
+            if generation_kwargs:
+                kwargs['generation_kwargs'] = generation_kwargs
+            return self.model_component.run(prompt, **kwargs)
+        finally:
+            self._release_model_injection()
+
+    def warm_up(self):
+        if hasattr(self.model_component, 'warm_up'):
+            self.model_component.warm_up()
+        self.token_enforcer_fn = self._prepare_token_enforcer_fn()
+
+    def _prepare_token_enforcer_fn(self) -> Optional[Callable]:
+        if not self.character_level_parser:
+            return None
+        if self._model_type == _ModelType.HUGGINGFACE:
+            tokenizer = self.model_component.pipeline.tokenizer
+            from lmformatenforcer.integrations.transformers import build_transformers_prefix_allowed_tokens_fn
+            return build_transformers_prefix_allowed_tokens_fn(tokenizer, self.character_level_parser)
+        raise NotImplementedError(f"Token enforcer not implemented for model type {self._model_type.name}")
+
+    def _resolve_model_type(self) -> _ModelType:
+        generator_component_name = self.model_component.__class__.__name__ 
+        try:
+            return _ModelType(generator_component_name)
+        except ValueError:
+            supported_strings = ",".join(str(t.name) for t in _ModelType)
+            raise ValueError(f"Unsupported local generator component layer: {generator_component_name}. "
+                             f"Must be one of {supported_strings}")
+        
+    def _inject_enforcer_into_model(self):
+        if not self.token_enforcer_fn:
+            return
+        if self._model_type == _ModelType.HUGGINGFACE:
+            self.model_component.generation_kwargs['prefix_allowed_tokens_fn'] = self.token_enforcer_fn
+        
+    
+    def _release_model_injection(self):
+        if not self.token_enforcer_fn:
+            return
+        if self._model_type == _ModelType.HUGGINGFACE:
+            del self.model_component.generation_kwargs['prefix_allowed_tokens_fn']

deepseek/lib/python3.10/site-packages/lmformatenforcer/integrations/trtllm.py

@@ -0,0 +1,84 @@
+import math
+from typing import List, Optional, Tuple, Union
+import torch
+from transformers import PreTrainedTokenizerBase
+from lmformatenforcer import CharacterLevelParser, FormatEnforcerAnalyzer
+from lmformatenforcer.tokenenforcer import TokenEnforcer, TokenEnforcerTokenizerData
+
+
+class TRTLLMLogitsProcessor:
+    def __init__(self, token_enforcer: TokenEnforcer, eos_token_id, analyze):
+        self.token_enforcer = token_enforcer
+        self.analyzer = FormatEnforcerAnalyzer(token_enforcer) if analyze else None
+        self.mask: Optional[torch.Tensor] = None
+        self.mask_val = -math.inf
+        self.eos_token_id = eos_token_id
+
+    def _trim(self, input):
+        return [x for x in input.tolist() if x not in \
+                (self.eos_token_id if isinstance(self.eos_token_id, list) else [self.eos_token_id])]
+
+    def __call__(self, step: int, batch_input_ids: List[List[int]], logits: torch.Tensor) -> torch.Tensor:
+        for idx in range(len(batch_input_ids)):
+            if self.analyzer:
+                self.analyzer.report_raw_logits(batch_input_ids[idx], logits[idx].tolist())
+
+            allowed_tokens = self.token_enforcer.get_allowed_tokens(self._trim(batch_input_ids[idx]))
+
+            if self.mask is not None:
+                self.mask.fill_(self.mask_val)
+            else:
+                # We create it here because full_like() also copies the device and dtype
+                self.mask = torch.full_like(logits[idx], self.mask_val)
+            self.mask[allowed_tokens] = 0
+            logits[idx] = logits[idx] + self.mask
+
+        return logits
+
+
+def _build_regular_tokens_list(tokenizer) -> List[Tuple[int, str, bool]]:
+    # There are many classes that can be passed here, this logic should work on all of them.
+    if hasattr(tokenizer, 'get_tokenizer'):
+        tokenizer = tokenizer.get_tokenizer()
+    if hasattr(tokenizer, 'tokenizer'):
+        tokenizer = tokenizer.tokenizer
+    token_0 = [tokenizer.encode("0")[-1]]
+    regular_tokens = []
+    vocab_size = tokenizer.vocab_size
+    for token_idx in range(vocab_size):
+        if token_idx in tokenizer.all_special_ids:
+            continue
+        # We prepend token 0 and skip the first letter of the result to get a space if the token is a start word.
+        tensor_after_0 = torch.tensor(token_0 + [token_idx], dtype=torch.long)
+        decoded_after_0 = tokenizer.decode(tensor_after_0)[1:]
+        decoded_regular = tokenizer.decode(token_0)
+        is_word_start_token = len(decoded_after_0) > len(decoded_regular)
+        regular_tokens.append((token_idx, decoded_after_0, is_word_start_token))
+    return regular_tokens
+
+
+def build_trtlmm_tokenizer_data(tokenizer: PreTrainedTokenizerBase) -> TokenEnforcerTokenizerData:
+    """Build the TokenEnforcerTokenizerData from a tokenizer in order to cache it between instances"""
+    regular_tokens = _build_regular_tokens_list(tokenizer)
+
+    def _decode(tokens: List[int]) -> str:
+        tensor = torch.tensor(tokens, dtype=torch.long)
+        return tokenizer.decode(tensor)
+
+    tokenizer_data = TokenEnforcerTokenizerData(regular_tokens, _decode, tokenizer.eos_token_id)
+    return tokenizer_data
+
+
+def build_trtllm_logits_processor(tokenizer: Union[PreTrainedTokenizerBase, TokenEnforcerTokenizerData],
+                                  character_level_parser: CharacterLevelParser,
+                                  analyze: bool = False) -> TRTLLMLogitsProcessor:
+    """
+    Build logits processor for feeding it into generate function (use_py_session should be True)
+    """
+    if isinstance(tokenizer, TokenEnforcerTokenizerData):
+        tokenizer_data = tokenizer
+    else:
+        tokenizer_data = build_trtlmm_tokenizer_data(tokenizer)
+
+    token_enforcer = TokenEnforcer(tokenizer_data, character_level_parser)
+    return TRTLLMLogitsProcessor(token_enforcer, tokenizer.eos_token_id, analyze)

deepseek/lib/python3.10/site-packages/lmformatenforcer/integrations/vllm.py

@@ -0,0 +1,60 @@
+try:
+    import torch
+    import vllm
+    from vllm.transformers_utils.tokenizer import MistralTokenizer
+    from transformers import PreTrainedTokenizerBase
+except ImportError:
+    raise ImportError('vllm is not installed. Please install it with "pip install vllm"')
+from lmformatenforcer import CharacterLevelParser, TokenEnforcer, FormatEnforcerAnalyzer, TokenEnforcerTokenizerData
+from lmformatenforcer.integrations.transformers import build_token_enforcer_tokenizer_data
+from typing import List, Optional, Union
+import math
+
+
+class VLLMLogitsProcessor:
+    def __init__(self, token_enforcer: TokenEnforcer, analyze):
+        self.token_enforcer = token_enforcer
+        self.analyzer = FormatEnforcerAnalyzer(token_enforcer) if analyze else None
+        self.mask: Optional[torch.Tensor] = None
+
+    def __call__(self, input_ids: List[int], scores: torch.Tensor) -> torch.Tensor:
+        token_sequence = input_ids
+        if self.analyzer:
+            self.analyzer.report_raw_logits(token_sequence, scores.tolist())
+        allowed_tokens = self.token_enforcer.get_allowed_tokens(token_sequence)
+        if self.mask is not None:
+            self.mask.fill_(-math.inf)
+        else:
+            # We create it here because full_like() also copies the device and dtype
+            self.mask = torch.full_like(scores, -math.inf)
+        self.mask[allowed_tokens] = 0
+        scores = scores + self.mask
+        return scores
+
+
+def build_vllm_token_enforcer_tokenizer_data(tokenizer: Union[vllm.LLM, PreTrainedTokenizerBase]) -> TokenEnforcerTokenizerData:
+    # There are many classes that can be passed here, this logic should work on all of them.
+    vocab_size = None
+    if hasattr(tokenizer, 'llm_engine'):
+        vocab_size = tokenizer.llm_engine.get_model_config().get_vocab_size()
+    if hasattr(tokenizer, 'get_tokenizer'):
+        tokenizer = tokenizer.get_tokenizer()
+    if isinstance(tokenizer, MistralTokenizer):
+        return build_token_enforcer_tokenizer_data(tokenizer, vocab_size)
+    if hasattr(tokenizer, 'tokenizer'):
+        tokenizer = tokenizer.tokenizer
+    return build_token_enforcer_tokenizer_data(tokenizer, vocab_size)
+
+
+def build_vllm_logits_processor(llm: Union[vllm.LLM, PreTrainedTokenizerBase, TokenEnforcerTokenizerData], 
+                                character_level_parser: CharacterLevelParser, 
+                                analyze: bool=False) -> VLLMLogitsProcessor:
+    """Build the logits processor function that llama.cpp will use to filter the tokens generated by the model. The result
+    can be passed in the logits_processor list that is sent to the call or generate() method of llama.cpp models."""
+    if not isinstance(llm, TokenEnforcerTokenizerData):
+        llm = build_vllm_token_enforcer_tokenizer_data(llm)
+    token_enforcer = TokenEnforcer(llm, character_level_parser)
+    return VLLMLogitsProcessor(token_enforcer, analyze)
+
+
+__all__ = ['build_vllm_logits_processor', 'build_vllm_token_enforcer_tokenizer_data']

