bartowski commited on
Commit
a7b564b
·
verified ·
1 Parent(s): 7c0197e

Upload folder using huggingface_hub

Browse files
Files changed (42) hide show
  1. layouts/InternScience_MLExpert-IQ2_M.layout.json +233 -0
  2. layouts/InternScience_MLExpert-IQ2_M.tensor-types.txt +432 -0
  3. layouts/InternScience_MLExpert-IQ2_S.layout.json +236 -0
  4. layouts/InternScience_MLExpert-IQ2_S.tensor-types.txt +432 -0
  5. layouts/InternScience_MLExpert-IQ2_XS.layout.json +229 -0
  6. layouts/InternScience_MLExpert-IQ2_XS.tensor-types.txt +432 -0
  7. layouts/InternScience_MLExpert-IQ2_XXS.layout.json +236 -0
  8. layouts/InternScience_MLExpert-IQ2_XXS.tensor-types.txt +432 -0
  9. layouts/InternScience_MLExpert-IQ3_M.layout.json +228 -0
  10. layouts/InternScience_MLExpert-IQ3_M.tensor-types.txt +432 -0
  11. layouts/InternScience_MLExpert-IQ3_XS.layout.json +228 -0
  12. layouts/InternScience_MLExpert-IQ3_XS.tensor-types.txt +432 -0
  13. layouts/InternScience_MLExpert-IQ3_XXS.layout.json +231 -0
  14. layouts/InternScience_MLExpert-IQ3_XXS.tensor-types.txt +432 -0
  15. layouts/InternScience_MLExpert-IQ4_NL.layout.json +226 -0
  16. layouts/InternScience_MLExpert-IQ4_NL.tensor-types.txt +432 -0
  17. layouts/InternScience_MLExpert-IQ4_XS.layout.json +224 -0
  18. layouts/InternScience_MLExpert-IQ4_XS.tensor-types.txt +432 -0
  19. layouts/InternScience_MLExpert-Q2_K.layout.json +229 -0
  20. layouts/InternScience_MLExpert-Q2_K.tensor-types.txt +432 -0
  21. layouts/InternScience_MLExpert-Q3_K_L.layout.json +229 -0
  22. layouts/InternScience_MLExpert-Q3_K_L.tensor-types.txt +432 -0
  23. layouts/InternScience_MLExpert-Q3_K_M.layout.json +228 -0
  24. layouts/InternScience_MLExpert-Q3_K_M.tensor-types.txt +432 -0
  25. layouts/InternScience_MLExpert-Q3_K_S.layout.json +228 -0
  26. layouts/InternScience_MLExpert-Q3_K_S.tensor-types.txt +432 -0
  27. layouts/InternScience_MLExpert-Q4_K_L.layout.json +226 -0
  28. layouts/InternScience_MLExpert-Q4_K_L.tensor-types.txt +432 -0
  29. layouts/InternScience_MLExpert-Q4_K_M.layout.json +226 -0
  30. layouts/InternScience_MLExpert-Q4_K_M.tensor-types.txt +432 -0
  31. layouts/InternScience_MLExpert-Q4_K_S.layout.json +222 -0
  32. layouts/InternScience_MLExpert-Q4_K_S.tensor-types.txt +432 -0
  33. layouts/InternScience_MLExpert-Q5_K_M.layout.json +224 -0
  34. layouts/InternScience_MLExpert-Q5_K_M.tensor-types.txt +432 -0
  35. layouts/InternScience_MLExpert-Q5_K_S.layout.json +223 -0
  36. layouts/InternScience_MLExpert-Q5_K_S.tensor-types.txt +432 -0
  37. layouts/InternScience_MLExpert-Q6_K.layout.json +223 -0
  38. layouts/InternScience_MLExpert-Q6_K.tensor-types.txt +432 -0
  39. layouts/InternScience_MLExpert-Q6_K_L.layout.json +223 -0
  40. layouts/InternScience_MLExpert-Q6_K_L.tensor-types.txt +432 -0
  41. layouts/InternScience_MLExpert-Q6_K_S.layout.json +223 -0
  42. layouts/InternScience_MLExpert-Q6_K_S.tensor-types.txt +432 -0
layouts/InternScience_MLExpert-IQ2_M.layout.json ADDED
@@ -0,0 +1,233 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ2_M",
5
+ "base_type": "iq2_s",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.70+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": 13240500224,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.029,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0181,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 11951144960,
159
+ "map_body+embd": 13118234624,
160
+ "ratio": 1.0977,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.7,
164
+ "body_bpw": 2.865,
165
+ "body_bpw_heuristic": 2.709,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 13222816256,
168
+ "file_bpw_pred": 3.052,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q4_k": 15,
172
+ "q5_k": 14,
173
+ "q8_0": 1
174
+ },
175
+ "attn_k": {
176
+ "q8_0": 7,
177
+ "q6_k": 3
178
+ },
179
+ "attn_output": {
180
+ "q6_k": 9,
181
+ "q8_0": 1
182
+ },
183
+ "attn_q": {
184
+ "q5_k": 10
185
+ },
186
+ "attn_qkv": {
187
+ "q5_k": 29,
188
+ "q6_k": 1
189
+ },
190
+ "attn_v": {
191
+ "q8_0": 10
192
+ },
193
+ "ffn_down_exps": {
194
+ "iq2_s": 30,
195
+ "iq3_s": 6,
196
+ "iq3_xxs": 3,
197
+ "iq4_xs": 1
198
+ },
199
+ "ffn_down_shexp": {
200
+ "q8_0": 40
201
+ },
202
+ "ffn_gate_exps": {
203
+ "iq2_s": 39,
204
+ "iq4_xs": 1
205
+ },
206
+ "ffn_gate_shexp": {
207
+ "q8_0": 40
208
+ },
209
+ "ffn_up_exps": {
210
+ "iq2_s": 29,
211
+ "iq3_s": 10,
212
+ "iq4_xs": 1
213
+ },
214
+ "ffn_up_shexp": {
215
+ "q8_0": 40
216
+ },
217
+ "ssm_out": {
218
+ "q5_k": 26,
219
+ "q6_k": 3,
220
+ "q8_0": 1
221
+ }
222
+ },
223
+ "prior": "layout/prior.json",
224
+ "pin_rule": "floor",
225
+ "embd_cap": "heur",
226
+ "generator": "auto_quant_v2.layout",
227
+ "generator_key": "3bf8b43e20a68f0b",
228
+ "quant_type": "IQ2_M",
229
+ "quantize_ftype": "IQ2_M",
230
+ "llama_cpp_version": "b11716",
231
+ "unknown_kinds": {},
232
+ "commit": "c3c0369c7e82d7b49e1472fe453a182650282f2d"
233
+ }
layouts/InternScience_MLExpert-IQ2_M.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q6_k
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq3_s
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq2_s
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq3_s
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q5_k
73
+ ^blk\.1\.attn_qkv\.weight$=q5_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq3_s
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq2_s
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq3_s
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q6_k
81
+ ^blk\.2\.attn_gate\.weight$=q5_k
82
+ ^blk\.2\.attn_qkv\.weight$=q5_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq3_xxs
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq2_s
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq3_s
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q6_k
90
+ ^blk\.3\.attn_k\.weight$=q6_k
91
+ ^blk\.3\.attn_output\.weight$=q6_k
92
+ ^blk\.3\.attn_q\.weight$=q5_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq3_s
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq2_s
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq3_s
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q5_k
101
+ ^blk\.4\.attn_qkv\.weight$=q5_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq3_xxs
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq2_s
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq3_s
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q6_k
109
+ ^blk\.5\.attn_gate\.weight$=q5_k
110
+ ^blk\.5\.attn_qkv\.weight$=q5_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq3_xxs
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq2_s
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq3_s
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q5_k
118
+ ^blk\.6\.attn_gate\.weight$=q5_k
119
+ ^blk\.6\.attn_qkv\.weight$=q5_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq2_s
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq2_s
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq3_s
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q5_k
127
+ ^blk\.7\.attn_k\.weight$=q6_k
128
+ ^blk\.7\.attn_output\.weight$=q6_k
129
+ ^blk\.7\.attn_q\.weight$=q5_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq3_s
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq2_s
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq3_s
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q5_k
138
+ ^blk\.8\.attn_qkv\.weight$=q5_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq2_s
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq2_s
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq3_s
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q5_k
146
+ ^blk\.9\.attn_gate\.weight$=q5_k
147
+ ^blk\.9\.attn_qkv\.weight$=q5_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq2_s
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq2_s
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq3_s
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q5_k
155
+ ^blk\.10\.attn_gate\.weight$=q4_k
156
+ ^blk\.10\.attn_qkv\.weight$=q5_k
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq2_s
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq2_s
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq2_s
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q5_k
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q6_k
166
+ ^blk\.11\.attn_q\.weight$=q5_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq2_s
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq2_s
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq2_s
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q4_k
175
+ ^blk\.12\.attn_qkv\.weight$=q5_k
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq2_s
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq2_s
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq2_s
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q5_k
183
+ ^blk\.13\.attn_gate\.weight$=q4_k
184
+ ^blk\.13\.attn_qkv\.weight$=q5_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq2_s
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq2_s
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq2_s
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q5_k
192
+ ^blk\.14\.attn_gate\.weight$=q4_k
193
+ ^blk\.14\.attn_qkv\.weight$=q5_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq2_s
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq2_s
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq2_s
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q5_k
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q6_k
203
+ ^blk\.15\.attn_q\.weight$=q5_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq2_s
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq2_s
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq2_s
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q4_k
212
+ ^blk\.16\.attn_qkv\.weight$=q5_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq2_s
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq2_s
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq2_s
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q5_k
220
+ ^blk\.17\.attn_gate\.weight$=q4_k
221
+ ^blk\.17\.attn_qkv\.weight$=q5_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq2_s
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq2_s
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq2_s
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q5_k
229
+ ^blk\.18\.attn_gate\.weight$=q4_k
230
+ ^blk\.18\.attn_qkv\.weight$=q5_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq2_s
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq2_s
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq2_s
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q5_k
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q6_k
240
+ ^blk\.19\.attn_q\.weight$=q5_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq2_s
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq2_s
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq2_s
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q4_k
249
+ ^blk\.20\.attn_qkv\.weight$=q5_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq2_s
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq2_s
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq2_s
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q5_k
257
+ ^blk\.21\.attn_gate\.weight$=q4_k
258
+ ^blk\.21\.attn_qkv\.weight$=q5_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq2_s
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq2_s
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq2_s
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q5_k
266
+ ^blk\.22\.attn_gate\.weight$=q4_k
267
+ ^blk\.22\.attn_qkv\.weight$=q5_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq2_s
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq2_s
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq2_s
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q5_k
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q6_k
277
+ ^blk\.23\.attn_q\.weight$=q5_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq2_s
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq2_s
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq2_s
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q4_k
286
+ ^blk\.24\.attn_qkv\.weight$=q5_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq2_s
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq2_s
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq2_s
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q5_k
294
+ ^blk\.25\.attn_gate\.weight$=q4_k
295
+ ^blk\.25\.attn_qkv\.weight$=q5_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq2_s
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq2_s
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq2_s
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q5_k
303
+ ^blk\.26\.attn_gate\.weight$=q4_k
304
+ ^blk\.26\.attn_qkv\.weight$=q5_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq2_s
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq2_s
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq2_s
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q5_k
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q6_k
314
+ ^blk\.27\.attn_q\.weight$=q5_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq2_s
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq2_s
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq2_s
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q4_k
323
+ ^blk\.28\.attn_qkv\.weight$=q5_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq2_s
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq2_s
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq2_s
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q5_k
331
+ ^blk\.29\.attn_gate\.weight$=q4_k
332
+ ^blk\.29\.attn_qkv\.weight$=q5_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq2_s
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq2_s
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq2_s
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q5_k
340
+ ^blk\.30\.attn_gate\.weight$=q5_k
341
+ ^blk\.30\.attn_qkv\.weight$=q5_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq2_s
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq2_s
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq2_s
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q5_k
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q6_k
351
+ ^blk\.31\.attn_q\.weight$=q5_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq3_s
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq2_s
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq2_s
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q5_k
360
+ ^blk\.32\.attn_qkv\.weight$=q5_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq2_s
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq2_s
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq2_s
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q5_k
368
+ ^blk\.33\.attn_gate\.weight$=q5_k
369
+ ^blk\.33\.attn_qkv\.weight$=q5_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq2_s
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq2_s
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq2_s
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q5_k
377
+ ^blk\.34\.attn_gate\.weight$=q5_k
378
+ ^blk\.34\.attn_qkv\.weight$=q5_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq2_s
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq2_s
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq2_s
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q5_k
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q6_k
388
+ ^blk\.35\.attn_q\.weight$=q5_k
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq3_s
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq2_s
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq2_s
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q5_k
397
+ ^blk\.36\.attn_qkv\.weight$=q5_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq2_s
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq2_s
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq2_s
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q5_k
405
+ ^blk\.37\.attn_gate\.weight$=q5_k
406
+ ^blk\.37\.attn_qkv\.weight$=q5_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq2_s
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq2_s
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq2_s
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q5_k
414
+ ^blk\.38\.attn_gate\.weight$=q5_k
415
+ ^blk\.38\.attn_qkv\.weight$=q5_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq2_s
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq2_s
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq2_s
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q5_k
423
+ ^blk\.39\.attn_k\.weight$=q6_k
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q5_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=iq4_xs
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq4_xs
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=iq4_xs
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ2_S.layout.json ADDED
@@ -0,0 +1,236 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ2_S",
5
+ "base_type": "iq2_xs",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.70+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": 11974737920,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0318,
17
+ "type": "q5_k"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0199,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q5_k",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 10894704640,
159
+ "map_body+embd": 11863973888,
160
+ "ratio": 1.089,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.7,
164
+ "body_bpw": 2.612,
165
+ "body_bpw_heuristic": 2.457,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 11968555520,
168
+ "file_bpw_pred": 2.762,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "iq4_xs": 15,
172
+ "q5_k": 7,
173
+ "q4_k": 7,
174
+ "q6_k": 1
175
+ },
176
+ "attn_k": {
177
+ "q6_k": 8,
178
+ "q8_0": 2
179
+ },
180
+ "attn_output": {
181
+ "q5_k": 9,
182
+ "q6_k": 1
183
+ },
184
+ "attn_q": {
185
+ "q4_k": 5,
186
+ "iq4_xs": 5
187
+ },
188
+ "attn_qkv": {
189
+ "q4_k": 22,
190
+ "iq4_xs": 7,
191
+ "q5_k": 1
192
+ },
193
+ "attn_v": {
194
+ "q8_0": 10
195
+ },
196
+ "ffn_down_exps": {
197
+ "iq2_xs": 31,
198
+ "iq3_xxs": 7,
199
+ "iq3_s": 1,
200
+ "iq4_xs": 1
201
+ },
202
+ "ffn_down_shexp": {
203
+ "q8_0": 40
204
+ },
205
+ "ffn_gate_exps": {
206
+ "iq2_xs": 39,
207
+ "iq3_s": 1
208
+ },
209
+ "ffn_gate_shexp": {
210
+ "q8_0": 25,
211
+ "q6_k": 15
212
+ },
213
+ "ffn_up_exps": {
214
+ "iq2_xs": 29,
215
+ "iq3_s": 10,
216
+ "iq4_xs": 1
217
+ },
218
+ "ffn_up_shexp": {
219
+ "q8_0": 40
220
+ },
221
+ "ssm_out": {
222
+ "q5_k": 29,
223
+ "q6_k": 1
224
+ }
225
+ },
226
+ "prior": "layout/prior.json",
227
+ "pin_rule": "floor",
228
+ "embd_cap": "heur",
229
+ "generator": "auto_quant_v2.layout",
230
+ "generator_key": "3bf8b43e20a68f0b",
231
+ "quant_type": "IQ2_S",
232
+ "quantize_ftype": "IQ2_S",
233
+ "llama_cpp_version": "b11716",
234
+ "unknown_kinds": {},
235
+ "commit": "c3c0369c7e82d7b49e1472fe453a182650282f2d"
236
+ }
layouts/InternScience_MLExpert-IQ2_S.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q5_k
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q6_k
64
+ ^blk\.0\.attn_qkv\.weight$=q5_k
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq3_s
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq2_xs
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq3_s
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q6_k
72
+ ^blk\.1\.attn_gate\.weight$=q5_k
73
+ ^blk\.1\.attn_qkv\.weight$=q4_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq3_xxs
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq2_xs
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq3_s
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q5_k
81
+ ^blk\.2\.attn_gate\.weight$=q5_k
82
+ ^blk\.2\.attn_qkv\.weight$=q4_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq3_xxs
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq2_xs
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq3_s
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q5_k
90
+ ^blk\.3\.attn_k\.weight$=q6_k
91
+ ^blk\.3\.attn_output\.weight$=q5_k
92
+ ^blk\.3\.attn_q\.weight$=q4_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq3_xxs
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq2_xs
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq3_s
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q5_k
101
+ ^blk\.4\.attn_qkv\.weight$=q4_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq3_xxs
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq2_xs
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq3_s
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q5_k
109
+ ^blk\.5\.attn_gate\.weight$=q5_k
110
+ ^blk\.5\.attn_qkv\.weight$=q4_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq2_xs
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq2_xs
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq3_s
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q5_k
118
+ ^blk\.6\.attn_gate\.weight$=q5_k
119
+ ^blk\.6\.attn_qkv\.weight$=q4_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq2_xs
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq2_xs
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq3_s
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q5_k
127
+ ^blk\.7\.attn_k\.weight$=q6_k
128
+ ^blk\.7\.attn_output\.weight$=q5_k
129
+ ^blk\.7\.attn_q\.weight$=q4_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq3_xxs
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq2_xs
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq3_s
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q5_k
138
+ ^blk\.8\.attn_qkv\.weight$=q4_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq2_xs
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq2_xs
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq3_s
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q5_k
146
+ ^blk\.9\.attn_gate\.weight$=q5_k
147
+ ^blk\.9\.attn_qkv\.weight$=q4_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq2_xs
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq2_xs
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq3_s
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q5_k
155
+ ^blk\.10\.attn_gate\.weight$=iq4_xs
156
+ ^blk\.10\.attn_qkv\.weight$=q4_k
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq2_xs
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq2_xs
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq2_xs
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q5_k
164
+ ^blk\.11\.attn_k\.weight$=q6_k
165
+ ^blk\.11\.attn_output\.weight$=q5_k
166
+ ^blk\.11\.attn_q\.weight$=iq4_xs
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq2_xs
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq2_xs
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q6_k
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq2_xs
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=iq4_xs
175
+ ^blk\.12\.attn_qkv\.weight$=q4_k
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq2_xs
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq2_xs
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq2_xs
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q5_k
183
+ ^blk\.13\.attn_gate\.weight$=iq4_xs
184
+ ^blk\.13\.attn_qkv\.weight$=q4_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq2_xs
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq2_xs
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq2_xs
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q5_k
192
+ ^blk\.14\.attn_gate\.weight$=iq4_xs
193
+ ^blk\.14\.attn_qkv\.weight$=q4_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq2_xs
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq2_xs
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq2_xs
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q5_k
201
+ ^blk\.15\.attn_k\.weight$=q6_k
202
+ ^blk\.15\.attn_output\.weight$=q5_k
203
+ ^blk\.15\.attn_q\.weight$=iq4_xs
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq2_xs
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq2_xs
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q6_k
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq2_xs
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=iq4_xs
212
+ ^blk\.16\.attn_qkv\.weight$=q4_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq2_xs
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq2_xs
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq2_xs
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q5_k
220
+ ^blk\.17\.attn_gate\.weight$=iq4_xs
221
+ ^blk\.17\.attn_qkv\.weight$=q4_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq2_xs
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq2_xs
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q6_k
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq2_xs
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q5_k
229
+ ^blk\.18\.attn_gate\.weight$=iq4_xs
230
+ ^blk\.18\.attn_qkv\.weight$=q4_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq2_xs
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq2_xs
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q6_k
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq2_xs
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q5_k
238
+ ^blk\.19\.attn_k\.weight$=q6_k
239
+ ^blk\.19\.attn_output\.weight$=q5_k
240
+ ^blk\.19\.attn_q\.weight$=iq4_xs
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq2_xs
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq2_xs
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q6_k
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq2_xs
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=iq4_xs
249
+ ^blk\.20\.attn_qkv\.weight$=q4_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq2_xs
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq2_xs
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q6_k
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq2_xs
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q5_k
257
+ ^blk\.21\.attn_gate\.weight$=iq4_xs
258
+ ^blk\.21\.attn_qkv\.weight$=q4_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq2_xs
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq2_xs
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q6_k
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq2_xs
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q5_k
266
+ ^blk\.22\.attn_gate\.weight$=iq4_xs
267
+ ^blk\.22\.attn_qkv\.weight$=q4_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq2_xs
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq2_xs
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q6_k
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq2_xs
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q5_k
275
+ ^blk\.23\.attn_k\.weight$=q6_k
276
+ ^blk\.23\.attn_output\.weight$=q5_k
277
+ ^blk\.23\.attn_q\.weight$=iq4_xs
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq2_xs
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq2_xs
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q6_k
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq2_xs
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=iq4_xs
286
+ ^blk\.24\.attn_qkv\.weight$=q4_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq2_xs
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq2_xs
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q6_k
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq2_xs
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q5_k
294
+ ^blk\.25\.attn_gate\.weight$=iq4_xs
295
+ ^blk\.25\.attn_qkv\.weight$=q4_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq2_xs
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq2_xs
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q6_k
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq2_xs
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q5_k
303
+ ^blk\.26\.attn_gate\.weight$=iq4_xs
304
+ ^blk\.26\.attn_qkv\.weight$=q4_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq2_xs
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq2_xs
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q6_k
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq2_xs
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q5_k
312
+ ^blk\.27\.attn_k\.weight$=q6_k
313
+ ^blk\.27\.attn_output\.weight$=q5_k
314
+ ^blk\.27\.attn_q\.weight$=iq4_xs
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq2_xs
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq2_xs
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q6_k
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq2_xs
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=iq4_xs
323
+ ^blk\.28\.attn_qkv\.weight$=q4_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq2_xs
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq2_xs
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q6_k
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq2_xs
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q5_k
331
+ ^blk\.29\.attn_gate\.weight$=iq4_xs
332
+ ^blk\.29\.attn_qkv\.weight$=q4_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq2_xs
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq2_xs
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q6_k
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq2_xs
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q5_k
340
+ ^blk\.30\.attn_gate\.weight$=q4_k
341
+ ^blk\.30\.attn_qkv\.weight$=iq4_xs
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq2_xs
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq2_xs
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq2_xs
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q5_k
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q5_k
351
+ ^blk\.31\.attn_q\.weight$=q4_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq3_xxs
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq2_xs
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq2_xs
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q4_k
360
+ ^blk\.32\.attn_qkv\.weight$=iq4_xs
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq2_xs
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq2_xs
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq2_xs
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q5_k
368
+ ^blk\.33\.attn_gate\.weight$=q4_k
369
+ ^blk\.33\.attn_qkv\.weight$=iq4_xs
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq2_xs
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq2_xs
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq2_xs
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q5_k
377
+ ^blk\.34\.attn_gate\.weight$=q4_k
378
+ ^blk\.34\.attn_qkv\.weight$=iq4_xs
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq2_xs
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq2_xs
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq2_xs
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q5_k
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q5_k
388
+ ^blk\.35\.attn_q\.weight$=q4_k
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq3_xxs
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq2_xs
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq2_xs
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q4_k
397
+ ^blk\.36\.attn_qkv\.weight$=iq4_xs
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq2_xs
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq2_xs
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq2_xs
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q5_k
405
+ ^blk\.37\.attn_gate\.weight$=q4_k
406
+ ^blk\.37\.attn_qkv\.weight$=iq4_xs
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq2_xs
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq2_xs
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq2_xs
