Upload 13 files
Browse files- config.json +1 -1
- generation_config.json +2 -10
- merge_report.json +242 -231
- model-00001-of-00004.safetensors +1 -1
- model-00002-of-00004.safetensors +1 -1
- model-00003-of-00004.safetensors +1 -1
- model-00004-of-00004.safetensors +1 -1
- perplexity.json +2 -2
- weight_averaging_stats.json +1 -0
config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "
|
3 |
"architectures": [
|
4 |
"Qwen2ForCausalLM"
|
5 |
],
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "",
|
3 |
"architectures": [
|
4 |
"Qwen2ForCausalLM"
|
5 |
],
|
generation_config.json
CHANGED
@@ -1,14 +1,6 @@
|
|
1 |
{
|
|
|
2 |
"bos_token_id": 151643,
|
3 |
-
"
|
4 |
-
"eos_token_id": [
|
5 |
-
151645,
|
6 |
-
151643
|
7 |
-
],
|
8 |
-
"pad_token_id": 151643,
|
9 |
-
"repetition_penalty": 1.05,
|
10 |
-
"temperature": 0.7,
|
11 |
-
"top_k": 20,
|
12 |
-
"top_p": 0.8,
|
13 |
"transformers_version": "4.46.2"
|
14 |
}
|
|
|
1 |
{
|
2 |
+
"_from_model_config": true,
|
3 |
"bos_token_id": 151643,
|
4 |
+
"eos_token_id": 151645,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"transformers_version": "4.46.2"
|
6 |
}
|
merge_report.json
CHANGED
@@ -1,271 +1,282 @@
|
|
1 |
{
|
2 |
-
"base_model": "
|
3 |
"fine_tuned_models": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
"sethuiyer/Qwen2.5-7B-Anvita",
|
|
|
5 |
"beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
|
|
|
|
|
|
6 |
"katanemo/Arch-Function-7B",
|
7 |
-
"
|
8 |
"Siheng99/Qwen2.5-7B-Instruct-SEALONG",
|
9 |
-
"
|
10 |
-
"
|
11 |
-
"opencompass/CompassJudger-1-7B-Instruct",
|
12 |
-
"scilo/qwen_ft_no_temp"
|
13 |
],
|
14 |
"layer_selection": {
|
15 |
-
"lm_head": "
|
16 |
-
"model.embed_tokens": "
|
17 |
-
"model.layers.0.input_layernorm": "
|
18 |
-
"model.layers.0.mlp.down_proj": "
|
19 |
-
"model.layers.0.mlp.gate_proj": "
|
20 |
-
"model.layers.0.mlp.up_proj": "
|
21 |
-
"model.layers.0.post_attention_layernorm": "
|
22 |
-
"model.layers.0.self_attn.k_proj": "
|
23 |
-
"model.layers.0.self_attn.o_proj": "
|
24 |
"model.layers.0.self_attn.q_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
25 |
-
"model.layers.0.self_attn.v_proj": "
|
26 |
-
"model.layers.1.input_layernorm": "
|
27 |
-
"model.layers.1.mlp.down_proj": "
|
28 |
-
"model.layers.1.mlp.gate_proj": "
|
29 |
-
"model.layers.1.mlp.up_proj": "
|
30 |
-
"model.layers.1.post_attention_layernorm": "
|
31 |
"model.layers.1.self_attn.k_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
32 |
-
"model.layers.1.self_attn.o_proj": "
|
33 |
-
"model.layers.1.self_attn.q_proj": "
|
34 |
-
"model.layers.1.self_attn.v_proj": "
|
35 |
-
"model.layers.10.input_layernorm": "
|
36 |
-
"model.layers.10.mlp.down_proj": "
|
37 |
-
"model.layers.10.mlp.gate_proj": "
|
38 |
-
"model.layers.10.mlp.up_proj": "
|
39 |
-
"model.layers.10.post_attention_layernorm": "
|
40 |
-
"model.layers.10.self_attn.k_proj": "
|
41 |
-
"model.layers.10.self_attn.o_proj": "
|
42 |
-
"model.layers.10.self_attn.q_proj": "
|
43 |
-
"model.layers.10.self_attn.v_proj": "
|
44 |
"model.layers.11.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
45 |
-
"model.layers.11.mlp.down_proj": "
|
46 |
-
"model.layers.11.mlp.gate_proj": "
|
47 |
-
"model.layers.11.mlp.up_proj": "
|
48 |
-
"model.layers.11.post_attention_layernorm": "
|
49 |
-
"model.layers.11.self_attn.k_proj": "
|
50 |
-
"model.layers.11.self_attn.o_proj": "
|
51 |
-
"model.layers.11.self_attn.q_proj": "
|
52 |
-
"model.layers.11.self_attn.v_proj": "
|
53 |
"model.layers.12.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
54 |
-
"model.layers.12.mlp.down_proj": "
|
55 |
-
"model.layers.12.mlp.gate_proj": "
|
56 |
-
"model.layers.12.mlp.up_proj": "
|
57 |
-
"model.layers.12.post_attention_layernorm": "
|
58 |
-
"model.layers.12.self_attn.k_proj": "
|
59 |
-
"model.layers.12.self_attn.o_proj": "
|
60 |
-
"model.layers.12.self_attn.q_proj": "
|
61 |
"model.layers.12.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
62 |
"model.layers.13.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
63 |
-
"model.layers.13.mlp.down_proj": "
|
64 |
-
"model.layers.13.mlp.gate_proj": "
|
65 |
-
"model.layers.13.mlp.up_proj": "
|
66 |
-
"model.layers.13.post_attention_layernorm": "
|
67 |
-
"model.layers.13.self_attn.k_proj": "
|
68 |
-
"model.layers.13.self_attn.o_proj": "
|
69 |
-
"model.layers.13.self_attn.q_proj": "
|
70 |
-
"model.layers.13.self_attn.v_proj": "
|
71 |
"model.layers.14.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
72 |
-
"model.layers.14.mlp.down_proj": "
|
73 |
-
"model.layers.14.mlp.gate_proj": "
|
74 |
-
"model.layers.14.mlp.up_proj": "
|
75 |
"model.layers.14.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