deepseek/lib/python3.10/site-packages/lmformatenforcer/jsonschemaparser.py

@@ -0,0 +1,710 @@
+from copy import deepcopy
+import enum
+import sys
+from typing import Dict, Hashable, List, Optional, Union, cast
+
+
+from .external.jsonschemaobject import JsonSchemaObject, json_schema_data_formats
+from .exceptions import LMFormatEnforcerException
+from .characterlevelparser import CharacterLevelParser, CharacterLevelParserConfig, ForceStopParser, SequenceParser, StringParser, UnionParser
+from .consts import BACKSLASH, BACKSLASH_ESCAPING_CHARACTERS, WHITESPACE_CHARACTERS
+from .regexparser import RegexParser
+
+# No need to include the 'integer' option in the anyOf, as it is a subset of 'number'
+_ANY_JSON_SCHEMA_DICT = {'anyOf': [{'type': type} for type in json_schema_data_formats.keys() if type != 'integer']}
+
+class JsonSchemaParser(CharacterLevelParser):
+    ANY_JSON_OBJECT_SCHEMA: JsonSchemaObject = JsonSchemaObject(**_ANY_JSON_SCHEMA_DICT)
+    class _Context:
+        model_class: JsonSchemaObject
+        # We store the active parser in the context, so that if a node adds to the stack, it knows
+        # to which parser's stack to add.
+        active_parser: "JsonSchemaParser"
+        alphabet_without_quotes: str
+        regex_parser_cache: Dict[str, RegexParser] = {}
+
+    object_stack: List[CharacterLevelParser]
+    context: _Context
+    num_consecutive_whitespaces: int
+    last_parsed_string: str  # Slight hack to allow communicating the parsed key to the object parser
+    last_non_whitespace_character: str  # Slight hack to allow list parser to know if there is an item on top
+
+    def __init__(self, 
+                 json_schema: Union[dict, _Context, None], 
+                 config: Optional[CharacterLevelParserConfig] = None, 
+                 existing_stack: Optional[List[CharacterLevelParser]] = None, 
+                 num_consecutive_whitespaces: int = 0):
+        """Create a CharacterLevelParser for parsing JSON.
+        :param json_schema: The json schema to parse. Can be a dict of a JSON schema, or None if any json output is allowed."""
+        super().__init__(config)
+        if isinstance(json_schema, JsonSchemaParser._Context):
+            self.context = json_schema
+        else:
+            self.context = JsonSchemaParser._Context()
+            json_schema = json_schema or _ANY_JSON_SCHEMA_DICT
+            self.context.model_class = JsonSchemaObject(**json_schema)
+            self.context.active_parser = self
+            self.context.alphabet_without_quotes = self.config.alphabet.replace('"', '')
+        
+        self.num_consecutive_whitespaces = num_consecutive_whitespaces
+        if existing_stack is None:
+            self.object_stack = [get_parser(self, self.context.model_class)]
+        else:
+            self.object_stack = existing_stack
+        self.last_parsed_string = ""
+        self.last_non_whitespace_character = ""
+    
+    def add_character(self, new_character: str) -> CharacterLevelParser:
+        self.context.active_parser = self
+        # Assumption: The top-most parser that can accept the character is the one that should accept it.
+        # This is different from the SequenceParser, in which we need to split (union) into all options.
+        receiving_idx = len(self.object_stack) - 1
+        last_parsed_string = self.last_parsed_string
+        while receiving_idx >= 0 and new_character not in self.object_stack[receiving_idx].get_allowed_characters():
+            finished_receiver = self.object_stack[receiving_idx]
+            if isinstance(finished_receiver, StringParsingState):
+                last_parsed_string = finished_receiver.parsed_string
+            receiving_idx -= 1
+        
+        updated_stack = self.object_stack[:receiving_idx + 1]
+        updated_parser = JsonSchemaParser(self.context, self.config, updated_stack, self.num_consecutive_whitespaces)
+        updated_parser.context.active_parser = updated_parser
+        updated_parser.last_parsed_string = last_parsed_string
+        if receiving_idx >= 0:
+            updated_parser.object_stack[receiving_idx] = updated_parser.object_stack[receiving_idx].add_character(new_character)
+        if new_character in WHITESPACE_CHARACTERS:
+            updated_parser.num_consecutive_whitespaces += 1
+            updated_parser.last_non_whitespace_character = self.last_non_whitespace_character
+        else:
+            updated_parser.num_consecutive_whitespaces = 0
+            updated_parser.last_non_whitespace_character = new_character
+
+        if updated_parser.object_stack and isinstance(updated_parser.object_stack[-1], UnionParser) and \
+            any(isinstance(parser, (ObjectParsingState, ListParsingState)) for parser in updated_parser.object_stack[-1].parsers):
+            # If the top parser is a union parser with "advanced" (=parsers that modify the object stack) parsers inside, 
+            # we need to split the top level parser into the different options,
+            # As each "fork" can live with a different object stack, and we need to make sure they have their own ones.
+            option_json_schema_parsers = []
+            for option_parser in updated_parser.object_stack[-1].parsers:
+                option_stack = updated_parser.object_stack[:-1] + [option_parser]
+                option_parser = JsonSchemaParser(self.context, self.config, option_stack, updated_parser.num_consecutive_whitespaces)
+                option_parser.context.active_parser = option_parser
+                option_parser.last_parsed_string = last_parsed_string
+                option_parser.last_non_whitespace_character = updated_parser.last_non_whitespace_character
+                option_json_schema_parsers.append(option_parser)
+            return UnionParser(option_json_schema_parsers)
+
+        # For some performance optimizations to work, we want to make sure we don't leave irrelevant
+        # objects at the top of the stack, which we know will be passed over next timestep
+        new_object_stack = updated_parser.object_stack
+        while new_object_stack and new_object_stack[-1].can_end() and new_object_stack[-1].get_allowed_characters() == '':
+            finished_receiver = new_object_stack[-1]
+            if isinstance(finished_receiver, StringParsingState):
+                updated_parser.last_parsed_string = finished_receiver.parsed_string
+            del new_object_stack[-1]
+            if new_object_stack:
+                new_top_parser = new_object_stack[-1]
+                if isinstance(new_top_parser, ListParsingState):
+                    new_top_parser = new_top_parser._clone()
+                    new_top_parser.num_items_seen += 1
+                    new_object_stack[-1] = new_top_parser
+                    
+
+        return updated_parser
+
+    def get_allowed_characters(self) -> str:
+        self.context.active_parser = self
+        
+        allowed_character_strs = []
+        for parser in reversed(self.object_stack):
+            # Similar to SequenceParser, if the top object can end, we need to know to accept the next character of parser below, etc.
+            allowed_character_strs.append(parser.get_allowed_characters())
+            if not parser.can_end():
+                break
+        if len(allowed_character_strs) > 0:
+            allowed_characters =  "".join(allowed_character_strs)
+        else:
+            # In certain cases, beam search / sample crashes when there are less legal 
+            # continuation tokens than there are beams. Therefore, we allow whitespace 
+            # characters when the object stack is empty (= we are done parsing)
+            allowed_characters = WHITESPACE_CHARACTERS
+        
+        if self.num_consecutive_whitespaces >= self.config.max_consecutive_whitespaces:
+            # print("Filtering whitespace characters")
+            allowed_characters = "".join(c for c in allowed_characters if c not in WHITESPACE_CHARACTERS)
+        return allowed_characters
+
+    def can_end(self) -> bool:
+        return all(parser.can_end() for parser in self.object_stack)
+
+    def shortcut_key(self) -> Optional[Hashable]:
+        if self.object_stack:
+            current_parser = self.object_stack[-1]
+            if isinstance(current_parser, StringParsingState):
+                if not current_parser.allowed_strings and current_parser.seen_opening_quote and not current_parser.seen_closing_quote and not current_parser.regex_parser:
+                    # Performance optimization: When we are parsing a string that is not from a list of allowed strings, most tokens
+                    # are legal. The exploration can be more costly than the LM itself for large tokenizers (because this is pure python),
+                    # so we signal that we are in a "freetext" mode, and reuse the allowed token list throughout the run.
+                    cur_len = len(current_parser.parsed_string)
+                    min_len = current_parser.min_length or 0
+                    max_len = current_parser.max_length or sys.maxsize
+                    assert min_len <= max_len, "Invalid schema for str: min length is larger than max length"
+                    if cur_len < max_len:
+                        return ('json_freetext', cur_len, min_len, max_len)
+        return None
+
+
+class BaseParsingState(CharacterLevelParser):
+    def __init__(self, root: JsonSchemaParser):
+        self.root = root
+
+
+def _merge_object_schemas(base_schema: JsonSchemaObject, option_schema: JsonSchemaObject) -> JsonSchemaObject:
+    base_schema_properties = base_schema.properties or {}
+    for property_name, property_value in base_schema_properties.items():
+        # We assume that if a property exists in both base and option, the option version will be
+        # more specific, therefore we only take missing entries
+        if property_name not in option_schema.properties:
+            option_schema.properties[property_name] = property_value
+    for required_property in base_schema.required:
+        if required_property not in option_schema.required:
+            option_schema.required.append(required_property)
+    return option_schema
+
+
+def get_parser(
+    parsing_state: JsonSchemaParser,
+    value_schema: JsonSchemaObject
+) -> CharacterLevelParser:
+    if value_schema is None:
+        raise Exception("JsonSchemaParser: Value schema is None")
+    if value_schema.anyOf:
+        parsers = [get_parser(parsing_state, schema) for schema in value_schema.anyOf]
+        return UnionParser(parsers)
+    if value_schema.allOf:
+        merged_schema = value_schema.allOf[0]
+        for schema in value_schema.allOf[1:]:
+            merged_schema = _merge_object_schemas(merged_schema, schema)
+        return get_parser(parsing_state, merged_schema)
+    if value_schema.extras and 'const' in value_schema.extras:
+        allowed_value = value_schema.extras['const']
+        is_string = type(allowed_value) == str
+        return StringParsingState(parsing_state, 
+                                  [allowed_value], 
+                                  require_opening_quote=is_string, 
+                                  require_closing_quote=is_string)
+    if value_schema.type == "string":
+        return StringParsingState(
+            parsing_state,
+            value_schema.enum,
+            require_opening_quote=True,
+            min_length=value_schema.minLength,
+            max_length=value_schema.maxLength,
+            pattern=value_schema.pattern,
+        )
+    if value_schema.oneOf:
+        # We create a combined object schema for each option that includes the information from the parent
+        # And then create a UnionParser based on the combined options
+        merged_schemas = [_merge_object_schemas(value_schema, option_schema) for option_schema in value_schema.oneOf]
+        object_parsing_options = [ObjectParsingState(merged_schema, parsing_state) for merged_schema in merged_schemas]
+        return UnionParser(object_parsing_options)
+    elif value_schema.type == "object":
+        return ObjectParsingState(value_schema, parsing_state)
+    elif value_schema.type == None and value_schema.ref:
+        value_class_name = value_schema.ref.split('/')[-1]
+        extras = parsing_state.context.model_class.extras
+        # Pydantic V1 and V2 have different names for the definitions field
+        if 'definitions' in extras:
+            definitions = extras['definitions']
+        elif '$defs' in extras:
+            definitions = extras['$defs']
+        else:
+            raise ValueError("No definitions found in schema")
+        class_dict = definitions[value_class_name]
+        value_schema = JsonSchemaObject(**class_dict)
+        return get_parser(parsing_state, value_schema)
+    elif value_schema.enum:
+        is_numeric = all(isinstance(i, (int, float)) for i in value_schema.enum)
+        is_string = all(isinstance(i, (str)) for i in value_schema.enum)
+        if is_string:
+            return StringParsingState(
+            parsing_state,
+            value_schema.enum,
+            require_opening_quote=True,
+        )
+        elif is_numeric:
+            return StringParsingState(
+                parsing_state,
+                [str(i) for i in value_schema.enum],
+                require_opening_quote=False,
+                require_closing_quote=False,
+            )
+        else:
+            raise Exception("Unsupported enum type " + str(value_schema.enum))
+    elif value_schema.type == "integer":
+        return NumberParsingState(parsing_state, False)
+    elif value_schema.type == "boolean":
+        return StringParsingState(
+            parsing_state,
+            ["true", "false"],
+            require_opening_quote=False,
+            require_closing_quote=False,
+        )
+    elif value_schema.type == "null":
+        return StringParsingState(
+            parsing_state,
+            ["null"],
+            require_opening_quote=False,
+            require_closing_quote=False,
+        )
+    elif value_schema.type == "number":
+        return NumberParsingState(parsing_state, True)
+    elif value_schema.type == "array":
+        item_schema = value_schema.items or JsonSchemaParser.ANY_JSON_OBJECT_SCHEMA
+        return ListParsingState(parsing_state, item_schema, value_schema.minItems, value_schema.maxItems)
+    else:
+        raise Exception("Unsupported type " + str(value_schema.type))
+
+
+class ObjectParsingStage(enum.Enum):
+    START_OBJECT = "StartObject"
+    PARSING_KEY_OR_END = "ParsingKey"
+    PARSING_KEY_VALUE_SEPARATOR = "ParsingKeyValueSeparator"
+    PARSING_VALUE = "ParsingValue"
+    PARSING_SEPARATOR_OR_END = "ParsingSeparatorOrEnd"
+    END_OBJECT = "EndObject"
+
+
+class ObjectParsingState(BaseParsingState):
+    schema_object: JsonSchemaObject
+    current_stage: ObjectParsingStage
+    existing_keys: List[str]
+    current_key: Optional[str]
+    is_dictionary: bool
+
+    def __init__(self, schema_object: JsonSchemaObject, root: JsonSchemaParser):
+        super().__init__(root)
+        self.schema_object = schema_object
+        self.current_stage = ObjectParsingStage.START_OBJECT
+        self.root = root
+        self.existing_keys = []
+        self.current_key = None
+        # Javascript objects represent both classes and dictionaries, so we need to know which one we are parsing
+        self.is_dictionary = self.schema_object.properties is None
+
+    def clone(self) -> 'ObjectParsingState':
+        clone = ObjectParsingState(self.schema_object, self.root)
+        clone.current_stage = self.current_stage
+        clone.existing_keys = self.existing_keys[:]
+        clone.current_key = self.current_key
+        clone.is_dictionary = self.is_dictionary
+        return clone
+
+    def add_character(self, new_character: str) -> CharacterLevelParser:
+        if new_character.strip() == "":
+            # In object scope, whitespaces can be ignored
+            return self
+        self = self.clone()  # Immutability requirement
+        if (
+            self.current_stage == ObjectParsingStage.START_OBJECT
+            and new_character == "{"
+        ):
+            self.current_stage = ObjectParsingStage.PARSING_KEY_OR_END
+        elif self.current_stage == ObjectParsingStage.PARSING_KEY_OR_END:
+            if new_character == "}":
+                self.current_stage = ObjectParsingStage.END_OBJECT
+            if new_character == '"':
+                possible_keys = None
+                if not self.is_dictionary:
+                    required_keys = self.schema_object.required or []
+                    next_required_key = next((key for key in required_keys if key not in self.existing_keys), None)
+                    if self.root.config.force_json_field_order and next_required_key:
+                        possible_keys = [next_required_key]
+                    else:
+                        possible_keys = list(self.schema_object.properties.keys())
+                        possible_keys = list(
+                            set(possible_keys).difference(self.existing_keys)
+                        )
+                # We send require_opening_quote=True and then add_character('"') instead of require_opening_quote=False
+                # Because there is a difference between "don't need a quote" and "received it before creating the parser"
+                key_parser = StringParsingState(
+                    self.root, possible_keys, require_opening_quote=True, require_closing_quote=True
+                )
+                key_parser = key_parser.add_character('"')
+                self.root.context.active_parser.object_stack.append(key_parser)
+                self.current_stage = ObjectParsingStage.PARSING_KEY_VALUE_SEPARATOR
+        elif self.current_stage == ObjectParsingStage.PARSING_KEY_VALUE_SEPARATOR:
+            if new_character == ":":
+                self.current_stage = ObjectParsingStage.PARSING_VALUE
+                self.current_key = self.root.context.active_parser.last_parsed_string
+                self.existing_keys.append(self.current_key)
+                if self.is_dictionary:
+                    if self.schema_object.additionalProperties:
+                        value_schema = self.schema_object.additionalProperties
+                    else:
+                        value_schema = JsonSchemaParser.ANY_JSON_OBJECT_SCHEMA
+                else:
+                    value_schema = self.schema_object.properties[self.current_key]
+                self.current_key_parser = get_parser(
+                    self.root, value_schema
+                )
+                self.root.context.active_parser.object_stack.append(self.current_key_parser)
+                self.current_key_parser = None
+        elif self.current_stage == ObjectParsingStage.PARSING_VALUE:
+            # If we recieve a character during parsing value, it means that its the finishing character
+            # of the value parser
+            if new_character == '"':
+                self.current_stage = ObjectParsingStage.PARSING_SEPARATOR_OR_END
+            elif new_character == ",":
+                self.current_stage = ObjectParsingStage.PARSING_KEY_OR_END
+            elif new_character == "}":
+                self.current_stage = ObjectParsingStage.END_OBJECT
+        elif self.current_stage == ObjectParsingStage.PARSING_SEPARATOR_OR_END:
+            if new_character == ",":
+                self.current_stage = ObjectParsingStage.PARSING_KEY_OR_END
+            elif new_character == "}":
+                self.current_stage = ObjectParsingStage.END_OBJECT
+        return self
+
+    def get_allowed_characters(self) -> str:
+        possible_keys = (
+            list(self.schema_object.properties.keys())
+            if not self.is_dictionary
+            else None
+        )
+        required_keys = self.schema_object.required or []
+        can_end = set(self.existing_keys).issuperset(required_keys)
+        can_parse_key = self.is_dictionary or set(possible_keys).difference(
+            self.existing_keys
+        )
+
+        possible_characters = [c for c in WHITESPACE_CHARACTERS]
+        if self.current_stage == ObjectParsingStage.START_OBJECT:
+            possible_characters.append('{')
+        elif self.current_stage == ObjectParsingStage.PARSING_KEY_OR_END:
+            if can_end:
+                possible_characters.append('}')
+            if can_parse_key:
+                possible_characters.append('"')
+        elif self.current_stage == ObjectParsingStage.PARSING_KEY_VALUE_SEPARATOR:
+            possible_characters.append(':')
+        elif self.current_stage == ObjectParsingStage.PARSING_VALUE:
+            # Sometimes the value parser considers finishing, so it needs to know which continuations are possible
+            if can_end:
+                possible_characters.append('}')
+            if can_parse_key:
+                possible_characters.append(',')
+        elif self.current_stage == ObjectParsingStage.PARSING_SEPARATOR_OR_END:
+            if can_end:
+                possible_characters.append('}')
+            if can_parse_key:
+                possible_characters.append(',')
+        return "".join(possible_characters)
+    
+    def can_end(self) -> bool:
+        return self.current_stage == ObjectParsingStage.END_OBJECT
+
+
+class StringParsingStage:
+    START_TOKEN = "StartToken"
+    PARSING_STRING = "ParsingString"
+    END_TOKEN = "EndToken"
+
+
+class PrimitiveParsingState(BaseParsingState):
+    def __init__(self, root: JsonSchemaParser):
+        super().__init__(root)
+        self.stage = StringParsingStage.START_TOKEN
+        self.parsed_string = ""
+
+    def _clone(self) -> "PrimitiveParsingState":
+        raise NotImplementedError()
+    
+    def add_character(self, new_character: str) -> "PrimitiveParsingState":
+        new = self._clone()
+        new.parsed_string += new_character
+        return new
+
+    def can_end(self) -> bool:
+        return True
+
+
+class NumberParsingState(PrimitiveParsingState):
+    def __init__(
+        self,
+        root: JsonSchemaParser,
+        allow_floating_point: bool,
+    ):
+        super().__init__(root)
+        self.allow_floating_point = allow_floating_point
+        self.seen_decimal_point = False
+        self.seen_whitespace_after_digits = False
+        self.seen_exponent = False
+        self.seen_digit = False
+
+    def _clone(self) -> "NumberParsingState":
+        clone = NumberParsingState(self.root, self.allow_floating_point)
+        clone.parsed_string = self.parsed_string
+        clone.seen_decimal_point = self.seen_decimal_point
+        clone.seen_whitespace_after_digits = self.seen_whitespace_after_digits
+        clone.seen_exponent = self.seen_exponent
+        clone.seen_digit = self.seen_digit
+        return clone
+    
+    def add_character(self, new_character: str) -> CharacterLevelParser:
+        if not self.parsed_string and new_character in WHITESPACE_CHARACTERS:
+            return self
+        self = cast(NumberParsingState, super().add_character(new_character))
+        if new_character in WHITESPACE_CHARACTERS:
+            if self.parsed_string:
+                self.seen_whitespace_after_digits = True
+            return self
+        if new_character == ".":
+            if not self.parsed_string or len(self.parsed_string) == 1:
+                raise LMFormatEnforcerException("Numbers cannot start with a decimal point.")
+            if self.seen_decimal_point:
+                raise LMFormatEnforcerException("Numbers cannot contain more than two decimal points.")
+            self.seen_decimal_point = True
+        elif new_character in "eE":
+            if self.seen_exponent or not self.seen_digit:
+                raise LMFormatEnforcerException("Invalid number format")
+            self.seen_exponent = True
+        elif new_character.isdigit():
+            self.seen_digit = True
+        return self
+
+    def get_allowed_characters(self) -> str:
+        if self.seen_whitespace_after_digits:
+            return WHITESPACE_CHARACTERS
+        allowed_characters = "0123456789"
+        if not self.parsed_string:
+            allowed_characters += "-" + WHITESPACE_CHARACTERS
+        if self.parsed_string and len(self.parsed_string) == 1 and self.parsed_string[0] == "0":
+            allowed_characters = WHITESPACE_CHARACTERS
+        if self.parsed_string and len(self.parsed_string) == 2 and self.parsed_string == "-0":
+            allowed_characters = "." + WHITESPACE_CHARACTERS
+        if self.parsed_string and self.parsed_string[-1] in "eE":
+            allowed_characters += "-+"
+        if self.seen_digit and not self.seen_exponent:
+            allowed_characters += "eE"
+        if self.allow_floating_point and not self.seen_decimal_point and self.seen_digit and not self.seen_exponent:
+            allowed_characters += "."
+        if self.parsed_string and self.parsed_string[-1].isdigit():
+            allowed_characters += WHITESPACE_CHARACTERS
+        return allowed_characters
+
+    def can_end(self) -> bool:
+        if self.seen_exponent and self.parsed_string[-1] in "eE+-":
+            return False
+        return bool(self.parsed_string) and (self.parsed_string[-1].isdigit() or self.seen_whitespace_after_digits)
+
+
+class StringParsingState(PrimitiveParsingState):
+    allowed_strings: List[str]
+    parsed_string: str
+    seen_closing_quote: bool
+    seen_opening_quote: bool
+    min_length: Optional[int]
+    max_length: Optional[int]
+    pattern: Optional[str]
+    regex_parser: Optional[RegexParser]
+
+    def __init__(
+        self,
+        root: JsonSchemaParser,
+        allowed_strings: List[str],
+        require_opening_quote: bool,
+        require_closing_quote: bool = True,
+        min_length: Optional[int]=None,
+        max_length: Optional[int]=None,
+        pattern: Optional[str]=None,
+        regex_parser: Optional[RegexParser]=None,
+    ):
+        super().__init__(root)
+        self.allowed_strings = allowed_strings
+        self.seen_closing_quote = False
+        self.seen_opening_quote = not require_opening_quote
+        self.require_closing_quote = require_closing_quote
+        self.require_opening_quote = require_opening_quote
+        self.min_length = min_length
+        self.max_length = max_length
+        self.pattern = pattern
+        if self.pattern and (self.min_length or self.max_length):
+            raise LMFormatEnforcerException("String schema contains both a pattern and a min/max length, which is not currently supported")
+        self.regex_parser = regex_parser
+        if self.pattern and not regex_parser:
+            if self.pattern not in self.root.context.regex_parser_cache:
+                self.root.context.regex_parser_cache[self.pattern] = RegexParser(self.pattern, self.root.config)
+            self.regex_parser = self.root.context.regex_parser_cache[self.pattern]
+
+
+    def _clone(self) -> "StringParsingState":
+        clone = StringParsingState(
+            self.root,
+            self.allowed_strings,
+            self.require_opening_quote,
+            self.require_closing_quote,
+            self.min_length,
+            self.max_length,
+            self.pattern,
+            self.regex_parser
+        )
+        clone.parsed_string = self.parsed_string
+        clone.seen_closing_quote = self.seen_closing_quote
+        clone.seen_opening_quote = self.seen_opening_quote
+        return clone
+
+    def add_character(self, new_character: str):
+        if (not self.parsed_string or self.seen_closing_quote) and new_character in WHITESPACE_CHARACTERS:
+            return self
+        self = cast(StringParsingState, super().add_character(new_character))
+        if new_character == '"':
+            if not self.seen_opening_quote:
+                self.seen_opening_quote = True
+                self.parsed_string = ""
+            else:
+                self.seen_closing_quote = True
+                self.parsed_string = self.parsed_string[:-1]
+        if self.regex_parser and new_character != '"' and self.seen_opening_quote and not self.seen_closing_quote:
+            self.regex_parser = self.regex_parser.add_character(new_character)
+        if new_character == BACKSLASH:
+            # After a backslack we immediately have the escaping character, and if its 'u', we have 4 hex digits
+            escaping_character_parsers: List[CharacterLevelParser] = [StringParser(c) for c in BACKSLASH_ESCAPING_CHARACTERS]
+            hex_digit_parser: CharacterLevelParser = UnionParser([StringParser(c) for c in "0123456789abcdefABCDEF"])
+            unicode_components: List[CharacterLevelParser] = list([StringParser("u")] + [hex_digit_parser] * 4)
+            unicode_escape_parser: CharacterLevelParser = SequenceParser(unicode_components)
+            json_escaping_parser = UnionParser(escaping_character_parsers + [unicode_escape_parser])
+            self.root.context.active_parser.object_stack.append(json_escaping_parser)
+        return self
+
+    def get_allowed_characters(self) -> str:
+        if not self.seen_opening_quote:
+            return '"' + WHITESPACE_CHARACTERS
+        if self.seen_closing_quote:
+            return WHITESPACE_CHARACTERS
+        if self.regex_parser:
+            regex_chars = self.regex_parser.get_allowed_characters()
+            # We don't currently support regexes with quotes or escaping backslashes, so we remove them from the allowed characters
+            regex_chars = regex_chars.replace('"', '').replace(BACKSLASH, '')
+            if self.regex_parser.can_end():
+                regex_chars += '"'
+            return regex_chars
+        if self.allowed_strings:
+            allowed_continuations = [
+                s[len(self.parsed_string) :]
+                for s in self.allowed_strings
+                if s.startswith(self.parsed_string)
+            ]
+            allowed_next_characters = [allowed_continuation[0] for allowed_continuation in allowed_continuations if len(allowed_continuation) > 0]
+            allowed_next_characters = list(set(allowed_next_characters))
+            if self.parsed_string in self.allowed_strings and self.require_closing_quote:
+                allowed_next_characters.append('"')
+            if (not self.parsed_string) and (not self.seen_opening_quote or not self.require_opening_quote):
+                allowed_next_characters.extend(WHITESPACE_CHARACTERS)
+            return "".join(allowed_next_characters)
+        else:
+            if self.min_length is not None and len(self.parsed_string) < self.min_length:
+                return self.root.context.alphabet_without_quotes + BACKSLASH
+            if self.max_length is not None and len(self.parsed_string) >= self.max_length:
+                return '"'
+            return self.root.config.alphabet + BACKSLASH
+
+    def can_end(self) -> bool:
+        if self.require_closing_quote:
+            return self.seen_closing_quote
+        else:
+            if self.allowed_strings:
+                return self.parsed_string in self.allowed_strings
+            else:
+                return bool(self.parsed_string)
+
+
+class ListParsingState(PrimitiveParsingState):
+    list_member_type: JsonSchemaObject
+    seen_list_opener: bool = False
+    seen_list_closer: bool = False
+    num_items_seen: int = 0
+
+    def __init__(
+        self,
+        root: JsonSchemaParser,
+        list_member_type: JsonSchemaObject,
+        min_items: Optional[int],
+        max_items: Optional[int],
+    ):
+        super().__init__(root)
+        self.list_member_type = list_member_type
+        self.min_items = min_items
+        self.max_items = max_items
+        default_max = root.config.max_json_array_length
+        if self.max_items is None and default_max > 0 and (min_items is None or min_items < default_max):
+            self.max_items = default_max
+
+    def _clone(self) -> PrimitiveParsingState:
+        new = ListParsingState(self.root, self.list_member_type, self.min_items, self.max_items)
+        new.parsed_string = self.parsed_string
+        new.num_items_seen = self.num_items_seen
+        new.seen_list_opener = self.seen_list_opener
+        new.seen_list_closer = self.seen_list_closer
+        return new
+
+    def add_character(self, new_character: str) -> "ListParsingState":
+        self = cast(ListParsingState, super().add_character(new_character))
+        if new_character == "[":
+            self.seen_list_opener = True
+            item_parser = get_parser(self.root, self.list_member_type)
+            requires_items = self.min_items is not None and self.min_items > 0
+            if requires_items:
+                parser_to_push = item_parser
+            else:
+                # If we don't require items, we can also end immediately, the Union + ForceStopParser combination achieves this
+                empty_list_parser = ForceStopParser(allow_whitespace=True)
+                if isinstance(item_parser, UnionParser):
+                    item_parser.parsers.append(empty_list_parser)
+                    parser_to_push = item_parser
+                else:
+                    parser_to_push = UnionParser([item_parser, empty_list_parser])
+            self.root.context.active_parser.object_stack.append(parser_to_push)
+        elif new_character == "]":
+            self.seen_list_closer = True
+        elif new_character == ",":
+            if not self.seen_list_closer:
+                self.num_items_seen += 1
+                
+                self.root.context.active_parser.object_stack.append(
+                    get_parser(
+                        self.root,
+                        self.list_member_type,
+                    )
+                )
+        return self
+
+    def get_allowed_characters(self) -> str:
+        if not self.seen_list_opener:
+            return "[" + WHITESPACE_CHARACTERS
+        elif not self.seen_list_closer:
+            return self.get_allowed_control_characters() + WHITESPACE_CHARACTERS
+        else:
+            return ""
+
+    def can_end(self) -> bool:
+        return self.seen_list_closer
+    
+    def get_allowed_control_characters(self):
+        num_items = self.num_items_seen
+        top_parser = self.root.context.active_parser.object_stack[-1]
+        is_on_top = top_parser == self or isinstance(top_parser, UnionParser) and self in top_parser.parsers
+        if (not is_on_top) and self.root.context.active_parser.last_non_whitespace_character != "[":
+            # If there is an active parser above us, and the last character is not [, 
+            # there is an active item parser on the stack that we did not count yet.
+            num_items += 1
+        control_characters = ""
+        has_enough_items = self.min_items is None or num_items >= self.min_items
+        can_add_another_item = self.max_items is None or num_items < self.max_items
+
+        if num_items > 0 and can_add_another_item:
+            control_characters += ","
+        if has_enough_items:
+            control_characters += "]"
+        return control_characters
+    

