Hugging Face's logo

iqrakiran
/
custom_mental-llama-7b-instruct-custom-ds-1-1000000

PEFT
llama
Model card Files Community
iqrakiran commited on
Commit
257651c
·
verified ·
1 Parent(s): 61c75e2

Upload model

Browse files
Files changed (3) hide show
  1. README.md +11 -0
  2. adapter_config.json +254 -0
  3. adapter_model.bin +3 -0
README.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ ---
4
+ ## Training procedure
5
+
6
+ ### Framework versions
7
+
8
+ - PEFT 0.5.0
9
+ - PEFT 0.5.0
10
+
11
+ - PEFT 0.5.0
adapter_config.json ADDED
@@ -0,0 +1,254 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "auto_mapping": null,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "fan_in_fan_out": false,
6
+ "inference_mode": true,
7
+ "init_lora_weights": true,
8
+ "layers_pattern": null,
9
+ "layers_to_transform": null,
10
+ "lora_alpha": 64,
11
+ "lora_dropout": 0.1,
12
+ "modules_to_save": null,
13
+ "peft_type": "LORA",
14
+ "r": 16,
15
+ "revision": null,
16
+ "target_modules": [
17
+ "base_model.model.model.layers.0.self_attn.q_proj",
18
+ "base_model.model.model.layers.0.self_attn.q_proj.lora_A.default",
19
+ "base_model.model.model.layers.0.self_attn.q_proj.lora_B.default",
20
+ "base_model.model.model.layers.0.self_attn.k_proj",
21
+ "base_model.model.model.layers.0.self_attn.k_proj.lora_A.default",
22
+ "base_model.model.model.layers.0.self_attn.k_proj.lora_B.default",
23
+ "base_model.model.model.layers.0.self_attn.v_proj",
24
+ "base_model.model.model.layers.0.self_attn.v_proj.lora_A.default",
25
+ "base_model.model.model.layers.0.self_attn.v_proj.lora_B.default",
26
+ "base_model.model.model.layers.0.self_attn.o_proj",
27
+ "base_model.model.model.layers.0.self_attn.o_proj.lora_A.default",
28
+ "base_model.model.model.layers.0.self_attn.o_proj.lora_B.default",
29
+ "base_model.model.model.layers.0.mlp.gate_proj",
30
+ "base_model.model.model.layers.0.mlp.gate_proj.lora_A.default",
31
+ "base_model.model.model.layers.0.mlp.gate_proj.lora_B.default",
32
+ "base_model.model.model.layers.0.mlp.down_proj",
33
+ "base_model.model.model.layers.0.mlp.up_proj",
34
+ "base_model.model.model.layers.1.self_attn.q_proj",
35
+ "base_model.model.model.layers.1.self_attn.k_proj",
36
+ "base_model.model.model.layers.1.self_attn.v_proj",
37
+ "base_model.model.model.layers.1.self_attn.o_proj",
38
+ "base_model.model.model.layers.1.mlp.gate_proj",
39
+ "base_model.model.model.layers.1.mlp.down_proj",
40
+ "base_model.model.model.layers.1.mlp.up_proj",
41
+ "base_model.model.model.layers.2.self_attn.q_proj",
42
+ "base_model.model.model.layers.2.self_attn.k_proj",
43
+ "base_model.model.model.layers.2.self_attn.v_proj",
44
+ "base_model.model.model.layers.2.self_attn.o_proj",
45
+ "base_model.model.model.layers.2.mlp.gate_proj",
46
+ "base_model.model.model.layers.2.mlp.down_proj",
47
+ "base_model.model.model.layers.2.mlp.up_proj",
48
+ "base_model.model.model.layers.3.self_attn.q_proj",
49
+ "base_model.model.model.layers.3.self_attn.k_proj",
50
+ "base_model.model.model.layers.3.self_attn.v_proj",
51
+ "base_model.model.model.layers.3.self_attn.o_proj",
52
+ "base_model.model.model.layers.3.mlp.gate_proj",
53
+ "base_model.model.model.layers.3.mlp.down_proj",
54
+ "base_model.model.model.layers.3.mlp.up_proj",
55
+ "base_model.model.model.layers.4.self_attn.q_proj",
56
+ "base_model.model.model.layers.4.self_attn.k_proj",
57
+ "base_model.model.model.layers.4.self_attn.v_proj",
58
+ "base_model.model.model.layers.4.self_attn.o_proj",
59
+ "base_model.model.model.layers.4.mlp.gate_proj",
60
+ "base_model.model.model.layers.4.mlp.down_proj",
61
+ "base_model.model.model.layers.4.mlp.up_proj",
62
