Hugging Face's logo

bartowski
/
UserRepGen-4B-GGUF

Text Generation
GGUF
English
user-representation
user-simulation
simulation
multi-turn
conversational
Model card Files
xet
 Community
bartowski commited on
Commit
6fcca44
·
verified ·
1 Parent(s): 76eac5b

Upload folder using huggingface_hub

Browse files
Files changed (36) hide show
  1. layouts/UserRepGen-4B-IQ2_M.layout.json +194 -0
  2. layouts/UserRepGen-4B-IQ2_M.tensor-types.txt +249 -0
  3. layouts/UserRepGen-4B-IQ3_M.layout.json +194 -0
  4. layouts/UserRepGen-4B-IQ3_M.tensor-types.txt +249 -0
  5. layouts/UserRepGen-4B-IQ3_XS.layout.json +189 -0
  6. layouts/UserRepGen-4B-IQ3_XS.tensor-types.txt +249 -0
  7. layouts/UserRepGen-4B-IQ3_XXS.layout.json +193 -0
  8. layouts/UserRepGen-4B-IQ3_XXS.tensor-types.txt +249 -0
  9. layouts/UserRepGen-4B-IQ4_NL.layout.json +190 -0
  10. layouts/UserRepGen-4B-IQ4_NL.tensor-types.txt +249 -0
  11. layouts/UserRepGen-4B-IQ4_XS.layout.json +190 -0
  12. layouts/UserRepGen-4B-IQ4_XS.tensor-types.txt +249 -0
  13. layouts/UserRepGen-4B-Q2_K.layout.json +191 -0
  14. layouts/UserRepGen-4B-Q2_K.tensor-types.txt +249 -0
  15. layouts/UserRepGen-4B-Q3_K_L.layout.json +196 -0
  16. layouts/UserRepGen-4B-Q3_K_L.tensor-types.txt +249 -0
  17. layouts/UserRepGen-4B-Q3_K_M.layout.json +194 -0
  18. layouts/UserRepGen-4B-Q3_K_M.tensor-types.txt +249 -0
  19. layouts/UserRepGen-4B-Q3_K_S.layout.json +190 -0
  20. layouts/UserRepGen-4B-Q3_K_S.tensor-types.txt +249 -0
  21. layouts/UserRepGen-4B-Q4_K_L.layout.json +194 -0
  22. layouts/UserRepGen-4B-Q4_K_L.tensor-types.txt +249 -0
  23. layouts/UserRepGen-4B-Q4_K_M.layout.json +191 -0
  24. layouts/UserRepGen-4B-Q4_K_M.tensor-types.txt +249 -0
  25. layouts/UserRepGen-4B-Q4_K_S.layout.json +190 -0
  26. layouts/UserRepGen-4B-Q4_K_S.tensor-types.txt +249 -0
  27. layouts/UserRepGen-4B-Q5_K_M.layout.json +187 -0
  28. layouts/UserRepGen-4B-Q5_K_M.tensor-types.txt +249 -0
  29. layouts/UserRepGen-4B-Q5_K_S.layout.json +189 -0
  30. layouts/UserRepGen-4B-Q5_K_S.tensor-types.txt +249 -0
  31. layouts/UserRepGen-4B-Q6_K.layout.json +187 -0
  32. layouts/UserRepGen-4B-Q6_K.tensor-types.txt +249 -0
  33. layouts/UserRepGen-4B-Q6_K_L.layout.json +188 -0
  34. layouts/UserRepGen-4B-Q6_K_L.tensor-types.txt +249 -0
  35. layouts/UserRepGen-4B-Q6_K_S.layout.json +187 -0
  36. layouts/UserRepGen-4B-Q6_K_S.tensor-types.txt +249 -0
layouts/UserRepGen-4B-IQ2_M.layout.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-imatrix.gguf",
4
+ "ftype": "IQ2_M",
5
+ "base_type": "iq2_s",
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
+ "token_embd.weight": {
16
+ "share": 0.2408,
17
+ "type": "q3_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q3_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 1795031040,
130
+ "map_body+embd": 1555445760,
131
+ "ratio": 0.8665,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 2.877,
136
+ "body_bpw_heuristic": 3.047,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 1575020544,
139
+ "file_bpw_pred": 2.996,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "iq2_s": 23,
143
+ "q5_k": 1
144
+ },
145
+ "attn_k": {
146
+ "iq4_xs": 4,
147
+ "q4_k": 4
148
+ },
149
+ "attn_output": {
150
+ "iq4_xs": 7,
151
+ "q5_k": 1
152
+ },
153
+ "attn_q": {
154
+ "iq2_s": 8
155
+ },
156
+ "attn_qkv": {
157
+ "iq2_s": 23,
158
+ "iq4_xs": 1
159
+ },
160
+ "attn_v": {
161
+ "q4_k": 5,
162
+ "q5_k": 3
163
+ },
164
+ "ffn_down": {
165
+ "iq2_s": 27,
166
+ "iq4_xs": 2,
167
+ "iq3_s": 2,
168
+ "iq3_xxs": 1
169
+ },
170
+ "ffn_gate": {
171
+ "iq2_s": 31,
172
+ "iq4_xs": 1
173
+ },
174
+ "ffn_up": {
175
+ "iq2_s": 23,
176
+ "iq3_s": 8,
177
+ "iq4_xs": 1
178
+ },
179
+ "ssm_out": {
180
+ "iq4_xs": 23,
181
+ "q5_k": 1
182
+ }
183
+ },
184
+ "prior": "layout/prior.json",
185
+ "pin_rule": "floor",
186
+ "embd_cap": "heur",
187
+ "generator": "auto_quant_v2.layout",
188
+ "generator_key": "3bf8b43e20a68f0b",
189
+ "quant_type": "IQ2_M",
190
+ "quantize_ftype": "IQ2_M",
191
+ "llama_cpp_version": "b11627",
192
+ "unknown_kinds": {},
193
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
194
+ }
layouts/UserRepGen-4B-IQ2_M.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q3_k
50
+ ^blk\.0\.attn_gate\.weight$=q5_k
51
+ ^blk\.0\.attn_qkv\.weight$=iq4_xs
52
+ ^blk\.0\.ffn_down\.weight$=iq4_xs
53
+ ^blk\.0\.ffn_gate\.weight$=iq2_s
54
+ ^blk\.0\.ffn_up\.weight$=iq3_s
55
+ ^blk\.0\.ssm_out\.weight$=q5_k
56
+ ^blk\.1\.attn_gate\.weight$=iq2_s
57
+ ^blk\.1\.attn_qkv\.weight$=iq2_s
58
+ ^blk\.1\.ffn_down\.weight$=iq2_s
59
+ ^blk\.1\.ffn_gate\.weight$=iq2_s
60
+ ^blk\.1\.ffn_up\.weight$=iq3_s
61
+ ^blk\.1\.ssm_out\.weight$=iq4_xs
62
+ ^blk\.2\.attn_gate\.weight$=iq2_s
63
+ ^blk\.2\.attn_qkv\.weight$=iq2_s
64
+ ^blk\.2\.ffn_down\.weight$=iq2_s
65
+ ^blk\.2\.ffn_gate\.weight$=iq2_s
66
+ ^blk\.2\.ffn_up\.weight$=iq3_s
67
+ ^blk\.2\.ssm_out\.weight$=iq4_xs
68
+ ^blk\.3\.attn_k\.weight$=iq4_xs
69
+ ^blk\.3\.attn_output\.weight$=iq4_xs
70
+ ^blk\.3\.attn_q\.weight$=iq2_s
71
+ ^blk\.3\.attn_v\.weight$=q4_k
72
+ ^blk\.3\.ffn_down\.weight$=iq3_s
73
+ ^blk\.3\.ffn_gate\.weight$=iq2_s
74
+ ^blk\.3\.ffn_up\.weight$=iq3_s
75
+ ^blk\.4\.attn_gate\.weight$=iq2_s
76
+ ^blk\.4\.attn_qkv\.weight$=iq2_s
77
+ ^blk\.4\.ffn_down\.weight$=iq2_s
78
+ ^blk\.4\.ffn_gate\.weight$=iq2_s
79
+ ^blk\.4\.ffn_up\.weight$=iq3_s
80
+ ^blk\.4\.ssm_out\.weight$=iq4_xs
81
+ ^blk\.5\.attn_gate\.weight$=iq2_s
82
+ ^blk\.5\.attn_qkv\.weight$=iq2_s
83
+ ^blk\.5\.ffn_down\.weight$=iq2_s
84
+ ^blk\.5\.ffn_gate\.weight$=iq2_s
85
+ ^blk\.5\.ffn_up\.weight$=iq3_s
86
+ ^blk\.5\.ssm_out\.weight$=iq4_xs
87
+ ^blk\.6\.attn_gate\.weight$=iq2_s
