MCPcopy Create free account
hub / github.com/OpenBMB/AgentCPM-GUI / __init__

Method __init__

sft/dataset.py:26–49  ·  view source on GitHub ↗
(
        self,
        raw_data,
        transform,
        tokenizer,
        slice_config,
        llm_type="minicpm",
        patch_size=14,
        query_nums=64,
        batch_vision=False,
        max_length=2048,
        max_line_res=1120
    )

Source from the content-addressed store, hash-verified

24 """Dataset for supervised fine-tuning."""
25
26 def __init__(
27 self,
28 raw_data,
29 transform,
30 tokenizer,
31 slice_config,
32 llm_type="minicpm",
33 patch_size=14,
34 query_nums=64,
35 batch_vision=False,
36 max_length=2048,
37 max_line_res=1120
38 ):
39 super(SupervisedDataset, self).__init__()
40 self.raw_data = raw_data
41 self.tokenizer = tokenizer
42 self.transform = transform
43 self.slice_config = slice_config
44 self.llm_type = llm_type
45 self.patch_size = patch_size
46 self.query_nums=query_nums
47 self.batch_vision = batch_vision
48 self.max_length = max_length
49 self.max_line_res = max_line_res
50
51 def __len__(self):
52 return len(self.raw_data)

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected