qwen.py 481 Bytes
Newer Older
yangql's avatar
yangql committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
from ._base import BaseGPTQForCausalLM


class QwenGPTQForCausalLM(BaseGPTQForCausalLM):
    layer_type = "QWenBlock"
    layers_block_name = "transformer.h"
    outside_layer_modules = [
        "transformer.wte",
        "transformer.wpe",
        "transformer.ln_f",
        "transformer.visual",
    ]
    inside_layer_modules = [
        ["attn.c_attn"],
        ["attn.c_proj"],
        ["mlp.w1", "mlp.w2"],
        ["mlp.c_proj"],
    ]


__all__ = ["QwenGPTQForCausalLM"]