88
+ ^blk\.6\.attn_qkv\.weight$=iq2_s
89
+ ^blk\.6\.ffn_down\.weight$=iq2_s
90
+ ^blk\.6\.ffn_gate\.weight$=iq2_s
91
+ ^blk\.6\.ffn_up\.weight$=iq3_s
92
+ ^blk\.6\.ssm_out\.weight$=iq4_xs
93
+ ^blk\.7\.attn_k\.weight$=iq4_xs
94
+ ^blk\.7\.attn_output\.weight$=iq4_xs
95
+ ^blk\.7\.attn_q\.weight$=iq2_s
96
+ ^blk\.7\.attn_v\.weight$=q4_k
97
+ ^blk\.7\.ffn_down\.weight$=iq3_xxs
98
+ ^blk\.7\.ffn_gate\.weight$=iq2_s
99
+ ^blk\.7\.ffn_up\.weight$=iq3_s
100
+ ^blk\.8\.attn_gate\.weight$=iq2_s
101
+ ^blk\.8\.attn_qkv\.weight$=iq2_s
102
+ ^blk\.8\.ffn_down\.weight$=iq2_s
103
+ ^blk\.8\.ffn_gate\.weight$=iq2_s
104
+ ^blk\.8\.ffn_up\.weight$=iq2_s
105
+ ^blk\.8\.ssm_out\.weight$=iq4_xs
106
+ ^blk\.9\.attn_gate\.weight$=iq2_s
107
+ ^blk\.9\.attn_qkv\.weight$=iq2_s
108
+ ^blk\.9\.ffn_down\.weight$=iq2_s
109
+ ^blk\.9\.ffn_gate\.weight$=iq2_s
110
+ ^blk\.9\.ffn_up\.weight$=iq2_s
111
+ ^blk\.9\.ssm_out\.weight$=iq4_xs
112
+ ^blk\.10\.attn_gate\.weight$=iq2_s
113
+ ^blk\.10\.attn_qkv\.weight$=iq2_s
114
+ ^blk\.10\.ffn_down\.weight$=iq2_s
115
+ ^blk\.10\.ffn_gate\.weight$=iq2_s
116
+ ^blk\.10\.ffn_up\.weight$=iq2_s
117
+ ^blk\.10\.ssm_out\.weight$=iq4_xs
118
+ ^blk\.11\.attn_k\.weight$=q4_k
119
+ ^blk\.11\.attn_output\.weight$=iq4_xs
120
+ ^blk\.11\.attn_q\.weight$=iq2_s
121
+ ^blk\.11\.attn_v\.weight$=q5_k
122
+ ^blk\.11\.ffn_down\.weight$=iq2_s
123
+ ^blk\.11\.ffn_gate\.weight$=iq2_s
124
+ ^blk\.11\.ffn_up\.weight$=iq2_s
125
+ ^blk\.12\.attn_gate\.weight$=iq2_s
126
+ ^blk\.12\.attn_qkv\.weight$=iq2_s
127
+ ^blk\.12\.ffn_down\.weight$=iq2_s
128
+ ^blk\.12\.ffn_gate\.weight$=iq2_s
129
+ ^blk\.12\.ffn_up\.weight$=iq2_s
130
+ ^blk\.12\.ssm_out\.weight$=iq4_xs
131
+ ^blk\.13\.attn_gate\.weight$=iq2_s
132
+ ^blk\.13\.attn_qkv\.weight$=iq2_s
133
+ ^blk\.13\.ffn_down\.weight$=iq2_s
134
+ ^blk\.13\.ffn_gate\.weight$=iq2_s
135
+ ^blk\.13\.ffn_up\.weight$=iq2_s
136
+ ^blk\.13\.ssm_out\.weight$=iq4_xs
137
+ ^blk\.14\.attn_gate\.weight$=iq2_s
138
+ ^blk\.14\.attn_qkv\.weight$=iq2_s
139
+ ^blk\.14\.ffn_down\.weight$=iq2_s
140
+ ^blk\.14\.ffn_gate\.weight$=iq2_s
141
+ ^blk\.14\.ffn_up\.weight$=iq2_s
142
+ ^blk\.14\.ssm_out\.weight$=iq4_xs
143
+ ^blk\.15\.attn_k\.weight$=q4_k
144
+ ^blk\.15\.attn_output\.weight$=iq4_xs
145
+ ^blk\.15\.attn_q\.weight$=iq2_s
146
+ ^blk\.15\.attn_v\.weight$=q4_k
147
+ ^blk\.15\.ffn_down\.weight$=iq2_s
148
+ ^blk\.15\.ffn_gate\.weight$=iq2_s
149
+ ^blk\.15\.ffn_up\.weight$=iq2_s
150
+ ^blk\.16\.attn_gate\.weight$=iq2_s
151
+ ^blk\.16\.attn_qkv\.weight$=iq2_s
152
+ ^blk\.16\.ffn_down\.weight$=iq2_s
153
+ ^blk\.16\.ffn_gate\.weight$=iq2_s
154
+ ^blk\.16\.ffn_up\.weight$=iq2_s
155
+ ^blk\.16\.ssm_out\.weight$=iq4_xs
156
+ ^blk\.17\.attn_gate\.weight$=iq2_s
157
+ ^blk\.17\.attn_qkv\.weight$=iq2_s
158
+ ^blk\.17\.ffn_down\.weight$=iq2_s
159
+ ^blk\.17\.ffn_gate\.weight$=iq2_s
160
+ ^blk\.17\.ffn_up\.weight$=iq2_s
161
+ ^blk\.17\.ssm_out\.weight$=iq4_xs
162
+ ^blk\.18\.attn_gate\.weight$=iq2_s
163
+ ^blk\.18\.attn_qkv\.weight$=iq2_s
164
+ ^blk\.18\.ffn_down\.weight$=iq2_s
165
+ ^blk\.18\.ffn_gate\.weight$=iq2_s
166
+ ^blk\.18\.ffn_up\.weight$=iq2_s
167
+ ^blk\.18\.ssm_out\.weight$=iq4_xs
168
+ ^blk\.19\.attn_k\.weight$=q4_k
169
+ ^blk\.19\.attn_output\.weight$=iq4_xs
170
+ ^blk\.19\.attn_q\.weight$=iq2_s
171
+ ^blk\.19\.attn_v\.weight$=q4_k
172
+ ^blk\.19\.ffn_down\.weight$=iq2_s
173
+ ^blk\.19\.ffn_gate\.weight$=iq2_s
174
+ ^blk\.19\.ffn_up\.weight$=iq2_s
175
+ ^blk\.20\.attn_gate\.weight$=iq2_s
176
+ ^blk\.20\.attn_qkv\.weight$=iq2_s
177
+ ^blk\.20\.ffn_down\.weight$=iq2_s
178
+ ^blk\.20\.ffn_gate\.weight$=iq2_s
179
+ ^blk\.20\.ffn_up\.weight$=iq2_s
180
+ ^blk\.20\.ssm_out\.weight$=iq4_xs
181
+ ^blk\.21\.attn_gate\.weight$=iq2_s
182
+ ^blk\.21\.attn_qkv\.weight$=iq2_s
183
+ ^blk\.21\.ffn_down\.weight$=iq2_s
184
+ ^blk\.21\.ffn_gate\.weight$=iq2_s
185
+ ^blk\.21\.ffn_up\.weight$=iq2_s
186
+ ^blk\.21\.ssm_out\.weight$=iq4_xs
187
+ ^blk\.22\.attn_gate\.weight$=iq2_s
188
+ ^blk\.22\.attn_qkv\.weight$=iq2_s
189
+ ^blk\.22\.ffn_down\.weight$=iq2_s
190
+ ^blk\.22\.ffn_gate\.weight$=iq2_s
191
+ ^blk\.22\.ffn_up\.weight$=iq2_s
192
+ ^blk\.22\.ssm_out\.weight$=iq4_xs
193
+ ^blk\.23\.attn_k\.weight$=iq4_xs
194
+ ^blk\.23\.attn_output\.weight$=iq4_xs
195
+ ^blk\.23\.attn_q\.weight$=iq2_s
196
+ ^blk\.23\.attn_v\.weight$=q4_k
197
+ ^blk\.23\.ffn_down\.weight$=iq2_s
198
+ ^blk\.23\.ffn_gate\.weight$=iq2_s
199
+ ^blk\.23\.ffn_up\.weight$=iq2_s
200
+ ^blk\.24\.attn_gate\.weight$=iq2_s
201
+ ^blk\.24\.attn_qkv\.weight$=iq2_s
202
+ ^blk\.24\.ffn_down\.weight$=iq2_s
203
+ ^blk\.24\.ffn_gate\.weight$=iq2_s
204
+ ^blk\.24\.ffn_up\.weight$=iq2_s
205
+ ^blk\.24\.ssm_out\.weight$=iq4_xs
206
+ ^blk\.25\.attn_gate\.weight$=iq2_s
207
+ ^blk\.25\.attn_qkv\.weight$=iq2_s
208
+ ^blk\.25\.ffn_down\.weight$=iq2_s
209
+ ^blk\.25\.ffn_gate\.weight$=iq2_s
210
+ ^blk\.25\.ffn_up\.weight$=iq2_s
211
+ ^blk\.25\.ssm_out\.weight$=iq4_xs
212
+ ^blk\.26\.attn_gate\.weight$=iq2_s
213
+ ^blk\.26\.attn_qkv\.weight$=iq2_s
214
+ ^blk\.26\.ffn_down\.weight$=iq2_s
215
+ ^blk\.26\.ffn_gate\.weight$=iq2_s
216
+ ^blk\.26\.ffn_up\.weight$=iq2_s
217
+ ^blk\.26\.ssm_out\.weight$=iq4_xs
218
+ ^blk\.27\.attn_k\.weight$=q4_k
219
+ ^blk\.27\.attn_output\.weight$=iq4_xs
220
+ ^blk\.27\.attn_q\.weight$=iq2_s
221
+ ^blk\.27\.attn_v\.weight$=q5_k
222
+ ^blk\.27\.ffn_down\.weight$=iq3_s
223
+ ^blk\.27\.ffn_gate\.weight$=iq2_s
224
+ ^blk\.27\.ffn_up\.weight$=iq2_s
225
+ ^blk\.28\.attn_gate\.weight$=iq2_s
226
+ ^blk\.28\.attn_qkv\.weight$=iq2_s
227
+ ^blk\.28\.ffn_down\.weight$=iq2_s
228
+ ^blk\.28\.ffn_gate\.weight$=iq2_s
229
+ ^blk\.28\.ffn_up\.weight$=iq2_s
230