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q5_k
414
+ ^blk\.38\.attn_gate\.weight$=q4_k
415
+ ^blk\.38\.attn_qkv\.weight$=iq4_xs
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq2_xs
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq2_xs
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq2_xs
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q5_k
423
+ ^blk\.39\.attn_k\.weight$=q6_k
424
+ ^blk\.39\.attn_output\.weight$=q6_k
425
+ ^blk\.39\.attn_q\.weight$=q4_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=iq4_xs
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq3_s
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=iq4_xs
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ2_XS.layout.json ADDED
@@ -0,0 +1,229 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ2_XS",
5
+ "base_type": "iq2_xs",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.90+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": 11149688832,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0324,
17
+ "type": "q5_k"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0155,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q5_k",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 10681573376,
159
+ "map_body+embd": 11149688832,
160
+ "ratio": 1.0438,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.9,
164
+ "body_bpw": 2.442,
165
+ "body_bpw_heuristic": 2.419,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 11254270464,
168
+ "file_bpw_pred": 2.598,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "iq4_xs": 21,
172
+ "q4_k": 8,
173
+ "q6_k": 1
174
+ },
175
+ "attn_k": {
176
+ "q6_k": 10
177
+ },
178
+ "attn_output": {
179
+ "q5_k": 9,
180
+ "q6_k": 1
181
+ },
182
+ "attn_q": {
183
+ "iq4_xs": 8,
184
+ "q4_k": 2
185
+ },
186
+ "attn_qkv": {
187
+ "iq4_xs": 22,
188
+ "q4_k": 7,
189
+ "q5_k": 1
190
+ },
191
+ "attn_v": {
192
+ "q8_0": 10
193
+ },
194
+ "ffn_down_exps": {
195
+ "iq2_xs": 39,
196
+ "iq4_xs": 1
197
+ },
198
+ "ffn_down_shexp": {
199
+ "q8_0": 40
200
+ },
201
+ "ffn_gate_exps": {
202
+ "iq2_xs": 40
203
+ },
204
+ "ffn_gate_shexp": {
205
+ "q8_0": 20,
206
+ "q6_k": 20
207
+ },
208
+ "ffn_up_exps": {
209
+ "iq2_xs": 40
210
+ },
211
+ "ffn_up_shexp": {
212
+ "q8_0": 40
213
+ },
214
+ "ssm_out": {
215
+ "q5_k": 29,
216
+ "q6_k": 1
217
+ }
218
+ },
219
+ "prior": "layout/prior.json",
220
+ "pin_rule": "floor",
221
+ "embd_cap": "heur",
222
+ "generator": "auto_quant_v2.layout",
223
+ "generator_key": "3bf8b43e20a68f0b",
224
+ "quant_type": "IQ2_XS",
225
+ "quantize_ftype": "IQ2_XS",
226
+ "llama_cpp_version": "b11716",
227
+ "unknown_kinds": {},
228
+ "commit": "1ac1b04da659470c625135a945d427b40b5da68b"
229
+ }
layouts/InternScience_MLExpert-IQ2_XS.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q5_k
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q6_k
64
+ ^blk\.0\.attn_qkv\.weight$=q5_k
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq2_xs
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq2_xs
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq2_xs
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q6_k
72
+ ^blk\.1\.attn_gate\.weight$=q4_k
73
+ ^blk\.1\.attn_qkv\.weight$=q4_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq2_xs
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq2_xs
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq2_xs
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q5_k
81
+ ^blk\.2\.attn_gate\.weight$=q4_k
82
+ ^blk\.2\.attn_qkv\.weight$=q4_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq2_xs
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq2_xs
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq2_xs
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q5_k
90
+ ^blk\.3\.attn_k\.weight$=q6_k
91
+ ^blk\.3\.attn_output\.weight$=q5_k
92
+ ^blk\.3\.attn_q\.weight$=iq4_xs
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq2_xs
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq2_xs
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq2_xs
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q4_k
101
+ ^blk\.4\.attn_qkv\.weight$=q4_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq2_xs
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq2_xs
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq2_xs
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q5_k
109
+ ^blk\.5\.attn_gate\.weight$=q4_k
110
+ ^blk\.5\.attn_qkv\.weight$=q4_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq2_xs
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq2_xs
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq2_xs
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q5_k
118
+ ^blk\.6\.attn_gate\.weight$=q4_k
119
+ ^blk\.6\.attn_qkv\.weight$=q4_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq2_xs
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq2_xs
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq2_xs
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q5_k
127
+ ^blk\.7\.attn_k\.weight$=q6_k
128
+ ^blk\.7\.attn_output\.weight$=q5_k
129
+ ^blk\.7\.attn_q\.weight$=iq4_xs
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq2_xs
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq2_xs
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq2_xs
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q4_k
138
+ ^blk\.8\.attn_qkv\.weight$=q4_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq2_xs
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq2_xs
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq2_xs
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q5_k
146
+ ^blk\.9\.attn_gate\.weight$=q4_k
147
+ ^blk\.9\.attn_qkv\.weight$=q4_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq2_xs
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq2_xs
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq2_xs
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q5_k
155
+ ^blk\.10\.attn_gate\.weight$=iq4_xs
156
+ ^blk\.10\.attn_qkv\.weight$=iq4_xs
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq2_xs
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq2_xs
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q6_k
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq2_xs
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q5_k
164
+ ^blk\.11\.attn_k\.weight$=q6_k
165
+ ^blk\.11\.attn_output\.weight$=q5_k
166
+ ^blk\.11\.attn_q\.weight$=iq4_xs
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq2_xs
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq2_xs
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q6_k
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq2_xs
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=iq4_xs
175
+ ^blk\.12\.attn_qkv\.weight$=iq4_xs
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq2_xs
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq2_xs
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q6_k
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq2_xs
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q5_k
183
+ ^blk\.13\.attn_gate\.weight$=iq4_xs
184
+ ^blk\.13\.attn_qkv\.weight$=iq4_xs
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq2_xs
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq2_xs
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q6_k
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq2_xs
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q5_k
192
+ ^blk\.14\.attn_gate\.weight$=iq4_xs
193
+ ^blk\.14\.attn_qkv\.weight$=iq4_xs
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq2_xs
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq2_xs
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q6_k
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq2_xs
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q5_k
201
+ ^blk\.15\.attn_k\.weight$=q6_k
202
+ ^blk\.15\.attn_output\.weight$=q5_k
203
+ ^blk\.15\.attn_q\.weight$=iq4_xs
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq2_xs
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq2_xs
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q6_k
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq2_xs
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=iq4_xs
212
+ ^blk\.16\.attn_qkv\.weight$=iq4_xs
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq2_xs
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq2_xs
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q6_k
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq2_xs
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q5_k
220
+ ^blk\.17\.attn_gate\.weight$=iq4_xs
221
+ ^blk\.17\.attn_qkv\.weight$=iq4_xs
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq2_xs
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq2_xs
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q6_k
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq2_xs
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q5_k
229
+ ^blk\.18\.attn_gate\.weight$=iq4_xs
230
+ ^blk\.18\.attn_qkv\.weight$=iq4_xs
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq2_xs
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq2_xs
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q6_k
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq2_xs
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q5_k
238
+ ^blk\.19\.attn_k\.weight$=q6_k
239
+ ^blk\.19\.attn_output\.weight$=q5_k
240
+ ^blk\.19\.attn_q\.weight$=iq4_xs
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq2_xs
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq2_xs
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q6_k
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq2_xs
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=iq4_xs
249
+ ^blk\.20\.attn_qkv\.weight$=iq4_xs
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq2_xs
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq2_xs
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q6_k
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq2_xs
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q5_k
257
+ ^blk\.21\.attn_gate\.weight$=iq4_xs
258
+ ^blk\.21\.attn_qkv\.weight$=iq4_xs
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq2_xs
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq2_xs
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q6_k
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq2_xs
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q5_k
266
+ ^blk\.22\.attn_gate\.weight$=iq4_xs
267
+ ^blk\.22\.attn_qkv\.weight$=iq4_xs
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq2_xs
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq2_xs
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q6_k
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq2_xs
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q5_k
275
+ ^blk\.23\.attn_k\.weight$=q6_k
276
+ ^blk\.23\.attn_output\.weight$=q5_k
277
+ ^blk\.23\.attn_q\.weight$=iq4_xs
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq2_xs
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq2_xs
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q6_k
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq2_xs
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=iq4_xs
286
+ ^blk\.24\.attn_qkv\.weight$=iq4_xs
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq2_xs
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq2_xs
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q6_k
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq2_xs
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q5_k
294
+ ^blk\.25\.attn_gate\.weight$=iq4_xs
295
+ ^blk\.25\.attn_qkv\.weight$=iq4_xs
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq2_xs
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq2_xs
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q6_k
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq2_xs
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q5_k
303
+ ^blk\.26\.attn_gate\.weight$=iq4_xs
304
+ ^blk\.26\.attn_qkv\.weight$=iq4_xs
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq2_xs
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq2_xs
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q6_k
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq2_xs
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q5_k
312
+ ^blk\.27\.attn_k\.weight$=q6_k
313
+ ^blk\.27\.attn_output\.weight$=q5_k
314
+ ^blk\.27\.attn_q\.weight$=iq4_xs
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq2_xs
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq2_xs
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q6_k
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq2_xs
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=iq4_xs
323
+ ^blk\.28\.attn_qkv\.weight$=iq4_xs
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq2_xs
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq2_xs
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q6_k
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq2_xs
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q5_k
331
+ ^blk\.29\.attn_gate\.weight$=iq4_xs
332
+ ^blk\.29\.attn_qkv\.weight$=iq4_xs
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq2_xs
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq2_xs
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q6_k
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq2_xs
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q5_k
340
+ ^blk\.30\.attn_gate\.weight$=q4_k
341
+ ^blk\.30\.attn_qkv\.weight$=iq4_xs
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq2_xs
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq2_xs
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq2_xs
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q5_k
349
+ ^blk\.31\.attn_k\.weight$=q6_k
350
+ ^blk\.31\.attn_output\.weight$=q5_k
351
+ ^blk\.31\.attn_q\.weight$=q4_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq2_xs
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq2_xs
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq2_xs
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=iq4_xs
360
+ ^blk\.32\.attn_qkv\.weight$=iq4_xs
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq2_xs
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq2_xs
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq2_xs
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q5_k
368
+ ^blk\.33\.attn_gate\.weight$=iq4_xs
369
+ ^blk\.33\.attn_qkv\.weight$=iq4_xs
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq2_xs
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq2_xs
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq2_xs
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q5_k
377
+ ^blk\.34\.attn_gate\.weight$=iq4_xs
378
+ ^blk\.34\.attn_qkv\.weight$=iq4_xs
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq2_xs
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq2_xs
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq2_xs
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q5_k
386
+ ^blk\.35\.attn_k\.weight$=q6_k
387
+ ^blk\.35\.attn_output\.weight$=q5_k
388
+ ^blk\.35\.attn_q\.weight$=iq4_xs
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq2_xs
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq2_xs
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq2_xs
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=iq4_xs
397
+ ^blk\.36\.attn_qkv\.weight$=iq4_xs
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq2_xs
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq2_xs
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq2_xs
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q5_k
405
+ ^blk\.37\.attn_gate\.weight$=iq4_xs
406
+ ^blk\.37\.attn_qkv\.weight$=iq4_xs
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq2_xs
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq2_xs
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq2_xs
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q5_k
414
+ ^blk\.38\.attn_gate\.weight$=iq4_xs
415
+ ^blk\.38\.attn_qkv\.weight$=iq4_xs
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq2_xs
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq2_xs
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq2_xs
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q5_k
423
+ ^blk\.39\.attn_k\.weight$=q6_k
424
+ ^blk\.39\.attn_output\.weight$=q6_k
425
+ ^blk\.39\.attn_q\.weight$=q4_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=iq4_xs
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq2_xs
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=iq2_xs
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ2_XXS.layout.json ADDED
@@ -0,0 +1,236 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ2_XXS",
5
+ "base_type": "iq2_xxs",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.70+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": 10776535040,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0358,
17
+ "type": "q5_k"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0171,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q5_k",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 9662881792,
159
+ "map_body+embd": 10677624832,
160
+ "ratio": 1.105,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.7,
164
+ "body_bpw": 2.329,
165
+ "body_bpw_heuristic": 2.177,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 10782206464,
168
+ "file_bpw_pred": 2.489,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "iq4_xs": 29,
172
+ "q5_k": 1
173
+ },
174
+ "attn_k": {
175
+ "q5_k": 8,
176
+ "q6_k": 2
177
+ },
178
+ "attn_output": {
179
+ "q5_k": 10
180
+ },
181
+ "attn_q": {
182
+ "iq4_xs": 10
183
+ },
184
+ "attn_qkv": {
185
+ "iq4_xs": 29,
186
+ "q5_k": 1
187
+ },
188
+ "attn_v": {
189
+ "q6_k": 10
190
+ },
191
+ "ffn_down_exps": {
192
+ "iq2_xxs": 31,
193
+ "iq3_xxs": 5,
194
+ "iq2_s": 2,
195
+ "iq2_xs": 1,
196
+ "iq3_s": 1
197
+ },
198
+ "ffn_down_shexp": {
199
+ "q6_k": 34,
200
+ "q8_0": 6
201
+ },
202
+ "ffn_gate_exps": {
203
+ "iq2_xxs": 39,
204
+ "iq3_xxs": 1
205
+ },
206
+ "ffn_gate_shexp": {
207
+ "q6_k": 26,
208
+ "q5_k": 13,
209
+ "q8_0": 1
210
+ },
211
+ "ffn_up_exps": {
212
+ "iq2_xxs": 29,
213
+ "iq3_xxs": 10,
214
+ "iq3_s": 1
215
+ },
216
+ "ffn_up_shexp": {
217
+ "q6_k": 29,
218
+ "q8_0": 11
219
+ },
220
+ "ssm_out": {
221
+ "iq4_xs": 22,
222
+ "q4_k": 7,
223
+ "q6_k": 1
224
+ }
225
+ },
226
+ "prior": "layout/prior.json",
227
+ "pin_rule": "floor",
228
+ "embd_cap": "heur",
229
+ "generator": "auto_quant_v2.layout",
230
+ "generator_key": "3bf8b43e20a68f0b",
231
+ "quant_type": "IQ2_XXS",
232
+ "quantize_ftype": "IQ2_XXS",
233
+ "llama_cpp_version": "b11716",
234
+ "unknown_kinds": {},
235
+ "commit": "1ac1b04da659470c625135a945d427b40b5da68b"
236
+ }
layouts/InternScience_MLExpert-IQ2_XXS.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q5_k
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q5_k
64
+ ^blk\.0\.attn_qkv\.weight$=q5_k
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq3_xxs
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq2_xxs
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q6_k
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq3_xxs
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q6_k
72
+ ^blk\.1\.attn_gate\.weight$=iq4_xs
73
+ ^blk\.1\.attn_qkv\.weight$=iq4_xs
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq2_s
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q6_k
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq2_xxs
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q6_k
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq3_xxs
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q4_k
81
+ ^blk\.2\.attn_gate\.weight$=iq4_xs
82
+ ^blk\.2\.attn_qkv\.weight$=iq4_xs
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq2_s
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q6_k
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq2_xxs
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q6_k
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq3_xxs
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q4_k
90
+ ^blk\.3\.attn_k\.weight$=q5_k
91
+ ^blk\.3\.attn_output\.weight$=q5_k
92
+ ^blk\.3\.attn_q\.weight$=iq4_xs
93
+ ^blk\.3\.attn_v\.weight$=q6_k
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq3_xxs
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq2_xxs
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q6_k
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq3_xxs
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=iq4_xs
101
+ ^blk\.4\.attn_qkv\.weight$=iq4_xs
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq2_xs
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q6_k
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq2_xxs
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q6_k
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq3_xxs
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q4_k
109
+ ^blk\.5\.attn_gate\.weight$=iq4_xs
110
+ ^blk\.5\.attn_qkv\.weight$=iq4_xs
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq2_xxs
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q6_k
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq2_xxs
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q6_k
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq3_xxs
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q4_k
118
+ ^blk\.6\.attn_gate\.weight$=iq4_xs
119
+ ^blk\.6\.attn_qkv\.weight$=iq4_xs
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq2_xxs
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q6_k
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq2_xxs
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q6_k
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq3_xxs
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q4_k
127
+ ^blk\.7\.attn_k\.weight$=q5_k
128
+ ^blk\.7\.attn_output\.weight$=q5_k
129
+ ^blk\.7\.attn_q\.weight$=iq4_xs
130
+ ^blk\.7\.attn_v\.weight$=q6_k
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq3_xxs
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq2_xxs
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q6_k
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq3_xxs
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=iq4_xs
138
+ ^blk\.8\.attn_qkv\.weight$=iq4_xs
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq2_xxs
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q6_k
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq2_xxs
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q6_k
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq3_xxs
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q4_k
146
+ ^blk\.9\.attn_gate\.weight$=iq4_xs
147
+ ^blk\.9\.attn_qkv\.weight$=iq4_xs
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq2_xxs
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q6_k
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq2_xxs
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q6_k
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq3_xxs
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q4_k
155
+ ^blk\.10\.attn_gate\.weight$=iq4_xs
156
+ ^blk\.10\.attn_qkv\.weight$=iq4_xs
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq2_xxs
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q6_k
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq2_xxs
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q6_k
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq2_xxs
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q6_k
163
+ ^blk\.10\.ssm_out\.weight$=iq4_xs
164
+ ^blk\.11\.attn_k\.weight$=q5_k
165
+ ^blk\.11\.attn_output\.weight$=q5_k
166
+ ^blk\.11\.attn_q\.weight$=iq4_xs
167
+ ^blk\.11\.attn_v\.weight$=q6_k
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq2_xxs
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q6_k
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq2_xxs
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q5_k
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq2_xxs
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q6_k
174
+ ^blk\.12\.attn_gate\.weight$=iq4_xs
175
+ ^blk\.12\.attn_qkv\.weight$=iq4_xs
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq2_xxs
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q6_k
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq2_xxs
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q6_k
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq2_xxs
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q6_k
182
+ ^blk\.12\.ssm_out\.weight$=iq4_xs
183
+ ^blk\.13\.attn_gate\.weight$=iq4_xs
184
+ ^blk\.13\.attn_qkv\.weight$=iq4_xs
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq2_xxs
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q6_k
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq2_xxs
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q6_k
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq2_xxs
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q6_k
191
+ ^blk\.13\.ssm_out\.weight$=iq4_xs
192
+ ^blk\.14\.attn_gate\.weight$=iq4_xs
193
+ ^blk\.14\.attn_qkv\.weight$=iq4_xs
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq2_xxs
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q6_k
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq2_xxs
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q6_k
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq2_xxs
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q6_k
200
+ ^blk\.14\.ssm_out\.weight$=iq4_xs
201
+ ^blk\.15\.attn_k\.weight$=q5_k
202
+ ^blk\.15\.attn_output\.weight$=q5_k
203
+ ^blk\.15\.attn_q\.weight$=iq4_xs
204
+ ^blk\.15\.attn_v\.weight$=q6_k
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq2_xxs
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q6_k
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq2_xxs
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q5_k
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq2_xxs
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q6_k
211
+ ^blk\.16\.attn_gate\.weight$=iq4_xs
212
+ ^blk\.16\.attn_qkv\.weight$=iq4_xs
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq2_xxs
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q6_k
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq2_xxs
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q6_k
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq2_xxs
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q6_k
219
+ ^blk\.16\.ssm_out\.weight$=iq4_xs
220
+ ^blk\.17\.attn_gate\.weight$=iq4_xs
221
+ ^blk\.17\.attn_qkv\.weight$=iq4_xs
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq2_xxs
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q6_k
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq2_xxs
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q6_k
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq2_xxs
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q6_k
228
+ ^blk\.17\.ssm_out\.weight$=iq4_xs
229
+ ^blk\.18\.attn_gate\.weight$=iq4_xs
230
+ ^blk\.18\.attn_qkv\.weight$=iq4_xs
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq2_xxs
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q6_k
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq2_xxs
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q6_k
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq2_xxs
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q6_k
237
+ ^blk\.18\.ssm_out\.weight$=iq4_xs
238
+ ^blk\.19\.attn_k\.weight$=q5_k
239
+ ^blk\.19\.attn_output\.weight$=q5_k
240
+ ^blk\.19\.attn_q\.weight$=iq4_xs
241
+ ^blk\.19\.attn_v\.weight$=q6_k
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq2_xxs
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q6_k
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq2_xxs
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q5_k
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq2_xxs
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q6_k
248
+ ^blk\.20\.attn_gate\.weight$=iq4_xs
249
+ ^blk\.20\.attn_qkv\.weight$=iq4_xs
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq2_xxs
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q6_k
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq2_xxs
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q5_k
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq2_xxs
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q6_k
256
+ ^blk\.20\.ssm_out\.weight$=iq4_xs
257
+ ^blk\.21\.attn_gate\.weight$=iq4_xs
258
+ ^blk\.21\.attn_qkv\.weight$=iq4_xs
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq2_xxs
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q6_k
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq2_xxs
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q5_k
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq2_xxs
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q6_k
265
+ ^blk\.21\.ssm_out\.weight$=iq4_xs
266
+ ^blk\.22\.attn_gate\.weight$=iq4_xs
267
+ ^blk\.22\.attn_qkv\.weight$=iq4_xs
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq2_xxs
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q6_k
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq2_xxs
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q5_k
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq2_xxs
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q6_k
274
+ ^blk\.22\.ssm_out\.weight$=iq4_xs
275
+ ^blk\.23\.attn_k\.weight$=q5_k
276
+ ^blk\.23\.attn_output\.weight$=q5_k
277
+ ^blk\.23\.attn_q\.weight$=iq4_xs
278
+ ^blk\.23\.attn_v\.weight$=q6_k
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq2_xxs
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q6_k
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq2_xxs
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q5_k
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq2_xxs
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q6_k
285
+ ^blk\.24\.attn_gate\.weight$=iq4_xs
286
+ ^blk\.24\.attn_qkv\.weight$=iq4_xs
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq2_xxs
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q6_k
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq2_xxs
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q5_k
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq2_xxs
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q6_k
293
+ ^blk\.24\.ssm_out\.weight$=iq4_xs
294
+ ^blk\.25\.attn_gate\.weight$=iq4_xs
295
+ ^blk\.25\.attn_qkv\.weight$=iq4_xs
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq2_xxs
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q6_k
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq2_xxs
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q5_k
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq2_xxs
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q6_k
302
+ ^blk\.25\.ssm_out\.weight$=iq4_xs
303
+ ^blk\.26\.attn_gate\.weight$=iq4_xs
304
+ ^blk\.26\.attn_qkv\.weight$=iq4_xs
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq2_xxs
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q6_k
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq2_xxs
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q5_k
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq2_xxs
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q6_k
311
+ ^blk\.26\.ssm_out\.weight$=iq4_xs
312
+ ^blk\.27\.attn_k\.weight$=q5_k
313
+ ^blk\.27\.attn_output\.weight$=q5_k
314
+ ^blk\.27\.attn_q\.weight$=iq4_xs
315
+ ^blk\.27\.attn_v\.weight$=q6_k
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq2_xxs
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q6_k
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq2_xxs
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q5_k
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq2_xxs
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q6_k
322
+ ^blk\.28\.attn_gate\.weight$=iq4_xs
323
+ ^blk\.28\.attn_qkv\.weight$=iq4_xs