76 |
-
"model.layers.14.self_attn.k_proj": "
|
77 |
-
"model.layers.14.self_attn.o_proj": "
|
78 |
-
"model.layers.14.self_attn.q_proj": "
|
79 |
-
"model.layers.14.self_attn.v_proj": "
|
80 |
-
"model.layers.15.input_layernorm": "
|
81 |
-
"model.layers.15.mlp.down_proj": "
|
82 |
-
"model.layers.15.mlp.gate_proj": "
|
83 |
-
"model.layers.15.mlp.up_proj": "
|
84 |
-
"model.layers.15.post_attention_layernorm": "
|
85 |
-
"model.layers.15.self_attn.k_proj": "
|
86 |
-
"model.layers.15.self_attn.o_proj": "
|
87 |
-
"model.layers.15.self_attn.q_proj": "
|
88 |
-
"model.layers.15.self_attn.v_proj": "
|
89 |
-
"model.layers.16.input_layernorm": "
|
90 |
-
"model.layers.16.mlp.down_proj": "
|
91 |
-
"model.layers.16.mlp.gate_proj": "
|
92 |
-
"model.layers.16.mlp.up_proj": "
|
93 |
"model.layers.16.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
94 |
-
"model.layers.16.self_attn.k_proj": "
|
95 |
-
"model.layers.16.self_attn.o_proj": "
|
96 |
-
"model.layers.16.self_attn.q_proj": "
|
97 |
-
"model.layers.16.self_attn.v_proj": "
|
98 |
"model.layers.17.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
99 |
-
"model.layers.17.mlp.down_proj": "
|
100 |
-
"model.layers.17.mlp.gate_proj": "
|
101 |
-
"model.layers.17.mlp.up_proj": "
|
102 |
"model.layers.17.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
103 |
-
"model.layers.17.self_attn.k_proj": "
|
104 |
-
"model.layers.17.self_attn.o_proj": "
|
105 |
-
"model.layers.17.self_attn.q_proj": "
|
106 |
-
"model.layers.17.self_attn.v_proj": "
|
107 |
-
"model.layers.18.input_layernorm": "
|
108 |
-
"model.layers.18.mlp.down_proj": "
|
109 |
-
"model.layers.18.mlp.gate_proj": "
|
110 |
-
"model.layers.18.mlp.up_proj": "
|
111 |
"model.layers.18.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
112 |
-
"model.layers.18.self_attn.k_proj": "
|
113 |
-
"model.layers.18.self_attn.o_proj": "
|
114 |
-
"model.layers.18.self_attn.q_proj": "
|
115 |
-
"model.layers.18.self_attn.v_proj": "
|
116 |
-
"model.layers.19.input_layernorm": "
|
117 |
-
"model.layers.19.mlp.down_proj": "
|
118 |
-
"model.layers.19.mlp.gate_proj": "
|
119 |
-
"model.layers.19.mlp.up_proj": "
|
120 |
"model.layers.19.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
121 |
-
"model.layers.19.self_attn.k_proj": "
|
122 |
-
"model.layers.19.self_attn.o_proj": "
|
123 |
-
"model.layers.19.self_attn.q_proj": "
|
124 |
"model.layers.19.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
125 |
-
"model.layers.2.input_layernorm": "
|
126 |
-
"model.layers.2.mlp.down_proj": "
|
127 |
-
"model.layers.2.mlp.gate_proj": "
|
128 |
-
"model.layers.2.mlp.up_proj": "
|
129 |
"model.layers.2.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
130 |
-
"model.layers.2.self_attn.k_proj": "
|
131 |
-
"model.layers.2.self_attn.o_proj": "
|
132 |
-
"model.layers.2.self_attn.q_proj": "
|
133 |
-
"model.layers.2.self_attn.v_proj": "
|
134 |
"model.layers.20.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
135 |
-
"model.layers.20.mlp.down_proj": "
|
136 |
-
"model.layers.20.mlp.gate_proj": "
|
137 |
-
"model.layers.20.mlp.up_proj": "
|
138 |
"model.layers.20.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
139 |
-
"model.layers.20.self_attn.k_proj": "
|
140 |
-
"model.layers.20.self_attn.o_proj": "
|
141 |
-
"model.layers.20.self_attn.q_proj": "
|
142 |
-
"model.layers.20.self_attn.v_proj": "
|
143 |
"model.layers.21.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
144 |
-
"model.layers.21.mlp.down_proj": "
|
145 |
-
"model.layers.21.mlp.gate_proj": "
|
146 |
-
"model.layers.21.mlp.up_proj": "
|
147 |
"model.layers.21.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
148 |
-
"model.layers.21.self_attn.k_proj": "
|
149 |
-
"model.layers.21.self_attn.o_proj": "
|
150 |
-
"model.layers.21.self_attn.q_proj": "
|
151 |
-
"model.layers.21.self_attn.v_proj": "
|
152 |
-
"model.layers.22.input_layernorm": "
|
153 |
-
"model.layers.22.mlp.down_proj": "
|
154 |
-
"model.layers.22.mlp.gate_proj": "
|
155 |
-
"model.layers.22.mlp.up_proj": "
|
156 |
-
"model.layers.22.post_attention_layernorm": "
|
157 |
-
"model.layers.22.self_attn.k_proj": "
|
158 |
-
"model.layers.22.self_attn.o_proj": "
|
159 |
-
"model.layers.22.self_attn.q_proj": "
|
160 |
-
"model.layers.22.self_attn.v_proj": "
|
161 |
-
"model.layers.23.input_layernorm": "
|
162 |
-
"model.layers.23.mlp.down_proj": "
|
163 |
-
"model.layers.23.mlp.gate_proj": "
|
164 |
-
"model.layers.23.mlp.up_proj": "
|
165 |
"model.layers.23.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
166 |
-
"model.layers.23.self_attn.k_proj": "
|
167 |
-
"model.layers.23.self_attn.o_proj": "
|
168 |
-
"model.layers.23.self_attn.q_proj": "
