(
self,
seq_id: int,
prompt: str,
prompt_token_ids: List[int],
block_size: int,
predict_output_len: int = 0,
)
| 109 | """ |
| 110 | |
| 111 | def __init__( |
| 112 | self, |
| 113 | seq_id: int, |
| 114 | prompt: str, |
| 115 | prompt_token_ids: List[int], |
| 116 | block_size: int, |
| 117 | predict_output_len: int = 0, |
| 118 | ) -> None: |
| 119 | self.seq_id = seq_id |
| 120 | self.prompt = prompt |
| 121 | self.block_size = block_size |
| 122 | |
| 123 | self.data = SequenceData(prompt_token_ids) |
| 124 | self.output_logprobs: SampleLogprobs = [] |
| 125 | self.output_text = "" |
| 126 | |
| 127 | self.logical_token_blocks: List[LogicalTokenBlock] = [] |
| 128 | # Initialize the logical token blocks with the prompt token ids. |
| 129 | self._append_tokens_to_blocks(prompt_token_ids) |
| 130 | self.status = SequenceStatus.WAITING |
| 131 | |
| 132 | # Used for incremental detokenization |
| 133 | self.prefix_offset = 0 |
| 134 | self.read_offset = 0 |
| 135 | # Input + output tokens |
| 136 | self.tokens: Optional[List[str]] = None |
| 137 | self.predict_output_len = predict_output_len |
| 138 | |
| 139 | def _append_logical_block(self) -> None: |
| 140 | block = LogicalTokenBlock( |
nothing calls this directly
no test coverage detected