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq2_xxs
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q6_k
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq2_xxs
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q5_k
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq2_xxs
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q6_k
330
+ ^blk\.28\.ssm_out\.weight$=iq4_xs
331
+ ^blk\.29\.attn_gate\.weight$=iq4_xs
332
+ ^blk\.29\.attn_qkv\.weight$=iq4_xs
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq2_xxs
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q6_k
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq2_xxs
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q5_k
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq2_xxs
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q6_k
339
+ ^blk\.29\.ssm_out\.weight$=iq4_xs
340
+ ^blk\.30\.attn_gate\.weight$=iq4_xs
341
+ ^blk\.30\.attn_qkv\.weight$=iq4_xs
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq2_xxs
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q6_k
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq2_xxs
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q6_k
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq2_xxs
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q6_k
348
+ ^blk\.30\.ssm_out\.weight$=iq4_xs
349
+ ^blk\.31\.attn_k\.weight$=q6_k
350
+ ^blk\.31\.attn_output\.weight$=q5_k
351
+ ^blk\.31\.attn_q\.weight$=iq4_xs
352
+ ^blk\.31\.attn_v\.weight$=q6_k
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq3_xxs
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq2_xxs
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q6_k
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq2_xxs
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q6_k
359
+ ^blk\.32\.attn_gate\.weight$=iq4_xs
360
+ ^blk\.32\.attn_qkv\.weight$=iq4_xs
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq2_xxs
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q6_k
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq2_xxs
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q6_k
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq2_xxs
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q6_k
367
+ ^blk\.32\.ssm_out\.weight$=iq4_xs
368
+ ^blk\.33\.attn_gate\.weight$=iq4_xs
369
+ ^blk\.33\.attn_qkv\.weight$=iq4_xs
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq2_xxs
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q6_k
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq2_xxs
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q6_k
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq2_xxs
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q6_k
376
+ ^blk\.33\.ssm_out\.weight$=iq4_xs
377
+ ^blk\.34\.attn_gate\.weight$=iq4_xs
378
+ ^blk\.34\.attn_qkv\.weight$=iq4_xs
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq2_xxs
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q6_k
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq2_xxs
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q6_k
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq2_xxs
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q6_k
385
+ ^blk\.34\.ssm_out\.weight$=iq4_xs
386
+ ^blk\.35\.attn_k\.weight$=q6_k
387
+ ^blk\.35\.attn_output\.weight$=q5_k
388
+ ^blk\.35\.attn_q\.weight$=iq4_xs
389
+ ^blk\.35\.attn_v\.weight$=q6_k
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq3_xxs
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq2_xxs
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q6_k
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq2_xxs
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q6_k
396
+ ^blk\.36\.attn_gate\.weight$=iq4_xs
397
+ ^blk\.36\.attn_qkv\.weight$=iq4_xs
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq2_xxs
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q6_k
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq2_xxs
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q6_k
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq2_xxs
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q6_k
404
+ ^blk\.36\.ssm_out\.weight$=iq4_xs
405
+ ^blk\.37\.attn_gate\.weight$=iq4_xs
406
+ ^blk\.37\.attn_qkv\.weight$=iq4_xs
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq2_xxs
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q6_k
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq2_xxs
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q6_k
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq2_xxs
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q6_k
413
+ ^blk\.37\.ssm_out\.weight$=iq4_xs
414
+ ^blk\.38\.attn_gate\.weight$=iq4_xs
415
+ ^blk\.38\.attn_qkv\.weight$=iq4_xs
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq2_xxs
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q6_k
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq2_xxs
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q6_k
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq2_xxs
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q6_k
422
+ ^blk\.38\.ssm_out\.weight$=iq4_xs
423
+ ^blk\.39\.attn_k\.weight$=q5_k
424
+ ^blk\.39\.attn_output\.weight$=q5_k
425
+ ^blk\.39\.attn_q\.weight$=iq4_xs
426
+ ^blk\.39\.attn_v\.weight$=q6_k
427
+ ^blk\.39\.ffn_down_exps\.weight$=iq3_s
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq3_xxs
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=iq3_s
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ3_M.layout.json ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ3_M",
5
+ "base_type": "iq3_s",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.50+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "L",
12
+ "rung_share": 0.5,
13
+ "crush_pass_bytes": 19073400832,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0247,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0129,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 16784687104,
159
+ "map_body+embd": 18842648576,
160
+ "ratio": 1.1226,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.5,
164
+ "body_bpw": 4.227,
165
+ "body_bpw_heuristic": 3.843,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 18947230208,
168
+ "file_bpw_pred": 4.373,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 16,
172
+ "q6_k": 14
173
+ },
174
+ "attn_k": {
175
+ "q8_0": 10
176
+ },
177
+ "attn_output": {
178
+ "q8_0": 10
179
+ },
180
+ "attn_q": {
181
+ "q8_0": 10
182
+ },
183
+ "attn_qkv": {
184
+ "q8_0": 26,
185
+ "q6_k": 4
186
+ },
187
+ "attn_v": {
188
+ "q8_0": 10
189
+ },
190
+ "ffn_down_exps": {
191
+ "iq3_s": 19,
192
+ "q5_k": 16,
193
+ "q4_k": 5
194
+ },
195
+ "ffn_down_shexp": {
196
+ "q8_0": 40
197
+ },
198
+ "ffn_gate_exps": {
199
+ "iq3_s": 38,
200
+ "q4_k": 1,
201
+ "q5_k": 1
202
+ },
203
+ "ffn_gate_shexp": {
204
+ "q8_0": 40
205
+ },
206
+ "ffn_up_exps": {
207
+ "iq3_s": 20,
208
+ "q5_k": 11,
209
+ "q4_k": 9
210
+ },
211
+ "ffn_up_shexp": {
212
+ "q8_0": 40
213
+ },
214
+ "ssm_out": {
215
+ "q8_0": 30
216
+ }
217
+ },
218
+ "prior": "layout/prior.json",
219
+ "pin_rule": "floor",
220
+ "embd_cap": "heur",
221
+ "generator": "auto_quant_v2.layout",
222
+ "generator_key": "3bf8b43e20a68f0b",
223
+ "quant_type": "IQ3_M",
224
+ "quantize_ftype": "IQ3_M",
225
+ "llama_cpp_version": "b11716",
226
+ "unknown_kinds": {},
227
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
228
+ }
layouts/InternScience_MLExpert-IQ3_M.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q5_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q4_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q5_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q5_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq3_s
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q5_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q5_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq3_s
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q5_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q5_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq3_s
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q5_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q5_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq3_s
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q5_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q5_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq3_s
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q5_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q5_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq3_s
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q5_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q5_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq3_s
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q5_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q5_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq3_s
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q5_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q5_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq3_s
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q5_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq3_s
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq3_s
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq3_s
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q4_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq3_s
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq3_s
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q6_k
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq3_s
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq3_s
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq3_s
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q6_k
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq3_s
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq3_s
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq3_s
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q6_k
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq3_s
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq3_s
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq3_s
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq3_s
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq3_s
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq3_s
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q6_k
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq3_s
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq3_s
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq3_s
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q6_k
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq3_s
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq3_s
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq3_s
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q6_k
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq3_s
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq3_s
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq3_s
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq3_s
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq3_s
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq3_s
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q6_k
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq3_s
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq3_s
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq3_s
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q6_k
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq3_s
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq3_s
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq3_s
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q6_k
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq3_s
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq3_s
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq3_s
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq3_s
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq3_s
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq3_s
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q6_k
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq3_s
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq3_s
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq3_s
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q6_k
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq3_s
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq3_s
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq3_s
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q6_k
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq3_s
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq3_s
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq3_s
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq3_s
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq3_s
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq3_s
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q6_k
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq3_s
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq3_s
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq3_s
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q6_k
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq3_s
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq3_s
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq3_s
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q5_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq3_s
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q4_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q5_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq3_s
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q4_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q5_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq3_s
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q4_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q5_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq3_s
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q4_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q6_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=q4_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq3_s
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q4_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q5_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq3_s
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q4_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q6_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=q4_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq3_s
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q4_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q6_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=q4_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq3_s
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q4_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q6_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=q4_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq3_s
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q4_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q5_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q5_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q5_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ3_XS.layout.json ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ3_XS",
5
+ "base_type": "iq3_s",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.90+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": 16203120640,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0257,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0135,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 16103112704,
159
+ "map_body+embd": 16203120640,
160
+ "ratio": 1.0062,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.9,
164
+ "body_bpw": 3.599,
165
+ "body_bpw_heuristic": 3.681,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 16307702272,
168
+ "file_bpw_pred": 3.764,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q5_k": 21,
172
+ "q6_k": 8,
173
+ "q8_0": 1
174
+ },
175
+ "attn_k": {
176
+ "q8_0": 10
177
+ },
178
+ "attn_output": {
179
+ "q8_0": 10
180
+ },
181
+ "attn_q": {
182
+ "q5_k": 7,
183
+ "q6_k": 3
184
+ },
185
+ "attn_qkv": {
186
+ "q5_k": 22,
187
+ "q6_k": 7,
188
+ "q8_0": 1
189
+ },
190
+ "attn_v": {
191
+ "q8_0": 10
192
+ },
193
+ "ffn_down_exps": {
194
+ "iq3_s": 39,
195
+ "q5_k": 1
196
+ },
197
+ "ffn_down_shexp": {
198
+ "q8_0": 40
199
+ },
200
+ "ffn_gate_exps": {
201
+ "iq3_s": 40
202
+ },
203
+ "ffn_gate_shexp": {
204
+ "q8_0": 40
205
+ },
206
+ "ffn_up_exps": {
207
+ "iq3_s": 39,
208
+ "q5_k": 1
209
+ },
210
+ "ffn_up_shexp": {
211
+ "q8_0": 40
212
+ },
213
+ "ssm_out": {
214
+ "q6_k": 22,
215
+ "q8_0": 8
216
+ }
217
+ },
218
+ "prior": "layout/prior.json",
219
+ "pin_rule": "floor",
220
+ "embd_cap": "heur",
221
+ "generator": "auto_quant_v2.layout",
222
+ "generator_key": "3bf8b43e20a68f0b",
223
+ "quant_type": "IQ3_XS",
224
+ "quantize_ftype": "IQ3_XS",
225
+ "llama_cpp_version": "b11716",
226
+ "unknown_kinds": {},
227
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
228
+ }
layouts/InternScience_MLExpert-IQ3_XS.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq3_s
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq3_s
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq3_s
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q6_k
73
+ ^blk\.1\.attn_qkv\.weight$=q6_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq3_s
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq3_s
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq3_s
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q6_k
82
+ ^blk\.2\.attn_qkv\.weight$=q6_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq3_s
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq3_s
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq3_s
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q5_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq3_s
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq3_s
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq3_s
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q6_k
101
+ ^blk\.4\.attn_qkv\.weight$=q6_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq3_s
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq3_s
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq3_s
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q6_k
110
+ ^blk\.5\.attn_qkv\.weight$=q6_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq3_s
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq3_s
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq3_s
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q6_k
119
+ ^blk\.6\.attn_qkv\.weight$=q6_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq3_s
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq3_s
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq3_s
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q5_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq3_s
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq3_s
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq3_s
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q6_k
138
+ ^blk\.8\.attn_qkv\.weight$=q6_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq3_s
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq3_s
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq3_s
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q6_k
147
+ ^blk\.9\.attn_qkv\.weight$=q6_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq3_s
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq3_s
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq3_s
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q5_k
156
+ ^blk\.10\.attn_qkv\.weight$=q5_k
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq3_s
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq3_s
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq3_s
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q6_k
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q5_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq3_s
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq3_s
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq3_s
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q5_k
175
+ ^blk\.12\.attn_qkv\.weight$=q5_k
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq3_s
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq3_s
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq3_s
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q6_k
183
+ ^blk\.13\.attn_gate\.weight$=q5_k
184
+ ^blk\.13\.attn_qkv\.weight$=q5_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq3_s
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq3_s
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq3_s
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q6_k
192
+ ^blk\.14\.attn_gate\.weight$=q5_k
193
+ ^blk\.14\.attn_qkv\.weight$=q5_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq3_s
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq3_s
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq3_s
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q6_k
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q5_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq3_s
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq3_s
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq3_s
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q5_k
212
+ ^blk\.16\.attn_qkv\.weight$=q5_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq3_s
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq3_s
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq3_s
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q6_k
220
+ ^blk\.17\.attn_gate\.weight$=q5_k
221
+ ^blk\.17\.attn_qkv\.weight$=q5_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq3_s
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq3_s
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq3_s
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q6_k
229
+ ^blk\.18\.attn_gate\.weight$=q5_k
230
+ ^blk\.18\.attn_qkv\.weight$=q5_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq3_s
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq3_s
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq3_s
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q6_k
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q5_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq3_s
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq3_s
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq3_s
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q5_k
249
+ ^blk\.20\.attn_qkv\.weight$=q5_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq3_s
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq3_s
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq3_s
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q6_k
257
+ ^blk\.21\.attn_gate\.weight$=q5_k
258
+ ^blk\.21\.attn_qkv\.weight$=q5_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq3_s
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq3_s
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq3_s
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q6_k
266
+ ^blk\.22\.attn_gate\.weight$=q5_k
267
+ ^blk\.22\.attn_qkv\.weight$=q5_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq3_s
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq3_s
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq3_s
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q6_k
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q5_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq3_s
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq3_s
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq3_s
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q5_k
286
+ ^blk\.24\.attn_qkv\.weight$=q5_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq3_s
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq3_s
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq3_s
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q6_k
294
+ ^blk\.25\.attn_gate\.weight$=q5_k
295
+ ^blk\.25\.attn_qkv\.weight$=q5_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq3_s
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq3_s
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq3_s
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q6_k
303
+ ^blk\.26\.attn_gate\.weight$=q5_k
304
+ ^blk\.26\.attn_qkv\.weight$=q5_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq3_s
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq3_s
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq3_s
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q6_k
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q5_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq3_s
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq3_s
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq3_s
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q5_k
323
+ ^blk\.28\.attn_qkv\.weight$=q5_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq3_s
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq3_s
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq3_s
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q6_k
331
+ ^blk\.29\.attn_gate\.weight$=q5_k
332
+ ^blk\.29\.attn_qkv\.weight$=q5_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq3_s
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq3_s
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq3_s
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q6_k
340
+ ^blk\.30\.attn_gate\.weight$=q6_k
341
+ ^blk\.30\.attn_qkv\.weight$=q5_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq3_s
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq3_s
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq3_s
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q6_k
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q6_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq3_s
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq3_s
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq3_s
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q5_k
360
+ ^blk\.32\.attn_qkv\.weight$=q5_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq3_s
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq3_s
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq3_s
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q6_k
368
+ ^blk\.33\.attn_gate\.weight$=q5_k
369
+ ^blk\.33\.attn_qkv\.weight$=q5_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq3_s
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq3_s
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq3_s
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q6_k
377
+ ^blk\.34\.attn_gate\.weight$=q5_k
378
+ ^blk\.34\.attn_qkv\.weight$=q5_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq3_s
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq3_s
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq3_s
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q6_k
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q6_k
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq3_s
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq3_s
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq3_s
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q5_k
397
+ ^blk\.36\.attn_qkv\.weight$=q5_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq3_s
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq3_s
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq3_s
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q6_k
405
+ ^blk\.37\.attn_gate\.weight$=q5_k
406
+ ^blk\.37\.attn_qkv\.weight$=q5_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq3_s
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq3_s
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq3_s
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q6_k
414
+ ^blk\.38\.attn_gate\.weight$=q5_k
415
+ ^blk\.38\.attn_qkv\.weight$=q5_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq3_s
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq3_s
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq3_s
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q6_k
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q6_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q5_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq3_s
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q5_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ3_XXS.layout.json ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ3_XXS",
5
+ "base_type": "iq3_xxs",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.70+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": 15613165568,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0235,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0147,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 14755258368,
159
+ "map_body+embd": 15466102784,
160
+ "ratio": 1.0482,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.7,
164
+ "body_bpw": 3.423,
165
+ "body_bpw_heuristic": 3.376,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 15570684416,
168
+ "file_bpw_pred": 3.594,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q5_k": 21,
172
+ "q6_k": 8,
173
+ "q8_0": 1
174
+ },
175
+ "attn_k": {
176
+ "q8_0": 10
177
+ },
178
+ "attn_output": {
179
+ "q8_0": 10
180
+ },
181
+ "attn_q": {
182
+ "q5_k": 9,
183
+ "q6_k": 1
184
+ },
185
+ "attn_qkv": {
186
+ "q5_k": 27,
187
+ "q6_k": 2,
188
+ "q8_0": 1
189
+ },
190
+ "attn_v": {
191
+ "q8_0": 10
192
+ },
193
+ "ffn_down_exps": {
194
+ "iq3_xxs": 30,
195
+ "iq4_xs": 9,
196
+ "q5_k": 1
197
+ },
198
+ "ffn_down_shexp": {
199
+ "q8_0": 40
200
+ },
201
+ "ffn_gate_exps": {
202
+ "iq3_xxs": 39,
203
+ "iq4_xs": 1
204
+ },
205
+ "ffn_gate_shexp": {
206
+ "q8_0": 40
207
+ },
208
+ "ffn_up_exps": {
209
+ "iq3_xxs": 29,
210
+ "iq4_xs": 10,
211
+ "q4_k": 1
212
+ },
213
+ "ffn_up_shexp": {
214
+ "q8_0": 40
215
+ },
216
+ "ssm_out": {
217
+ "q6_k": 22,
218
+ "q8_0": 8
219
+ }
220
+ },
221
+ "prior": "layout/prior.json",
222
+ "pin_rule": "floor",
223
+ "embd_cap": "heur",
224
+ "generator": "auto_quant_v2.layout",
225
+ "generator_key": "3bf8b43e20a68f0b",
226
+ "quant_type": "IQ3_XXS",
227
+ "quantize_ftype": "IQ3_XXS",
228
+ "llama_cpp_version": "b11716",
229
+ "unknown_kinds": {},
230
+ "commit": "c3c0369c7e82d7b49e1472fe453a182650282f2d"
231
+ }
layouts/InternScience_MLExpert-IQ3_XXS.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq4_xs
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq3_xxs
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq4_xs
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q6_k
73
+ ^blk\.1\.attn_qkv\.weight$=q6_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq4_xs
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq3_xxs
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq4_xs
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q6_k
82
+ ^blk\.2\.attn_qkv\.weight$=q6_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq4_xs
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq3_xxs
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq4_xs
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q5_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq4_xs
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq3_xxs
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq4_xs
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q6_k
101
+ ^blk\.4\.attn_qkv\.weight$=q5_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq4_xs
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq3_xxs
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq4_xs
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q6_k
110
+ ^blk\.5\.attn_qkv\.weight$=q5_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq4_xs
