File size: 970 Bytes
2c55a32
ffe9bd8
2c55a32
3c18b84
2c55a32
 
 
 
 
 
11d1ffd
55fcde9
28c9739
8376fe9
f4f04db
1afbd9f
2c55a32
3c18b84
2c55a32
 
dc12aad
2c55a32
 
 
 
 
 
 
185b654
 
af0db75
 
dc12aad
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
# custom_model.py
from transformers import PreTrainedModel, PretrainedConfig, AutoConfig, AutoModel
import torch.nn as nn
import subprocess

class CustomModelConfig(PretrainedConfig):
    model_type = "custom-model"
    def __init__(self, hidden_size=128, **kwargs):
        super().__init__(**kwargs)
        self.hidden_size = hidden_size
        self.num_attention_heads = 1
        self.vocab_size = 1000
        self.num_hidden_layers = 1
        self.intermediate_size = 256
        self.hidden_act = "silu"
        self.rms_norm_eps = 1e-8


class CustomModel(PreTrainedModel):
    config_class = CustomModelConfig

    def __init__(self, config):
        super().__init__(config)
        self.linear = nn.Linear(config.hidden_size, config.hidden_size)
    
    def forward(self, input_ids):
        output = self.linear(input_ids)
        return output

AutoConfig.register("custom-model", CustomModelConfig)
AutoModel.register(CustomModelConfig, CustomModel)