|
169 |
-
"model.layers.23.self_attn.v_proj": "
|
170 |
"model.layers.24.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
171 |
-
"model.layers.24.mlp.down_proj": "
|
172 |
-
"model.layers.24.mlp.gate_proj": "
|
173 |
-
"model.layers.24.mlp.up_proj": "
|
174 |
-
"model.layers.24.post_attention_layernorm": "
|
175 |
-
"model.layers.24.self_attn.k_proj": "
|
176 |
-
"model.layers.24.self_attn.o_proj": "
|
177 |
-
"model.layers.24.self_attn.q_proj": "
|
178 |
-
"model.layers.24.self_attn.v_proj": "
|
179 |
-
"model.layers.25.input_layernorm": "
|
180 |
-
"model.layers.25.mlp.down_proj": "
|
181 |
-
"model.layers.25.mlp.gate_proj": "
|
182 |
-
"model.layers.25.mlp.up_proj": "
|
183 |
-
"model.layers.25.post_attention_layernorm": "
|
184 |
-
"model.layers.25.self_attn.k_proj": "
|
185 |
-
"model.layers.25.self_attn.o_proj": "
|
186 |
-
"model.layers.25.self_attn.q_proj": "
|
187 |
-
"model.layers.25.self_attn.v_proj": "
|
188 |
"model.layers.26.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
189 |
-
"model.layers.26.mlp.down_proj": "
|
190 |
-
"model.layers.26.mlp.gate_proj": "
|
191 |
-
"model.layers.26.mlp.up_proj": "
|
192 |
"model.layers.26.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
193 |
-
"model.layers.26.self_attn.k_proj": "
|
194 |
-
"model.layers.26.self_attn.o_proj": "
|
195 |
-
"model.layers.26.self_attn.q_proj": "
|
196 |
-
"model.layers.26.self_attn.v_proj": "
|
197 |
-
"model.layers.27.input_layernorm": "
|
198 |
-
"model.layers.27.mlp.down_proj": "
|
199 |
-
"model.layers.27.mlp.gate_proj": "
|
200 |
-
"model.layers.27.mlp.up_proj": "
|
201 |
-
"model.layers.27.post_attention_layernorm": "
|
202 |
-
"model.layers.27.self_attn.k_proj": "
|
203 |
-
"model.layers.27.self_attn.o_proj": "
|
204 |
-
"model.layers.27.self_attn.q_proj": "
|
205 |
-
"model.layers.27.self_attn.v_proj": "
|
206 |
-
"model.layers.3.input_layernorm": "
|
207 |
-
"model.layers.3.mlp.down_proj": "
|
208 |
-
"model.layers.3.mlp.gate_proj": "
|
209 |
-
"model.layers.3.mlp.up_proj": "
|
210 |
-
"model.layers.3.post_attention_layernorm": "
|
211 |
-
"model.layers.3.self_attn.k_proj": "
|
212 |
-
"model.layers.3.self_attn.o_proj": "
|
213 |
-
"model.layers.3.self_attn.q_proj": "
|
214 |
-
"model.layers.3.self_attn.v_proj": "
|
215 |
-
"model.layers.4.input_layernorm": "
|
216 |
-
"model.layers.4.mlp.down_proj": "
|
217 |
-
"model.layers.4.mlp.gate_proj": "
|
218 |
-
"model.layers.4.mlp.up_proj": "
|
219 |
-
"model.layers.4.post_attention_layernorm": "
|
220 |
-
"model.layers.4.self_attn.k_proj": "
|
221 |
-
"model.layers.4.self_attn.o_proj": "
|
222 |
-
"model.layers.4.self_attn.q_proj": "
|
223 |
-
"model.layers.4.self_attn.v_proj": "
|
224 |
"model.layers.5.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
225 |
-
"model.layers.5.mlp.down_proj": "
|
226 |
-
"model.layers.5.mlp.gate_proj": "
|
227 |
-
"model.layers.5.mlp.up_proj": "
|
228 |
-
"model.layers.5.post_attention_layernorm": "
|
229 |
-
"model.layers.5.self_attn.k_proj": "
|
230 |
-
"model.layers.5.self_attn.o_proj": "
|
231 |
-
"model.layers.5.self_attn.q_proj": "
|
232 |
-
"model.layers.5.self_attn.v_proj": "
|
233 |
"model.layers.6.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
234 |
-
"model.layers.6.mlp.down_proj": "
|
235 |
-
"model.layers.6.mlp.gate_proj": "
|
236 |
-
"model.layers.6.mlp.up_proj": "
|
237 |
"model.layers.6.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
238 |
-
"model.layers.6.self_attn.k_proj": "
|
239 |
-
"model.layers.6.self_attn.o_proj": "
|
240 |
-
"model.layers.6.self_attn.q_proj": "
|
241 |
-
"model.layers.6.self_attn.v_proj": "
|
242 |
"model.layers.7.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
243 |
-
"model.layers.7.mlp.down_proj": "
|
244 |
-
"model.layers.7.mlp.gate_proj": "
|
245 |
-
"model.layers.7.mlp.up_proj": "
|
246 |
-
"model.layers.7.post_attention_layernorm": "
|
247 |
-
"model.layers.7.self_attn.k_proj": "
|
248 |
-
"model.layers.7.self_attn.o_proj": "
|
249 |
-
"model.layers.7.self_attn.q_proj": "
|
250 |
-
"model.layers.7.self_attn.v_proj": "
|
251 |
-
"model.layers.8.input_layernorm": "
|
252 |
-
"model.layers.8.mlp.down_proj": "
|
253 |
-
"model.layers.8.mlp.gate_proj": "
|
254 |
-
"model.layers.8.mlp.up_proj": "
|
255 |
"model.layers.8.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
256 |
-
"model.layers.8.self_attn.k_proj": "
|
257 |
-
"model.layers.8.self_attn.o_proj": "
|
258 |
-
"model.layers.8.self_attn.q_proj": "
|
259 |
-
"model.layers.8.self_attn.v_proj": "
|
260 |
-
"model.layers.9.input_layernorm": "
|
261 |
-
"model.layers.9.mlp.down_proj": "
|
262 |
-
"model.layers.9.mlp.gate_proj": "
|
263 |
-
"model.layers.9.mlp.up_proj": "
|
264 |