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq3_xxs
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq4_xs
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q6_k
119
+ ^blk\.6\.attn_qkv\.weight$=q5_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq3_xxs
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq3_xxs
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq4_xs
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q5_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq4_xs
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq3_xxs
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq4_xs
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q6_k
138
+ ^blk\.8\.attn_qkv\.weight$=q5_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq3_xxs
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq3_xxs
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq4_xs
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q6_k
147
+ ^blk\.9\.attn_qkv\.weight$=q5_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq3_xxs
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq3_xxs
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq4_xs
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q5_k
156
+ ^blk\.10\.attn_qkv\.weight$=q5_k
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq3_xxs
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq3_xxs
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq3_xxs
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q6_k
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q5_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq3_xxs
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq3_xxs
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq3_xxs
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q5_k
175
+ ^blk\.12\.attn_qkv\.weight$=q5_k
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq3_xxs
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq3_xxs
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq3_xxs
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q6_k
183
+ ^blk\.13\.attn_gate\.weight$=q5_k
184
+ ^blk\.13\.attn_qkv\.weight$=q5_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq3_xxs
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq3_xxs
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq3_xxs
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q6_k
192
+ ^blk\.14\.attn_gate\.weight$=q5_k
193
+ ^blk\.14\.attn_qkv\.weight$=q5_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq3_xxs
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq3_xxs
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq3_xxs
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q6_k
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q5_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq3_xxs
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq3_xxs
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq3_xxs
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q5_k
212
+ ^blk\.16\.attn_qkv\.weight$=q5_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq3_xxs
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq3_xxs
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq3_xxs
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q6_k
220
+ ^blk\.17\.attn_gate\.weight$=q5_k
221
+ ^blk\.17\.attn_qkv\.weight$=q5_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq3_xxs
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq3_xxs
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq3_xxs
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q6_k
229
+ ^blk\.18\.attn_gate\.weight$=q5_k
230
+ ^blk\.18\.attn_qkv\.weight$=q5_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq3_xxs
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq3_xxs
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq3_xxs
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q6_k
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q5_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq3_xxs
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq3_xxs
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq3_xxs
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q5_k
249
+ ^blk\.20\.attn_qkv\.weight$=q5_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq3_xxs
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq3_xxs
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq3_xxs
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q6_k
257
+ ^blk\.21\.attn_gate\.weight$=q5_k
258
+ ^blk\.21\.attn_qkv\.weight$=q5_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq3_xxs
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq3_xxs
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq3_xxs
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q6_k
266
+ ^blk\.22\.attn_gate\.weight$=q5_k
267
+ ^blk\.22\.attn_qkv\.weight$=q5_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq3_xxs
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq3_xxs
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq3_xxs
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q6_k
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q5_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq3_xxs
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq3_xxs
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq3_xxs
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q5_k
286
+ ^blk\.24\.attn_qkv\.weight$=q5_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq3_xxs
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq3_xxs
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq3_xxs
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q6_k
294
+ ^blk\.25\.attn_gate\.weight$=q5_k
295
+ ^blk\.25\.attn_qkv\.weight$=q5_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq3_xxs
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq3_xxs
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq3_xxs
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q6_k
303
+ ^blk\.26\.attn_gate\.weight$=q5_k
304
+ ^blk\.26\.attn_qkv\.weight$=q5_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq3_xxs
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq3_xxs
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq3_xxs
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q6_k
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q5_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq3_xxs
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq3_xxs
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq3_xxs
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q5_k
323
+ ^blk\.28\.attn_qkv\.weight$=q5_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq3_xxs
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq3_xxs
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq3_xxs
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q6_k
331
+ ^blk\.29\.attn_gate\.weight$=q5_k
332
+ ^blk\.29\.attn_qkv\.weight$=q5_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq3_xxs
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq3_xxs
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq3_xxs
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q6_k
340
+ ^blk\.30\.attn_gate\.weight$=q6_k
341
+ ^blk\.30\.attn_qkv\.weight$=q5_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq3_xxs
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq3_xxs
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq3_xxs
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q6_k
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q5_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq4_xs
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq3_xxs
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq3_xxs
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q5_k
360
+ ^blk\.32\.attn_qkv\.weight$=q5_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq3_xxs
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq3_xxs
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq3_xxs
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q6_k
368
+ ^blk\.33\.attn_gate\.weight$=q5_k
369
+ ^blk\.33\.attn_qkv\.weight$=q5_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq3_xxs
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq3_xxs
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq3_xxs
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q6_k
377
+ ^blk\.34\.attn_gate\.weight$=q5_k
378
+ ^blk\.34\.attn_qkv\.weight$=q5_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq3_xxs
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq3_xxs
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq3_xxs
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q6_k
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q5_k
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq4_xs
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq3_xxs
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq3_xxs
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q5_k
397
+ ^blk\.36\.attn_qkv\.weight$=q5_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq3_xxs
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq3_xxs
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq3_xxs
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q6_k
405
+ ^blk\.37\.attn_gate\.weight$=q5_k
406
+ ^blk\.37\.attn_qkv\.weight$=q5_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq3_xxs
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq3_xxs
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq3_xxs
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q6_k
414
+ ^blk\.38\.attn_gate\.weight$=q5_k
415
+ ^blk\.38\.attn_qkv\.weight$=q5_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq3_xxs
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq3_xxs
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq3_xxs
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q6_k
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q6_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q5_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq4_xs
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q4_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ4_NL.layout.json ADDED
@@ -0,0 +1,226 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ4_NL",
5
+ "base_type": "iq4_nl",
6
+ "family": "iq",
7
+ "recipe": "ident",
8
+ "variant": "shr0.70",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.021,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0144,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 19745775616,
159
+ "map_body+embd": 22202941440,
160
+ "ratio": 1.1244,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.701,
164
+ "body_bpw": 5.027,
165
+ "body_bpw_heuristic": 4.532,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 22307523072,
168
+ "file_bpw_pred": 5.149,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "iq4_nl": 30,
190
+ "q6_k": 5,
191
+ "q5_k": 4,
192
+ "q8_0": 1
193
+ },
194
+ "ffn_down_shexp": {
195
+ "q8_0": 40
196
+ },
197
+ "ffn_gate_exps": {
198
+ "iq4_nl": 39,
199
+ "q6_k": 1
200
+ },
201
+ "ffn_gate_shexp": {
202
+ "q8_0": 40
203
+ },
204
+ "ffn_up_exps": {
205
+ "iq4_nl": 29,
206
+ "q6_k": 10,
207
+ "q8_0": 1
208
+ },
209
+ "ffn_up_shexp": {
210
+ "q8_0": 40
211
+ },
212
+ "ssm_out": {
213
+ "q8_0": 30
214
+ }
215
+ },
216
+ "prior": "layout/prior.json",
217
+ "pin_rule": "floor",
218
+ "embd_cap": "heur",
219
+ "generator": "auto_quant_v2.layout",
220
+ "generator_key": "3bf8b43e20a68f0b",
221
+ "quant_type": "IQ4_NL",
222
+ "quantize_ftype": "IQ4_NL",
223
+ "llama_cpp_version": "b11716",
224
+ "unknown_kinds": {},
225
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
226
+ }
layouts/InternScience_MLExpert-IQ4_NL.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q6_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq4_nl
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q6_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q5_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq4_nl
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q6_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q5_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq4_nl
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q6_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q6_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq4_nl
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q6_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q5_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq4_nl
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q6_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q5_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq4_nl
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q6_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq4_nl
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq4_nl
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q6_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q6_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq4_nl
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q6_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq4_nl
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq4_nl
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q6_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq4_nl
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq4_nl
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q6_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq4_nl
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq4_nl
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq4_nl
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq4_nl
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq4_nl
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq4_nl
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq4_nl
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq4_nl
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq4_nl
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq4_nl
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq4_nl
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq4_nl
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq4_nl
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq4_nl
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq4_nl
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq4_nl
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq4_nl
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq4_nl
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq4_nl
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq4_nl
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq4_nl
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq4_nl
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq4_nl
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq4_nl
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq4_nl
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq4_nl
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq4_nl
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq4_nl
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq4_nl
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq4_nl
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq4_nl
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq4_nl
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq4_nl
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq4_nl
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq4_nl
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq4_nl
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq4_nl
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq4_nl
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq4_nl
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq4_nl
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq4_nl
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq4_nl
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq4_nl
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq4_nl
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq4_nl
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq4_nl
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq4_nl
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq4_nl
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq4_nl
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq4_nl
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq4_nl
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq4_nl
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq4_nl
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq4_nl
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq4_nl
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq4_nl
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq4_nl
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq4_nl
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq4_nl
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq4_nl
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq4_nl
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq4_nl
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq4_nl
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q6_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq4_nl
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq4_nl
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq4_nl
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq4_nl
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq4_nl
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq4_nl
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq4_nl
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq4_nl
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq4_nl
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq4_nl
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq4_nl
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q6_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq4_nl
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq4_nl
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq4_nl
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq4_nl
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq4_nl
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq4_nl
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq4_nl
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq4_nl
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq4_nl
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq4_nl
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq4_nl
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q6_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-IQ4_XS.layout.json ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "IQ4_XS",
5
+ "base_type": "iq4_xs",
6
+ "family": "iq",
7
+ "recipe": "ident+crush",
8
+ "variant": "shr0.90+crush",
9
+ "auto_crush": true,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": 19772342272,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0222,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0144,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 18690875392,
159
+ "map_body+embd": 19771359232,
160
+ "ratio": 1.0578,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.9,
164
+ "body_bpw": 4.448,
165
+ "body_bpw_heuristic": 4.284,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 19875940864,
168
+ "file_bpw_pred": 4.588,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 15,
172
+ "q6_k": 15
173
+ },
174
+ "attn_k": {
175
+ "q8_0": 10
176
+ },
177
+ "attn_output": {
178
+ "q8_0": 10
179
+ },
180
+ "attn_q": {
181
+ "q8_0": 10
182
+ },
183
+ "attn_qkv": {
184
+ "q8_0": 29,
185
+ "q6_k": 1
186
+ },
187
+ "attn_v": {
188
+ "q8_0": 10
189
+ },
190
+ "ffn_down_exps": {
191
+ "iq4_xs": 39,
192
+ "q6_k": 1
193
+ },
194
+ "ffn_down_shexp": {
195
+ "q8_0": 40
196
+ },
197
+ "ffn_gate_exps": {
198
+ "iq4_xs": 40
199
+ },
200
+ "ffn_gate_shexp": {
201
+ "q8_0": 40
202
+ },
203
+ "ffn_up_exps": {
204
+ "iq4_xs": 39,
205
+ "q5_k": 1
206
+ },
207
+ "ffn_up_shexp": {
208
+ "q8_0": 40
209
+ },
210
+ "ssm_out": {
211
+ "q8_0": 30
212
+ }
213
+ },
214
+ "prior": "layout/prior.json",
215
+ "pin_rule": "floor",
216
+ "embd_cap": "heur",
217
+ "generator": "auto_quant_v2.layout",
218
+ "generator_key": "3bf8b43e20a68f0b",
219
+ "quant_type": "IQ4_XS",
220
+ "quantize_ftype": "IQ4_XS",
221
+ "llama_cpp_version": "b11716",
222
+ "unknown_kinds": {},
223
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
224
+ }
layouts/InternScience_MLExpert-IQ4_XS.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=iq4_xs
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=iq4_xs
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=iq4_xs
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=iq4_xs
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=iq4_xs
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=iq4_xs
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=iq4_xs
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=iq4_xs
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=iq4_xs
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=iq4_xs
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=iq4_xs
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=iq4_xs
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=iq4_xs
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=iq4_xs
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=iq4_xs
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=iq4_xs
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=iq4_xs
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=iq4_xs
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=iq4_xs
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=iq4_xs
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=iq4_xs
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=iq4_xs
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=iq4_xs
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=iq4_xs
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=iq4_xs
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=iq4_xs
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=iq4_xs
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=iq4_xs
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=iq4_xs
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=iq4_xs
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q6_k
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=iq4_xs
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=iq4_xs
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=iq4_xs
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=iq4_xs
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=iq4_xs
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=iq4_xs
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q6_k
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=iq4_xs
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=iq4_xs
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=iq4_xs
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q6_k
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=iq4_xs
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=iq4_xs
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=iq4_xs
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q6_k
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=iq4_xs
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=iq4_xs
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=iq4_xs
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=iq4_xs
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=iq4_xs
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=iq4_xs
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q6_k
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=iq4_xs
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=iq4_xs
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=iq4_xs
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q6_k
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=iq4_xs
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=iq4_xs
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=iq4_xs
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q6_k
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=iq4_xs
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=iq4_xs
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=iq4_xs
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=iq4_xs
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=iq4_xs
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=iq4_xs
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q6_k
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=iq4_xs
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=iq4_xs
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=iq4_xs
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q6_k
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=iq4_xs
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=iq4_xs
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=iq4_xs
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q6_k
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=iq4_xs
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=iq4_xs
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=iq4_xs
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=iq4_xs
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=iq4_xs
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=iq4_xs
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q6_k
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=iq4_xs
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=iq4_xs
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=iq4_xs
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q6_k
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=iq4_xs
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=iq4_xs
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=iq4_xs
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q6_k
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=iq4_xs
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=iq4_xs
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=iq4_xs
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=iq4_xs
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=iq4_xs
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=iq4_xs
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q6_k
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=iq4_xs
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=iq4_xs
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=iq4_xs
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q6_k
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=iq4_xs
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=iq4_xs
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=iq4_xs
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=iq4_xs
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=iq4_xs
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=iq4_xs
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=iq4_xs
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=iq4_xs
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=iq4_xs
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=iq4_xs
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=iq4_xs
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=iq4_xs
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=iq4_xs
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=iq4_xs
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=iq4_xs
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=iq4_xs
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=iq4_xs
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=iq4_xs
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=iq4_xs
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=iq4_xs
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=iq4_xs
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=iq4_xs
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=iq4_xs
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=iq4_xs
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=iq4_xs
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=iq4_xs
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=iq4_xs
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q6_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=iq4_xs
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=iq4_xs
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=iq4_xs
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q6_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=iq4_xs
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q5_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q2_K.layout.json ADDED
@@ -0,0 +1,229 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q2_K",
5
+ "base_type": "q2_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.70",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0331,
17
+ "type": "q5_k"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0132,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q5_k",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 12501508096,
159
+ "map_body+embd": 13723762688,
160
+ "ratio": 1.0978,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.7,
164
+ "body_bpw": 3.054,
165
+ "body_bpw_heuristic": 2.836,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 13828344320,
168
+ "file_bpw_pred": 3.192,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q5_k": 21,
172
+ "q6_k": 8,
173
+ "q8_0": 1
174
+ },
175
+ "attn_k": {
176
+ "q8_0": 10
177
+ },
178
+ "attn_output": {
179
+ "q8_0": 10
180
+ },
181
+ "attn_q": {
182
+ "q5_k": 9,
183
+ "q6_k": 1
184
+ },
185
+ "attn_qkv": {
186
+ "q5_k": 25,
187
+ "q6_k": 4,
188
+ "q8_0": 1
189
+ },
190
+ "attn_v": {
191
+ "q8_0": 10
192
+ },
193
+ "ffn_down_exps": {
194
+ "q2_k": 34,
195
+ "q4_k": 6
196
+ },
197
+ "ffn_down_shexp": {
198
+ "q8_0": 40
199
+ },
200
+ "ffn_gate_exps": {
201
+ "q2_k": 39,
202
+ "q4_k": 1
203
+ },
204
+ "ffn_gate_shexp": {
205
+ "q8_0": 40
206
+ },
207
+ "ffn_up_exps": {
208
+ "q2_k": 29,
209
+ "q4_k": 11
210
+ },
211
+ "ffn_up_shexp": {
212
+ "q8_0": 40
213
+ },
214
+ "ssm_out": {
215
+ "q6_k": 22,
216
+ "q8_0": 8
217
+ }
218
+ },
219
+ "prior": "layout/prior.json",
220
+ "pin_rule": "floor",
221
+ "embd_cap": "heur",
222
+ "generator": "auto_quant_v2.layout",
223
+ "generator_key": "3bf8b43e20a68f0b",
224
+ "quant_type": "Q2_K",
225
+ "quantize_ftype": "Q2_K",
226
+ "llama_cpp_version": "b11716",
227
+ "unknown_kinds": {},
228
+ "commit": "c3c0369c7e82d7b49e1472fe453a182650282f2d"
229
+ }
layouts/InternScience_MLExpert-Q2_K.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q5_k
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q4_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q2_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q4_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q6_k
73
+ ^blk\.1\.attn_qkv\.weight$=q6_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=q2_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q2_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q4_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q6_k
82
+ ^blk\.2\.attn_qkv\.weight$=q6_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=q2_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q2_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q4_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q5_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q4_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q2_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q4_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q6_k
101