+ "base_model.model.model.layers.5.self_attn.q_proj",
63
+ "base_model.model.model.layers.5.self_attn.k_proj",
64
+ "base_model.model.model.layers.5.self_attn.v_proj",
65
+ "base_model.model.model.layers.5.self_attn.o_proj",
66
+ "base_model.model.model.layers.5.mlp.gate_proj",
67
+ "base_model.model.model.layers.5.mlp.down_proj",
68
+ "base_model.model.model.layers.5.mlp.up_proj",
69
+ "base_model.model.model.layers.6.self_attn.q_proj",
70
+ "base_model.model.model.layers.6.self_attn.k_proj",
71
+ "base_model.model.model.layers.6.self_attn.v_proj",
72
+ "base_model.model.model.layers.6.self_attn.o_proj",
73
+ "base_model.model.model.layers.6.mlp.gate_proj",
74
+ "base_model.model.model.layers.6.mlp.down_proj",
75
+ "base_model.model.model.layers.6.mlp.up_proj",
76
+ "base_model.model.model.layers.7.self_attn.q_proj",
77
+ "base_model.model.model.layers.7.self_attn.k_proj",
78
+ "base_model.model.model.layers.7.self_attn.v_proj",
79
+ "base_model.model.model.layers.7.self_attn.o_proj",
80
+ "base_model.model.model.layers.7.mlp.gate_proj",
81
+ "base_model.model.model.layers.7.mlp.down_proj",
82
+ "base_model.model.model.layers.7.mlp.up_proj",
83
+ "base_model.model.model.layers.8.self_attn.q_proj",
84
+ "base_model.model.model.layers.8.self_attn.k_proj",
85
+ "base_model.model.model.layers.8.self_attn.v_proj",
86
+ "base_model.model.model.layers.8.self_attn.o_proj",
87
+ "base_model.model.model.layers.8.mlp.gate_proj",
88
+ "base_model.model.model.layers.8.mlp.down_proj",
89
+ "base_model.model.model.layers.8.mlp.up_proj",
90
+ "base_model.model.model.layers.9.self_attn.q_proj",
91
+ "base_model.model.model.layers.9.self_attn.k_proj",
92
+ "base_model.model.model.layers.9.self_attn.v_proj",
93
+ "base_model.model.model.layers.9.self_attn.o_proj",
94
+ "base_model.model.model.layers.9.mlp.gate_proj",
95
+ "base_model.model.model.layers.9.mlp.down_proj",
96
+ "base_model.model.model.layers.9.mlp.up_proj",
97
+ "base_model.model.model.layers.10.self_attn.q_proj",
98
+ "base_model.model.model.layers.10.self_attn.k_proj",
99
+ "base_model.model.model.layers.10.self_attn.v_proj",
100
+ "base_model.model.model.layers.10.self_attn.o_proj",
101
+ "base_model.model.model.layers.10.mlp.gate_proj",
102
+ "base_model.model.model.layers.10.mlp.down_proj",
103
+ "base_model.model.model.layers.10.mlp.up_proj",
104
+ "base_model.model.model.layers.11.self_attn.q_proj",
105
+ "base_model.model.model.layers.11.self_attn.k_proj",
106
+ "base_model.model.model.layers.11.self_attn.v_proj",
107
+ "base_model.model.model.layers.11.self_attn.o_proj",
108
+ "base_model.model.model.layers.11.mlp.gate_proj",
109
+ "base_model.model.model.layers.11.mlp.down_proj",
110
+ "base_model.model.model.layers.11.mlp.up_proj",
111
+ "base_model.model.model.layers.12.self_attn.q_proj",
112
+ "base_model.model.model.layers.12.self_attn.k_proj",
113
+ "base_model.model.model.layers.12.self_attn.v_proj",
114
+ "base_model.model.model.layers.12.self_attn.o_proj",
115
+ "base_model.model.model.layers.12.mlp.gate_proj",
116
+ "base_model.model.model.layers.12.mlp.down_proj",
117
+ "base_model.model.model.layers.12.mlp.up_proj",
118
+ "base_model.model.model.layers.13.self_attn.q_proj",
119
+ "base_model.model.model.layers.13.self_attn.k_proj",
120
+ "base_model.model.model.layers.13.self_attn.v_proj",
121
+ "base_model.model.model.layers.13.self_attn.o_proj",
122
+ "base_model.model.model.layers.13.mlp.gate_proj",
123
+ "base_model.model.model.layers.13.mlp.down_proj",
124
+ "base_model.model.model.layers.13.mlp.up_proj",
125
+ "base_model.model.model.layers.14.self_attn.q_proj",
126
+ "base_model.model.model.layers.14.self_attn.k_proj",
127