"model.layers.9.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
265 |
-
"model.layers.9.self_attn.k_proj": "
|
266 |
-
"model.layers.9.self_attn.o_proj": "
|
267 |
-
"model.layers.9.self_attn.q_proj": "
|
268 |
-
"model.layers.9.self_attn.v_proj": "
|
269 |
"model.norm": "sethuiyer/Qwen2.5-7B-Anvita"
|
270 |
}
|
271 |
}
|
|
|
1 |
{
|
2 |
+
"base_model": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
3 |
"fine_tuned_models": [
|
4 |
+
"Qwen/Qwen2.5-7B",
|
5 |
+
"EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
6 |
+
"FourOhFour/Vapor_v2_7B",
|
7 |
+
"Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
8 |
+
"happzy2633/qwen2.5-7b-ins-v3",
|
9 |
+
"huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2",
|
10 |
+
"HumanLLMs/Humanish-Qwen2.5-7B-Instruct",
|
11 |
+
"Orion-zhen/Qwen2.5-7B-Instruct-Uncensored",
|
12 |
+
"Orion-zhen/Meissa-Qwen2.5-7B-Instruct",
|
13 |
"sethuiyer/Qwen2.5-7B-Anvita",
|
14 |
+
"rombodawg/Rombos-LLM-V2.5-Qwen-7b",
|
15 |
"beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
16 |
+
"Orion-zhen/Qwen2.5-7B-Gutenberg-KTO",
|
17 |
+
"fblgit/cybertron-v4-qw7B-MGS",
|
18 |
+
"edgerunner-ai/EdgeRunner-Command-Nested",
|
19 |
"katanemo/Arch-Function-7B",
|
20 |
+
"DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
21 |
"Siheng99/Qwen2.5-7B-Instruct-SEALONG",
|
22 |
+
"nhyha/merge_Qwen2.5-7B-Instruct_20241023_0314",
|
23 |
+
"thomas-yanxin/XinYuan-Qwen2.5-7B-0917"
|
|
|
|
|
24 |
],
|
25 |
"layer_selection": {
|
26 |
+
"lm_head": "sethuiyer/Qwen2.5-7B-Anvita",
|
27 |
+
"model.embed_tokens": "sethuiyer/Qwen2.5-7B-Anvita",
|
28 |
+
"model.layers.0.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
29 |
+
"model.layers.0.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
30 |
+
"model.layers.0.mlp.gate_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
31 |
+
"model.layers.0.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
32 |
+
"model.layers.0.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
33 |
+
"model.layers.0.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
34 |
+
"model.layers.0.self_attn.o_proj": "huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2",
|
35 |
"model.layers.0.self_attn.q_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
36 |
+
"model.layers.0.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
37 |
+
"model.layers.1.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
38 |
+
"model.layers.1.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
39 |
+
"model.layers.1.mlp.gate_proj": "thomas-yanxin/XinYuan-Qwen2.5-7B-0917",
|
40 |
+
"model.layers.1.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
41 |
+
"model.layers.1.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
42 |
"model.layers.1.self_attn.k_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
43 |
+
"model.layers.1.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
44 |
+
"model.layers.1.self_attn.q_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
45 |
+
"model.layers.1.self_attn.v_proj": "katanemo/Arch-Function-7B",
|
46 |
+
"model.layers.10.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
47 |
+
"model.layers.10.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
48 |
+
"model.layers.10.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
49 |
+
"model.layers.10.mlp.up_proj": "rombodawg/Rombos-LLM-V2.5-Qwen-7b",
|
50 |
+
"model.layers.10.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
51 |
+
"model.layers.10.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
52 |
+
"model.layers.10.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
53 |
+
"model.layers.10.self_attn.q_proj": "Orion-zhen/Meissa-Qwen2.5-7B-Instruct",
|
54 |
+
"model.layers.10.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
55 |
"model.layers.11.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
56 |
+
"model.layers.11.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
57 |
+
"model.layers.11.mlp.gate_proj": "nhyha/merge_Qwen2.5-7B-Instruct_20241023_0314",
|
58 |
+
"model.layers.11.mlp.up_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
59 |
+
"model.layers.11.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
60 |
+
"model.layers.11.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
61 |
+
"model.layers.11.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
62 |
+
"model.layers.11.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
63 |
+
"model.layers.11.self_attn.v_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
64 |
"model.layers.12.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
65 |
+
"model.layers.12.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
66 |
+