+ ^blk\.28\.ssm_out\.weight$=iq4_xs
231
+ ^blk\.29\.attn_gate\.weight$=iq2_s
232
+ ^blk\.29\.attn_qkv\.weight$=iq2_s
233
+ ^blk\.29\.ffn_down\.weight$=iq2_s
234
+ ^blk\.29\.ffn_gate\.weight$=iq2_s
235
+ ^blk\.29\.ffn_up\.weight$=iq2_s
236
+ ^blk\.29\.ssm_out\.weight$=iq4_xs
237
+ ^blk\.30\.attn_gate\.weight$=iq2_s
238
+ ^blk\.30\.attn_qkv\.weight$=iq2_s
239
+ ^blk\.30\.ffn_down\.weight$=iq2_s
240
+ ^blk\.30\.ffn_gate\.weight$=iq2_s
241
+ ^blk\.30\.ffn_up\.weight$=iq2_s
242
+ ^blk\.30\.ssm_out\.weight$=iq4_xs
243
+ ^blk\.31\.attn_k\.weight$=iq4_xs
244
+ ^blk\.31\.attn_output\.weight$=q5_k
245
+ ^blk\.31\.attn_q\.weight$=iq2_s
246
+ ^blk\.31\.attn_v\.weight$=q5_k
247
+ ^blk\.31\.ffn_down\.weight$=iq4_xs
248
+ ^blk\.31\.ffn_gate\.weight$=iq4_xs
249
+ ^blk\.31\.ffn_up\.weight$=iq4_xs
layouts/UserRepGen-4B-IQ3_M.layout.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-imatrix.gguf",
4
+ "ftype": "IQ3_M",
5
+ "base_type": "iq3_s",
6
+ "family": "iq",
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
+ "token_embd.weight": {
16
+ "share": 0.2225,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2324613120,
130
+ "map_body+embd": 2247454720,
131
+ "ratio": 0.9668,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.5,
135
+ "body_bpw": 4.241,
136
+ "body_bpw_heuristic": 4.046,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2267029504,
139
+ "file_bpw_pred": 4.312,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "iq3_s": 17,
143
+ "q5_k": 6,
144
+ "q6_k": 1
145
+ },
146
+ "attn_k": {
147
+ "q5_k": 7,
148
+ "q6_k": 1
149
+ },
150
+ "attn_output": {
151
+ "q5_k": 7,
152
+ "q6_k": 1
153
+ },
154
+ "attn_q": {
155
+ "iq3_s": 7,
156
+ "q5_k": 1
157
+ },
158
+ "attn_qkv": {
159
+ "iq3_s": 23,
160
+ "q5_k": 1
161
+ },
162
+ "attn_v": {
163
+ "q6_k": 5,
164
+ "q8_0": 3
165
+ },
166
+ "ffn_down": {
167
+ "q5_k": 17,
168
+ "iq3_s": 13,
169
+ "q4_k": 2
170
+ },
171
+ "ffn_gate": {
172
+ "iq3_s": 30,
173
+ "q5_k": 2
174
+ },
175
+ "ffn_up": {
176
+ "q5_k": 16,
177
+ "iq3_s": 16
178
+ },
179
+ "ssm_out": {
180
+ "q5_k": 23,
181
+ "q8_0": 1
182
+ }
183
+ },
184
+ "prior": "layout/prior.json",
185
+ "pin_rule": "floor",
186
+ "embd_cap": "heur",
187
+ "generator": "auto_quant_v2.layout",
188
+ "generator_key": "3bf8b43e20a68f0b",
189
+ "quant_type": "IQ3_M",
190
+ "quantize_ftype": "IQ3_M",
191
+ "llama_cpp_version": "b11627",
192
+ "unknown_kinds": {},
193
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
194
+ }
layouts/UserRepGen-4B-IQ3_M.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q6_k
51
+ ^blk\.0\.attn_qkv\.weight$=q5_k
52
+ ^blk\.0\.ffn_down\.weight$=q5_k
53
+ ^blk\.0\.ffn_gate\.weight$=q5_k
54
+ ^blk\.0\.ffn_up\.weight$=q5_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q5_k
57
+ ^blk\.1\.attn_qkv\.weight$=iq3_s
58
+ ^blk\.1\.ffn_down\.weight$=q5_k
59
+ ^blk\.1\.ffn_gate\.weight$=iq3_s
60
+ ^blk\.1\.ffn_up\.weight$=q5_k
61
+ ^blk\.1\.ssm_out\.weight$=q5_k
62
+ ^blk\.2\.attn_gate\.weight$=q5_k
63
+ ^blk\.2\.attn_qkv\.weight$=iq3_s
64
+ ^blk\.2\.ffn_down\.weight$=q5_k
65
+ ^blk\.2\.ffn_gate\.weight$=iq3_s
66
+ ^blk\.2\.ffn_up\.weight$=q5_k
67
+ ^blk\.2\.ssm_out\.weight$=q5_k
68
+ ^blk\.3\.attn_k\.weight$=q5_k
69
+ ^blk\.3\.attn_output\.weight$=q5_k
70
+ ^blk\.3\.attn_q\.weight$=iq3_s
71
+ ^blk\.3\.attn_v\.weight$=q6_k
72
+ ^blk\.3\.ffn_down\.weight$=q5_k
73
+ ^blk\.3\.ffn_gate\.weight$=iq3_s
74
+ ^blk\.3\.ffn_up\.weight$=q5_k
75
+ ^blk\.4\.attn_gate\.weight$=q5_k
76
+ ^blk\.4\.attn_qkv\.weight$=iq3_s
77
+ ^blk\.4\.ffn_down\.weight$=q5_k
78
+ ^blk\.4\.ffn_gate\.weight$=iq3_s
79
+ ^blk\.4\.ffn_up\.weight$=q5_k
80
+ ^blk\.4\.ssm_out\.weight$=q5_k
81
+ ^blk\.5\.attn_gate\.weight$=q5_k
82
+ ^blk\.5\.attn_qkv\.weight$=iq3_s
83
+ ^blk\.5\.ffn_down\.weight$=q5_k
84
+ ^blk\.5\.ffn_gate\.weight$=iq3_s
85
+ ^blk\.5\.ffn_up\.weight$=q5_k
86
+ ^blk\.5\.ssm_out\.weight$=q5_k
87
+ ^blk\.6\.attn_gate\.weight$=q5_k
88
+ ^blk\.6\.attn_qkv\.weight$=iq3_s
89
+ ^blk\.6\.ffn_down\.weight$=q5_k
90
+ ^blk\.6\.ffn_gate\.weight$=iq3_s
91
+ ^blk\.6\.ffn_up\.weight$=q5_k
92
+ ^blk\.6\.ssm_out\.weight$=q5_k
93
+ ^blk\.7\.attn_k\.weight$=q5_k
94
+ ^blk\.7\.attn_output\.weight$=q5_k
95
+ ^blk\.7\.attn_q\.weight$=iq3_s
96
+ ^blk\.7\.attn_v\.weight$=q6_k
97
+ ^blk\.7\.ffn_down\.weight$=q5_k
98
+ ^blk\.7\.ffn_gate\.weight$=iq3_s
99
+ ^blk\.7\.ffn_up\.weight$=q5_k
100
+ ^blk\.8\.attn_gate\.weight$=iq3_s
101
+ ^blk\.8\.attn_qkv\.weight$=iq3_s
102
+ ^blk\.8\.ffn_down\.weight$=iq3_s
103
+ ^blk\.8\.ffn_gate\.weight$=iq3_s
104
+ ^blk\.8\.ffn_up\.weight$=iq3_s
105
+ ^blk\.8\.ssm_out\.weight$=q5_k
106
+ ^blk\.9\.attn_gate\.weight$=iq3_s
107
+ ^blk\.9\.attn_qkv\.weight$=iq3_s
108
+ ^blk\.9\.ffn_down\.weight$=iq3_s
109
+ ^blk\.9\.ffn_gate\.weight$=iq3_s
110
+ ^blk\.9\.ffn_up\.weight$=iq3_s
111
+ ^blk\.9\.ssm_out\.weight$=q5_k
112
+ ^blk\.10\.attn_gate\.weight$=iq3_s
113
+ ^blk\.10\.attn_qkv\.weight$=iq3_s
114
+ ^blk\.10\.ffn_down\.weight$=iq3_s
115
+ ^blk\.10\.ffn_gate\.weight$=iq3_s
116
+ ^blk\.10\.ffn_up\.weight$=iq3_s
117
+ ^blk\.10\.ssm_out\.weight$=q5_k
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q5_k
120
+ ^blk\.11\.attn_q\.weight$=iq3_s
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q5_k
123
+ ^blk\.11\.ffn_gate\.weight$=iq3_s
124
+ ^blk\.11\.ffn_up\.weight$=iq3_s
125
+ ^blk\.12\.attn_gate\.weight$=iq3_s
126
+ ^blk\.12\.attn_qkv\.weight$=iq3_s
127
+ ^blk\.12\.ffn_down\.weight$=iq3_s
128
+ ^blk\.12\.ffn_gate\.weight$=iq3_s
129
+ ^blk\.12\.ffn_up\.weight$=iq3_s
130
+ ^blk\.12\.ssm_out\.weight$=q5_k
131
+ ^blk\.13\.attn_gate\.weight$=iq3_s
132
+ ^blk\.13\.attn_qkv\.weight$=iq3_s
133
+ ^blk\.13\.ffn_down\.weight$=iq3_s
134
+ ^blk\.13\.ffn_gate\.weight$=iq3_s
135
+ ^blk\.13\.ffn_up\.weight$=iq3_s
136
+ ^blk\.13\.ssm_out\.weight$=q5_k
137
+ ^blk\.14\.attn_gate\.weight$=iq3_s
138
+ ^blk\.14\.attn_qkv\.weight$=iq3_s
139
+ ^blk\.14\.ffn_down\.weight$=iq3_s
140
+ ^blk\.14\.ffn_gate\.weight$=iq3_s
141
+ ^blk\.14\.ffn_up\.weight$=iq3_s
142
+ ^blk\.14\.ssm_out\.weight$=q5_k
143
+ ^blk\.15\.attn_k\.weight$=q5_k
144
+ ^blk\.15\.attn_output\.weight$=q5_k
145
+ ^blk\.15\.attn_q\.weight$=iq3_s
146
+ ^blk\.15\.attn_v\.weight$=q6_k
147
+ ^blk\.15\.ffn_down\.weight$=q4_k
148
+ ^blk\.15\.ffn_gate\.weight$=iq3_s
149
+ ^blk\.15\.ffn_up\.weight$=iq3_s
150
+ ^blk\.16\.attn_gate\.weight$=iq3_s
151
+ ^blk\.16\.attn_qkv\.weight$=iq3_s
152
+ ^blk\.16\.ffn_down\.weight$=iq3_s
153
+ ^blk\.16\.ffn_gate\.weight$=iq3_s
154
+ ^blk\.16\.ffn_up\.weight$=iq3_s
155
+ ^blk\.16\.ssm_out\.weight$=q5_k
156
+ ^blk\.17\.attn_gate\.weight$=iq3_s
157
+ ^blk\.17\.attn_qkv\.weight$=iq3_s
158
+ ^blk\.17\.ffn_down\.weight$=iq3_s
159
+ ^blk\.17\.ffn_gate\.weight$=iq3_s
160
+ ^blk\.17\.ffn_up\.weight$=iq3_s
161
+ ^blk\.17\.ssm_out\.weight$=q5_k
162
+ ^blk\.18\.attn_gate\.weight$=iq3_s
163
+ ^blk\.18\.attn_qkv\.weight$=iq3_s
164
+ ^blk\.18\.ffn_down\.weight$=iq3_s
165
+ ^blk\.18\.ffn_gate\.weight$=iq3_s
166
+ ^blk\.18\.ffn_up\.weight$=iq3_s
167
+ ^blk\.18\.ssm_out\.weight$=q5_k
168
+ ^blk\.19\.attn_k\.weight$=q5_k
169
+ ^blk\.19\.attn_output\.weight$=q5_k
170
+ ^blk\.19\.attn_q\.weight$=iq3_s
171
+ ^blk\.19\.attn_v\.weight$=q6_k
172
+ ^blk\.19\.ffn_down\.weight$=q4_k
173
+ ^blk\.19\.ffn_gate\.weight$=iq3_s
174
+ ^blk\.19\.ffn_up\.weight$=iq3_s
175
+ ^blk\.20\.attn_gate\.weight$=iq3_s
176
+ ^blk\.20\.attn_qkv\.weight$=iq3_s
177
+ ^blk\.20\.ffn_down\.weight$=iq3_s
178
+ ^blk\.20\.ffn_gate\.weight$=iq3_s
179
+ ^blk\.20\.ffn_up\.weight$=iq3_s
180
+ ^blk\.20\.ssm_out\.weight$=q5_k
181
+ ^blk\.21\.attn_gate\.weight$=iq3_s
182
+ ^blk\.21\.attn_qkv\.weight$=iq3_s
183
+ ^blk\.21\.ffn_down\.weight$=iq3_s
184
+ ^blk\.21\.ffn_gate\.weight$=iq3_s
185
+ ^blk\.21\.ffn_up\.weight$=iq3_s
186
+ ^blk\.21\.ssm_out\.weight$=q5_k
187
+ ^blk\.22\.attn_gate\.weight$=iq3_s
188
+ ^blk\.22\.attn_qkv\.weight$=iq3_s
189
+ ^blk\.22\.ffn_down\.weight$=iq3_s
190
+ ^blk\.22\.ffn_gate\.weight$=iq3_s
191
+ ^blk\.22\.ffn_up\.weight$=iq3_s
192
+ ^blk\.22\.ssm_out\.weight$=q5_k
193
+ ^blk\.23\.attn_k\.weight$=q5_k
194
+ ^blk\.23\.attn_output\.weight$=q5_k
195
+ ^blk\.23\.attn_q\.weight$=iq3_s
196
+ ^blk\.23\.attn_v\.weight$=q6_k
197
+ ^blk\.23\.ffn_down\.weight$=iq3_s
198
+ ^blk\.23\.ffn_gate\.weight$=iq3_s
199
+ ^blk\.23\.ffn_up\.weight$=iq3_s
200
+ ^blk\.24\.attn_gate\.weight$=q5_k
201
+ ^blk\.24\.attn_qkv\.weight$=iq3_s
202
+ ^blk\.24\.ffn_down\.weight$=q5_k
203
+ ^blk\.24\.ffn_gate\.weight$=iq3_s
204
+ ^blk\.24\.ffn_up\.weight$=q5_k
205
+ ^blk\.24\.ssm_out\.weight$=q5_k
206
+ ^blk\.25\.attn_gate\.weight$=iq3_s
207
+ ^blk\.25\.attn_qkv\.weight$=iq3_s
208
+ ^blk\.25\.ffn_down\.weight$=q5_k
209
+ ^blk\.25\.ffn_gate\.weight$=iq3_s
210
+ ^blk\.25\.ffn_up\.weight$=q5_k
211
+ ^blk\.25\.ssm_out\.weight$=q5_k
212
+ ^blk\.26\.attn_gate\.weight$=iq3_s
213
+ ^blk\.26\.attn_qkv\.weight$=iq3_s
214
+ ^blk\.26\.ffn_down\.weight$=q5_k
215
+ ^blk\.26\.ffn_gate\.weight$=iq3_s
216
+ ^blk\.26\.ffn_up\.weight$=q5_k
217
+ ^blk\.26\.ssm_out\.weight$=q5_k
218
+ ^blk\.27\.attn_k\.weight$=q6_k
219
+ ^blk\.27\.attn_output\.weight$=q5_k
220
+ ^blk\.27\.attn_q\.weight$=iq3_s
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q5_k
223
+ ^blk\.27\.ffn_gate\.weight$=iq3_s
224
+ ^blk\.27\.ffn_up\.weight$=q5_k
225
+ ^blk\.28\.attn_gate\.weight$=iq3_s
226
+ ^blk\.28\.attn_qkv\.weight$=iq3_s
227
+ ^blk\.28\.ffn_down\.weight$=q5_k
228
+ ^blk\.28\.ffn_gate\.weight$=iq3_s
229
+ ^blk\.28\.ffn_up\.weight$=q5_k
230
+ ^blk\.28\.ssm_out\.weight$=q5_k
231
+ ^blk\.29\.attn_gate\.weight$=iq3_s
232
+ ^blk\.29\.attn_qkv\.weight$=iq3_s
233
+ ^blk\.29\.ffn_down\.weight$=q5_k
234
+ ^blk\.29\.ffn_gate\.weight$=iq3_s
235
+ ^blk\.29\.ffn_up\.weight$=q5_k
236
+ ^blk\.29\.ssm_out\.weight$=q5_k
237
+ ^blk\.30\.attn_gate\.weight$=iq3_s
238
+ ^blk\.30\.attn_qkv\.weight$=iq3_s
239
+ ^blk\.30\.ffn_down\.weight$=q5_k
240
+ ^blk\.30\.ffn_gate\.weight$=iq3_s
241
+ ^blk\.30\.ffn_up\.weight$=q5_k
242
+ ^blk\.30\.ssm_out\.weight$=q5_k
243
+ ^blk\.31\.attn_k\.weight$=q5_k
244
+ ^blk\.31\.attn_output\.weight$=q6_k
245
+ ^blk\.31\.attn_q\.weight$=q5_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q5_k
248
+ ^blk\.31\.ffn_gate\.weight$=q5_k
249
+ ^blk\.31\.ffn_up\.weight$=q5_k
layouts/UserRepGen-4B-IQ3_XS.layout.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-imatrix.gguf",
4
+ "ftype": "IQ3_XS",
5
+ "base_type": "iq3_s",
6
+ "family": "iq",
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
+ "token_embd.weight": {
16
+ "share": 0.2303,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2245232640,
130
+ "map_body+embd": 1951784960,
131
+ "ratio": 0.8693,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.9,
135
+ "body_bpw": 3.577,
136
+ "body_bpw_heuristic": 3.868,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 1971359744,
139
+ "file_bpw_pred": 3.75,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "iq3_s": 23,
143
+ "q6_k": 1
144
+ },
145
+ "attn_k": {
146
+ "q5_k": 8
147
+ },
148
+ "attn_output": {
149
+ "q5_k": 8
150
+ },
151
+ "attn_q": {
152
+ "iq3_s": 8
153
+ },
154
+ "attn_qkv": {
155
+ "iq3_s": 23,
156
+ "q5_k": 1
157
+ },
158
+ "attn_v": {
159
+ "q5_k": 4,
160
+ "q6_k": 4
161
+ },
162
+ "ffn_down": {
163
+ "iq3_s": 31,
164
+ "q5_k": 1
165
+ },
166
+ "ffn_gate": {
167
+ "iq3_s": 32
168
+ },
169
+ "ffn_up": {
170
+ "iq3_s": 31,
171
+ "q5_k": 1
172
+ },
173
+ "ssm_out": {
174
+ "iq3_s": 22,
175
+ "q6_k": 1,
176
+ "q5_k": 1
177
+ }
178
+ },
179
+ "prior": "layout/prior.json",
180
+ "pin_rule": "floor",
181
+ "embd_cap": "heur",
182
+ "generator": "auto_quant_v2.layout",
183
+ "generator_key": "3bf8b43e20a68f0b",
184
+ "quant_type": "IQ3_XS",
185
+ "quantize_ftype": "IQ3_XS",
186
+ "llama_cpp_version": "b11627",
187
+ "unknown_kinds": {},
188
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
189
+ }
layouts/UserRepGen-4B-IQ3_XS.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q6_k
51
+ ^blk\.0\.attn_qkv\.weight$=q5_k
52
+ ^blk\.0\.ffn_down\.weight$=iq3_s
53
+ ^blk\.0\.ffn_gate\.weight$=iq3_s
54
+ ^blk\.0\.ffn_up\.weight$=iq3_s
55
+ ^blk\.0\.ssm_out\.weight$=q6_k
56
+ ^blk\.1\.attn_gate\.weight$=iq3_s
57
+ ^blk\.1\.attn_qkv\.weight$=iq3_s
58
+ ^blk\.1\.ffn_down\.weight$=iq3_s
59
+ ^blk\.1\.ffn_gate\.weight$=iq3_s
60
+ ^blk\.1\.ffn_up\.weight$=iq3_s
61
+ ^blk\.1\.ssm_out\.weight$=q5_k
62
+ ^blk\.2\.attn_gate\.weight$=iq3_s
63
+ ^blk\.2\.attn_qkv\.weight$=iq3_s
64
+ ^blk\.2\.ffn_down\.weight$=iq3_s
65
+ ^blk\.2\.ffn_gate\.weight$=iq3_s
66
+ ^blk\.2\.ffn_up\.weight$=iq3_s
67
+ ^blk\.2\.ssm_out\.weight$=iq3_s
68
+ ^blk\.3\.attn_k\.weight$=q5_k
69
+ ^blk\.3\.attn_output\.weight$=q5_k
70
+ ^blk\.3\.attn_q\.weight$=iq3_s
71
+ ^blk\.3\.attn_v\.weight$=q5_k
72
+ ^blk\.3\.ffn_down\.weight$=iq3_s
73
+ ^blk\.3\.ffn_gate\.weight$=iq3_s
74
+ ^blk\.3\.ffn_up\.weight$=iq3_s
75
+ ^blk\.4\.attn_gate\.weight$=iq3_s
76
+ ^blk\.4\.attn_qkv\.weight$=iq3_s
77
+ ^blk\.4\.ffn_down\.weight$=iq3_s
78
+ ^blk\.4\.ffn_gate\.weight$=iq3_s
79
+ ^blk\.4\.ffn_up\.weight$=iq3_s
80
+ ^blk\.4\.ssm_out\.weight$=iq3_s
81
+ ^blk\.5\.attn_gate\.weight$=iq3_s
82
+ ^blk\.5\.attn_qkv\.weight$=iq3_s
83
+ ^blk\.5\.ffn_down\.weight$=iq3_s
84
+ ^blk\.5\.ffn_gate\.weight$=iq3_s
85
+ ^blk\.5\.ffn_up\.weight$=iq3_s
86
+ ^blk\.5\.ssm_out\.weight$=iq3_s
87
+ ^blk\.6\.attn_gate\.weight$=iq3_s
88
+ ^blk\.6\.attn_qkv\.weight$=iq3_s
89
+ ^blk\.6\.ffn_down\.weight$=iq3_s
90
+ ^blk\.6\.ffn_gate\.weight$=iq3_s
91
+ ^blk\.6\.ffn_up\.weight$=iq3_s
92
+ ^blk\.6\.ssm_out\.weight$=iq3_s
93
+ ^blk\.7\.attn_k\.weight$=q5_k
94
+ ^blk\.7\.attn_output\.weight$=q5_k
95
+ ^blk\.7\.attn_q\.weight$=iq3_s
96
+ ^blk\.7\.attn_v\.weight$=q5_k
97
+ ^blk\.7\.ffn_down\.weight$=iq3_s
98
+ ^blk\.7\.ffn_gate\.weight$=iq3_s
99
+ ^blk\.7\.ffn_up\.weight$=iq3_s
100
+ ^blk\.8\.attn_gate\.weight$=iq3_s
101
+ ^blk\.8\.attn_qkv\.weight$=iq3_s
102
+ ^blk\.8\.ffn_down\.weight$=iq3_s
103
+ ^blk\.8\.ffn_gate\.weight$=iq3_s
104
+ ^blk\.8\.ffn_up\.weight$=iq3_s
105
+ ^blk\.8\.ssm_out\.weight$=iq3_s
106
+ ^blk\.9\.attn_gate\.weight$=iq3_s
107
+ ^blk\.9\.attn_qkv\.weight$=iq3_s
108
+ ^blk\.9\.ffn_down\.weight$=iq3_s
109
+ ^blk\.9\.ffn_gate\.weight$=iq3_s
110
+ ^blk\.9\.ffn_up\.weight$=iq3_s
111
+ ^blk\.9\.ssm_out\.weight$=iq3_s
112
+ ^blk\.10\.attn_gate\.weight$=iq3_s
113
+ ^blk\.10\.attn_qkv\.weight$=iq3_s
114
+ ^blk\.10\.ffn_down\.weight$=iq3_s
115
+ ^blk\.10\.ffn_gate\.weight$=iq3_s
116
+ ^blk\.10\.ffn_up\.weight$=iq3_s
117
+ ^blk\.10\.ssm_out\.weight$=iq3_s
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q5_k
120
+ ^blk\.11\.attn_q\.weight$=iq3_s
121
+ ^blk\.11\.attn_v\.weight$=q6_k
122
+ ^blk\.11\.ffn_down\.weight$=iq3_s
123
+ ^blk\.11\.ffn_gate\.weight$=iq3_s
124
+ ^blk\.11\.ffn_up\.weight$=iq3_s
125
+ ^blk\.12\.attn_gate\.weight$=iq3_s
126
+ ^blk\.12\.attn_qkv\.weight$=iq3_s
127
+ ^blk\.12\.ffn_down\.weight$=iq3_s
128
+ ^blk\.12\.ffn_gate\.weight$=iq3_s
129
+ ^blk\.12\.ffn_up\.weight$=iq3_s
130
+ ^blk\.12\.ssm_out\.weight$=iq3_s
131
+ ^blk\.13\.attn_gate\.weight$=iq3_s
132
+ ^blk\.13\.attn_qkv\.weight$=iq3_s
133
+ ^blk\.13\.ffn_down\.weight$=iq3_s
134
+ ^blk\.13\.ffn_gate\.weight$=iq3_s
135
+ ^blk\.13\.ffn_up\.weight$=iq3_s
136
+ ^blk\.13\.ssm_out\.weight$=iq3_s
137
+ ^blk\.14\.attn_gate\.weight$=iq3_s
138
+ ^blk\.14\.attn_qkv\.weight$=iq3_s
139
+ ^blk\.14\.ffn_down\.weight$=iq3_s
140
+ ^blk\.14\.ffn_gate\.weight$=iq3_s
141
+ ^blk\.14\.ffn_up\.weight$=iq3_s
142
+ ^blk\.14\.ssm_out\.weight$=iq3_s
143
+ ^blk\.15\.attn_k\.weight$=q5_k
144
+ ^blk\.15\.attn_output\.weight$=q5_k
145
+ ^blk\.15\.attn_q\.weight$=iq3_s
146
+ ^blk\.15\.attn_v\.weight$=q6_k
147
+ ^blk\.15\.ffn_down\.weight$=iq3_s
148
+ ^blk\.15\.ffn_gate\.weight$=iq3_s
149
+ ^blk\.15\.ffn_up\.weight$=iq3_s
150
+ ^blk\.16\.attn_gate\.weight$=iq3_s
151
+ ^blk\.16\.attn_qkv\.weight$=iq3_s
152
+ ^blk\.16\.ffn_down\.weight$=iq3_s
153
+ ^blk\.16\.ffn_gate\.weight$=iq3_s
154
+ ^blk\.16\.ffn_up\.weight$=iq3_s
155
+ ^blk\.16\.ssm_out\.weight$=iq3_s
156
+ ^blk\.17\.attn_gate\.weight$=iq3_s
157
+ ^blk\.17\.attn_qkv\.weight$=iq3_s
158
+ ^blk\.17\.ffn_down\.weight$=iq3_s
159
+ ^blk\.17\.ffn_gate\.weight$=iq3_s
160
+ ^blk\.17\.ffn_up\.weight$=iq3_s
161
+ ^blk\.17\.ssm_out\.weight$=iq3_s
162
+ ^blk\.18\.attn_gate\.weight$=iq3_s
163
+ ^blk\.18\.attn_qkv\.weight$=iq3_s
164
+ ^blk\.18\.ffn_down\.weight$=iq3_s
165
+ ^blk\.18\.ffn_gate\.weight$=iq3_s
166
+ ^blk\.18\.ffn_up\.weight$=iq3_s
167
+ ^blk\.18\.ssm_out\.weight$=iq3_s
168
+ ^blk\.19\.attn_k\.weight$=q5_k
169
+ ^blk\.19\.attn_output\.weight$=q5_k
170
+ ^blk\.19\.attn_q\.weight$=iq3_s
171
+ ^blk\.19\.attn_v\.weight$=q5_k
172
+ ^blk\.19\.ffn_down\.weight$=iq3_s
173
+ ^blk\.19\.ffn_gate\.weight$=iq3_s
174
+ ^blk\.19\.ffn_up\.weight$=iq3_s
175
+ ^blk\.20\.attn_gate\.weight$=iq3_s
176
+ ^blk\.20\.attn_qkv\.weight$=iq3_s
177
+ ^blk\.20\.ffn_down\.weight$=iq3_s
178
+ ^blk\.20\.ffn_gate\.weight$=iq3_s
179
+ ^blk\.20\.ffn_up\.weight$=iq3_s
180
+ ^blk\.20\.ssm_out\.weight$=iq3_s
181
+ ^blk\.21\.attn_gate\.weight$=iq3_s
182
+ ^blk\.21\.attn_qkv\.weight$=iq3_s
183
+ ^blk\.21\.ffn_down\.weight$=iq3_s
184
+ ^blk\.21\.ffn_gate\.weight$=iq3_s
185
+ ^blk\.21\.ffn_up\.weight$=iq3_s
186
+ ^blk\.21\.ssm_out\.weight$=iq3_s
187
+ ^blk\.22\.attn_gate\.weight$=iq3_s
188
+ ^blk\.22\.attn_qkv\.weight$=iq3_s
189
+ ^blk\.22\.ffn_down\.weight$=iq3_s
190
+ ^blk\.22\.ffn_gate\.weight$=iq3_s
191
+ ^blk\.22\.ffn_up\.weight$=iq3_s
192
+ ^blk\.22\.ssm_out\.weight$=iq3_s
193
+ ^blk\.23\.attn_k\.weight$=q5_k
194
+ ^blk\.23\.attn_output\.weight$=q5_k
195
+ ^blk\.23\.attn_q\.weight$=iq3_s
196
+ ^blk\.23\.attn_v\.weight$=q5_k
197
+ ^blk\.23\.ffn_down\.weight$=iq3_s
198
+ ^blk\.23\.ffn_gate\.weight$=iq3_s
199
+ ^blk\.23\.ffn_up\.weight$=iq3_s
200
+ ^blk\.24\.attn_gate\.weight$=iq3_s
201
+ ^blk\.24\.attn_qkv\.weight$=iq3_s
202
+ ^blk\.24\.ffn_down\.weight$=iq3_s
203
+ ^blk\.24\.ffn_gate\.weight$=iq3_s
204
+ ^blk\.24\.ffn_up\.weight$=iq3_s
205
+ ^blk\.24\.ssm_out\.weight$=iq3_s
206
+ ^blk\.25\.attn_gate\.weight$=iq3_s
207
+ ^blk\.25\.attn_qkv\.weight$=iq3_s
208
+ ^blk\.25\.ffn_down\.weight$=iq3_s
209
+ ^blk\.25\.ffn_gate\.weight$=iq3_s
210
+ ^blk\.25\.ffn_up\.weight$=iq3_s
211
+ ^blk\.25\.ssm_out\.weight$=iq3_s
212
+ ^blk\.26\.attn_gate\.weight$=iq3_s
213
+ ^blk\.26\.attn_qkv\.weight$=iq3_s
214
+ ^blk\.26\.ffn_down\.weight$=iq3_s
215
+ ^blk\.26\.ffn_gate\.weight$=iq3_s
216
+ ^blk\.26\.ffn_up\.weight$=iq3_s
217
+ ^blk\.26\.ssm_out\.weight$=iq3_s
218
+ ^blk\.27\.attn_k\.weight$=q5_k
219
+ ^blk\.27\.attn_output\.weight$=q5_k
220
+ ^blk\.27\.attn_q\.weight$=iq3_s
221
+ ^blk\.27\.attn_v\.weight$=q6_k
222
+ ^blk\.27\.ffn_down\.weight$=iq3_s
223
+ ^blk\.27\.ffn_gate\.weight$=iq3_s
224
+ ^blk\.27\.ffn_up\.weight$=iq3_s
225
+ ^blk\.28\.attn_gate\.weight$=iq3_s
226
+ ^blk\.28\.attn_qkv\.weight$=iq3_s
227
+ ^blk\.28\.ffn_down\.weight$=iq3_s
228
+ ^blk\.28\.ffn_gate\.weight$=iq3_s
229
+ ^blk\.28\.ffn_up\.weight$=iq3_s
230
+ ^blk\.28\.ssm_out\.weight$=iq3_s
231
+ ^blk\.29\.attn_gate\.weight$=iq3_s
232
+ ^blk\.29\.attn_qkv\.weight$=iq3_s
233
+ ^blk\.29\.ffn_down\.weight$=iq3_s
234
+ ^blk\.29\.ffn_gate\.weight$=iq3_s
235
+ ^blk\.29\.ffn_up\.weight$=iq3_s
236
+ ^blk\.29\.ssm_out\.weight$=iq3_s
237
+ ^blk\.30\.attn_gate\.weight$=iq3_s
238
+ ^blk\.30\.attn_qkv\.weight$=iq3_s
239
+ ^blk\.30\.ffn_down\.weight$=iq3_s
240
+ ^blk\.30\.ffn_gate\.weight$=iq3_s
241
+ ^blk\.30\.ffn_up\.weight$=iq3_s
242
+ ^blk\.30\.ssm_out\.weight$=iq3_s
243
+ ^blk\.31\.attn_k\.weight$=q5_k
244
+ ^blk\.31\.attn_output\.weight$=q5_k
245
+ ^blk\.31\.attn_q\.weight$=iq3_s
246
+ ^blk\.31\.attn_v\.weight$=q6_k
247
+ ^blk\.31\.ffn_down\.weight$=q5_k
248
+ ^blk\.31\.ffn_gate\.weight$=iq3_s
249
+ ^blk\.31\.ffn_up\.weight$=q5_k
layouts/UserRepGen-4B-IQ3_XXS.layout.json ADDED
@@ -0,0 +1,193 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-imatrix.gguf",
4
+ "ftype": "IQ3_XXS",
5
+ "base_type": "iq3_xxs",
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
+ "token_embd.weight": {
16
+ "share": 0.2077,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2085109760,
130
+ "map_body+embd": 1877155840,
131
+ "ratio": 0.9003,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 3.41,
136
+ "body_bpw_heuristic": 3.698,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 1896730624,
139
+ "file_bpw_pred": 3.608,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "iq3_xxs": 23,
143
+ "q5_k": 1
144
+ },
145
+ "attn_k": {
146
+ "q5_k": 7,
147
+ "q4_k": 1
148
+ },
149
+ "attn_output": {
150
+ "q5_k": 8
151
+ },
152
+ "attn_q": {
153
+ "iq3_xxs": 8
154
+ },
155
+ "attn_qkv": {
156
+ "iq3_xxs": 23,
157
+ "q5_k": 1
158
+ },
159
+ "attn_v": {
160
+ "q5_k": 6,
161
+ "q6_k": 2
162
+ },
163
+ "ffn_down": {
164
+ "iq3_xxs": 26,
165
+ "iq4_xs": 5,
166
+ "q5_k": 1
167
+ },
168
+ "ffn_gate": {
169
+ "iq3_xxs": 31,
170
+ "q4_k": 1
171
+ },
172
+ "ffn_up": {
173
+ "iq3_xxs": 23,
174
+ "iq4_xs": 8,
175
+ "q5_k": 1
176
+ },
177
+ "ssm_out": {
178
+ "iq4_xs": 17,
179
+ "q4_k": 6,
180
+ "q6_k": 1
181
+ }
182
+ },
183
+ "prior": "layout/prior.json",
184
+ "pin_rule": "floor",
185
+ "embd_cap": "heur",
186
+ "generator": "auto_quant_v2.layout",
187
+ "generator_key": "3bf8b43e20a68f0b",
188
+ "quant_type": "IQ3_XXS",
189
+ "quantize_ftype": "IQ3_XXS",
190
+ "llama_cpp_version": "b11627",
191
+ "unknown_kinds": {},
192
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
193
+ }
layouts/UserRepGen-4B-IQ3_XXS.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q5_k
51
+ ^blk\.0\.attn_qkv\.weight$=q5_k
52
+ ^blk\.0\.ffn_down\.weight$=iq4_xs
53
+ ^blk\.0\.ffn_gate\.weight$=iq3_xxs
54
+ ^blk\.0\.ffn_up\.weight$=iq4_xs
55
+ ^blk\.0\.ssm_out\.weight$=q6_k
56
+ ^blk\.1\.attn_gate\.weight$=iq3_xxs
57
+ ^blk\.1\.attn_qkv\.weight$=iq3_xxs
58
+ ^blk\.1\.ffn_down\.weight$=iq4_xs
59
+ ^blk\.1\.ffn_gate\.weight$=iq3_xxs
60
+ ^blk\.1\.ffn_up\.weight$=iq4_xs
61
+ ^blk\.1\.ssm_out\.weight$=q4_k
62
+ ^blk\.2\.attn_gate\.weight$=iq3_xxs
63
+ ^blk\.2\.attn_qkv\.weight$=iq3_xxs
64
+ ^blk\.2\.ffn_down\.weight$=iq3_xxs
65
+ ^blk\.2\.ffn_gate\.weight$=iq3_xxs
66
+ ^blk\.2\.ffn_up\.weight$=iq4_xs
67
+ ^blk\.2\.ssm_out\.weight$=q4_k
68
+ ^blk\.3\.attn_k\.weight$=q5_k
69
+ ^blk\.3\.attn_output\.weight$=q5_k
70
+ ^blk\.3\.attn_q\.weight$=iq3_xxs
71
+ ^blk\.3\.attn_v\.weight$=q5_k
72
+ ^blk\.3\.ffn_down\.weight$=iq4_xs
73
+ ^blk\.3\.ffn_gate\.weight$=iq3_xxs
74
+ ^blk\.3\.ffn_up\.weight$=iq4_xs
75
+ ^blk\.4\.attn_gate\.weight$=iq3_xxs
76
+ ^blk\.4\.attn_qkv\.weight$=iq3_xxs
77
+ ^blk\.4\.ffn_down\.weight$=iq3_xxs
78
+ ^blk\.4\.ffn_gate\.weight$=iq3_xxs
79
+ ^blk\.4\.ffn_up\.weight$=iq4_xs
80
+ ^blk\.4\.ssm_out\.weight$=q4_k
81
+ ^blk\.5\.attn_gate\.weight$=iq3_xxs
82
+ ^blk\.5\.attn_qkv\.weight$=iq3_xxs
83