+ "base_model.model.model.layers.14.self_attn.v_proj",
128
+ "base_model.model.model.layers.14.self_attn.o_proj",
129
+ "base_model.model.model.layers.14.mlp.gate_proj",
130
+ "base_model.model.model.layers.14.mlp.down_proj",
131
+ "base_model.model.model.layers.14.mlp.up_proj",
132
+ "base_model.model.model.layers.15.self_attn.q_proj",
133
+ "base_model.model.model.layers.15.self_attn.k_proj",
134
+ "base_model.model.model.layers.15.self_attn.v_proj",
135
+ "base_model.model.model.layers.15.self_attn.o_proj",
136
+ "base_model.model.model.layers.15.mlp.gate_proj",
137
+ "base_model.model.model.layers.15.mlp.down_proj",
138
+ "base_model.model.model.layers.15.mlp.up_proj",
139
+ "base_model.model.model.layers.16.self_attn.q_proj",
140
+ "base_model.model.model.layers.16.self_attn.k_proj",
141
+ "base_model.model.model.layers.16.self_attn.v_proj",
142
+ "base_model.model.model.layers.16.self_attn.o_proj",
143
+ "base_model.model.model.layers.16.mlp.gate_proj",
144
+ "base_model.model.model.layers.16.mlp.down_proj",
145
+ "base_model.model.model.layers.16.mlp.up_proj",
146
+ "base_model.model.model.layers.17.self_attn.q_proj",
147
+ "base_model.model.model.layers.17.self_attn.k_proj",
148
+ "base_model.model.model.layers.17.self_attn.v_proj",
149
+ "base_model.model.model.layers.17.self_attn.o_proj",
150
+ "base_model.model.model.layers.17.mlp.gate_proj",
151
+ "base_model.model.model.layers.17.mlp.down_proj",
152
+ "base_model.model.model.layers.17.mlp.up_proj",
153
+ "base_model.model.model.layers.18.self_attn.q_proj",
154
+ "base_model.model.model.layers.18.self_attn.k_proj",
155
+ "base_model.model.model.layers.18.self_attn.v_proj",
156
+ "base_model.model.model.layers.18.self_attn.o_proj",
157
+ "base_model.model.model.layers.18.mlp.gate_proj",
158
+ "base_model.model.model.layers.18.mlp.down_proj",
159
+ "base_model.model.model.layers.18.mlp.up_proj",
160
+ "base_model.model.model.layers.19.self_attn.q_proj",
161
+ "base_model.model.model.layers.19.self_attn.k_proj",
162
+ "base_model.model.model.layers.19.self_attn.v_proj",
163
+ "base_model.model.model.layers.19.self_attn.o_proj",
164
+ "base_model.model.model.layers.19.mlp.gate_proj",
165
+ "base_model.model.model.layers.19.mlp.down_proj",
166
+ "base_model.model.model.layers.19.mlp.up_proj",
167
+ "base_model.model.model.layers.20.self_attn.q_proj",
168
+ "base_model.model.model.layers.20.self_attn.k_proj",
169
+ "base_model.model.model.layers.20.self_attn.v_proj",
170
+ "base_model.model.model.layers.20.self_attn.o_proj",
171
+ "base_model.model.model.layers.20.mlp.gate_proj",
172
+ "base_model.model.model.layers.20.mlp.down_proj",
173
+ "base_model.model.model.layers.20.mlp.up_proj",
174
+ "base_model.model.model.layers.21.self_attn.q_proj",
175
+ "base_model.model.model.layers.21.self_attn.k_proj",
176
+ "base_model.model.model.layers.21.self_attn.v_proj",
177
+ "base_model.model.model.layers.21.self_attn.o_proj",
178
+ "base_model.model.model.layers.21.mlp.gate_proj",
179
+ "base_model.model.model.layers.21.mlp.down_proj",
180
+ "base_model.model.model.layers.21.mlp.up_proj",
181
+ "base_model.model.model.layers.22.self_attn.q_proj",
182
+ "base_model.model.model.layers.22.self_attn.k_proj",
183
+ "base_model.model.model.layers.22.self_attn.v_proj",
184
+ "base_model.model.model.layers.22.self_attn.o_proj",
185
+ "base_model.model.model.layers.22.mlp.gate_proj",
186
+ "base_model.model.model.layers.22.mlp.down_proj",
187
+ "base_model.model.model.layers.22.mlp.up_proj",
188
+ "base_model.model.model.layers.23.self_attn.q_proj",
189
+ "base_model.model.model.layers.23.self_attn.k_proj",
190
+ "base_model.model.model.layers.23.self_attn.v_proj",