"model.layers.12.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
67 |
+
"model.layers.12.mlp.up_proj": "thomas-yanxin/XinYuan-Qwen2.5-7B-0917",
|
68 |
+
"model.layers.12.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
69 |
+
"model.layers.12.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
70 |
+
"model.layers.12.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
71 |
+
"model.layers.12.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
72 |
"model.layers.12.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
73 |
"model.layers.13.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
74 |
+
"model.layers.13.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
75 |
+
"model.layers.13.mlp.gate_proj": "Orion-zhen/Qwen2.5-7B-Instruct-Uncensored",
|
76 |
+
"model.layers.13.mlp.up_proj": "fblgit/cybertron-v4-qw7B-MGS",
|
77 |
+
"model.layers.13.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
78 |
+
"model.layers.13.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
79 |
+
"model.layers.13.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
80 |
+
"model.layers.13.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
81 |
+
"model.layers.13.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
82 |
"model.layers.14.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
83 |
+
"model.layers.14.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
84 |
+
"model.layers.14.mlp.gate_proj": "HumanLLMs/Humanish-Qwen2.5-7B-Instruct",
|
85 |
+
"model.layers.14.mlp.up_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
86 |
"model.layers.14.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
87 |
+
"model.layers.14.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
88 |
+
"model.layers.14.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
89 |
+
"model.layers.14.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
90 |
+
"model.layers.14.self_attn.v_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
91 |
+
"model.layers.15.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
92 |
+
"model.layers.15.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
93 |
+
"model.layers.15.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
94 |
+
"model.layers.15.mlp.up_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
95 |
+
"model.layers.15.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
96 |
+
"model.layers.15.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
97 |
+
"model.layers.15.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
98 |
+
"model.layers.15.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
99 |
+
"model.layers.15.self_attn.v_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
100 |
+
"model.layers.16.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
101 |
+
"model.layers.16.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
102 |
+
"model.layers.16.mlp.gate_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
103 |
+
"model.layers.16.mlp.up_proj": "Orion-zhen/Meissa-Qwen2.5-7B-Instruct",
|
104 |
"model.layers.16.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
105 |
+
"model.layers.16.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
106 |
+
"model.layers.16.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
107 |
+
"model.layers.16.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
108 |
+
"model.layers.16.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
109 |
"model.layers.17.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
110 |
+
"model.layers.17.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
111 |
+
"model.layers.17.mlp.gate_proj": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
112 |
+
"model.layers.17.mlp.up_proj": "Orion-zhen/Qwen2.5-7B-Gutenberg-KTO",
|
113 |
"model.layers.17.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
114 |
+
"model.layers.17.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
115 |
+
"model.layers.17.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
116 |
+
"model.layers.17.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
117 |
+
"model.layers.17.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
118 |
+
"model.layers.18.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
119 |
+
"model.layers.18.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
120 |
+
"model.layers.18.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
121 |
+
"model.layers.18.mlp.up_proj": "Orion-zhen/Qwen2.5-7B-Instruct-Uncensored",
|
122 |
"model.layers.18.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
123 |
+
"model.layers.18.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
124 |
+
"model.layers.18.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
125 |
+