deepseek/lib/python3.10/site-packages/lmformatenforcer/tokenenforcer.py

@@ -0,0 +1,166 @@
+from dataclasses import dataclass, field
+import sys
+from typing import Callable, Dict, Hashable, List, Optional, Tuple, Union
+import logging
+
+from .exceptions import LMFormatEnforcerException
+from .characterlevelparser import CharacterLevelParser, ForceStopParser, CharacterLevelParserConfig
+from .tokenizerprefixtree import TokenizerPrefixTree, TokenizerPrefixTreeNode
+
+
+class TokenEnforcerTokenizerData:
+    """TokenEnforcerTokenizerData contains all of the preprocessing for preparing the TokenEnforcer to work with a 
+    specific tokenizer. It does some calculations, so it is recommended to reuse it for multiple TokenEnforcers"""
+    def __init__(self, 
+                 regular_tokens: List[Tuple[int, str, bool]], 
+                 decoder: Callable[[List[int]], str],
+                 eos_token_id: Union[int, List[int]]):
+        """
+        Create the tokenizer data that the TokenEnforcer needs. This can be reused for multiple TokenEnforcers if they work with the same tokenizer.
+        :param regular_tokens: A list of tuples (token_id, token_string, is_new_word_token) for all the regular (not special) tokens in the tokenizer vocabulary.
+        Note that token_string is expected to include leading / trailing whitespaces if relevant.
+        :param decoder: A function that decodes a list of token ids into a string.
+        :param eos_token_id: The token id(s) of the end-of-string token(s).
+        """
+        self.regular_tokens = regular_tokens
+        self.tokenizer_tree = TokenizerPrefixTree(regular_tokens)
+        self.decoder = decoder
+        self.eos_token_id = eos_token_id
+        self.tokenizer_alphabet = "".join(token_str for token_str in self.tokenizer_tree.root.children.keys() if len(token_str) == 1)
+
+
+class TokenEnforcer:
+    """TokenEnforcer provides a token filtering mechanism, given a CharacterLevelParser and some information about the tokenizer.
+    It is the main entry point for extending lm-format-enforcer to new inference libraries. See __init__() and get_allowed_tokens()"""
+    @dataclass
+    class OutputTensorState:
+        parser: CharacterLevelParser
+        allowed_tokens: List[int] = field(default_factory=list)
+        current_word_tokens: List[int] = field(default_factory=list)
+
+    def __init__(self, tokenizer_data: TokenEnforcerTokenizerData, parser: CharacterLevelParser):
+        """
+        Create a new TokenEnforcer.
+        :param tokenizer_data: Per tokenizer data that the token enforcer needs in order to operate.
+        :param parser: A CharacterLevelParser that defines the allowed strings.
+        """
+        self.prefix_states: Dict[Tuple, TokenEnforcer.OutputTensorState] = {}
+        self.root_parser = parser
+        self.tokenizer_tree = tokenizer_data.tokenizer_tree
+        self.decoder = tokenizer_data.decoder
+        self.eos_token_id = tokenizer_data.eos_token_id
+        self.regular_tokens = tokenizer_data.regular_tokens
+        self.allowed_token_cache: Dict[Hashable, List[int]] = {}
+        
+        config = CharacterLevelParserConfig(alphabet=tokenizer_data.tokenizer_alphabet)
+        parser.config = config
+
+    def get_allowed_tokens(self, token_sequence: List[int]) -> List[int]:
+        """
+        Get a list of allowed tokens, given a list of tokens that were already generated.
+        :param token_sequence: The tokens that were already generated, and the next token will be generated for.
+        :return: A list of token ids that are allowed to be selected next.
+        """
+        # In order to elegantly support beam search and batching, we don't store per-batch information.
+        # Instead, we store a hash of all the states (unique token tensors) we encountered so far.
+        # When we encounter a new unique token tensor, we find the token tensor that led to it, and continue from there.
+        sent_tuple = tuple(token_sequence)
+        prev_step_tuple = sent_tuple[:-1]
+
+        if sent_tuple in self.prefix_states:
+            # We already calculated for this node, return cached list
+            return self.prefix_states[sent_tuple].allowed_tokens
+        elif prev_step_tuple not in self.prefix_states:
+            # We have not encountered the tensor up to the before-last entry. This means that this is the first call - the instruction / prompt tensor.
+            # Initialize the root node
+            state = TokenEnforcer.OutputTensorState(parser=self.root_parser)
+            self.prefix_states[sent_tuple] = state
+            self._compute_allowed_tokens(sent_tuple, state)
+            return state.allowed_tokens
+        else:
+            # Find the state that led to this node. We explicitly don't use the concept of "timestep" because of beam search        
+            prev_step_state = self.prefix_states[prev_step_tuple]
+            new_state = self._apply_new_characters(prev_step_state, token_sequence)
+            self.prefix_states[sent_tuple] = new_state
+            self._compute_allowed_tokens(sent_tuple, new_state)
+            return new_state.allowed_tokens
+
+    def _compute_allowed_tokens(self, state_tokens: Tuple, state: 'TokenEnforcer.OutputTensorState'):
+        try:
+            allowed_tokens: List[int] = []
+            cache_key = state.parser.cache_key()
+            if cache_key is not None and cache_key in self.allowed_token_cache:
+                state.allowed_tokens = self.allowed_token_cache[cache_key]
+                return
+            shortcut_key = state.parser.shortcut_key()
+            self._collect_allowed_tokens(state.parser, self.tokenizer_tree.root, allowed_tokens, shortcut_key)
+            if state.parser.can_end():
+                allowed_tokens.extend(self.eos_token_id if isinstance(self.eos_token_id, list) else [self.eos_token_id])
+            if not allowed_tokens:
+                raise ValueError(f"Parser reached state with no allowed tokens")
+            # root_state = next(state for state in self.prefix_states.values() if state.parser == self.root_parser)
+            # print(f"Allowing {len(allowed_tokens)} tokens after {state.str_so_far[len(root_state.str_so_far):]}")
+            state.allowed_tokens = allowed_tokens
+            if cache_key is not None:
+                self.allowed_token_cache[cache_key] = allowed_tokens
+        except LMFormatEnforcerException:
+            # Getting an LMFormatEnforcerException means that we know what the user did wrong, 
+            # and we can give a nice error message for them to fix.
+            raise
+        except Exception:
+            # Other exceptions are potential bugs and should be reported
+            logging.basicConfig(level=logging.ERROR)  # Initialize if no loggers
+            prefix = self.decoder(list(state_tokens))
+            logging.exception(f"Unknown LMFormatEnforcer Problem. Prefix: '{prefix}'\n"
+                              "Terminating the parser. Please open an issue at \n"
+                              "https://github.com/noamgat/lm-format-enforcer/issues with the prefix and "
+                              "CharacterLevelParser parameters")
+            state.allowed_tokens = self.eos_token_id if isinstance(self.eos_token_id, list) else [self.eos_token_id]
+
+    def _collect_allowed_tokens(self, parser: CharacterLevelParser, tree_node: TokenizerPrefixTreeNode, allowed_tokens: List[int], shortcut_key: Optional[Hashable]):
+        allowed_tokens.extend(tree_node.tokens)
+        allowed_characters = parser.get_allowed_characters()
+        relevant_characters = tree_node.children.keys()
+        # This next line is the heart of the traversal algorithm. We only explore paths that are shared by both the parser and the tokenizer.
+        characters_to_explore = set(relevant_characters).intersection(allowed_characters)
+        
+        # Performance optimization: If we are in JSON freetext, all of the tokens that don't contain quote, or end with quote, are legal, so we take
+        # their cached list. If the quote character is allowed, we only need to dynamically explore the cases where the string starts with a quote.
+        # This breaks the elegance of the API, but otherwise it is a huge performance hit.
+        if isinstance(shortcut_key, tuple) and shortcut_key[0] == 'json_freetext':
+            assert len(shortcut_key) == 4
+            _, cur_len, min_len, max_len = shortcut_key
+            cache = self.tokenizer_tree.json_freetext_tokens
+
+            min_remaining = min(cache.max_token_len, max(0, min_len - cur_len))  # no " allowed before this many chars
+            max_allowed_len = min(cache.max_token_len, max_len - cur_len)  # max new characters allowed (before ")
+
+            allowed_tokens.extend(cache.lookup_allowed_tokens(min_remaining, max_allowed_len))
+            characters_to_explore = characters_to_explore.intersection(['"'])
+
+        for character in characters_to_explore:
+            next_parser = parser.add_character(character)
+            next_tree_node = tree_node.children[character]
+            self._collect_allowed_tokens(next_parser, next_tree_node, allowed_tokens, None)
+            
+    def _apply_new_characters(self, state: 'TokenEnforcer.OutputTensorState', token_sequence: List[int]):
+        new_state = TokenEnforcer.OutputTensorState(parser=state.parser)
+        new_token = token_sequence[-1]
+        if new_token in self.tokenizer_tree.new_word_tokens:
+            new_state.current_word_tokens = [new_token]
+            new_characters = self.tokenizer_tree.tokens_to_strs[new_token]
+        else:
+            new_state.current_word_tokens = state.current_word_tokens + [new_token]
+            prev_decoded = self.decoder(state.current_word_tokens)
+            new_decoded = self.decoder(new_state.current_word_tokens)
+            new_characters = new_decoded[len(prev_decoded):]
+        for character in new_characters:
+            try:
+                new_state.parser = new_state.parser.add_character(character)
+            except Exception as e:
+                # This can happen in beam / batch scenarios, when some of the batches finished but others are continuing.
+                logging.debug(f"Received an invalid character '{character}', switching to ForceStopParser (Exception:{e})")
+                new_state.parser = ForceStopParser()
+        return new_state
+        
+

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/chaos_pendulum.al

@@ -0,0 +1,33 @@
+CONSTANTS G,LB,W,H
+MOTIONVARIABLES' THETA'',PHI'',OMEGA',ALPHA'
+NEWTONIAN N
+BODIES A,B
+SIMPROT(N,A,2,THETA)
+SIMPROT(A,B,3,PHI)
+POINT O
+LA = (LB-H/2)/2
+P_O_AO> = LA*A3>
+P_O_BO> = LB*A3>
+OMEGA = THETA'
+ALPHA = PHI'
+W_A_N> = OMEGA*N2>
+W_B_A> = ALPHA*A3>
+V_O_N> = 0>
+V2PTS(N, A, O, AO)
+V2PTS(N, A, O, BO)
+MASS A=MA, B=MB
+IAXX = 1/12*MA*(2*LA)^2
+IAYY = IAXX
+IAZZ = 0
+IBXX = 1/12*MB*H^2
+IBYY = 1/12*MB*(W^2+H^2)
+IBZZ = 1/12*MB*W^2
+INERTIA A, IAXX, IAYY, IAZZ
+INERTIA B, IBXX, IBYY, IBZZ
+GRAVITY(G*N3>)
+ZERO = FR() + FRSTAR()
+KANE()
+INPUT LB=0.2,H=0.1,W=0.2,MA=0.01,MB=0.1,G=9.81
+INPUT THETA = 90 DEG, PHI = 0.5 DEG, OMEGA=0, ALPHA=0
+INPUT TFINAL=10, INTEGSTP=0.02
+CODE DYNAMICS() some_filename.c

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/chaos_pendulum.py

@@ -0,0 +1,55 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+g, lb, w, h = _sm.symbols('g lb w h', real=True)
+theta, phi, omega, alpha = _me.dynamicsymbols('theta phi omega alpha')
+theta_d, phi_d, omega_d, alpha_d = _me.dynamicsymbols('theta_ phi_ omega_ alpha_', 1)
+theta_dd, phi_dd = _me.dynamicsymbols('theta_ phi_', 2)
+frame_n = _me.ReferenceFrame('n')
+body_a_cm = _me.Point('a_cm')
+body_a_cm.set_vel(frame_n, 0)
+body_a_f = _me.ReferenceFrame('a_f')
+body_a = _me.RigidBody('a', body_a_cm, body_a_f, _sm.symbols('m'), (_me.outer(body_a_f.x,body_a_f.x),body_a_cm))
+body_b_cm = _me.Point('b_cm')
+body_b_cm.set_vel(frame_n, 0)
+body_b_f = _me.ReferenceFrame('b_f')
+body_b = _me.RigidBody('b', body_b_cm, body_b_f, _sm.symbols('m'), (_me.outer(body_b_f.x,body_b_f.x),body_b_cm))
+body_a_f.orient(frame_n, 'Axis', [theta, frame_n.y])
+body_b_f.orient(body_a_f, 'Axis', [phi, body_a_f.z])
+point_o = _me.Point('o')
+la = (lb-h/2)/2
+body_a_cm.set_pos(point_o, la*body_a_f.z)
+body_b_cm.set_pos(point_o, lb*body_a_f.z)
+body_a_f.set_ang_vel(frame_n, omega*frame_n.y)
+body_b_f.set_ang_vel(body_a_f, alpha*body_a_f.z)
+point_o.set_vel(frame_n, 0)
+body_a_cm.v2pt_theory(point_o,frame_n,body_a_f)
+body_b_cm.v2pt_theory(point_o,frame_n,body_a_f)
+ma = _sm.symbols('ma')
+body_a.mass = ma
+mb = _sm.symbols('mb')
+body_b.mass = mb
+iaxx = 1/12*ma*(2*la)**2
+iayy = iaxx
+iazz = 0
+ibxx = 1/12*mb*h**2
+ibyy = 1/12*mb*(w**2+h**2)
+ibzz = 1/12*mb*w**2
+body_a.inertia = (_me.inertia(body_a_f, iaxx, iayy, iazz, 0, 0, 0), body_a_cm)
+body_b.inertia = (_me.inertia(body_b_f, ibxx, ibyy, ibzz, 0, 0, 0), body_b_cm)
+force_a = body_a.mass*(g*frame_n.z)
+force_b = body_b.mass*(g*frame_n.z)
+kd_eqs = [theta_d - omega, phi_d - alpha]
+forceList = [(body_a.masscenter,body_a.mass*(g*frame_n.z)), (body_b.masscenter,body_b.mass*(g*frame_n.z))]
+kane = _me.KanesMethod(frame_n, q_ind=[theta,phi], u_ind=[omega, alpha], kd_eqs = kd_eqs)
+fr, frstar = kane.kanes_equations([body_a, body_b], forceList)
+zero = fr+frstar
+from pydy.system import System
+sys = System(kane, constants = {g:9.81, lb:0.2, w:0.2, h:0.1, ma:0.01, mb:0.1},
+specifieds={},
+initial_conditions={theta:_np.deg2rad(90), phi:_np.deg2rad(0.5), omega:0, alpha:0},
+times = _np.linspace(0.0, 10, 10/0.02))
+
+y=sys.integrate()

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/double_pendulum.al

@@ -0,0 +1,25 @@
+MOTIONVARIABLES' Q{2}', U{2}'
+CONSTANTS L,M,G
+NEWTONIAN N
+FRAMES A,B
+SIMPROT(N, A, 3, Q1)
+SIMPROT(N, B, 3, Q2)
+W_A_N>=U1*N3>
+W_B_N>=U2*N3>
+POINT O
+PARTICLES P,R
+P_O_P> = L*A1>
+P_P_R> = L*B1>
+V_O_N> = 0>
+V2PTS(N, A, O, P)
+V2PTS(N, B, P, R)
+MASS P=M, R=M
+Q1' = U1
+Q2' = U2
+GRAVITY(G*N1>)
+ZERO = FR() + FRSTAR()
+KANE()
+INPUT M=1,G=9.81,L=1
+INPUT Q1=.1,Q2=.2,U1=0,U2=0
+INPUT TFINAL=10, INTEGSTP=.01
+CODE DYNAMICS() some_filename.c

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/double_pendulum.py

@@ -0,0 +1,39 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+q1, q2, u1, u2 = _me.dynamicsymbols('q1 q2 u1 u2')
+q1_d, q2_d, u1_d, u2_d = _me.dynamicsymbols('q1_ q2_ u1_ u2_', 1)
+l, m, g = _sm.symbols('l m g', real=True)
+frame_n = _me.ReferenceFrame('n')
+frame_a = _me.ReferenceFrame('a')
+frame_b = _me.ReferenceFrame('b')
+frame_a.orient(frame_n, 'Axis', [q1, frame_n.z])
+frame_b.orient(frame_n, 'Axis', [q2, frame_n.z])
+frame_a.set_ang_vel(frame_n, u1*frame_n.z)
+frame_b.set_ang_vel(frame_n, u2*frame_n.z)
+point_o = _me.Point('o')
+particle_p = _me.Particle('p', _me.Point('p_pt'), _sm.Symbol('m'))
+particle_r = _me.Particle('r', _me.Point('r_pt'), _sm.Symbol('m'))
+particle_p.point.set_pos(point_o, l*frame_a.x)
+particle_r.point.set_pos(particle_p.point, l*frame_b.x)
+point_o.set_vel(frame_n, 0)
+particle_p.point.v2pt_theory(point_o,frame_n,frame_a)
+particle_r.point.v2pt_theory(particle_p.point,frame_n,frame_b)
+particle_p.mass = m
+particle_r.mass = m
+force_p = particle_p.mass*(g*frame_n.x)
+force_r = particle_r.mass*(g*frame_n.x)
+kd_eqs = [q1_d - u1, q2_d - u2]
+forceList = [(particle_p.point,particle_p.mass*(g*frame_n.x)), (particle_r.point,particle_r.mass*(g*frame_n.x))]
+kane = _me.KanesMethod(frame_n, q_ind=[q1,q2], u_ind=[u1, u2], kd_eqs = kd_eqs)
+fr, frstar = kane.kanes_equations([particle_p, particle_r], forceList)
+zero = fr+frstar
+from pydy.system import System
+sys = System(kane, constants = {l:1, m:1, g:9.81},
+specifieds={},
+initial_conditions={q1:.1, q2:.2, u1:0, u2:0},
+times = _np.linspace(0.0, 10, 10/.01))
+
+y=sys.integrate()

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/mass_spring_damper.al

@@ -0,0 +1,19 @@
+CONSTANTS M,K,B,G
+MOTIONVARIABLES' POSITION',SPEED'
+VARIABLES O
+FORCE = O*SIN(T)
+NEWTONIAN CEILING
+POINTS ORIGIN
+V_ORIGIN_CEILING> = 0>
+PARTICLES BLOCK
+P_ORIGIN_BLOCK> = POSITION*CEILING1>
+MASS BLOCK=M
+V_BLOCK_CEILING>=SPEED*CEILING1>
+POSITION' = SPEED
+FORCE_MAGNITUDE = M*G-K*POSITION-B*SPEED+FORCE
+FORCE_BLOCK>=EXPLICIT(FORCE_MAGNITUDE*CEILING1>)
+ZERO = FR() + FRSTAR()
+KANE()
+INPUT TFINAL=10.0, INTEGSTP=0.01
+INPUT M=1.0, K=1.0, B=0.2, G=9.8, POSITION=0.1, SPEED=-1.0, O=2
+CODE DYNAMICS() dummy_file.c

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/mass_spring_damper.py

@@ -0,0 +1,31 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+m, k, b, g = _sm.symbols('m k b g', real=True)
+position, speed = _me.dynamicsymbols('position speed')
+position_d, speed_d = _me.dynamicsymbols('position_ speed_', 1)
+o = _me.dynamicsymbols('o')
+force = o*_sm.sin(_me.dynamicsymbols._t)
+frame_ceiling = _me.ReferenceFrame('ceiling')
+point_origin = _me.Point('origin')
+point_origin.set_vel(frame_ceiling, 0)
+particle_block = _me.Particle('block', _me.Point('block_pt'), _sm.Symbol('m'))
+particle_block.point.set_pos(point_origin, position*frame_ceiling.x)
+particle_block.mass = m
+particle_block.point.set_vel(frame_ceiling, speed*frame_ceiling.x)
+force_magnitude = m*g-k*position-b*speed+force
+force_block = (force_magnitude*frame_ceiling.x).subs({position_d:speed})
+kd_eqs = [position_d - speed]
+forceList = [(particle_block.point,(force_magnitude*frame_ceiling.x).subs({position_d:speed}))]
+kane = _me.KanesMethod(frame_ceiling, q_ind=[position], u_ind=[speed], kd_eqs = kd_eqs)
+fr, frstar = kane.kanes_equations([particle_block], forceList)
+zero = fr+frstar
+from pydy.system import System
+sys = System(kane, constants = {m:1.0, k:1.0, b:0.2, g:9.8},
+specifieds={_me.dynamicsymbols('t'):lambda x, t: t, o:2},
+initial_conditions={position:0.1, speed:-1*1.0},
+times = _np.linspace(0.0, 10.0, 10.0/0.01))
+
+y=sys.integrate()

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/non_min_pendulum.al

@@ -0,0 +1,20 @@
+MOTIONVARIABLES' Q{2}''
+CONSTANTS L,M,G
+NEWTONIAN N
+POINT PN
+V_PN_N> = 0>
+THETA1 = ATAN(Q2/Q1)
+FRAMES A
+SIMPROT(N, A, 3, THETA1)
+PARTICLES P
+P_PN_P> = Q1*N1>+Q2*N2>
+MASS P=M
+V_P_N>=DT(P_P_PN>, N)
+F_V = DOT(EXPRESS(V_P_N>,A), A1>)
+GRAVITY(G*N1>)
+DEPENDENT[1] = F_V
+CONSTRAIN(DEPENDENT[Q1'])
+ZERO=FR()+FRSTAR()
+F_C = MAG(P_P_PN>)-L
+CONFIG[1]=F_C
+ZERO[2]=CONFIG[1]

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/pydy-example-repo/non_min_pendulum.py

@@ -0,0 +1,36 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+q1, q2 = _me.dynamicsymbols('q1 q2')
+q1_d, q2_d = _me.dynamicsymbols('q1_ q2_', 1)
+q1_dd, q2_dd = _me.dynamicsymbols('q1_ q2_', 2)
+l, m, g = _sm.symbols('l m g', real=True)
+frame_n = _me.ReferenceFrame('n')
+point_pn = _me.Point('pn')
+point_pn.set_vel(frame_n, 0)
+theta1 = _sm.atan(q2/q1)
+frame_a = _me.ReferenceFrame('a')
+frame_a.orient(frame_n, 'Axis', [theta1, frame_n.z])
+particle_p = _me.Particle('p', _me.Point('p_pt'), _sm.Symbol('m'))
+particle_p.point.set_pos(point_pn, q1*frame_n.x+q2*frame_n.y)
+particle_p.mass = m
+particle_p.point.set_vel(frame_n, (point_pn.pos_from(particle_p.point)).dt(frame_n))
+f_v = _me.dot((particle_p.point.vel(frame_n)).express(frame_a), frame_a.x)
+force_p = particle_p.mass*(g*frame_n.x)
+dependent = _sm.Matrix([[0]])
+dependent[0] = f_v
+velocity_constraints = [i for i in dependent]
+u_q1_d = _me.dynamicsymbols('u_q1_d')
+u_q2_d = _me.dynamicsymbols('u_q2_d')
+kd_eqs = [q1_d-u_q1_d, q2_d-u_q2_d]
+forceList = [(particle_p.point,particle_p.mass*(g*frame_n.x))]
+kane = _me.KanesMethod(frame_n, q_ind=[q1,q2], u_ind=[u_q2_d], u_dependent=[u_q1_d], kd_eqs = kd_eqs, velocity_constraints = velocity_constraints)
+fr, frstar = kane.kanes_equations([particle_p], forceList)
+zero = fr+frstar
+f_c = point_pn.pos_from(particle_p.point).magnitude()-l
+config = _sm.Matrix([[0]])
+config[0] = f_c
+zero = zero.row_insert(zero.shape[0], _sm.Matrix([[0]]))
+zero[zero.shape[0]-1] = config[0]

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest1.al

@@ -0,0 +1,8 @@
+% ruletest1.al
+CONSTANTS F = 3, G = 9.81
+CONSTANTS A, B
+CONSTANTS S, S1, S2+, S3+, S4-
+CONSTANTS K{4}, L{1:3}, P{1:2,1:3}
+CONSTANTS C{2,3}
+E1 = A*F + S2 - G
+E2 = F^2 + K3*K2*G

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest1.py

@@ -0,0 +1,15 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+f = _sm.S(3)
+g = _sm.S(9.81)
+a, b = _sm.symbols('a b', real=True)
+s, s1 = _sm.symbols('s s1', real=True)
+s2, s3 = _sm.symbols('s2 s3', real=True, nonnegative=True)
+s4 = _sm.symbols('s4', real=True, nonpositive=True)
+k1, k2, k3, k4, l1, l2, l3, p11, p12, p13, p21, p22, p23 = _sm.symbols('k1 k2 k3 k4 l1 l2 l3 p11 p12 p13 p21 p22 p23', real=True)
+c11, c12, c13, c21, c22, c23 = _sm.symbols('c11 c12 c13 c21 c22 c23', real=True)
+e1 = a*f+s2-g
+e2 = f**2+k3*k2*g

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest10.py

@@ -0,0 +1,64 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+x, y = _me.dynamicsymbols('x y')
+a, b = _sm.symbols('a b', real=True)
+e = a*(b*x+y)**2
+m = _sm.Matrix([e,e]).reshape(2, 1)
+e = e.expand()
+m = _sm.Matrix([i.expand() for i in m]).reshape((m).shape[0], (m).shape[1])
+e = _sm.factor(e, x)
+m = _sm.Matrix([_sm.factor(i,x) for i in m]).reshape((m).shape[0], (m).shape[1])
+eqn = _sm.Matrix([[0]])
+eqn[0] = a*x+b*y
+eqn = eqn.row_insert(eqn.shape[0], _sm.Matrix([[0]]))
+eqn[eqn.shape[0]-1] = 2*a*x-3*b*y
+print(_sm.solve(eqn,x,y))
+rhs_y = _sm.solve(eqn,x,y)[y]
+e = (x+y)**2+2*x**2
+e.collect(x)
+a, b, c = _sm.symbols('a b c', real=True)
+m = _sm.Matrix([a,b,c,0]).reshape(2, 2)
+m2 = _sm.Matrix([i.subs({a:1,b:2,c:3}) for i in m]).reshape((m).shape[0], (m).shape[1])
+eigvalue = _sm.Matrix([i.evalf() for i in (m2).eigenvals().keys()])
+eigvec = _sm.Matrix([i[2][0].evalf() for i in (m2).eigenvects()]).reshape(m2.shape[0], m2.shape[1])
+frame_n = _me.ReferenceFrame('n')
+frame_a = _me.ReferenceFrame('a')
+frame_a.orient(frame_n, 'Axis', [x, frame_n.x])
+frame_a.orient(frame_n, 'Axis', [_sm.pi/2, frame_n.x])
+c1, c2, c3 = _sm.symbols('c1 c2 c3', real=True)
+v = c1*frame_a.x+c2*frame_a.y+c3*frame_a.z
+point_o = _me.Point('o')
+point_p = _me.Point('p')
+point_o.set_pos(point_p, c1*frame_a.x)
+v = (v).express(frame_n)
+point_o.set_pos(point_p, (point_o.pos_from(point_p)).express(frame_n))
+frame_a.set_ang_vel(frame_n, c3*frame_a.z)
+print(frame_n.ang_vel_in(frame_a))
+point_p.v2pt_theory(point_o,frame_n,frame_a)
+particle_p1 = _me.Particle('p1', _me.Point('p1_pt'), _sm.Symbol('m'))
+particle_p2 = _me.Particle('p2', _me.Point('p2_pt'), _sm.Symbol('m'))
+particle_p2.point.v2pt_theory(particle_p1.point,frame_n,frame_a)
+point_p.a2pt_theory(particle_p1.point,frame_n,frame_a)
+body_b1_cm = _me.Point('b1_cm')
+body_b1_cm.set_vel(frame_n, 0)
+body_b1_f = _me.ReferenceFrame('b1_f')
+body_b1 = _me.RigidBody('b1', body_b1_cm, body_b1_f, _sm.symbols('m'), (_me.outer(body_b1_f.x,body_b1_f.x),body_b1_cm))
+body_b2_cm = _me.Point('b2_cm')
+body_b2_cm.set_vel(frame_n, 0)
+body_b2_f = _me.ReferenceFrame('b2_f')
+body_b2 = _me.RigidBody('b2', body_b2_cm, body_b2_f, _sm.symbols('m'), (_me.outer(body_b2_f.x,body_b2_f.x),body_b2_cm))
+g = _sm.symbols('g', real=True)
+force_p1 = particle_p1.mass*(g*frame_n.x)
+force_p2 = particle_p2.mass*(g*frame_n.x)
+force_b1 = body_b1.mass*(g*frame_n.x)
+force_b2 = body_b2.mass*(g*frame_n.x)
+z = _me.dynamicsymbols('z')
+v = x*frame_a.x+y*frame_a.z
+point_o.set_pos(point_p, x*frame_a.x+y*frame_a.y)
+v = (v).subs({x:2*z, y:z})
+point_o.set_pos(point_p, (point_o.pos_from(point_p)).subs({x:2*z, y:z}))
+force_o = -1*(x*y*frame_a.x)
+force_p1 = particle_p1.mass*(g*frame_n.x)+ x*y*frame_a.x

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest11.al

@@ -0,0 +1,6 @@
+VARIABLES X, Y
+CONSTANTS A{1:2, 1:2}, B{1:2}
+EQN[1] = A11*x + A12*y - B1
+EQN[2] = A21*x + A22*y - B2
+INPUT A11=2, A12=5, A21=3, A22=4, B1=7, B2=6
+CODE ALGEBRAIC(EQN, X, Y) some_filename.c

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest2.py

@@ -0,0 +1,22 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+x1, x2 = _me.dynamicsymbols('x1 x2')
+f1 = x1*x2+3*x1**2
+f2 = x1*_me.dynamicsymbols._t+x2*_me.dynamicsymbols._t**2
+x, y = _me.dynamicsymbols('x y')
+x_d, y_d = _me.dynamicsymbols('x_ y_', 1)
+y_dd = _me.dynamicsymbols('y_', 2)
+q1, q2, q3, u1, u2 = _me.dynamicsymbols('q1 q2 q3 u1 u2')
+p1, p2 = _me.dynamicsymbols('p1 p2')
+p1_d, p2_d = _me.dynamicsymbols('p1_ p2_', 1)
+w1, w2, w3, r1, r2 = _me.dynamicsymbols('w1 w2 w3 r1 r2')
+w1_d, w2_d, w3_d, r1_d, r2_d = _me.dynamicsymbols('w1_ w2_ w3_ r1_ r2_', 1)
+r1_dd, r2_dd = _me.dynamicsymbols('r1_ r2_', 2)
+c11, c12, c21, c22 = _me.dynamicsymbols('c11 c12 c21 c22')
+d11, d12, d13 = _me.dynamicsymbols('d11 d12 d13')
+j1, j2 = _me.dynamicsymbols('j1 j2')
+n = _sm.symbols('n')
+n = _sm.I