+ ^blk\.4\.attn_qkv\.weight$=q6_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=q2_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q2_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q4_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q6_k
110
+ ^blk\.5\.attn_qkv\.weight$=q6_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=q2_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q2_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q4_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q6_k
119
+ ^blk\.6\.attn_qkv\.weight$=q5_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=q2_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q2_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q4_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q5_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q4_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q2_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q4_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q6_k
138
+ ^blk\.8\.attn_qkv\.weight$=q5_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=q2_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q2_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q4_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q6_k
147
+ ^blk\.9\.attn_qkv\.weight$=q5_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=q2_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q2_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q4_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q5_k
156
+ ^blk\.10\.attn_qkv\.weight$=q5_k
157
+ ^blk\.10\.ffn_down_exps\.weight$=q2_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q2_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q2_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q6_k
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q5_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q2_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q2_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q2_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q5_k
175
+ ^blk\.12\.attn_qkv\.weight$=q5_k
176
+ ^blk\.12\.ffn_down_exps\.weight$=q2_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q2_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q2_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q6_k
183
+ ^blk\.13\.attn_gate\.weight$=q5_k
184
+ ^blk\.13\.attn_qkv\.weight$=q5_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=q2_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q2_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q2_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q6_k
192
+ ^blk\.14\.attn_gate\.weight$=q5_k
193
+ ^blk\.14\.attn_qkv\.weight$=q5_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=q2_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q2_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q2_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q6_k
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q5_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q2_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q2_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q2_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q5_k
212
+ ^blk\.16\.attn_qkv\.weight$=q5_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=q2_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q2_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q2_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q6_k
220
+ ^blk\.17\.attn_gate\.weight$=q5_k
221
+ ^blk\.17\.attn_qkv\.weight$=q5_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=q2_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q2_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q2_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q6_k
229
+ ^blk\.18\.attn_gate\.weight$=q5_k
230
+ ^blk\.18\.attn_qkv\.weight$=q5_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=q2_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q2_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q2_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q6_k
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q5_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q2_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q2_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q2_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q5_k
249
+ ^blk\.20\.attn_qkv\.weight$=q5_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=q2_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q2_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q2_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q6_k
257
+ ^blk\.21\.attn_gate\.weight$=q5_k
258
+ ^blk\.21\.attn_qkv\.weight$=q5_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=q2_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q2_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q2_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q6_k
266
+ ^blk\.22\.attn_gate\.weight$=q5_k
267
+ ^blk\.22\.attn_qkv\.weight$=q5_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=q2_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q2_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q2_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q6_k
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q5_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q2_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q2_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q2_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q5_k
286
+ ^blk\.24\.attn_qkv\.weight$=q5_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=q2_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q2_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q2_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q6_k
294
+ ^blk\.25\.attn_gate\.weight$=q5_k
295
+ ^blk\.25\.attn_qkv\.weight$=q5_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=q2_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q2_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q2_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q6_k
303
+ ^blk\.26\.attn_gate\.weight$=q5_k
304
+ ^blk\.26\.attn_qkv\.weight$=q5_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=q2_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q2_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q2_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q6_k
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q5_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q2_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q2_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q2_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q5_k
323
+ ^blk\.28\.attn_qkv\.weight$=q5_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=q2_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q2_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q2_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q6_k
331
+ ^blk\.29\.attn_gate\.weight$=q5_k
332
+ ^blk\.29\.attn_qkv\.weight$=q5_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=q2_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q2_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q2_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q6_k
340
+ ^blk\.30\.attn_gate\.weight$=q6_k
341
+ ^blk\.30\.attn_qkv\.weight$=q5_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=q2_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q2_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q2_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q6_k
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q5_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q4_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q2_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q2_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q5_k
360
+ ^blk\.32\.attn_qkv\.weight$=q5_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=q2_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q2_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q2_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q6_k
368
+ ^blk\.33\.attn_gate\.weight$=q5_k
369
+ ^blk\.33\.attn_qkv\.weight$=q5_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=q2_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q2_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q2_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q6_k
377
+ ^blk\.34\.attn_gate\.weight$=q5_k
378
+ ^blk\.34\.attn_qkv\.weight$=q5_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=q2_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q2_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q2_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q6_k
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q5_k
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q4_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q2_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q2_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q5_k
397
+ ^blk\.36\.attn_qkv\.weight$=q5_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=q2_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q2_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q2_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q6_k
405
+ ^blk\.37\.attn_gate\.weight$=q5_k
406
+ ^blk\.37\.attn_qkv\.weight$=q5_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=q2_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q2_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q2_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q6_k
414
+ ^blk\.38\.attn_gate\.weight$=q5_k
415
+ ^blk\.38\.attn_qkv\.weight$=q5_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=q2_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q2_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q2_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q6_k
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q6_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q4_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q4_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q4_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q3_K_L.layout.json ADDED
@@ -0,0 +1,229 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q3_K_L",
5
+ "base_type": "q3_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.50",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "L",
12
+ "rung_share": 0.5,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0247,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0129,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 16770924544,
159
+ "map_body+embd": 18151505920,
160
+ "ratio": 1.0823,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.5,
164
+ "body_bpw": 4.062,
165
+ "body_bpw_heuristic": 3.84,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 18256087552,
168
+ "file_bpw_pred": 4.214,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 16,
172
+ "q6_k": 14
173
+ },
174
+ "attn_k": {
175
+ "q8_0": 10
176
+ },
177
+ "attn_output": {
178
+ "q8_0": 10
179
+ },
180
+ "attn_q": {
181
+ "q6_k": 7,
182
+ "q8_0": 3
183
+ },
184
+ "attn_qkv": {
185
+ "q6_k": 20,
186
+ "q8_0": 10
187
+ },
188
+ "attn_v": {
189
+ "q8_0": 10
190
+ },
191
+ "ffn_down_exps": {
192
+ "q4_k": 22,
193
+ "q3_k": 16,
194
+ "q5_k": 2
195
+ },
196
+ "ffn_down_shexp": {
197
+ "q8_0": 40
198
+ },
199
+ "ffn_gate_exps": {
200
+ "q3_k": 38,
201
+ "q4_k": 1,
202
+ "q5_k": 1
203
+ },
204
+ "ffn_gate_shexp": {
205
+ "q8_0": 40
206
+ },
207
+ "ffn_up_exps": {
208
+ "q3_k": 20,
209
+ "q4_k": 16,
210
+ "q5_k": 4
211
+ },
212
+ "ffn_up_shexp": {
213
+ "q8_0": 40
214
+ },
215
+ "ssm_out": {
216
+ "q8_0": 30
217
+ }
218
+ },
219
+ "prior": "layout/prior.json",
220
+ "pin_rule": "floor",
221
+ "embd_cap": "heur",
222
+ "generator": "auto_quant_v2.layout",
223
+ "generator_key": "3bf8b43e20a68f0b",
224
+ "quant_type": "Q3_K_L",
225
+ "quantize_ftype": "Q3_K_L",
226
+ "llama_cpp_version": "b11716",
227
+ "unknown_kinds": {},
228
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
229
+ }
layouts/InternScience_MLExpert-Q3_K_L.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q5_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q4_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q4_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q4_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q3_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q5_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q4_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q3_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q4_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q6_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q4_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q3_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q5_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q4_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q3_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q4_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q4_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q3_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q4_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q4_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q3_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q4_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q6_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q4_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q3_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q5_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q4_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q3_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q4_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q4_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q3_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q4_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q3_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q3_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q3_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q6_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q4_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q3_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q3_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q6_k
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q3_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q3_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q3_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q6_k
184
+ ^blk\.13\.attn_qkv\.weight$=q6_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=q3_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q3_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q3_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q6_k
193
+ ^blk\.14\.attn_qkv\.weight$=q6_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=q3_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q3_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q3_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q6_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q4_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q3_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q3_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q6_k
212
+ ^blk\.16\.attn_qkv\.weight$=q6_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=q3_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q3_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q3_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q6_k
221
+ ^blk\.17\.attn_qkv\.weight$=q6_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=q3_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q3_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q3_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q6_k
230
+ ^blk\.18\.attn_qkv\.weight$=q6_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=q3_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q3_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q3_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q6_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q4_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q3_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q3_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q6_k
249
+ ^blk\.20\.attn_qkv\.weight$=q6_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=q3_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q3_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q3_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q6_k
258
+ ^blk\.21\.attn_qkv\.weight$=q6_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=q3_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q3_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q3_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q6_k
267
+ ^blk\.22\.attn_qkv\.weight$=q6_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=q3_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q3_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q3_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q6_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q4_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q3_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q3_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q6_k
286
+ ^blk\.24\.attn_qkv\.weight$=q6_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=q3_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q3_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q3_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q6_k
295
+ ^blk\.25\.attn_qkv\.weight$=q6_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=q3_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q3_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q3_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q6_k
304
+ ^blk\.26\.attn_qkv\.weight$=q6_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=q3_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q3_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q3_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q6_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q3_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q3_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q3_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q6_k
323
+ ^blk\.28\.attn_qkv\.weight$=q6_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=q3_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q3_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q3_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q6_k
332
+ ^blk\.29\.attn_qkv\.weight$=q6_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=q3_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q3_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q3_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q6_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=q4_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q3_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q4_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q4_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q3_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q4_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q6_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=q4_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q3_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q4_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q6_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=q4_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q3_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q4_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q6_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=q4_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q3_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q4_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q4_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q3_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q4_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q6_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=q4_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q3_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q4_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q6_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=q4_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q3_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q4_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q6_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=q4_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q3_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q4_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q5_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q5_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q5_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q3_K_M.layout.json ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q3_K_M",
5
+ "base_type": "q3_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.70",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0257,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0135,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 16110977024,
159
+ "map_body+embd": 17228759040,
160
+ "ratio": 1.0694,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.7,
164
+ "body_bpw": 3.843,
165
+ "body_bpw_heuristic": 3.683,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 17333340672,
168
+ "file_bpw_pred": 4.001,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 16,
172
+ "q6_k": 14
173
+ },
174
+ "attn_k": {
175
+ "q8_0": 10
176
+ },
177
+ "attn_output": {
178
+ "q8_0": 10
179
+ },
180
+ "attn_q": {
181
+ "q6_k": 7,
182
+ "q8_0": 3
183
+ },
184
+ "attn_qkv": {
185
+ "q6_k": 20,
186
+ "q8_0": 10
187
+ },
188
+ "attn_v": {
189
+ "q8_0": 10
190
+ },
191
+ "ffn_down_exps": {
192
+ "q3_k": 30,
193
+ "q4_k": 8,
194
+ "q5_k": 2
195
+ },
196
+ "ffn_down_shexp": {
197
+ "q8_0": 40
198
+ },
199
+ "ffn_gate_exps": {
200
+ "q3_k": 39,
201
+ "q5_k": 1
202
+ },
203
+ "ffn_gate_shexp": {
204
+ "q8_0": 40
205
+ },
206
+ "ffn_up_exps": {
207
+ "q3_k": 29,
208
+ "q4_k": 9,
209
+ "q5_k": 2
210
+ },
211
+ "ffn_up_shexp": {
212
+ "q8_0": 40
213
+ },
214
+ "ssm_out": {
215
+ "q8_0": 30
216
+ }
217
+ },
218
+ "prior": "layout/prior.json",
219
+ "pin_rule": "floor",
220
+ "embd_cap": "heur",
221
+ "generator": "auto_quant_v2.layout",
222
+ "generator_key": "3bf8b43e20a68f0b",
223
+ "quant_type": "Q3_K_M",
224
+ "quantize_ftype": "Q3_K_M",
225
+ "llama_cpp_version": "b11716",
226
+ "unknown_kinds": {},
227
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
228
+ }
layouts/InternScience_MLExpert-Q3_K_M.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q5_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q3_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q4_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q4_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q3_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q4_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q4_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q3_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q4_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q6_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q4_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q3_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q5_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q4_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q3_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q4_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q4_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q3_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q4_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q3_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q3_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q4_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q6_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q4_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q3_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q4_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q3_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q3_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q4_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q3_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q3_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q4_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q3_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q3_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q3_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q6_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q3_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q3_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q3_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q6_k
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q3_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q3_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q3_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q6_k
184
+ ^blk\.13\.attn_qkv\.weight$=q6_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=q3_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q3_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q3_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q6_k
193
+ ^blk\.14\.attn_qkv\.weight$=q6_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=q3_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q3_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q3_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q6_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q3_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q3_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q3_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q6_k
212
+ ^blk\.16\.attn_qkv\.weight$=q6_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=q3_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q3_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q3_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q6_k
221
+ ^blk\.17\.attn_qkv\.weight$=q6_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=q3_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q3_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q3_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q6_k
230
+ ^blk\.18\.attn_qkv\.weight$=q6_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=q3_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q3_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q3_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q6_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q3_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q3_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q3_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q6_k
249
+ ^blk\.20\.attn_qkv\.weight$=q6_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=q3_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q3_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q3_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q6_k
258
+ ^blk\.21\.attn_qkv\.weight$=q6_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=q3_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q3_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q3_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q6_k
267
+ ^blk\.22\.attn_qkv\.weight$=q6_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=q3_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q3_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q3_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q6_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q3_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q3_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q3_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q6_k
286
+ ^blk\.24\.attn_qkv\.weight$=q6_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=q3_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q3_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q3_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q6_k
295
+ ^blk\.25\.attn_qkv\.weight$=q6_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=q3_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q3_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q3_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q6_k
304
+ ^blk\.26\.attn_qkv\.weight$=q6_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=q3_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q3_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q3_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q6_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q3_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q3_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q3_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q6_k
323
+ ^blk\.28\.attn_qkv\.weight$=q6_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=q3_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q3_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q3_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q6_k
332
+ ^blk\.29\.attn_qkv\.weight$=q6_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=q3_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q3_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q3_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q6_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=q3_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q3_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q3_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q4_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q3_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q3_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q6_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=q3_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q3_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q3_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q6_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=q3_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q3_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q3_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q6_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=q3_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q3_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q3_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q4_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q3_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q3_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q6_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=q3_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q3_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q3_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q6_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=q3_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q3_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q3_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q6_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=q3_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q3_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q3_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q5_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q5_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q5_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q3_K_S.layout.json ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q3_K_S",
5
+ "base_type": "q3_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.90",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0269,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0141,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 15397945344,
159
+ "map_body+embd": 16203120640,
160
+ "ratio": 1.0523,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.9,
