final-lora-r64-ep5 / adapter_config.json
ycchen's picture
Upload folder using huggingface_hub
79cebda
raw
history blame
4.32 kB
{
"auto_mapping": null,
"base_model_name_or_path": "ycchen/yc-test1",
"bias": "none",
"fan_in_fan_out": false,
"inference_mode": true,
"init_lora_weights": true,
"layers_pattern": null,
"layers_to_transform": null,
"lora_alpha": 16,
"lora_dropout": 0.05,
"modules_to_save": null,
"peft_type": "LORA",
"r": 16,
"revision": null,
"target_modules": [
"0.attn.c_attn",
"0.attn.c_proj",
"0.mlp.c_proj",
"0.mlp.w1",
"0.mlp.w2",
"1.attn.c_attn",
"1.attn.c_proj",
"1.mlp.c_proj",
"1.mlp.w1",
"1.mlp.w2",
"2.attn.c_attn",
"2.attn.c_proj",
"2.mlp.c_proj",
"2.mlp.w1",
"2.mlp.w2",
"3.attn.c_attn",
"3.attn.c_proj",
"3.mlp.c_proj",
"3.mlp.w1",
"3.mlp.w2",
"4.attn.c_attn",
"4.attn.c_proj",
"4.mlp.c_proj",
"4.mlp.w1",
"4.mlp.w2",
"5.attn.c_attn",
"5.attn.c_proj",
"5.mlp.c_proj",
"5.mlp.w1",
"5.mlp.w2",
"6.attn.c_attn",
"6.attn.c_proj",
"6.mlp.c_proj",
"6.mlp.w1",
"6.mlp.w2",
"7.attn.c_attn",
"7.attn.c_proj",
"7.mlp.c_proj",
"7.mlp.w1",
"7.mlp.w2",
"8.attn.c_attn",
"8.attn.c_proj",
"8.mlp.c_proj",
"8.mlp.w1",
"8.mlp.w2",
"9.attn.c_attn",
"9.attn.c_proj",
"9.mlp.c_proj",
"9.mlp.w1",
"9.mlp.w2",
"10.attn.c_attn",
"10.attn.c_proj",
"10.mlp.c_proj",
"10.mlp.w1",
"10.mlp.w2",
"11.attn.c_attn",
"11.attn.c_proj",
"11.mlp.c_proj",
"11.mlp.w1",
"11.mlp.w2",
"12.attn.c_attn",
"12.attn.c_proj",
"12.mlp.c_proj",
"12.mlp.w1",
"12.mlp.w2",
"13.attn.c_attn",
"13.attn.c_proj",
"13.mlp.c_proj",
"13.mlp.w1",
"13.mlp.w2",
"14.attn.c_attn",
"14.attn.c_proj",
"14.mlp.c_proj",
"14.mlp.w1",
"14.mlp.w2",
"15.attn.c_attn",
"15.attn.c_proj",
"15.mlp.c_proj",
"15.mlp.w1",
"15.mlp.w2",
"16.attn.c_attn",
"16.attn.c_proj",
"16.mlp.c_proj",
"16.mlp.w1",
"16.mlp.w2",
"17.attn.c_attn",
"17.attn.c_proj",
"17.mlp.c_proj",
"17.mlp.w1",
"17.mlp.w2",
"18.attn.c_attn",
"18.attn.c_proj",
"18.mlp.c_proj",
"18.mlp.w1",
"18.mlp.w2",
"19.attn.c_attn",
"19.attn.c_proj",
"19.mlp.c_proj",
"19.mlp.w1",
"19.mlp.w2",
"20.attn.c_attn",
"20.attn.c_proj",
"20.mlp.c_proj",
"20.mlp.w1",
"20.mlp.w2",
"21.attn.c_attn",
"21.attn.c_proj",
"21.mlp.c_proj",
"21.mlp.w1",
"21.mlp.w2",
"22.attn.c_attn",
"22.attn.c_proj",
"22.mlp.c_proj",
"22.mlp.w1",
"22.mlp.w2",
"23.attn.c_attn",
"23.attn.c_proj",
"23.mlp.c_proj",
"23.mlp.w1",
"23.mlp.w2",
"24.attn.c_attn",
"24.attn.c_proj",
"24.mlp.c_proj",
"24.mlp.w1",
"24.mlp.w2",
"25.attn.c_attn",
"25.attn.c_proj",
"25.mlp.c_proj",
"25.mlp.w1",
"25.mlp.w2",
"26.attn.c_attn",
"26.attn.c_proj",
"26.mlp.c_proj",
"26.mlp.w1",
"26.mlp.w2",
"27.attn.c_attn",
"27.attn.c_proj",
"27.mlp.c_proj",
"27.mlp.w1",
"27.mlp.w2",
"28.attn.c_attn",
"28.attn.c_proj",
"28.mlp.c_proj",
"28.mlp.w1",
"28.mlp.w2",
"29.attn.c_attn",
"29.attn.c_proj",
"29.mlp.c_proj",
"29.mlp.w1",
"29.mlp.w2",
"30.attn.c_attn",
"30.attn.c_proj",
"30.mlp.c_proj",
"30.mlp.w1",
"30.mlp.w2",
"31.attn.c_attn",
"31.attn.c_proj",
"31.mlp.c_proj",
"31.mlp.w1",
"31.mlp.w2",
"32.attn.c_attn",
"32.attn.c_proj",
"32.mlp.c_proj",
"32.mlp.w1",
"32.mlp.w2",
"33.attn.c_attn",
"33.attn.c_proj",
"33.mlp.c_proj",
"33.mlp.w1",
"33.mlp.w2",
"34.attn.c_attn",
"34.attn.c_proj",
"34.mlp.c_proj",
"34.mlp.w1",
"34.mlp.w2",
"35.attn.c_attn",
"35.attn.c_proj",
"35.mlp.c_proj",
"35.mlp.w1",
"35.mlp.w2",
"36.attn.c_attn",
"36.attn.c_proj",
"36.mlp.c_proj",
"36.mlp.w1",
"36.mlp.w2",
"37.attn.c_attn",
"37.attn.c_proj",
"37.mlp.c_proj",
"37.mlp.w1",
"37.mlp.w2",
"38.attn.c_attn",
"38.attn.c_proj",
"38.mlp.c_proj",
"38.mlp.w1",
"38.mlp.w2",
"39.attn.c_attn",
"39.attn.c_proj",
"39.mlp.c_proj",
"39.mlp.w1",
"39.mlp.w2"
],
"task_type": "CAUSAL_LM"
}