"model.layers.18.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
126 |
+
"model.layers.18.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
127 |
+
"model.layers.19.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
128 |
+
"model.layers.19.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
129 |
+
"model.layers.19.mlp.gate_proj": "Siheng99/Qwen2.5-7B-Instruct-SEALONG",
|
130 |
+
"model.layers.19.mlp.up_proj": "fblgit/cybertron-v4-qw7B-MGS",
|
131 |
"model.layers.19.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
132 |
+
"model.layers.19.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
133 |
+
"model.layers.19.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
134 |
+
"model.layers.19.self_attn.q_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
135 |
"model.layers.19.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
136 |
+
"model.layers.2.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
137 |
+
"model.layers.2.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
138 |
+
"model.layers.2.mlp.gate_proj": "HumanLLMs/Humanish-Qwen2.5-7B-Instruct",
|
139 |
+
"model.layers.2.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
140 |
"model.layers.2.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
141 |
+
"model.layers.2.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
142 |
+
"model.layers.2.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
143 |
+
"model.layers.2.self_attn.q_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
144 |
+
"model.layers.2.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
145 |
"model.layers.20.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
146 |
+
"model.layers.20.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
147 |
+
"model.layers.20.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
148 |
+
"model.layers.20.mlp.up_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
149 |
"model.layers.20.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
150 |
+
"model.layers.20.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
151 |
+
"model.layers.20.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
152 |
+
"model.layers.20.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
153 |
+
"model.layers.20.self_attn.v_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
154 |
"model.layers.21.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
155 |
+
"model.layers.21.mlp.down_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
156 |
+
"model.layers.21.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
157 |
+
"model.layers.21.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
158 |
"model.layers.21.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
159 |
+
"model.layers.21.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
160 |
+
"model.layers.21.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
161 |
+
"model.layers.21.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
162 |
+
"model.layers.21.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
163 |
+
"model.layers.22.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
164 |
+
"model.layers.22.mlp.down_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
165 |
+
"model.layers.22.mlp.gate_proj": "fblgit/cybertron-v4-qw7B-MGS",
|
166 |
+
"model.layers.22.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
167 |
+
"model.layers.22.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
168 |
+
"model.layers.22.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
169 |
+
"model.layers.22.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
170 |
+
"model.layers.22.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
171 |
+
"model.layers.22.self_attn.v_proj": "katanemo/Arch-Function-7B",
|
172 |
+
"model.layers.23.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
173 |
+
"model.layers.23.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
174 |
+
"model.layers.23.mlp.gate_proj": "FourOhFour/Vapor_v2_7B",
|
175 |
+
"model.layers.23.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
176 |
"model.layers.23.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
177 |
+
"model.layers.23.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
178 |
+
"model.layers.23.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
179 |
+
"model.layers.23.self_attn.q_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
180 |
+
"model.layers.23.self_attn.v_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
181 |
"model.layers.24.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
182 |
+
"model.layers.24.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
183 |
+
"model.layers.24.mlp.gate_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
184 |
+
"model.layers.24.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
185 |
+
"model.layers.24.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
186 |
+
"model.layers.24.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
187 |
+
"model.layers.24.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
188 |
+
"model.layers.24.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
189 |
+
"model.layers.24.self_attn.v_proj": "edgerunner-ai/EdgeRunner-Command-Nested",
|
190 |
+
"model.layers.25.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
191 |
+
"model.layers.25.mlp.down_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
192 |
+
"model.layers.25.mlp.gate_proj": "fblgit/cybertron-v4-qw7B-MGS",
|
193 |
+
"model.layers.25.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
194 |
+
"model.layers.25.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
195 |
+
"model.layers.25.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
196 |
+
"model.layers.25.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
197 |
+
"model.layers.25.self_attn.q_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
198 |
+
"model.layers.25.self_attn.v_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
199 |
"model.layers.26.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
200 |
+
"model.layers.26.mlp.down_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
201 |
+
"model.layers.26.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
202 |
+
"model.layers.26.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
203 |
"model.layers.26.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
204 |
+
"model.layers.26.self_attn.k_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
205 |
+
"model.layers.26.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
206 |
+
"model.layers.26.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
207 |
+
"model.layers.26.self_attn.v_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
208 |
+
"model.layers.27.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
209 |
+
"model.layers.27.mlp.down_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
210 |
+
"model.layers.27.mlp.gate_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
211 |
+
"model.layers.27.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
212 |
+
"model.layers.27.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
213 |
+
"model.layers.27.self_attn.k_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
214 |
+
"model.layers.27.self_attn.o_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
215 |
+
"model.layers.27.self_attn.q_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
216 |
+
"model.layers.27.self_attn.v_proj": "FourOhFour/Vapor_v2_7B",
|
217 |
+
"model.layers.3.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
218 |
+
"model.layers.3.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
219 |
+
"model.layers.3.mlp.gate_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
220 |
+
"model.layers.3.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
221 |
+
"model.layers.3.post_attention_layernorm": "Qwen/Qwen2.5-7B",
|
222 |
+
"model.layers.3.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
223 |
+
"model.layers.3.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
224 |
+
"model.layers.3.self_attn.q_proj": "HumanLLMs/Humanish-Qwen2.5-7B-Instruct",
|
225 |
+
"model.layers.3.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
226 |
+
"model.layers.4.input_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
227 |
+
"model.layers.4.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
228 |
+
"model.layers.4.mlp.gate_proj": "Orion-zhen/Meissa-Qwen2.5-7B-Instruct",
|
229 |
+
"model.layers.4.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
230 |
+
"model.layers.4.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
231 |
+
"model.layers.4.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
232 |
+
"model.layers.4.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
233 |
+
"model.layers.4.self_attn.q_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
234 |
+
"model.layers.4.self_attn.v_proj": "katanemo/Arch-Function-7B",
|
235 |
"model.layers.5.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
236 |
+
"model.layers.5.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
237 |
+
"model.layers.5.mlp.gate_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
238 |
+
"model.layers.5.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
239 |
+
"model.layers.5.post_attention_layernorm": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
240 |
+
"model.layers.5.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
241 |