+ ^blk\.5\.ffn_down\.weight$=iq3_xxs
84
+ ^blk\.5\.ffn_gate\.weight$=iq3_xxs
85
+ ^blk\.5\.ffn_up\.weight$=iq4_xs
86
+ ^blk\.5\.ssm_out\.weight$=q4_k
87
+ ^blk\.6\.attn_gate\.weight$=iq3_xxs
88
+ ^blk\.6\.attn_qkv\.weight$=iq3_xxs
89
+ ^blk\.6\.ffn_down\.weight$=iq3_xxs
90
+ ^blk\.6\.ffn_gate\.weight$=iq3_xxs
91
+ ^blk\.6\.ffn_up\.weight$=iq4_xs
92
+ ^blk\.6\.ssm_out\.weight$=q4_k
93
+ ^blk\.7\.attn_k\.weight$=q5_k
94
+ ^blk\.7\.attn_output\.weight$=q5_k
95
+ ^blk\.7\.attn_q\.weight$=iq3_xxs
96
+ ^blk\.7\.attn_v\.weight$=q5_k
97
+ ^blk\.7\.ffn_down\.weight$=iq4_xs
98
+ ^blk\.7\.ffn_gate\.weight$=iq3_xxs
99
+ ^blk\.7\.ffn_up\.weight$=iq4_xs
100
+ ^blk\.8\.attn_gate\.weight$=iq3_xxs
101
+ ^blk\.8\.attn_qkv\.weight$=iq3_xxs
102
+ ^blk\.8\.ffn_down\.weight$=iq3_xxs
103
+ ^blk\.8\.ffn_gate\.weight$=iq3_xxs
104
+ ^blk\.8\.ffn_up\.weight$=iq3_xxs
105
+ ^blk\.8\.ssm_out\.weight$=iq4_xs
106
+ ^blk\.9\.attn_gate\.weight$=iq3_xxs
107
+ ^blk\.9\.attn_qkv\.weight$=iq3_xxs
108
+ ^blk\.9\.ffn_down\.weight$=iq3_xxs
109
+ ^blk\.9\.ffn_gate\.weight$=iq3_xxs
110
+ ^blk\.9\.ffn_up\.weight$=iq3_xxs
111
+ ^blk\.9\.ssm_out\.weight$=iq4_xs
112
+ ^blk\.10\.attn_gate\.weight$=iq3_xxs
113
+ ^blk\.10\.attn_qkv\.weight$=iq3_xxs
114
+ ^blk\.10\.ffn_down\.weight$=iq3_xxs
115
+ ^blk\.10\.ffn_gate\.weight$=iq3_xxs
116
+ ^blk\.10\.ffn_up\.weight$=iq3_xxs
117
+ ^blk\.10\.ssm_out\.weight$=iq4_xs
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q5_k
120
+ ^blk\.11\.attn_q\.weight$=iq3_xxs
121
+ ^blk\.11\.attn_v\.weight$=q5_k
122
+ ^blk\.11\.ffn_down\.weight$=iq3_xxs
123
+ ^blk\.11\.ffn_gate\.weight$=iq3_xxs
124
+ ^blk\.11\.ffn_up\.weight$=iq3_xxs
125
+ ^blk\.12\.attn_gate\.weight$=iq3_xxs
126
+ ^blk\.12\.attn_qkv\.weight$=iq3_xxs
127
+ ^blk\.12\.ffn_down\.weight$=iq3_xxs
128
+ ^blk\.12\.ffn_gate\.weight$=iq3_xxs
129
+ ^blk\.12\.ffn_up\.weight$=iq3_xxs
130
+ ^blk\.12\.ssm_out\.weight$=iq4_xs
131
+ ^blk\.13\.attn_gate\.weight$=iq3_xxs
132
+ ^blk\.13\.attn_qkv\.weight$=iq3_xxs
133
+ ^blk\.13\.ffn_down\.weight$=iq3_xxs
134
+ ^blk\.13\.ffn_gate\.weight$=iq3_xxs
135
+ ^blk\.13\.ffn_up\.weight$=iq3_xxs
136
+ ^blk\.13\.ssm_out\.weight$=iq4_xs
137
+ ^blk\.14\.attn_gate\.weight$=iq3_xxs
138
+ ^blk\.14\.attn_qkv\.weight$=iq3_xxs
139
+ ^blk\.14\.ffn_down\.weight$=iq3_xxs
140
+ ^blk\.14\.ffn_gate\.weight$=iq3_xxs
141
+ ^blk\.14\.ffn_up\.weight$=iq3_xxs
142
+ ^blk\.14\.ssm_out\.weight$=iq4_xs
143
+ ^blk\.15\.attn_k\.weight$=q5_k
144
+ ^blk\.15\.attn_output\.weight$=q5_k
145
+ ^blk\.15\.attn_q\.weight$=iq3_xxs
146
+ ^blk\.15\.attn_v\.weight$=q5_k
147
+ ^blk\.15\.ffn_down\.weight$=iq3_xxs
148
+ ^blk\.15\.ffn_gate\.weight$=iq3_xxs
149
+ ^blk\.15\.ffn_up\.weight$=iq3_xxs
150
+ ^blk\.16\.attn_gate\.weight$=iq3_xxs
151
+ ^blk\.16\.attn_qkv\.weight$=iq3_xxs
152
+ ^blk\.16\.ffn_down\.weight$=iq3_xxs
153
+ ^blk\.16\.ffn_gate\.weight$=iq3_xxs
154
+ ^blk\.16\.ffn_up\.weight$=iq3_xxs
155
+ ^blk\.16\.ssm_out\.weight$=iq4_xs
156
+ ^blk\.17\.attn_gate\.weight$=iq3_xxs
157
+ ^blk\.17\.attn_qkv\.weight$=iq3_xxs
158
+ ^blk\.17\.ffn_down\.weight$=iq3_xxs
159
+ ^blk\.17\.ffn_gate\.weight$=iq3_xxs
160
+ ^blk\.17\.ffn_up\.weight$=iq3_xxs
161
+ ^blk\.17\.ssm_out\.weight$=iq4_xs
162
+ ^blk\.18\.attn_gate\.weight$=iq3_xxs
163
+ ^blk\.18\.attn_qkv\.weight$=iq3_xxs
164
+ ^blk\.18\.ffn_down\.weight$=iq3_xxs
165
+ ^blk\.18\.ffn_gate\.weight$=iq3_xxs
166
+ ^blk\.18\.ffn_up\.weight$=iq3_xxs
167
+ ^blk\.18\.ssm_out\.weight$=iq4_xs
168
+ ^blk\.19\.attn_k\.weight$=q5_k
169
+ ^blk\.19\.attn_output\.weight$=q5_k
170
+ ^blk\.19\.attn_q\.weight$=iq3_xxs
171
+ ^blk\.19\.attn_v\.weight$=q5_k
172
+ ^blk\.19\.ffn_down\.weight$=iq3_xxs
173
+ ^blk\.19\.ffn_gate\.weight$=iq3_xxs
174
+ ^blk\.19\.ffn_up\.weight$=iq3_xxs
175
+ ^blk\.20\.attn_gate\.weight$=iq3_xxs
176
+ ^blk\.20\.attn_qkv\.weight$=iq3_xxs
177
+ ^blk\.20\.ffn_down\.weight$=iq3_xxs
178
+ ^blk\.20\.ffn_gate\.weight$=iq3_xxs
179
+ ^blk\.20\.ffn_up\.weight$=iq3_xxs
180
+ ^blk\.20\.ssm_out\.weight$=iq4_xs
181
+ ^blk\.21\.attn_gate\.weight$=iq3_xxs
182
+ ^blk\.21\.attn_qkv\.weight$=iq3_xxs
183
+ ^blk\.21\.ffn_down\.weight$=iq3_xxs
184
+ ^blk\.21\.ffn_gate\.weight$=iq3_xxs
185
+ ^blk\.21\.ffn_up\.weight$=iq3_xxs
186
+ ^blk\.21\.ssm_out\.weight$=iq4_xs
187
+ ^blk\.22\.attn_gate\.weight$=iq3_xxs
188
+ ^blk\.22\.attn_qkv\.weight$=iq3_xxs
189
+ ^blk\.22\.ffn_down\.weight$=iq3_xxs
190
+ ^blk\.22\.ffn_gate\.weight$=iq3_xxs
191
+ ^blk\.22\.ffn_up\.weight$=iq3_xxs
192
+ ^blk\.22\.ssm_out\.weight$=iq4_xs
193
+ ^blk\.23\.attn_k\.weight$=q5_k
194
+ ^blk\.23\.attn_output\.weight$=q5_k
195
+ ^blk\.23\.attn_q\.weight$=iq3_xxs
196
+ ^blk\.23\.attn_v\.weight$=q5_k
197
+ ^blk\.23\.ffn_down\.weight$=iq3_xxs
198
+ ^blk\.23\.ffn_gate\.weight$=iq3_xxs
199
+ ^blk\.23\.ffn_up\.weight$=iq3_xxs
200
+ ^blk\.24\.attn_gate\.weight$=iq3_xxs
201
+ ^blk\.24\.attn_qkv\.weight$=iq3_xxs
202
+ ^blk\.24\.ffn_down\.weight$=iq3_xxs
203
+ ^blk\.24\.ffn_gate\.weight$=iq3_xxs
204
+ ^blk\.24\.ffn_up\.weight$=iq3_xxs
205
+ ^blk\.24\.ssm_out\.weight$=q4_k
206
+ ^blk\.25\.attn_gate\.weight$=iq3_xxs
207
+ ^blk\.25\.attn_qkv\.weight$=iq3_xxs
208
+ ^blk\.25\.ffn_down\.weight$=iq3_xxs
209
+ ^blk\.25\.ffn_gate\.weight$=iq3_xxs
210
+ ^blk\.25\.ffn_up\.weight$=iq3_xxs
211
+ ^blk\.25\.ssm_out\.weight$=iq4_xs
212
+ ^blk\.26\.attn_gate\.weight$=iq3_xxs
213
+ ^blk\.26\.attn_qkv\.weight$=iq3_xxs
214
+ ^blk\.26\.ffn_down\.weight$=iq3_xxs
215
+ ^blk\.26\.ffn_gate\.weight$=iq3_xxs
216
+ ^blk\.26\.ffn_up\.weight$=iq3_xxs
217
+ ^blk\.26\.ssm_out\.weight$=iq4_xs
218
+ ^blk\.27\.attn_k\.weight$=q5_k
219
+ ^blk\.27\.attn_output\.weight$=q5_k
220
+ ^blk\.27\.attn_q\.weight$=iq3_xxs
221
+ ^blk\.27\.attn_v\.weight$=q6_k
222