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest3.py

@@ -0,0 +1,37 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+frame_a = _me.ReferenceFrame('a')
+frame_b = _me.ReferenceFrame('b')
+frame_n = _me.ReferenceFrame('n')
+x1, x2, x3 = _me.dynamicsymbols('x1 x2 x3')
+l = _sm.symbols('l', real=True)
+v1 = x1*frame_a.x+x2*frame_a.y+x3*frame_a.z
+v2 = x1*frame_b.x+x2*frame_b.y+x3*frame_b.z
+v3 = x1*frame_n.x+x2*frame_n.y+x3*frame_n.z
+v = v1+v2+v3
+point_c = _me.Point('c')
+point_d = _me.Point('d')
+point_po1 = _me.Point('po1')
+point_po2 = _me.Point('po2')
+point_po3 = _me.Point('po3')
+particle_l = _me.Particle('l', _me.Point('l_pt'), _sm.Symbol('m'))
+particle_p1 = _me.Particle('p1', _me.Point('p1_pt'), _sm.Symbol('m'))
+particle_p2 = _me.Particle('p2', _me.Point('p2_pt'), _sm.Symbol('m'))
+particle_p3 = _me.Particle('p3', _me.Point('p3_pt'), _sm.Symbol('m'))
+body_s_cm = _me.Point('s_cm')
+body_s_cm.set_vel(frame_n, 0)
+body_s_f = _me.ReferenceFrame('s_f')
+body_s = _me.RigidBody('s', body_s_cm, body_s_f, _sm.symbols('m'), (_me.outer(body_s_f.x,body_s_f.x),body_s_cm))
+body_r1_cm = _me.Point('r1_cm')
+body_r1_cm.set_vel(frame_n, 0)
+body_r1_f = _me.ReferenceFrame('r1_f')
+body_r1 = _me.RigidBody('r1', body_r1_cm, body_r1_f, _sm.symbols('m'), (_me.outer(body_r1_f.x,body_r1_f.x),body_r1_cm))
+body_r2_cm = _me.Point('r2_cm')
+body_r2_cm.set_vel(frame_n, 0)
+body_r2_f = _me.ReferenceFrame('r2_f')
+body_r2 = _me.RigidBody('r2', body_r2_cm, body_r2_f, _sm.symbols('m'), (_me.outer(body_r2_f.x,body_r2_f.x),body_r2_cm))
+v4 = x1*body_s_f.x+x2*body_s_f.y+x3*body_s_f.z
+body_s_cm.set_pos(point_c, l*frame_n.x)

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest5.al

@@ -0,0 +1,32 @@
+% ruletest5.al
+VARIABLES X', Y'
+
+E1 = (X+Y)^2 + (X-Y)^3
+E2 = (X-Y)^2
+E3 = X^2 + Y^2 + 2*X*Y
+
+M1 = [E1;E2]
+M2 = [(X+Y)^2,(X-Y)^2]
+M3 = M1 + [X;Y]
+
+AM = EXPAND(M1)
+CM = EXPAND([(X+Y)^2,(X-Y)^2])
+EM = EXPAND(M1 + [X;Y])
+F = EXPAND(E1)
+G = EXPAND(E2)
+
+A = FACTOR(E3, X)
+BM = FACTOR(M1, X)
+CM = FACTOR(M1 + [X;Y], X)
+
+A = D(E3, X)
+B = D(E3, Y)
+CM = D(M2, X)
+DM = D(M1 + [X;Y], X)
+FRAMES A, B
+A_B = [1,0,0;1,0,0;1,0,0]
+V1> = X*A1> + Y*A2> + X*Y*A3>
+E> = D(V1>, X, B)
+FM = DT(M1)
+GM = DT([(X+Y)^2,(X-Y)^2])
+H> = DT(V1>, B)

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest6.al

@@ -0,0 +1,41 @@
+% ruletest6.al
+VARIABLES Q{2}
+VARIABLES X,Y,Z
+Q1 = X^2 + Y^2
+Q2 = X-Y
+E = Q1 + Q2
+A = EXPLICIT(E)
+E2 = COS(X)
+E3 = COS(X*Y)
+A = TAYLOR(E2, 0:2, X=0)
+B = TAYLOR(E3, 0:2, X=0, Y=0)
+
+E = EXPAND((X+Y)^2)
+A = EVALUATE(E, X=1, Y=Z)
+BM = EVALUATE([E;2*E], X=1, Y=Z)
+
+E = Q1 + Q2
+A = EVALUATE(E, X=2, Y=Z^2)
+
+CONSTANTS J,K,L
+P1 = POLYNOMIAL([J,K,L],X)
+P2 = POLYNOMIAL(J*X+K,X,1)
+
+ROOT1 = ROOTS(P1, X, 2)
+ROOT2 = ROOTS([1;2;3])
+
+M = [1,2,3,4;5,6,7,8;9,10,11,12;13,14,15,16]
+
+AM = TRANSPOSE(M) + M
+BM = EIG(M)
+C1 = DIAGMAT(4, 1)
+C2 = DIAGMAT(3, 4, 2)
+DM = INV(M+C1)
+E = DET(M+C1) + TRACE([1,0;0,1])
+F = ELEMENT(M, 2, 3)
+
+A = COLS(M)
+BM = COLS(M, 1)
+CM = COLS(M, 1, 2:4, 3)
+DM = ROWS(M, 1)
+EM = ROWS(M, 1, 2:4, 3)

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest6.py

@@ -0,0 +1,36 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+q1, q2 = _me.dynamicsymbols('q1 q2')
+x, y, z = _me.dynamicsymbols('x y z')
+e = q1+q2
+a = (e).subs({q1:x**2+y**2, q2:x-y})
+e2 = _sm.cos(x)
+e3 = _sm.cos(x*y)
+a = (e2).series(x, 0, 2).removeO()
+b = (e3).series(x, 0, 2).removeO().series(y, 0, 2).removeO()
+e = ((x+y)**2).expand()
+a = (e).subs({q1:x**2+y**2,q2:x-y}).subs({x:1,y:z})
+bm = _sm.Matrix([i.subs({x:1,y:z}) for i in _sm.Matrix([e,2*e]).reshape(2, 1)]).reshape((_sm.Matrix([e,2*e]).reshape(2, 1)).shape[0], (_sm.Matrix([e,2*e]).reshape(2, 1)).shape[1])
+e = q1+q2
+a = (e).subs({q1:x**2+y**2,q2:x-y}).subs({x:2,y:z**2})
+j, k, l = _sm.symbols('j k l', real=True)
+p1 = _sm.Poly(_sm.Matrix([j,k,l]).reshape(1, 3), x)
+p2 = _sm.Poly(j*x+k, x)
+root1 = [i.evalf() for i in _sm.solve(p1, x)]
+root2 = [i.evalf() for i in _sm.solve(_sm.Poly(_sm.Matrix([1,2,3]).reshape(3, 1), x),x)]
+m = _sm.Matrix([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]).reshape(4, 4)
+am = (m).T+m
+bm = _sm.Matrix([i.evalf() for i in (m).eigenvals().keys()])
+c1 = _sm.diag(1,1,1,1)
+c2 = _sm.Matrix([2 if i==j else 0 for i in range(3) for j in range(4)]).reshape(3, 4)
+dm = (m+c1)**(-1)
+e = (m+c1).det()+(_sm.Matrix([1,0,0,1]).reshape(2, 2)).trace()
+f = (m)[1,2]
+a = (m).cols
+bm = (m).col(0)
+cm = _sm.Matrix([(m).T.row(0),(m).T.row(1),(m).T.row(2),(m).T.row(3),(m).T.row(2)])
+dm = (m).row(0)
+em = _sm.Matrix([(m).row(0),(m).row(1),(m).row(2),(m).row(3),(m).row(2)])

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/autolev/test-examples/ruletest9.py

@@ -0,0 +1,55 @@
+import sympy.physics.mechanics as _me
+import sympy as _sm
+import math as m
+import numpy as _np
+
+frame_n = _me.ReferenceFrame('n')
+frame_a = _me.ReferenceFrame('a')
+a = 0
+d = _me.inertia(frame_a, 1, 1, 1)
+point_po1 = _me.Point('po1')
+point_po2 = _me.Point('po2')
+particle_p1 = _me.Particle('p1', _me.Point('p1_pt'), _sm.Symbol('m'))
+particle_p2 = _me.Particle('p2', _me.Point('p2_pt'), _sm.Symbol('m'))
+c1, c2, c3 = _me.dynamicsymbols('c1 c2 c3')
+c1_d, c2_d, c3_d = _me.dynamicsymbols('c1_ c2_ c3_', 1)
+body_r_cm = _me.Point('r_cm')
+body_r_cm.set_vel(frame_n, 0)
+body_r_f = _me.ReferenceFrame('r_f')
+body_r = _me.RigidBody('r', body_r_cm, body_r_f, _sm.symbols('m'), (_me.outer(body_r_f.x,body_r_f.x),body_r_cm))
+point_po2.set_pos(particle_p1.point, c1*frame_a.x)
+v = 2*point_po2.pos_from(particle_p1.point)+c2*frame_a.y
+frame_a.set_ang_vel(frame_n, c3*frame_a.z)
+v = 2*frame_a.ang_vel_in(frame_n)+c2*frame_a.y
+body_r_f.set_ang_vel(frame_n, c3*frame_a.z)
+v = 2*body_r_f.ang_vel_in(frame_n)+c2*frame_a.y
+frame_a.set_ang_acc(frame_n, (frame_a.ang_vel_in(frame_n)).dt(frame_a))
+v = 2*frame_a.ang_acc_in(frame_n)+c2*frame_a.y
+particle_p1.point.set_vel(frame_a, c1*frame_a.x+c3*frame_a.y)
+body_r_cm.set_acc(frame_n, c2*frame_a.y)
+v_a = _me.cross(body_r_cm.acc(frame_n), particle_p1.point.vel(frame_a))
+x_b_c = v_a
+x_b_d = 2*x_b_c
+a_b_c_d_e = x_b_d*2
+a_b_c = 2*c1*c2*c3
+a_b_c += 2*c1
+a_b_c  =  3*c1
+q1, q2, u1, u2 = _me.dynamicsymbols('q1 q2 u1 u2')
+q1_d, q2_d, u1_d, u2_d = _me.dynamicsymbols('q1_ q2_ u1_ u2_', 1)
+x, y = _me.dynamicsymbols('x y')
+x_d, y_d = _me.dynamicsymbols('x_ y_', 1)
+x_dd, y_dd = _me.dynamicsymbols('x_ y_', 2)
+yy = _me.dynamicsymbols('yy')
+yy = x*x_d**2+1
+m = _sm.Matrix([[0]])
+m[0] = 2*x
+m = m.row_insert(m.shape[0], _sm.Matrix([[0]]))
+m[m.shape[0]-1] = 2*y
+a = 2*m[0]
+m = _sm.Matrix([1,2,3,4,5,6,7,8,9]).reshape(3, 3)
+m[0,1] = 5
+a = m[0, 1]*2
+force_ro = q1*frame_n.x
+torque_a = q2*frame_n.z
+force_ro = q1*frame_n.x + q2*frame_n.y
+f = force_ro*2

deepseekvl2/lib/python3.10/site-packages/sympy/parsing/fortran/__init__.py

@@ -0,0 +1 @@
+"""Used for translating Fortran source code into a SymPy expression. """