+
"model.layers.5.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
242 |
+
"model.layers.5.self_attn.q_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
243 |
+
"model.layers.5.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
244 |
"model.layers.6.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
245 |
+
"model.layers.6.mlp.down_proj": "huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2",
|
246 |
+
"model.layers.6.mlp.gate_proj": "happzy2633/qwen2.5-7b-ins-v3",
|
247 |
+
"model.layers.6.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
248 |
"model.layers.6.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
249 |
+
"model.layers.6.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
250 |
+
"model.layers.6.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
251 |
+
"model.layers.6.self_attn.q_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
252 |
+
"model.layers.6.self_attn.v_proj": "thomas-yanxin/XinYuan-Qwen2.5-7B-0917",
|
253 |
"model.layers.7.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
254 |
+
"model.layers.7.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
255 |
+
"model.layers.7.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
256 |
+
"model.layers.7.mlp.up_proj": "katanemo/Arch-Function-7B",
|
257 |
+
"model.layers.7.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
258 |
+
"model.layers.7.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
259 |
+
"model.layers.7.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
260 |
+
"model.layers.7.self_attn.q_proj": "thomas-yanxin/XinYuan-Qwen2.5-7B-0917",
|
261 |
+
"model.layers.7.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
262 |
+
"model.layers.8.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
263 |
+
"model.layers.8.mlp.down_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
264 |
+
"model.layers.8.mlp.gate_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
265 |
+
"model.layers.8.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
266 |
"model.layers.8.post_attention_layernorm": "beomi/Qwen2.5-7B-Instruct-kowiki-qa",
|
267 |
+
"model.layers.8.self_attn.k_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
268 |
+
"model.layers.8.self_attn.o_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
269 |
+
"model.layers.8.self_attn.q_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
270 |
+
"model.layers.8.self_attn.v_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
271 |
+
"model.layers.9.input_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
272 |
+
"model.layers.9.mlp.down_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
273 |
+
"model.layers.9.mlp.gate_proj": "Orion-zhen/Meissa-Qwen2.5-7B-Instruct",
|
274 |
+
"model.layers.9.mlp.up_proj": "sethuiyer/Qwen2.5-7B-Anvita",
|
275 |
"model.layers.9.post_attention_layernorm": "sethuiyer/Qwen2.5-7B-Anvita",
|
276 |
+
"model.layers.9.self_attn.k_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
277 |
+
"model.layers.9.self_attn.o_proj": "Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2",
|
278 |
+
"model.layers.9.self_attn.q_proj": "DeepGlint-AI/llava-mlcd-qwen2.5-7b",
|
279 |
+
"model.layers.9.self_attn.v_proj": "EVA-UNIT-01/EVA-Qwen2.5-7B-v0.1",
|
280 |
"model.norm": "sethuiyer/Qwen2.5-7B-Anvita"
|
281 |
}
|
282 |
}
|
model-00001-of-00004.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4877660776
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9bceae4c3b50bcb2e59bce40cd140fca093d5e9ab764dbea658badef924ea674
|
3 |
size 4877660776
|
model-00002-of-00004.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4932751008
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b53af0044a3d64ab0e34b7c4d344bfe697de4157b9fbd4be1a1f602e19336954
|
3 |
size 4932751008
|
model-00003-of-00004.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4330865200
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0b63ae87a2a16097e612d1a12b67a857290469fbbc1ced1d75f8b641ff3d5498
|
3 |
size 4330865200
|
model-00004-of-00004.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1089994880
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3994de9777084e198ebcf0a0414b5c56b326c902a0e5040b66077eaa23138211
|
3 |
size 1089994880
|
perplexity.json
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
-
"model_path": "E
|
3 |
-
"perplexity": 9.
|
4 |
"eval_samples": 1000,
|
5 |
"max_length": 512,
|
6 |
"dataset": "wikitext",
|
|
|
1 |
{
|
2 |
+
"model_path": "E:/merged_model/composite_model_20241117_160240",
|
3 |
+
"perplexity": 9.0625,
|
4 |
"eval_samples": 1000,
|
5 |
"max_length": 512,
|
6 |
"dataset": "wikitext",
|
weight_averaging_stats.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{}
|