164
+ "body_bpw": 3.599,
165
+ "body_bpw_heuristic": 3.513,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 16307702272,
168
+ "file_bpw_pred": 3.764,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q5_k": 21,
172
+ "q6_k": 8,
173
+ "q8_0": 1
174
+ },
175
+ "attn_k": {
176
+ "q8_0": 10
177
+ },
178
+ "attn_output": {
179
+ "q8_0": 10
180
+ },
181
+ "attn_q": {
182
+ "q5_k": 7,
183
+ "q6_k": 3
184
+ },
185
+ "attn_qkv": {
186
+ "q5_k": 22,
187
+ "q6_k": 7,
188
+ "q8_0": 1
189
+ },
190
+ "attn_v": {
191
+ "q8_0": 10
192
+ },
193
+ "ffn_down_exps": {
194
+ "q3_k": 39,
195
+ "q5_k": 1
196
+ },
197
+ "ffn_down_shexp": {
198
+ "q8_0": 40
199
+ },
200
+ "ffn_gate_exps": {
201
+ "q3_k": 40
202
+ },
203
+ "ffn_gate_shexp": {
204
+ "q8_0": 40
205
+ },
206
+ "ffn_up_exps": {
207
+ "q3_k": 39,
208
+ "q5_k": 1
209
+ },
210
+ "ffn_up_shexp": {
211
+ "q8_0": 40
212
+ },
213
+ "ssm_out": {
214
+ "q6_k": 22,
215
+ "q8_0": 8
216
+ }
217
+ },
218
+ "prior": "layout/prior.json",
219
+ "pin_rule": "floor",
220
+ "embd_cap": "heur",
221
+ "generator": "auto_quant_v2.layout",
222
+ "generator_key": "3bf8b43e20a68f0b",
223
+ "quant_type": "Q3_K_S",
224
+ "quantize_ftype": "Q3_K_S",
225
+ "llama_cpp_version": "b11716",
226
+ "unknown_kinds": {},
227
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
228
+ }
layouts/InternScience_MLExpert-Q3_K_S.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q3_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q3_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q3_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q6_k
73
+ ^blk\.1\.attn_qkv\.weight$=q6_k
74
+ ^blk\.1\.ffn_down_exps\.weight$=q3_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q3_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q3_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q6_k
82
+ ^blk\.2\.attn_qkv\.weight$=q6_k
83
+ ^blk\.2\.ffn_down_exps\.weight$=q3_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q3_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q3_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q5_k
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q3_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q3_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q3_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q6_k
101
+ ^blk\.4\.attn_qkv\.weight$=q6_k
102
+ ^blk\.4\.ffn_down_exps\.weight$=q3_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q3_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q3_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q6_k
110
+ ^blk\.5\.attn_qkv\.weight$=q6_k
111
+ ^blk\.5\.ffn_down_exps\.weight$=q3_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q3_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q3_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q6_k
119
+ ^blk\.6\.attn_qkv\.weight$=q6_k
120
+ ^blk\.6\.ffn_down_exps\.weight$=q3_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q3_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q3_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q5_k
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q3_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q3_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q3_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q6_k
138
+ ^blk\.8\.attn_qkv\.weight$=q6_k
139
+ ^blk\.8\.ffn_down_exps\.weight$=q3_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q3_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q3_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q6_k
147
+ ^blk\.9\.attn_qkv\.weight$=q6_k
148
+ ^blk\.9\.ffn_down_exps\.weight$=q3_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q3_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q3_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q5_k
156
+ ^blk\.10\.attn_qkv\.weight$=q5_k
157
+ ^blk\.10\.ffn_down_exps\.weight$=q3_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q3_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q3_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q6_k
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q5_k
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q3_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q3_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q3_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q5_k
175
+ ^blk\.12\.attn_qkv\.weight$=q5_k
176
+ ^blk\.12\.ffn_down_exps\.weight$=q3_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q3_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q3_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q6_k
183
+ ^blk\.13\.attn_gate\.weight$=q5_k
184
+ ^blk\.13\.attn_qkv\.weight$=q5_k
185
+ ^blk\.13\.ffn_down_exps\.weight$=q3_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q3_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q3_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q6_k
192
+ ^blk\.14\.attn_gate\.weight$=q5_k
193
+ ^blk\.14\.attn_qkv\.weight$=q5_k
194
+ ^blk\.14\.ffn_down_exps\.weight$=q3_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q3_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q3_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q6_k
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q5_k
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q3_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q3_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q3_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q5_k
212
+ ^blk\.16\.attn_qkv\.weight$=q5_k
213
+ ^blk\.16\.ffn_down_exps\.weight$=q3_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q3_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q3_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q6_k
220
+ ^blk\.17\.attn_gate\.weight$=q5_k
221
+ ^blk\.17\.attn_qkv\.weight$=q5_k
222
+ ^blk\.17\.ffn_down_exps\.weight$=q3_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q3_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q3_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q6_k
229
+ ^blk\.18\.attn_gate\.weight$=q5_k
230
+ ^blk\.18\.attn_qkv\.weight$=q5_k
231
+ ^blk\.18\.ffn_down_exps\.weight$=q3_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q3_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q3_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q6_k
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q5_k
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q3_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q3_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q3_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q5_k
249
+ ^blk\.20\.attn_qkv\.weight$=q5_k
250
+ ^blk\.20\.ffn_down_exps\.weight$=q3_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q3_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q3_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q6_k
257
+ ^blk\.21\.attn_gate\.weight$=q5_k
258
+ ^blk\.21\.attn_qkv\.weight$=q5_k
259
+ ^blk\.21\.ffn_down_exps\.weight$=q3_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q3_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q3_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q6_k
266
+ ^blk\.22\.attn_gate\.weight$=q5_k
267
+ ^blk\.22\.attn_qkv\.weight$=q5_k
268
+ ^blk\.22\.ffn_down_exps\.weight$=q3_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q3_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q3_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q6_k
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q5_k
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q3_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q3_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q3_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q5_k
286
+ ^blk\.24\.attn_qkv\.weight$=q5_k
287
+ ^blk\.24\.ffn_down_exps\.weight$=q3_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q3_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q3_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q6_k
294
+ ^blk\.25\.attn_gate\.weight$=q5_k
295
+ ^blk\.25\.attn_qkv\.weight$=q5_k
296
+ ^blk\.25\.ffn_down_exps\.weight$=q3_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q3_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q3_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q6_k
303
+ ^blk\.26\.attn_gate\.weight$=q5_k
304
+ ^blk\.26\.attn_qkv\.weight$=q5_k
305
+ ^blk\.26\.ffn_down_exps\.weight$=q3_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q3_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q3_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q6_k
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q5_k
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q3_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q3_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q3_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q5_k
323
+ ^blk\.28\.attn_qkv\.weight$=q5_k
324
+ ^blk\.28\.ffn_down_exps\.weight$=q3_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q3_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q3_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q6_k
331
+ ^blk\.29\.attn_gate\.weight$=q5_k
332
+ ^blk\.29\.attn_qkv\.weight$=q5_k
333
+ ^blk\.29\.ffn_down_exps\.weight$=q3_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q3_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q3_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q6_k
340
+ ^blk\.30\.attn_gate\.weight$=q6_k
341
+ ^blk\.30\.attn_qkv\.weight$=q5_k
342
+ ^blk\.30\.ffn_down_exps\.weight$=q3_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q3_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q3_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q6_k
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q6_k
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q3_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q3_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q3_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q5_k
360
+ ^blk\.32\.attn_qkv\.weight$=q5_k
361
+ ^blk\.32\.ffn_down_exps\.weight$=q3_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q3_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q3_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q6_k
368
+ ^blk\.33\.attn_gate\.weight$=q5_k
369
+ ^blk\.33\.attn_qkv\.weight$=q5_k
370
+ ^blk\.33\.ffn_down_exps\.weight$=q3_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q3_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q3_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q6_k
377
+ ^blk\.34\.attn_gate\.weight$=q5_k
378
+ ^blk\.34\.attn_qkv\.weight$=q5_k
379
+ ^blk\.34\.ffn_down_exps\.weight$=q3_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q3_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q3_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q6_k
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q6_k
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q3_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q3_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q3_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q5_k
397
+ ^blk\.36\.attn_qkv\.weight$=q5_k
398
+ ^blk\.36\.ffn_down_exps\.weight$=q3_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q3_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q3_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q6_k
405
+ ^blk\.37\.attn_gate\.weight$=q5_k
406
+ ^blk\.37\.attn_qkv\.weight$=q5_k
407
+ ^blk\.37\.ffn_down_exps\.weight$=q3_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q3_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q3_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q6_k
414
+ ^blk\.38\.attn_gate\.weight$=q5_k
415
+ ^blk\.38\.attn_qkv\.weight$=q5_k
416
+ ^blk\.38\.ffn_down_exps\.weight$=q3_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q3_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q3_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q6_k
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q6_k
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q5_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q3_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q5_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q4_K_L.layout.json ADDED
@@ -0,0 +1,226 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q4_K_L",
5
+ "base_type": "q4_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.50",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "L",
12
+ "rung_share": 0.5,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0195,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0134,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 21275877376,
159
+ "map_body+embd": 23998103552,
160
+ "ratio": 1.1279,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.501,
164
+ "body_bpw": 5.454,
165
+ "body_bpw_heuristic": 4.896,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 24102685184,
168
+ "file_bpw_pred": 5.563,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q6_k": 20,
190
+ "q4_k": 18,
191
+ "q8_0": 2
192
+ },
193
+ "ffn_down_shexp": {
194
+ "q8_0": 40
195
+ },
196
+ "ffn_gate_exps": {
197
+ "q4_k": 38,
198
+ "q6_k": 1,
199
+ "q8_0": 1
200
+ },
201
+ "ffn_gate_shexp": {
202
+ "q8_0": 40
203
+ },
204
+ "ffn_up_exps": {
205
+ "q4_k": 20,
206
+ "q6_k": 19,
207
+ "q8_0": 1
208
+ },
209
+ "ffn_up_shexp": {
210
+ "q8_0": 40
211
+ },
212
+ "ssm_out": {
213
+ "q8_0": 30
214
+ }
215
+ },
216
+ "prior": "layout/prior.json",
217
+ "pin_rule": "floor",
218
+ "embd_cap": "heur",
219
+ "generator": "auto_quant_v2.layout",
220
+ "generator_key": "3bf8b43e20a68f0b",
221
+ "quant_type": "Q4_K_L",
222
+ "quantize_ftype": "Q4_K_M",
223
+ "llama_cpp_version": "b11716",
224
+ "unknown_kinds": {},
225
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
226
+ }
layouts/InternScience_MLExpert-Q4_K_L.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q8_0
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q6_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q6_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q6_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q4_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q6_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q6_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q4_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q6_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q6_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q4_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q6_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q6_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q4_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q6_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q6_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q4_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q6_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q6_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q4_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q6_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q6_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q4_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q6_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q6_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q4_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q6_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q6_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q4_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q6_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q4_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q4_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q4_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q6_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q4_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q4_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q4_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q4_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q4_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q4_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q4_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q4_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q4_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q4_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q4_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q6_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q4_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q4_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q4_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q4_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q4_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q4_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q4_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q4_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q4_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q4_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q4_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q4_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q4_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q4_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q4_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q4_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q4_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q4_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q4_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q4_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q4_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q4_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q4_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q4_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q4_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q4_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q4_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q4_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q4_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q4_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q4_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q4_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q4_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q4_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q4_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q4_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q4_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q4_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q4_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q4_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q4_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q4_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q4_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q4_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q6_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q4_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q6_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q6_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q4_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q6_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q6_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q4_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q6_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q6_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q4_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q6_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q6_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q4_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q6_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q6_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q4_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q6_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q6_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q4_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q6_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q6_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q4_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q6_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q6_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q4_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q6_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q8_0
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q4_K_M.layout.json ADDED
@@ -0,0 +1,226 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q4_K_M",
5
+ "base_type": "q4_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.70",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0195,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0134,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 21275877376,
159
+ "map_body+embd": 22202941440,
160
+ "ratio": 1.0436,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.701,
164
+ "body_bpw": 5.027,
165
+ "body_bpw_heuristic": 4.896,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 22307523072,
168
+ "file_bpw_pred": 5.149,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q4_k": 30,
190
+ "q6_k": 5,
191
+ "q5_k": 4,
192
+ "q8_0": 1
193
+ },
194
+ "ffn_down_shexp": {
195
+ "q8_0": 40
196
+ },
197
+ "ffn_gate_exps": {
198
+ "q4_k": 39,
199
+ "q6_k": 1
200
+ },
201
+ "ffn_gate_shexp": {
202
+ "q8_0": 40
203
+ },
204
+ "ffn_up_exps": {
205
+ "q4_k": 29,
206
+ "q6_k": 10,
207
+ "q8_0": 1
208
+ },
209
+ "ffn_up_shexp": {
210
+ "q8_0": 40
211
+ },
212
+ "ssm_out": {
213
+ "q8_0": 30
214
+ }
215
+ },
216
+ "prior": "layout/prior.json",
217
+ "pin_rule": "floor",
218
+ "embd_cap": "heur",
219
+ "generator": "auto_quant_v2.layout",
220
+ "generator_key": "3bf8b43e20a68f0b",
221
+ "quant_type": "Q4_K_M",
222
+ "quantize_ftype": "Q4_K_M",
223
+ "llama_cpp_version": "b11716",
224
+ "unknown_kinds": {},
225
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
226
+ }
layouts/InternScience_MLExpert-Q4_K_M.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q6_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q4_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q6_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q5_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q4_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q6_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q5_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q4_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q6_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q6_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q4_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q6_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q5_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q4_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q6_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q5_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q4_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q6_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q4_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q4_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q6_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q6_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q4_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q6_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q4_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q4_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q6_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q4_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q4_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q6_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q4_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q4_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q4_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q4_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q4_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q4_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q4_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q4_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q4_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q4_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q4_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q4_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q4_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q4_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q4_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q4_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q4_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q4_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q4_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q4_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q4_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q4_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q4_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q4_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q4_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q4_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q4_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q4_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q4_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q4_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q4_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q4_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q4_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q4_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q4_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q4_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q4_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q4_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q4_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q4_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q4_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q4_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q4_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q4_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q4_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q4_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q4_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q4_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q4_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q4_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q4_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q4_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q4_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q4_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q4_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q4_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q4_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q4_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q4_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q4_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q4_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q4_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q4_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q6_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q4_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q4_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q4_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q4_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q4_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q4_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q4_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q4_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q4_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q4_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q4_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q6_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q4_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q4_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q4_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q4_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q4_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q4_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q4_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q4_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q4_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q4_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q4_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q6_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q4_K_S.layout.json ADDED
@@ -0,0 +1,222 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q4_K_S",
5
+ "base_type": "q4_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.90",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0203,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0139,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 20477157376,
159
+ "map_body+embd": 20831404032,
160
+ "ratio": 1.0173,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.902,
164
+ "body_bpw": 4.7,
165
+ "body_bpw_heuristic": 4.706,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 20935985664,
168
+ "file_bpw_pred": 4.832,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q4_k": 39,
190
+ "q6_k": 1
191
+ },
192
+ "ffn_down_shexp": {
193
+ "q8_0": 40
194
+ },
195
+ "ffn_gate_exps": {
196
+ "q4_k": 40
197
+ },
198
+ "ffn_gate_shexp": {
199
+ "q8_0": 40
200
+ },
201
+ "ffn_up_exps": {
202
+ "q4_k": 39,
203
+ "q6_k": 1
204
+ },
205
+ "ffn_up_shexp": {
206
+ "q8_0": 40
207
+ },
208
+ "ssm_out": {
209
+ "q8_0": 30
210
+ }
211
+ },
212
+ "prior": "layout/prior.json",
213
+ "pin_rule": "floor",
214
+ "embd_cap": "heur",
215
+ "generator": "auto_quant_v2.layout",
216
+ "generator_key": "3bf8b43e20a68f0b",
217
+ "quant_type": "Q4_K_S",
218
+ "quantize_ftype": "Q4_K_S",
219
+ "llama_cpp_version": "b11716",
220
+ "unknown_kinds": {},
221
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
222
+ }