191
+ "base_model.model.model.layers.23.self_attn.o_proj",
192
+ "base_model.model.model.layers.23.mlp.gate_proj",
193
+ "base_model.model.model.layers.23.mlp.down_proj",
194
+ "base_model.model.model.layers.23.mlp.up_proj",
195
+ "base_model.model.model.layers.24.self_attn.q_proj",
196
+ "base_model.model.model.layers.24.self_attn.k_proj",
197
+ "base_model.model.model.layers.24.self_attn.v_proj",
198
+ "base_model.model.model.layers.24.self_attn.o_proj",
199
+ "base_model.model.model.layers.24.mlp.gate_proj",
200
+ "base_model.model.model.layers.24.mlp.down_proj",
201
+ "base_model.model.model.layers.24.mlp.up_proj",
202
+ "base_model.model.model.layers.25.self_attn.q_proj",
203
+ "base_model.model.model.layers.25.self_attn.k_proj",
204
+ "base_model.model.model.layers.25.self_attn.v_proj",
205
+ "base_model.model.model.layers.25.self_attn.o_proj",
206
+ "base_model.model.model.layers.25.mlp.gate_proj",
207
+ "base_model.model.model.layers.25.mlp.down_proj",
208
+ "base_model.model.model.layers.25.mlp.up_proj",
209
+ "base_model.model.model.layers.26.self_attn.q_proj",
210
+ "base_model.model.model.layers.26.self_attn.k_proj",
211
+ "base_model.model.model.layers.26.self_attn.v_proj",
212
+ "base_model.model.model.layers.26.self_attn.o_proj",
213
+ "base_model.model.model.layers.26.mlp.gate_proj",
214
+ "base_model.model.model.layers.26.mlp.down_proj",
215
+ "base_model.model.model.layers.26.mlp.up_proj",
216
+ "base_model.model.model.layers.27.self_attn.q_proj",
217
+ "base_model.model.model.layers.27.self_attn.k_proj",
218
+ "base_model.model.model.layers.27.self_attn.v_proj",
219
+ "base_model.model.model.layers.27.self_attn.o_proj",
220
+ "base_model.model.model.layers.27.mlp.gate_proj",
221
+ "base_model.model.model.layers.27.mlp.down_proj",
222
+ "base_model.model.model.layers.27.mlp.up_proj",
223
+ "base_model.model.model.layers.28.self_attn.q_proj",
224
+ "base_model.model.model.layers.28.self_attn.k_proj",
225
+ "base_model.model.model.layers.28.self_attn.v_proj",
226
+ "base_model.model.model.layers.28.self_attn.o_proj",
227
+ "base_model.model.model.layers.28.mlp.gate_proj",
228
+ "base_model.model.model.layers.28.mlp.down_proj",
229
+ "base_model.model.model.layers.28.mlp.up_proj",
230
+ "base_model.model.model.layers.29.self_attn.q_proj",
231
+ "base_model.model.model.layers.29.self_attn.k_proj",
232
+ "base_model.model.model.layers.29.self_attn.v_proj",
233
+ "base_model.model.model.layers.29.self_attn.o_proj",
234
+ "base_model.model.model.layers.29.mlp.gate_proj",
235
+ "base_model.model.model.layers.29.mlp.down_proj",
236
+ "base_model.model.model.layers.29.mlp.up_proj",
237
+ "base_model.model.model.layers.30.self_attn.q_proj",
238
+ "base_model.model.model.layers.30.self_attn.k_proj",
239
+ "base_model.model.model.layers.30.self_attn.v_proj",
240
+ "base_model.model.model.layers.30.self_attn.o_proj",
241
+ "base_model.model.model.layers.30.mlp.gate_proj",
242
+ "base_model.model.model.layers.30.mlp.down_proj",
243
+ "base_model.model.model.layers.30.mlp.up_proj",
244
+ "base_model.model.model.layers.31.self_attn.q_proj",
245
+ "base_model.model.model.layers.31.self_attn.k_proj",
246
+ "base_model.model.model.layers.31.self_attn.v_proj",
247
+ "base_model.model.model.layers.31.self_attn.o_proj",
248
+ "base_model.model.model.layers.31.mlp.gate_proj",
249
+ "base_model.model.model.layers.31.mlp.down_proj",
250
+ "base_model.model.model.layers.31.mlp.up_proj",
251
+ "base_model.model.lm_head"
252
+ ],
253
+ "task_type": "CAUSAL_LM"
254
+ }
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ed6fd381ebe010a006d68d0b7ebbc5c2863f9e8dff1222cd342978175568e12
3
+ size 165462050