layouts/InternScience_MLExpert-Q4_K_S.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q4_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q4_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q4_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q4_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q4_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q4_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q4_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q4_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q4_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q4_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q4_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q4_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q4_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q4_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q4_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q4_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q4_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q4_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q4_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q4_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q4_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q4_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q4_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q4_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q4_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q4_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q4_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q4_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q4_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q4_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q4_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q4_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q4_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q4_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q4_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q4_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q4_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q4_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q4_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q4_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q4_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q4_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q4_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q4_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q4_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q4_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q4_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q4_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q4_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q4_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q4_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q4_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q4_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q4_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q4_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q4_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q4_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q4_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q4_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q4_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q4_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q4_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q4_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q4_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q4_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q4_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q4_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q4_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q4_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q4_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q4_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q4_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q4_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q4_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q4_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q4_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q4_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q4_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q4_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q4_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q4_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q4_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q4_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q4_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q4_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q4_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q4_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q4_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q4_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q4_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q4_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q4_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q4_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q4_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q4_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q4_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q4_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q4_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q4_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q4_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q4_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q4_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q4_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q4_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q4_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q4_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q4_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q4_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q4_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q4_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q4_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q4_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q4_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q4_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q4_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q4_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q4_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q6_k
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q4_k
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q6_k
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q5_K_M.layout.json ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q5_K_M",
5
+ "base_type": "q5_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.70",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0167,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.014,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 24901885952,
159
+ "map_body+embd": 26869104640,
160
+ "ratio": 1.079,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.701,
164
+ "body_bpw": 6.137,
165
+ "body_bpw_heuristic": 5.743,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 26973686272,
168
+ "file_bpw_pred": 6.226,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q5_k": 30,
190
+ "q8_0": 9,
191
+ "q6_k": 1
192
+ },
193
+ "ffn_down_shexp": {
194
+ "q8_0": 40
195
+ },
196
+ "ffn_gate_exps": {
197
+ "q5_k": 39,
198
+ "q8_0": 1
199
+ },
200
+ "ffn_gate_shexp": {
201
+ "q8_0": 40
202
+ },
203
+ "ffn_up_exps": {
204
+ "q5_k": 29,
205
+ "q8_0": 11
206
+ },
207
+ "ffn_up_shexp": {
208
+ "q8_0": 40
209
+ },
210
+ "ssm_out": {
211
+ "q8_0": 30
212
+ }
213
+ },
214
+ "prior": "layout/prior.json",
215
+ "pin_rule": "floor",
216
+ "embd_cap": "heur",
217
+ "generator": "auto_quant_v2.layout",
218
+ "generator_key": "3bf8b43e20a68f0b",
219
+ "quant_type": "Q5_K_M",
220
+ "quantize_ftype": "Q5_K_M",
221
+ "llama_cpp_version": "b11716",
222
+ "unknown_kinds": {},
223
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
224
+ }
layouts/InternScience_MLExpert-Q5_K_M.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q8_0
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q5_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q8_0
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q8_0
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q5_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q8_0
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q8_0
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q5_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q8_0
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q8_0
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q5_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q8_0
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q8_0
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q5_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q8_0
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q6_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q5_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q8_0
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q5_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q5_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q8_0
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q8_0
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q5_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q8_0
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q5_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q5_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q8_0
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q5_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q5_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q8_0
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q5_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q5_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q5_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q5_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q5_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q5_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q5_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q5_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q5_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q5_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q5_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q5_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q5_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q5_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q5_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q5_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q5_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q5_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q5_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q5_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q5_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q5_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q5_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q5_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q5_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q5_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q5_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q5_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q5_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q5_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q5_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q5_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q5_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q5_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q5_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q5_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q5_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q5_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q5_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q5_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q5_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q5_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q5_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q5_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q5_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q5_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q5_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q5_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q5_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q5_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q5_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q5_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q5_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q5_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q5_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q5_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q5_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q5_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q5_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q5_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q5_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q5_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q5_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q8_0
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q5_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q5_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q5_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q5_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q5_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q5_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q5_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q5_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q5_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q5_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q5_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q8_0
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q5_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q5_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q5_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q5_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q5_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q5_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q5_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q5_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q5_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q5_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q5_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q8_0
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q5_K_S.layout.json ADDED
@@ -0,0 +1,223 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q5_K_S",
5
+ "base_type": "q5_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.90",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0173,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0145,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 24040579072,
159
+ "map_body+embd": 25021513728,
160
+ "ratio": 1.0408,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.902,
164
+ "body_bpw": 5.697,
165
+ "body_bpw_heuristic": 5.539,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 25126095360,
168
+ "file_bpw_pred": 5.799,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q5_k": 39,
190
+ "q8_0": 1
191
+ },
192
+ "ffn_down_shexp": {
193
+ "q8_0": 40
194
+ },
195
+ "ffn_gate_exps": {
196
+ "q5_k": 39,
197
+ "q8_0": 1
198
+ },
199
+ "ffn_gate_shexp": {
200
+ "q8_0": 40
201
+ },
202
+ "ffn_up_exps": {
203
+ "q5_k": 39,
204
+ "q8_0": 1
205
+ },
206
+ "ffn_up_shexp": {
207
+ "q8_0": 40
208
+ },
209
+ "ssm_out": {
210
+ "q8_0": 30
211
+ }
212
+ },
213
+ "prior": "layout/prior.json",
214
+ "pin_rule": "floor",
215
+ "embd_cap": "heur",
216
+ "generator": "auto_quant_v2.layout",
217
+ "generator_key": "3bf8b43e20a68f0b",
218
+ "quant_type": "Q5_K_S",
219
+ "quantize_ftype": "Q5_K_S",
220
+ "llama_cpp_version": "b11716",
221
+ "unknown_kinds": {},
222
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
223
+ }
layouts/InternScience_MLExpert-Q5_K_S.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q5_k
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q5_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q5_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q5_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q5_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q5_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q5_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q5_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q5_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q5_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q5_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q5_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q5_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q5_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q5_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q5_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q5_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q5_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q5_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q5_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q5_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q5_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q5_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q5_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q5_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q5_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q5_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q5_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q5_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q5_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q5_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q5_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q5_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q5_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q5_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q5_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q5_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q5_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q5_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q5_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q5_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q5_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q5_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q5_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q5_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q5_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q5_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q5_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q5_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q5_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q5_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q5_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q5_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q5_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q5_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q5_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q5_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q5_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q5_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q5_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q5_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q5_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q5_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q5_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q5_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q5_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q5_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q5_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q5_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q5_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q5_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q5_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q5_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q5_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q5_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q5_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q5_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q5_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q5_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q5_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q5_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q5_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q5_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q5_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q5_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q5_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q5_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q5_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q5_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q5_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q5_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q5_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q5_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q5_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q5_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q5_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q5_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q5_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q5_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q5_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q5_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q5_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q5_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q5_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q5_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q5_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q5_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q5_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q5_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q5_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q5_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q5_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q5_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q5_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q5_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q5_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q5_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q8_0
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q6_K.layout.json ADDED
@@ -0,0 +1,223 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q6_K",
5
+ "base_type": "q6_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.70",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "M",
12
+ "rung_share": 0.7,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0139,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0139,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 29937844224,
159
+ "map_body+embd": 30623006720,
160
+ "ratio": 1.0229,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.708,
164
+ "body_bpw": 7.03,
165
+ "body_bpw_heuristic": 6.926,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 30727588352,
168
+ "file_bpw_pred": 7.092,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q6_k": 27,
190
+ "q8_0": 13
191
+ },
192
+ "ffn_down_shexp": {
193
+ "q8_0": 40
194
+ },
195
+ "ffn_gate_exps": {
196
+ "q6_k": 39,
197
+ "q8_0": 1
198
+ },
199
+ "ffn_gate_shexp": {
200
+ "q8_0": 40
201
+ },
202
+ "ffn_up_exps": {
203
+ "q6_k": 29,
204
+ "q8_0": 11
205
+ },
206
+ "ffn_up_shexp": {
207
+ "q8_0": 40
208
+ },
209
+ "ssm_out": {
210
+ "q8_0": 30
211
+ }
212
+ },
213
+ "prior": "layout/prior.json",
214
+ "pin_rule": "floor",
215
+ "embd_cap": "heur",
216
+ "generator": "auto_quant_v2.layout",
217
+ "generator_key": "3bf8b43e20a68f0b",
218
+ "quant_type": "Q6_K",
219
+ "quantize_ftype": "Q6_K",
220
+ "llama_cpp_version": "b11716",
221
+ "unknown_kinds": {},
222
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
223
+ }
layouts/InternScience_MLExpert-Q6_K.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q8_0
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q6_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q8_0
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q8_0
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q6_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q8_0
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q8_0
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q6_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q8_0
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q8_0
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q6_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q8_0
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q8_0
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q6_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q8_0
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q8_0
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q6_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q8_0
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q8_0
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q6_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q8_0
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q8_0
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q6_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q8_0
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q8_0
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q6_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q8_0
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q8_0
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q6_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q8_0
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q6_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q6_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q6_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q6_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q6_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q6_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q6_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q6_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q6_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q6_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q6_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q6_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q6_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q6_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q6_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q6_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q6_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q6_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q6_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q6_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q6_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q6_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q6_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q6_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q6_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q6_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q6_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q6_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q6_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q6_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q6_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q6_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q6_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q6_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q6_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q6_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q6_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q6_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q6_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q6_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q6_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q6_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q6_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q6_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q6_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q6_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q6_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q6_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q6_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q6_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q6_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q6_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q6_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q6_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q6_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q6_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q6_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q6_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q6_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q6_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q6_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q6_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q6_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q8_0
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q6_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q6_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q6_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q6_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q6_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q6_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q6_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q6_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q6_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q6_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q6_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q8_0
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q6_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q6_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q6_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q6_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q6_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q6_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q6_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q6_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q6_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q6_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q6_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q8_0
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q6_K_L.layout.json ADDED
@@ -0,0 +1,223 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q6_K_L",
5
+ "base_type": "q6_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.50",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "L",
12
+ "rung_share": 0.5,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0139,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0139,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 29937844224,
159
+ "map_body+embd": 32183287808,
160
+ "ratio": 1.075,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.503,
164
+ "body_bpw": 7.402,
165
+ "body_bpw_heuristic": 6.926,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 32287869440,
168
+ "file_bpw_pred": 7.452,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q8_0": 27,
190
+ "q6_k": 13
191
+ },
192
+ "ffn_down_shexp": {
193
+ "q8_0": 40
194
+ },
195
+ "ffn_gate_exps": {
196
+ "q6_k": 38,
197
+ "q8_0": 2
198
+ },
199
+ "ffn_gate_shexp": {
200
+ "q8_0": 40
201
+ },
202
+ "ffn_up_exps": {
203
+ "q8_0": 20,
204
+ "q6_k": 20
205
+ },
206
+ "ffn_up_shexp": {
207
+ "q8_0": 40
208
+ },
209
+ "ssm_out": {
210
+ "q8_0": 30
211
+ }
212
+ },
213
+ "prior": "layout/prior.json",
214
+ "pin_rule": "floor",
215
+ "embd_cap": "heur",
216
+ "generator": "auto_quant_v2.layout",
217
+ "generator_key": "3bf8b43e20a68f0b",
218
+ "quant_type": "Q6_K_L",
219
+ "quantize_ftype": "Q6_K",
220
+ "llama_cpp_version": "b11716",
221
+ "unknown_kinds": {},
222
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
223
+ }
layouts/InternScience_MLExpert-Q6_K_L.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q8_0
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q8_0
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q8_0
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q8_0
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q6_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q8_0
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q8_0
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q6_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q8_0
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q8_0
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q6_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q8_0
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q8_0
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q6_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q8_0
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q8_0
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q6_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q8_0
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q8_0
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q6_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q8_0
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q8_0
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q6_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q8_0
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q8_0
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q6_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q8_0
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q8_0
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q6_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q8_0
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q8_0
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q6_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q6_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q8_0
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q6_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q6_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q8_0
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q6_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q6_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q6_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q6_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q6_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q6_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q6_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q6_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q8_0
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q6_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q6_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q6_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q6_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q6_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q6_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q6_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q6_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q6_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q6_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q6_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q8_0
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q6_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q6_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q6_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q6_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q6_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q6_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q6_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q6_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q6_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q6_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q6_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q8_0
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q6_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q6_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q6_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q6_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q6_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q6_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q6_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q6_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q6_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q6_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q6_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q8_0
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q6_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q6_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q6_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q6_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q6_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q6_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q6_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q6_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q8_0
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q6_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q8_0
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q8_0
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q6_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q8_0
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q8_0
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q6_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q8_0
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q8_0
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q6_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q8_0
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q8_0
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q6_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q8_0
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q8_0
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q6_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q8_0
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q8_0
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q6_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q8_0
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q8_0
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q6_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q8_0
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q8_0
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q6_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q8_0
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q8_0
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0
layouts/InternScience_MLExpert-Q6_K_S.layout.json ADDED
@@ -0,0 +1,223 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/MLExpert-GGUF/InternScience_MLExpert-imatrix.gguf",
4
+ "ftype": "Q6_K_S",
5
+ "base_type": "q6_k",
6
+ "family": "k",
7
+ "recipe": "ident",
8
+ "variant": "shr0.90",
9
+ "auto_crush": false,
10
+ "rung_mode": "share",
11
+ "tier": "S",
12
+ "rung_share": 0.9,
13
+ "crush_pass_bytes": null,
14
+ "embd_shares": {
15
+ "output.weight": {
16
+ "share": 0.0139,
17
+ "type": "q8_0"
18
+ },
19
+ "token_embd.weight": {
20
+ "share": 0.0139,
21
+ "type": "q8_0"
22
+ }
23
+ },
24
+ "arch": "moe",
25
+ "tied": false,
26
+ "n_layer": 40,
27
+ "body_rule": true,
28
+ "pins": {
29
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
30
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
31
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
32
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
33
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
34
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
35
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
36
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
37
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
38
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
39
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
40
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
41
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
42
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
43
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
44
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
45
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
46
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
47
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
48
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
49
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
50
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
51
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
52
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
53
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
54
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
55
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
56
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
57
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
58
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
59
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
60
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
61
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
62
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
63
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
64
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
65
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
66
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
67
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
68
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
69
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
70
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
71
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
72
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
73
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
74
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
75
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
76
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
77
+ "^blk\\.32\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
78
+ "^blk\\.32\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
79
+ "^blk\\.33\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
80
+ "^blk\\.33\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
81
+ "^blk\\.34\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
82
+ "^blk\\.34\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
83
+ "^blk\\.36\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
84
+ "^blk\\.36\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
85
+ "^blk\\.37\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
86
+ "^blk\\.37\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32",
87
+ "^blk\\.38\\.ssm_alpha\\.weight$": "tiny: 65,536 params, kept at f32",
88
+ "^blk\\.38\\.ssm_beta\\.weight$": "tiny: 65,536 params, kept at f32"
89
+ },
90
+ "shape_pins": {},
91
+ "fixed_pins": {
92
+ "blk.0.ssm_alpha.weight": "f32",
93
+ "blk.0.ssm_beta.weight": "f32",
94
+ "blk.1.ssm_alpha.weight": "f32",
95
+ "blk.1.ssm_beta.weight": "f32",
96
+ "blk.2.ssm_alpha.weight": "f32",
97
+ "blk.2.ssm_beta.weight": "f32",
98
+ "blk.4.ssm_alpha.weight": "f32",
99
+ "blk.4.ssm_beta.weight": "f32",
100
+ "blk.5.ssm_alpha.weight": "f32",
101
+ "blk.5.ssm_beta.weight": "f32",
102
+ "blk.6.ssm_alpha.weight": "f32",
103
+ "blk.6.ssm_beta.weight": "f32",
104
+ "blk.8.ssm_alpha.weight": "f32",
105
+ "blk.8.ssm_beta.weight": "f32",
106
+ "blk.9.ssm_alpha.weight": "f32",
107
+ "blk.9.ssm_beta.weight": "f32",
108
+ "blk.10.ssm_alpha.weight": "f32",
109
+ "blk.10.ssm_beta.weight": "f32",
110
+ "blk.12.ssm_alpha.weight": "f32",
111
+ "blk.12.ssm_beta.weight": "f32",
112
+ "blk.13.ssm_alpha.weight": "f32",
113
+ "blk.13.ssm_beta.weight": "f32",
114
+ "blk.14.ssm_alpha.weight": "f32",
115
+ "blk.14.ssm_beta.weight": "f32",
116
+ "blk.16.ssm_alpha.weight": "f32",
117
+ "blk.16.ssm_beta.weight": "f32",
118
+ "blk.17.ssm_alpha.weight": "f32",
119
+ "blk.17.ssm_beta.weight": "f32",
120
+ "blk.18.ssm_alpha.weight": "f32",
121
+ "blk.18.ssm_beta.weight": "f32",
122
+ "blk.20.ssm_alpha.weight": "f32",
123
+ "blk.20.ssm_beta.weight": "f32",
124
+ "blk.21.ssm_alpha.weight": "f32",
125
+ "blk.21.ssm_beta.weight": "f32",
126
+ "blk.22.ssm_alpha.weight": "f32",
127
+ "blk.22.ssm_beta.weight": "f32",
128
+ "blk.24.ssm_alpha.weight": "f32",
129
+ "blk.24.ssm_beta.weight": "f32",
130
+ "blk.25.ssm_alpha.weight": "f32",
131
+ "blk.25.ssm_beta.weight": "f32",
132
+ "blk.26.ssm_alpha.weight": "f32",
133
+ "blk.26.ssm_beta.weight": "f32",
134
+ "blk.28.ssm_alpha.weight": "f32",
135
+ "blk.28.ssm_beta.weight": "f32",
136
+ "blk.29.ssm_alpha.weight": "f32",
137
+ "blk.29.ssm_beta.weight": "f32",
138
+ "blk.30.ssm_alpha.weight": "f32",
139
+ "blk.30.ssm_beta.weight": "f32",
140
+ "blk.32.ssm_alpha.weight": "f32",
141
+ "blk.32.ssm_beta.weight": "f32",
142
+ "blk.33.ssm_alpha.weight": "f32",
143
+ "blk.33.ssm_beta.weight": "f32",
144
+ "blk.34.ssm_alpha.weight": "f32",
145
+ "blk.34.ssm_beta.weight": "f32",
146
+ "blk.36.ssm_alpha.weight": "f32",
147
+ "blk.36.ssm_beta.weight": "f32",
148
+ "blk.37.ssm_alpha.weight": "f32",
149
+ "blk.37.ssm_beta.weight": "f32",
150
+ "blk.38.ssm_alpha.weight": "f32",
151
+ "blk.38.ssm_beta.weight": "f32"
152
+ },
153
+ "embeddings": {
154
+ "output.weight": "q8_0",
155
+ "token_embd.weight": "q8_0"
156
+ },
157
+ "bytes": {
158
+ "heuristic_body+embd": 29937844224,
159
+ "map_body+embd": 29257760768,
160
+ "ratio": 0.9773,
161
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
162
+ },
163
+ "body_base_share": 0.907,
164
+ "body_bpw": 6.705,
165
+ "body_bpw_heuristic": 6.926,
166
+ "total_params": 34660610688,
167
+ "file_bytes_pred": 29362342400,
168
+ "file_bpw_pred": 6.777,
169
+ "histogram": {
170
+ "attn_gate": {
171
+ "q8_0": 30
172
+ },
173
+ "attn_k": {
174
+ "q8_0": 10
175
+ },
176
+ "attn_output": {
177
+ "q8_0": 10
178
+ },
179
+ "attn_q": {
180
+ "q8_0": 10
181
+ },
182
+ "attn_qkv": {
183
+ "q8_0": 30
184
+ },
185
+ "attn_v": {
186
+ "q8_0": 10
187
+ },
188
+ "ffn_down_exps": {
189
+ "q6_k": 38,
190
+ "q8_0": 2
191
+ },
192
+ "ffn_down_shexp": {
193
+ "q8_0": 40
194
+ },
195
+ "ffn_gate_exps": {
196
+ "q6_k": 39,
197
+ "q8_0": 1
198
+ },
199
+ "ffn_gate_shexp": {
200
+ "q8_0": 40
201
+ },
202
+ "ffn_up_exps": {
203
+ "q6_k": 39,
204
+ "q8_0": 1
205
+ },
206
+ "ffn_up_shexp": {
207
+ "q8_0": 40
208
+ },
209
+ "ssm_out": {
210
+ "q8_0": 30
211
+ }
212
+ },
213
+ "prior": "layout/prior.json",
214
+ "pin_rule": "floor",
215
+ "embd_cap": "heur",
216
+ "generator": "auto_quant_v2.layout",
217
+ "generator_key": "3bf8b43e20a68f0b",
218
+ "quant_type": "Q6_K_S",
219
+ "quantize_ftype": "Q6_K",
220
+ "llama_cpp_version": "b11716",
221
+ "unknown_kinds": {},
222
+ "commit": "0933d56cdc75d35ccd1973a9eebfe2cafbd935d5"
223
+ }
layouts/InternScience_MLExpert-Q6_K_S.tensor-types.txt ADDED
@@ -0,0 +1,432 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ^blk\.0\.ssm_alpha\.weight$=f32
2
+ ^blk\.0\.ssm_beta\.weight$=f32
3
+ ^blk\.1\.ssm_alpha\.weight$=f32
4
+ ^blk\.1\.ssm_beta\.weight$=f32
5
+ ^blk\.2\.ssm_alpha\.weight$=f32
6
+ ^blk\.2\.ssm_beta\.weight$=f32
7
+ ^blk\.4\.ssm_alpha\.weight$=f32
8
+ ^blk\.4\.ssm_beta\.weight$=f32
9
+ ^blk\.5\.ssm_alpha\.weight$=f32
10
+ ^blk\.5\.ssm_beta\.weight$=f32
11
+ ^blk\.6\.ssm_alpha\.weight$=f32
12
+ ^blk\.6\.ssm_beta\.weight$=f32
13
+ ^blk\.8\.ssm_alpha\.weight$=f32
14
+ ^blk\.8\.ssm_beta\.weight$=f32
15
+ ^blk\.9\.ssm_alpha\.weight$=f32
16
+ ^blk\.9\.ssm_beta\.weight$=f32
17
+ ^blk\.10\.ssm_alpha\.weight$=f32
18
+ ^blk\.10\.ssm_beta\.weight$=f32
19
+ ^blk\.12\.ssm_alpha\.weight$=f32
20
+ ^blk\.12\.ssm_beta\.weight$=f32
21
+ ^blk\.13\.ssm_alpha\.weight$=f32
22
+ ^blk\.13\.ssm_beta\.weight$=f32
23
+ ^blk\.14\.ssm_alpha\.weight$=f32
24
+ ^blk\.14\.ssm_beta\.weight$=f32
25
+ ^blk\.16\.ssm_alpha\.weight$=f32
26
+ ^blk\.16\.ssm_beta\.weight$=f32
27
+ ^blk\.17\.ssm_alpha\.weight$=f32
28
+ ^blk\.17\.ssm_beta\.weight$=f32
29
+ ^blk\.18\.ssm_alpha\.weight$=f32
30
+ ^blk\.18\.ssm_beta\.weight$=f32
31
+ ^blk\.20\.ssm_alpha\.weight$=f32
32
+ ^blk\.20\.ssm_beta\.weight$=f32
33
+ ^blk\.21\.ssm_alpha\.weight$=f32
34
+ ^blk\.21\.ssm_beta\.weight$=f32
35
+ ^blk\.22\.ssm_alpha\.weight$=f32
36
+ ^blk\.22\.ssm_beta\.weight$=f32
37
+ ^blk\.24\.ssm_alpha\.weight$=f32
38
+ ^blk\.24\.ssm_beta\.weight$=f32
39
+ ^blk\.25\.ssm_alpha\.weight$=f32
40
+ ^blk\.25\.ssm_beta\.weight$=f32
41
+ ^blk\.26\.ssm_alpha\.weight$=f32
42
+ ^blk\.26\.ssm_beta\.weight$=f32
43
+ ^blk\.28\.ssm_alpha\.weight$=f32
44
+ ^blk\.28\.ssm_beta\.weight$=f32
45
+ ^blk\.29\.ssm_alpha\.weight$=f32
46
+ ^blk\.29\.ssm_beta\.weight$=f32
47
+ ^blk\.30\.ssm_alpha\.weight$=f32
48
+ ^blk\.30\.ssm_beta\.weight$=f32
49
+ ^blk\.32\.ssm_alpha\.weight$=f32
50
+ ^blk\.32\.ssm_beta\.weight$=f32
51
+ ^blk\.33\.ssm_alpha\.weight$=f32
52
+ ^blk\.33\.ssm_beta\.weight$=f32
53
+ ^blk\.34\.ssm_alpha\.weight$=f32
54
+ ^blk\.34\.ssm_beta\.weight$=f32
55
+ ^blk\.36\.ssm_alpha\.weight$=f32
56
+ ^blk\.36\.ssm_beta\.weight$=f32
57
+ ^blk\.37\.ssm_alpha\.weight$=f32
58
+ ^blk\.37\.ssm_beta\.weight$=f32
59
+ ^blk\.38\.ssm_alpha\.weight$=f32
60
+ ^blk\.38\.ssm_beta\.weight$=f32
61
+ ^output\.weight$=q8_0
62
+ ^token_embd\.weight$=q8_0
63
+ ^blk\.0\.attn_gate\.weight$=q8_0
64
+ ^blk\.0\.attn_qkv\.weight$=q8_0
65
+ ^blk\.0\.ffn_down_exps\.weight$=q8_0
66
+ ^blk\.0\.ffn_down_shexp\.weight$=q8_0
67
+ ^blk\.0\.ffn_gate_exps\.weight$=q6_k
68
+ ^blk\.0\.ffn_gate_shexp\.weight$=q8_0
69
+ ^blk\.0\.ffn_up_exps\.weight$=q6_k
70
+ ^blk\.0\.ffn_up_shexp\.weight$=q8_0
71
+ ^blk\.0\.ssm_out\.weight$=q8_0
72
+ ^blk\.1\.attn_gate\.weight$=q8_0
73
+ ^blk\.1\.attn_qkv\.weight$=q8_0
74
+ ^blk\.1\.ffn_down_exps\.weight$=q6_k
75
+ ^blk\.1\.ffn_down_shexp\.weight$=q8_0
76
+ ^blk\.1\.ffn_gate_exps\.weight$=q6_k
77
+ ^blk\.1\.ffn_gate_shexp\.weight$=q8_0
78
+ ^blk\.1\.ffn_up_exps\.weight$=q6_k
79
+ ^blk\.1\.ffn_up_shexp\.weight$=q8_0
80
+ ^blk\.1\.ssm_out\.weight$=q8_0
81
+ ^blk\.2\.attn_gate\.weight$=q8_0
82
+ ^blk\.2\.attn_qkv\.weight$=q8_0
83
+ ^blk\.2\.ffn_down_exps\.weight$=q6_k
84
+ ^blk\.2\.ffn_down_shexp\.weight$=q8_0
85
+ ^blk\.2\.ffn_gate_exps\.weight$=q6_k
86
+ ^blk\.2\.ffn_gate_shexp\.weight$=q8_0
87
+ ^blk\.2\.ffn_up_exps\.weight$=q6_k
88
+ ^blk\.2\.ffn_up_shexp\.weight$=q8_0
89
+ ^blk\.2\.ssm_out\.weight$=q8_0
90
+ ^blk\.3\.attn_k\.weight$=q8_0
91
+ ^blk\.3\.attn_output\.weight$=q8_0
92
+ ^blk\.3\.attn_q\.weight$=q8_0
93
+ ^blk\.3\.attn_v\.weight$=q8_0
94
+ ^blk\.3\.ffn_down_exps\.weight$=q6_k
95
+ ^blk\.3\.ffn_down_shexp\.weight$=q8_0
96
+ ^blk\.3\.ffn_gate_exps\.weight$=q6_k
97
+ ^blk\.3\.ffn_gate_shexp\.weight$=q8_0
98
+ ^blk\.3\.ffn_up_exps\.weight$=q6_k
99
+ ^blk\.3\.ffn_up_shexp\.weight$=q8_0
100
+ ^blk\.4\.attn_gate\.weight$=q8_0
101
+ ^blk\.4\.attn_qkv\.weight$=q8_0
102
+ ^blk\.4\.ffn_down_exps\.weight$=q6_k
103
+ ^blk\.4\.ffn_down_shexp\.weight$=q8_0
104
+ ^blk\.4\.ffn_gate_exps\.weight$=q6_k
105
+ ^blk\.4\.ffn_gate_shexp\.weight$=q8_0
106
+ ^blk\.4\.ffn_up_exps\.weight$=q6_k
107
+ ^blk\.4\.ffn_up_shexp\.weight$=q8_0
108
+ ^blk\.4\.ssm_out\.weight$=q8_0
109
+ ^blk\.5\.attn_gate\.weight$=q8_0
110
+ ^blk\.5\.attn_qkv\.weight$=q8_0
111
+ ^blk\.5\.ffn_down_exps\.weight$=q6_k
112
+ ^blk\.5\.ffn_down_shexp\.weight$=q8_0
113
+ ^blk\.5\.ffn_gate_exps\.weight$=q6_k
114
+ ^blk\.5\.ffn_gate_shexp\.weight$=q8_0
115
+ ^blk\.5\.ffn_up_exps\.weight$=q6_k
116
+ ^blk\.5\.ffn_up_shexp\.weight$=q8_0
117
+ ^blk\.5\.ssm_out\.weight$=q8_0
118
+ ^blk\.6\.attn_gate\.weight$=q8_0
119
+ ^blk\.6\.attn_qkv\.weight$=q8_0
120
+ ^blk\.6\.ffn_down_exps\.weight$=q6_k
121
+ ^blk\.6\.ffn_down_shexp\.weight$=q8_0
122
+ ^blk\.6\.ffn_gate_exps\.weight$=q6_k
123
+ ^blk\.6\.ffn_gate_shexp\.weight$=q8_0
124
+ ^blk\.6\.ffn_up_exps\.weight$=q6_k
125
+ ^blk\.6\.ffn_up_shexp\.weight$=q8_0
126
+ ^blk\.6\.ssm_out\.weight$=q8_0
127
+ ^blk\.7\.attn_k\.weight$=q8_0
128
+ ^blk\.7\.attn_output\.weight$=q8_0
129
+ ^blk\.7\.attn_q\.weight$=q8_0
130
+ ^blk\.7\.attn_v\.weight$=q8_0
131
+ ^blk\.7\.ffn_down_exps\.weight$=q6_k
132
+ ^blk\.7\.ffn_down_shexp\.weight$=q8_0
133
+ ^blk\.7\.ffn_gate_exps\.weight$=q6_k
134
+ ^blk\.7\.ffn_gate_shexp\.weight$=q8_0
135
+ ^blk\.7\.ffn_up_exps\.weight$=q6_k
136
+ ^blk\.7\.ffn_up_shexp\.weight$=q8_0
137
+ ^blk\.8\.attn_gate\.weight$=q8_0
138
+ ^blk\.8\.attn_qkv\.weight$=q8_0
139
+ ^blk\.8\.ffn_down_exps\.weight$=q6_k
140
+ ^blk\.8\.ffn_down_shexp\.weight$=q8_0
141
+ ^blk\.8\.ffn_gate_exps\.weight$=q6_k
142
+ ^blk\.8\.ffn_gate_shexp\.weight$=q8_0
143
+ ^blk\.8\.ffn_up_exps\.weight$=q6_k
144
+ ^blk\.8\.ffn_up_shexp\.weight$=q8_0
145
+ ^blk\.8\.ssm_out\.weight$=q8_0
146
+ ^blk\.9\.attn_gate\.weight$=q8_0
147
+ ^blk\.9\.attn_qkv\.weight$=q8_0
148
+ ^blk\.9\.ffn_down_exps\.weight$=q6_k
149
+ ^blk\.9\.ffn_down_shexp\.weight$=q8_0
150
+ ^blk\.9\.ffn_gate_exps\.weight$=q6_k
151
+ ^blk\.9\.ffn_gate_shexp\.weight$=q8_0
152
+ ^blk\.9\.ffn_up_exps\.weight$=q6_k
153
+ ^blk\.9\.ffn_up_shexp\.weight$=q8_0
154
+ ^blk\.9\.ssm_out\.weight$=q8_0
155
+ ^blk\.10\.attn_gate\.weight$=q8_0
156
+ ^blk\.10\.attn_qkv\.weight$=q8_0
157
+ ^blk\.10\.ffn_down_exps\.weight$=q6_k
158
+ ^blk\.10\.ffn_down_shexp\.weight$=q8_0
159
+ ^blk\.10\.ffn_gate_exps\.weight$=q6_k
160
+ ^blk\.10\.ffn_gate_shexp\.weight$=q8_0
161
+ ^blk\.10\.ffn_up_exps\.weight$=q6_k
162
+ ^blk\.10\.ffn_up_shexp\.weight$=q8_0
163
+ ^blk\.10\.ssm_out\.weight$=q8_0
164
+ ^blk\.11\.attn_k\.weight$=q8_0
165
+ ^blk\.11\.attn_output\.weight$=q8_0
166
+ ^blk\.11\.attn_q\.weight$=q8_0
167
+ ^blk\.11\.attn_v\.weight$=q8_0
168
+ ^blk\.11\.ffn_down_exps\.weight$=q6_k
169
+ ^blk\.11\.ffn_down_shexp\.weight$=q8_0
170
+ ^blk\.11\.ffn_gate_exps\.weight$=q6_k
171
+ ^blk\.11\.ffn_gate_shexp\.weight$=q8_0
172
+ ^blk\.11\.ffn_up_exps\.weight$=q6_k
173
+ ^blk\.11\.ffn_up_shexp\.weight$=q8_0
174
+ ^blk\.12\.attn_gate\.weight$=q8_0
175
+ ^blk\.12\.attn_qkv\.weight$=q8_0
176
+ ^blk\.12\.ffn_down_exps\.weight$=q6_k
177
+ ^blk\.12\.ffn_down_shexp\.weight$=q8_0
178
+ ^blk\.12\.ffn_gate_exps\.weight$=q6_k
179
+ ^blk\.12\.ffn_gate_shexp\.weight$=q8_0
180
+ ^blk\.12\.ffn_up_exps\.weight$=q6_k
181
+ ^blk\.12\.ffn_up_shexp\.weight$=q8_0
182
+ ^blk\.12\.ssm_out\.weight$=q8_0
183
+ ^blk\.13\.attn_gate\.weight$=q8_0
184
+ ^blk\.13\.attn_qkv\.weight$=q8_0
185
+ ^blk\.13\.ffn_down_exps\.weight$=q6_k
186
+ ^blk\.13\.ffn_down_shexp\.weight$=q8_0
187
+ ^blk\.13\.ffn_gate_exps\.weight$=q6_k
188
+ ^blk\.13\.ffn_gate_shexp\.weight$=q8_0
189
+ ^blk\.13\.ffn_up_exps\.weight$=q6_k
190
+ ^blk\.13\.ffn_up_shexp\.weight$=q8_0
191
+ ^blk\.13\.ssm_out\.weight$=q8_0
192
+ ^blk\.14\.attn_gate\.weight$=q8_0
193
+ ^blk\.14\.attn_qkv\.weight$=q8_0
194
+ ^blk\.14\.ffn_down_exps\.weight$=q6_k
195
+ ^blk\.14\.ffn_down_shexp\.weight$=q8_0
196
+ ^blk\.14\.ffn_gate_exps\.weight$=q6_k
197
+ ^blk\.14\.ffn_gate_shexp\.weight$=q8_0
198
+ ^blk\.14\.ffn_up_exps\.weight$=q6_k
199
+ ^blk\.14\.ffn_up_shexp\.weight$=q8_0
200
+ ^blk\.14\.ssm_out\.weight$=q8_0
201
+ ^blk\.15\.attn_k\.weight$=q8_0
202
+ ^blk\.15\.attn_output\.weight$=q8_0
203
+ ^blk\.15\.attn_q\.weight$=q8_0
204
+ ^blk\.15\.attn_v\.weight$=q8_0
205
+ ^blk\.15\.ffn_down_exps\.weight$=q6_k
206
+ ^blk\.15\.ffn_down_shexp\.weight$=q8_0
207
+ ^blk\.15\.ffn_gate_exps\.weight$=q6_k
208
+ ^blk\.15\.ffn_gate_shexp\.weight$=q8_0
209
+ ^blk\.15\.ffn_up_exps\.weight$=q6_k
210
+ ^blk\.15\.ffn_up_shexp\.weight$=q8_0
211
+ ^blk\.16\.attn_gate\.weight$=q8_0
212
+ ^blk\.16\.attn_qkv\.weight$=q8_0
213
+ ^blk\.16\.ffn_down_exps\.weight$=q6_k
214
+ ^blk\.16\.ffn_down_shexp\.weight$=q8_0
215
+ ^blk\.16\.ffn_gate_exps\.weight$=q6_k
216
+ ^blk\.16\.ffn_gate_shexp\.weight$=q8_0
217
+ ^blk\.16\.ffn_up_exps\.weight$=q6_k
218
+ ^blk\.16\.ffn_up_shexp\.weight$=q8_0
219
+ ^blk\.16\.ssm_out\.weight$=q8_0
220
+ ^blk\.17\.attn_gate\.weight$=q8_0
221
+ ^blk\.17\.attn_qkv\.weight$=q8_0
222
+ ^blk\.17\.ffn_down_exps\.weight$=q6_k
223
+ ^blk\.17\.ffn_down_shexp\.weight$=q8_0
224
+ ^blk\.17\.ffn_gate_exps\.weight$=q6_k
225
+ ^blk\.17\.ffn_gate_shexp\.weight$=q8_0
226
+ ^blk\.17\.ffn_up_exps\.weight$=q6_k
227
+ ^blk\.17\.ffn_up_shexp\.weight$=q8_0
228
+ ^blk\.17\.ssm_out\.weight$=q8_0
229
+ ^blk\.18\.attn_gate\.weight$=q8_0
230
+ ^blk\.18\.attn_qkv\.weight$=q8_0
231
+ ^blk\.18\.ffn_down_exps\.weight$=q6_k
232
+ ^blk\.18\.ffn_down_shexp\.weight$=q8_0
233
+ ^blk\.18\.ffn_gate_exps\.weight$=q6_k
234
+ ^blk\.18\.ffn_gate_shexp\.weight$=q8_0
235
+ ^blk\.18\.ffn_up_exps\.weight$=q6_k
236
+ ^blk\.18\.ffn_up_shexp\.weight$=q8_0
237
+ ^blk\.18\.ssm_out\.weight$=q8_0
238
+ ^blk\.19\.attn_k\.weight$=q8_0
239
+ ^blk\.19\.attn_output\.weight$=q8_0
240
+ ^blk\.19\.attn_q\.weight$=q8_0
241
+ ^blk\.19\.attn_v\.weight$=q8_0
242
+ ^blk\.19\.ffn_down_exps\.weight$=q6_k
243
+ ^blk\.19\.ffn_down_shexp\.weight$=q8_0
244
+ ^blk\.19\.ffn_gate_exps\.weight$=q6_k
245
+ ^blk\.19\.ffn_gate_shexp\.weight$=q8_0
246
+ ^blk\.19\.ffn_up_exps\.weight$=q6_k
247
+ ^blk\.19\.ffn_up_shexp\.weight$=q8_0
248
+ ^blk\.20\.attn_gate\.weight$=q8_0
249
+ ^blk\.20\.attn_qkv\.weight$=q8_0
250
+ ^blk\.20\.ffn_down_exps\.weight$=q6_k
251
+ ^blk\.20\.ffn_down_shexp\.weight$=q8_0
252
+ ^blk\.20\.ffn_gate_exps\.weight$=q6_k
253
+ ^blk\.20\.ffn_gate_shexp\.weight$=q8_0
254
+ ^blk\.20\.ffn_up_exps\.weight$=q6_k
255
+ ^blk\.20\.ffn_up_shexp\.weight$=q8_0
256
+ ^blk\.20\.ssm_out\.weight$=q8_0
257
+ ^blk\.21\.attn_gate\.weight$=q8_0
258
+ ^blk\.21\.attn_qkv\.weight$=q8_0
259
+ ^blk\.21\.ffn_down_exps\.weight$=q6_k
260
+ ^blk\.21\.ffn_down_shexp\.weight$=q8_0
261
+ ^blk\.21\.ffn_gate_exps\.weight$=q6_k
262
+ ^blk\.21\.ffn_gate_shexp\.weight$=q8_0
263
+ ^blk\.21\.ffn_up_exps\.weight$=q6_k
264
+ ^blk\.21\.ffn_up_shexp\.weight$=q8_0
265
+ ^blk\.21\.ssm_out\.weight$=q8_0
266
+ ^blk\.22\.attn_gate\.weight$=q8_0
267
+ ^blk\.22\.attn_qkv\.weight$=q8_0
268
+ ^blk\.22\.ffn_down_exps\.weight$=q6_k
269
+ ^blk\.22\.ffn_down_shexp\.weight$=q8_0
270
+ ^blk\.22\.ffn_gate_exps\.weight$=q6_k
271
+ ^blk\.22\.ffn_gate_shexp\.weight$=q8_0
272
+ ^blk\.22\.ffn_up_exps\.weight$=q6_k
273
+ ^blk\.22\.ffn_up_shexp\.weight$=q8_0
274
+ ^blk\.22\.ssm_out\.weight$=q8_0
275
+ ^blk\.23\.attn_k\.weight$=q8_0
276
+ ^blk\.23\.attn_output\.weight$=q8_0
277
+ ^blk\.23\.attn_q\.weight$=q8_0
278
+ ^blk\.23\.attn_v\.weight$=q8_0
279
+ ^blk\.23\.ffn_down_exps\.weight$=q6_k
280
+ ^blk\.23\.ffn_down_shexp\.weight$=q8_0
281
+ ^blk\.23\.ffn_gate_exps\.weight$=q6_k
282
+ ^blk\.23\.ffn_gate_shexp\.weight$=q8_0
283
+ ^blk\.23\.ffn_up_exps\.weight$=q6_k
284
+ ^blk\.23\.ffn_up_shexp\.weight$=q8_0
285
+ ^blk\.24\.attn_gate\.weight$=q8_0
286
+ ^blk\.24\.attn_qkv\.weight$=q8_0
287
+ ^blk\.24\.ffn_down_exps\.weight$=q6_k
288
+ ^blk\.24\.ffn_down_shexp\.weight$=q8_0
289
+ ^blk\.24\.ffn_gate_exps\.weight$=q6_k
290
+ ^blk\.24\.ffn_gate_shexp\.weight$=q8_0
291
+ ^blk\.24\.ffn_up_exps\.weight$=q6_k
292
+ ^blk\.24\.ffn_up_shexp\.weight$=q8_0
293
+ ^blk\.24\.ssm_out\.weight$=q8_0
294
+ ^blk\.25\.attn_gate\.weight$=q8_0
295
+ ^blk\.25\.attn_qkv\.weight$=q8_0
296
+ ^blk\.25\.ffn_down_exps\.weight$=q6_k
297
+ ^blk\.25\.ffn_down_shexp\.weight$=q8_0
298
+ ^blk\.25\.ffn_gate_exps\.weight$=q6_k
299
+ ^blk\.25\.ffn_gate_shexp\.weight$=q8_0
300
+ ^blk\.25\.ffn_up_exps\.weight$=q6_k
301
+ ^blk\.25\.ffn_up_shexp\.weight$=q8_0
302
+ ^blk\.25\.ssm_out\.weight$=q8_0
303
+ ^blk\.26\.attn_gate\.weight$=q8_0
304
+ ^blk\.26\.attn_qkv\.weight$=q8_0
305
+ ^blk\.26\.ffn_down_exps\.weight$=q6_k
306
+ ^blk\.26\.ffn_down_shexp\.weight$=q8_0
307
+ ^blk\.26\.ffn_gate_exps\.weight$=q6_k
308
+ ^blk\.26\.ffn_gate_shexp\.weight$=q8_0
309
+ ^blk\.26\.ffn_up_exps\.weight$=q6_k
310
+ ^blk\.26\.ffn_up_shexp\.weight$=q8_0
311
+ ^blk\.26\.ssm_out\.weight$=q8_0
312
+ ^blk\.27\.attn_k\.weight$=q8_0
313
+ ^blk\.27\.attn_output\.weight$=q8_0
314
+ ^blk\.27\.attn_q\.weight$=q8_0
315
+ ^blk\.27\.attn_v\.weight$=q8_0
316
+ ^blk\.27\.ffn_down_exps\.weight$=q6_k
317
+ ^blk\.27\.ffn_down_shexp\.weight$=q8_0
318
+ ^blk\.27\.ffn_gate_exps\.weight$=q6_k
319
+ ^blk\.27\.ffn_gate_shexp\.weight$=q8_0
320
+ ^blk\.27\.ffn_up_exps\.weight$=q6_k
321
+ ^blk\.27\.ffn_up_shexp\.weight$=q8_0
322
+ ^blk\.28\.attn_gate\.weight$=q8_0
323
+ ^blk\.28\.attn_qkv\.weight$=q8_0
324
+ ^blk\.28\.ffn_down_exps\.weight$=q6_k
325
+ ^blk\.28\.ffn_down_shexp\.weight$=q8_0
326
+ ^blk\.28\.ffn_gate_exps\.weight$=q6_k
327
+ ^blk\.28\.ffn_gate_shexp\.weight$=q8_0
328
+ ^blk\.28\.ffn_up_exps\.weight$=q6_k
329
+ ^blk\.28\.ffn_up_shexp\.weight$=q8_0
330
+ ^blk\.28\.ssm_out\.weight$=q8_0
331
+ ^blk\.29\.attn_gate\.weight$=q8_0
332
+ ^blk\.29\.attn_qkv\.weight$=q8_0
333
+ ^blk\.29\.ffn_down_exps\.weight$=q6_k
334
+ ^blk\.29\.ffn_down_shexp\.weight$=q8_0
335
+ ^blk\.29\.ffn_gate_exps\.weight$=q6_k
336
+ ^blk\.29\.ffn_gate_shexp\.weight$=q8_0
337
+ ^blk\.29\.ffn_up_exps\.weight$=q6_k
338
+ ^blk\.29\.ffn_up_shexp\.weight$=q8_0
339
+ ^blk\.29\.ssm_out\.weight$=q8_0
340
+ ^blk\.30\.attn_gate\.weight$=q8_0
341
+ ^blk\.30\.attn_qkv\.weight$=q8_0
342
+ ^blk\.30\.ffn_down_exps\.weight$=q6_k
343
+ ^blk\.30\.ffn_down_shexp\.weight$=q8_0
344
+ ^blk\.30\.ffn_gate_exps\.weight$=q6_k
345
+ ^blk\.30\.ffn_gate_shexp\.weight$=q8_0
346
+ ^blk\.30\.ffn_up_exps\.weight$=q6_k
347
+ ^blk\.30\.ffn_up_shexp\.weight$=q8_0
348
+ ^blk\.30\.ssm_out\.weight$=q8_0
349
+ ^blk\.31\.attn_k\.weight$=q8_0
350
+ ^blk\.31\.attn_output\.weight$=q8_0
351
+ ^blk\.31\.attn_q\.weight$=q8_0
352
+ ^blk\.31\.attn_v\.weight$=q8_0
353
+ ^blk\.31\.ffn_down_exps\.weight$=q6_k
354
+ ^blk\.31\.ffn_down_shexp\.weight$=q8_0
355
+ ^blk\.31\.ffn_gate_exps\.weight$=q6_k
356
+ ^blk\.31\.ffn_gate_shexp\.weight$=q8_0
357
+ ^blk\.31\.ffn_up_exps\.weight$=q6_k
358
+ ^blk\.31\.ffn_up_shexp\.weight$=q8_0
359
+ ^blk\.32\.attn_gate\.weight$=q8_0
360
+ ^blk\.32\.attn_qkv\.weight$=q8_0
361
+ ^blk\.32\.ffn_down_exps\.weight$=q6_k
362
+ ^blk\.32\.ffn_down_shexp\.weight$=q8_0
363
+ ^blk\.32\.ffn_gate_exps\.weight$=q6_k
364
+ ^blk\.32\.ffn_gate_shexp\.weight$=q8_0
365
+ ^blk\.32\.ffn_up_exps\.weight$=q6_k
366
+ ^blk\.32\.ffn_up_shexp\.weight$=q8_0
367
+ ^blk\.32\.ssm_out\.weight$=q8_0
368
+ ^blk\.33\.attn_gate\.weight$=q8_0
369
+ ^blk\.33\.attn_qkv\.weight$=q8_0
370
+ ^blk\.33\.ffn_down_exps\.weight$=q6_k
371
+ ^blk\.33\.ffn_down_shexp\.weight$=q8_0
372
+ ^blk\.33\.ffn_gate_exps\.weight$=q6_k
373
+ ^blk\.33\.ffn_gate_shexp\.weight$=q8_0
374
+ ^blk\.33\.ffn_up_exps\.weight$=q6_k
375
+ ^blk\.33\.ffn_up_shexp\.weight$=q8_0
376
+ ^blk\.33\.ssm_out\.weight$=q8_0
377
+ ^blk\.34\.attn_gate\.weight$=q8_0
378
+ ^blk\.34\.attn_qkv\.weight$=q8_0
379
+ ^blk\.34\.ffn_down_exps\.weight$=q6_k
380
+ ^blk\.34\.ffn_down_shexp\.weight$=q8_0
381
+ ^blk\.34\.ffn_gate_exps\.weight$=q6_k
382
+ ^blk\.34\.ffn_gate_shexp\.weight$=q8_0
383
+ ^blk\.34\.ffn_up_exps\.weight$=q6_k
384
+ ^blk\.34\.ffn_up_shexp\.weight$=q8_0
385
+ ^blk\.34\.ssm_out\.weight$=q8_0
386
+ ^blk\.35\.attn_k\.weight$=q8_0
387
+ ^blk\.35\.attn_output\.weight$=q8_0
388
+ ^blk\.35\.attn_q\.weight$=q8_0
389
+ ^blk\.35\.attn_v\.weight$=q8_0
390
+ ^blk\.35\.ffn_down_exps\.weight$=q6_k
391
+ ^blk\.35\.ffn_down_shexp\.weight$=q8_0
392
+ ^blk\.35\.ffn_gate_exps\.weight$=q6_k
393
+ ^blk\.35\.ffn_gate_shexp\.weight$=q8_0
394
+ ^blk\.35\.ffn_up_exps\.weight$=q6_k
395
+ ^blk\.35\.ffn_up_shexp\.weight$=q8_0
396
+ ^blk\.36\.attn_gate\.weight$=q8_0
397
+ ^blk\.36\.attn_qkv\.weight$=q8_0
398
+ ^blk\.36\.ffn_down_exps\.weight$=q6_k
399
+ ^blk\.36\.ffn_down_shexp\.weight$=q8_0
400
+ ^blk\.36\.ffn_gate_exps\.weight$=q6_k
401
+ ^blk\.36\.ffn_gate_shexp\.weight$=q8_0
402
+ ^blk\.36\.ffn_up_exps\.weight$=q6_k
403
+ ^blk\.36\.ffn_up_shexp\.weight$=q8_0
404
+ ^blk\.36\.ssm_out\.weight$=q8_0
405
+ ^blk\.37\.attn_gate\.weight$=q8_0
406
+ ^blk\.37\.attn_qkv\.weight$=q8_0
407
+ ^blk\.37\.ffn_down_exps\.weight$=q6_k
408
+ ^blk\.37\.ffn_down_shexp\.weight$=q8_0
409
+ ^blk\.37\.ffn_gate_exps\.weight$=q6_k
410
+ ^blk\.37\.ffn_gate_shexp\.weight$=q8_0
411
+ ^blk\.37\.ffn_up_exps\.weight$=q6_k
412
+ ^blk\.37\.ffn_up_shexp\.weight$=q8_0
413
+ ^blk\.37\.ssm_out\.weight$=q8_0
414
+ ^blk\.38\.attn_gate\.weight$=q8_0
415
+ ^blk\.38\.attn_qkv\.weight$=q8_0
416
+ ^blk\.38\.ffn_down_exps\.weight$=q6_k
417
+ ^blk\.38\.ffn_down_shexp\.weight$=q8_0
418
+ ^blk\.38\.ffn_gate_exps\.weight$=q6_k
419
+ ^blk\.38\.ffn_gate_shexp\.weight$=q8_0
420
+ ^blk\.38\.ffn_up_exps\.weight$=q6_k
421
+ ^blk\.38\.ffn_up_shexp\.weight$=q8_0
422
+ ^blk\.38\.ssm_out\.weight$=q8_0
423
+ ^blk\.39\.attn_k\.weight$=q8_0
424
+ ^blk\.39\.attn_output\.weight$=q8_0
425
+ ^blk\.39\.attn_q\.weight$=q8_0
426
+ ^blk\.39\.attn_v\.weight$=q8_0
427
+ ^blk\.39\.ffn_down_exps\.weight$=q8_0
428
+ ^blk\.39\.ffn_down_shexp\.weight$=q8_0
429
+ ^blk\.39\.ffn_gate_exps\.weight$=q8_0
430
+ ^blk\.39\.ffn_gate_shexp\.weight$=q8_0
431
+ ^blk\.39\.ffn_up_exps\.weight$=q8_0
432
+ ^blk\.39\.ffn_up_shexp\.weight$=q8_0