+ ^blk\.27\.ffn_down\.weight$=iq4_xs
223
+ ^blk\.27\.ffn_gate\.weight$=iq3_xxs
224
+ ^blk\.27\.ffn_up\.weight$=iq3_xxs
225
+ ^blk\.28\.attn_gate\.weight$=iq3_xxs
226
+ ^blk\.28\.attn_qkv\.weight$=iq3_xxs
227
+ ^blk\.28\.ffn_down\.weight$=iq3_xxs
228
+ ^blk\.28\.ffn_gate\.weight$=iq3_xxs
229
+ ^blk\.28\.ffn_up\.weight$=iq3_xxs
230
+ ^blk\.28\.ssm_out\.weight$=iq4_xs
231
+ ^blk\.29\.attn_gate\.weight$=iq3_xxs
232
+ ^blk\.29\.attn_qkv\.weight$=iq3_xxs
233
+ ^blk\.29\.ffn_down\.weight$=iq3_xxs
234
+ ^blk\.29\.ffn_gate\.weight$=iq3_xxs
235
+ ^blk\.29\.ffn_up\.weight$=iq3_xxs
236
+ ^blk\.29\.ssm_out\.weight$=iq4_xs
237
+ ^blk\.30\.attn_gate\.weight$=iq3_xxs
238
+ ^blk\.30\.attn_qkv\.weight$=iq3_xxs
239
+ ^blk\.30\.ffn_down\.weight$=iq3_xxs
240
+ ^blk\.30\.ffn_gate\.weight$=iq3_xxs
241
+ ^blk\.30\.ffn_up\.weight$=iq3_xxs
242
+ ^blk\.30\.ssm_out\.weight$=iq4_xs
243
+ ^blk\.31\.attn_k\.weight$=q4_k
244
+ ^blk\.31\.attn_output\.weight$=q5_k
245
+ ^blk\.31\.attn_q\.weight$=iq3_xxs
246
+ ^blk\.31\.attn_v\.weight$=q6_k
247
+ ^blk\.31\.ffn_down\.weight$=q5_k
248
+ ^blk\.31\.ffn_gate\.weight$=q4_k
249
+ ^blk\.31\.ffn_up\.weight$=q5_k
layouts/UserRepGen-4B-IQ4_NL.layout.json ADDED
@@ -0,0 +1,190 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1978,
17
+ "type": "q5_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q5_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2616821760,
130
+ "map_body+embd": 2688634880,
131
+ "ratio": 1.0274,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 5.052,
136
+ "body_bpw_heuristic": 4.702,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2708209664,
139
+ "file_bpw_pred": 5.151,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "iq4_nl": 23,
143
+ "q8_0": 1
144
+ },
145
+ "attn_k": {
146
+ "q8_0": 8
147
+ },
148
+ "attn_output": {
149
+ "q8_0": 8
150
+ },
151
+ "attn_q": {
152
+ "iq4_nl": 8
153
+ },
154
+ "attn_qkv": {
155
+ "iq4_nl": 23,
156
+ "q6_k": 1
157
+ },
158
+ "attn_v": {
159
+ "q8_0": 8
160
+ },
161
+ "ffn_down": {
162
+ "iq4_nl": 27,
163
+ "q6_k": 4,
164
+ "q8_0": 1
165
+ },
166
+ "ffn_gate": {
167
+ "iq4_nl": 31,
168
+ "q6_k": 1
169
+ },
170
+ "ffn_up": {
171
+ "iq4_nl": 23,
172
+ "q6_k": 8,
173
+ "q8_0": 1
174
+ },
175
+ "ssm_out": {
176
+ "q6_k": 22,
177
+ "q8_0": 2
178
+ }
179
+ },
180
+ "prior": "layout/prior.json",
181
+ "pin_rule": "floor",
182
+ "embd_cap": "heur",
183
+ "generator": "auto_quant_v2.layout",
184
+ "generator_key": "3bf8b43e20a68f0b",
185
+ "quant_type": "IQ4_NL",
186
+ "quantize_ftype": "IQ4_NL",
187
+ "llama_cpp_version": "b11627",
188
+ "unknown_kinds": {},
189
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
190
+ }
layouts/UserRepGen-4B-IQ4_NL.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q5_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q6_k
52
+ ^blk\.0\.ffn_down\.weight$=q6_k
53
+ ^blk\.0\.ffn_gate\.weight$=iq4_nl
54
+ ^blk\.0\.ffn_up\.weight$=q6_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=iq4_nl
57
+ ^blk\.1\.attn_qkv\.weight$=iq4_nl
58
+ ^blk\.1\.ffn_down\.weight$=iq4_nl
59
+ ^blk\.1\.ffn_gate\.weight$=iq4_nl
60
+ ^blk\.1\.ffn_up\.weight$=q6_k
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=iq4_nl
63
+ ^blk\.2\.attn_qkv\.weight$=iq4_nl
64
+ ^blk\.2\.ffn_down\.weight$=iq4_nl
65
+ ^blk\.2\.ffn_gate\.weight$=iq4_nl
66
+ ^blk\.2\.ffn_up\.weight$=q6_k
67
+ ^blk\.2\.ssm_out\.weight$=q6_k
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=iq4_nl
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q6_k
73
+ ^blk\.3\.ffn_gate\.weight$=iq4_nl
74
+ ^blk\.3\.ffn_up\.weight$=q6_k
75
+ ^blk\.4\.attn_gate\.weight$=iq4_nl
76
+ ^blk\.4\.attn_qkv\.weight$=iq4_nl
77
+ ^blk\.4\.ffn_down\.weight$=iq4_nl
78
+ ^blk\.4\.ffn_gate\.weight$=iq4_nl
79
+ ^blk\.4\.ffn_up\.weight$=q6_k
80
+ ^blk\.4\.ssm_out\.weight$=q6_k
81
+ ^blk\.5\.attn_gate\.weight$=iq4_nl
82
+ ^blk\.5\.attn_qkv\.weight$=iq4_nl
83
+ ^blk\.5\.ffn_down\.weight$=iq4_nl
84
+ ^blk\.5\.ffn_gate\.weight$=iq4_nl
85
+ ^blk\.5\.ffn_up\.weight$=q6_k
86
+ ^blk\.5\.ssm_out\.weight$=q6_k
87
+ ^blk\.6\.attn_gate\.weight$=iq4_nl
88
+ ^blk\.6\.attn_qkv\.weight$=iq4_nl
89
+ ^blk\.6\.ffn_down\.weight$=iq4_nl
90
+ ^blk\.6\.ffn_gate\.weight$=iq4_nl
91
+ ^blk\.6\.ffn_up\.weight$=q6_k
92
+ ^blk\.6\.ssm_out\.weight$=q6_k
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=iq4_nl
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q6_k
98
+ ^blk\.7\.ffn_gate\.weight$=iq4_nl
99
+ ^blk\.7\.ffn_up\.weight$=q6_k
100
+ ^blk\.8\.attn_gate\.weight$=iq4_nl
101
+ ^blk\.8\.attn_qkv\.weight$=iq4_nl
102
+ ^blk\.8\.ffn_down\.weight$=iq4_nl
103
+ ^blk\.8\.ffn_gate\.weight$=iq4_nl
104
+ ^blk\.8\.ffn_up\.weight$=iq4_nl
105
+ ^blk\.8\.ssm_out\.weight$=q6_k
106
+ ^blk\.9\.attn_gate\.weight$=iq4_nl
107
+ ^blk\.9\.attn_qkv\.weight$=iq4_nl
108
+ ^blk\.9\.ffn_down\.weight$=iq4_nl
109
+ ^blk\.9\.ffn_gate\.weight$=iq4_nl
110
+ ^blk\.9\.ffn_up\.weight$=iq4_nl
111
+ ^blk\.9\.ssm_out\.weight$=q6_k
112
+ ^blk\.10\.attn_gate\.weight$=iq4_nl
113
+ ^blk\.10\.attn_qkv\.weight$=iq4_nl
114
+ ^blk\.10\.ffn_down\.weight$=iq4_nl
115
+ ^blk\.10\.ffn_gate\.weight$=iq4_nl
116
+ ^blk\.10\.ffn_up\.weight$=iq4_nl
117
+ ^blk\.10\.ssm_out\.weight$=q6_k
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=iq4_nl
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=iq4_nl
123
+ ^blk\.11\.ffn_gate\.weight$=iq4_nl
124
+ ^blk\.11\.ffn_up\.weight$=iq4_nl
125
+ ^blk\.12\.attn_gate\.weight$=iq4_nl
126
+ ^blk\.12\.attn_qkv\.weight$=iq4_nl
127
+ ^blk\.12\.ffn_down\.weight$=iq4_nl
128
+ ^blk\.12\.ffn_gate\.weight$=iq4_nl
129
+ ^blk\.12\.ffn_up\.weight$=iq4_nl
130
+ ^blk\.12\.ssm_out\.weight$=q6_k
131
+ ^blk\.13\.attn_gate\.weight$=iq4_nl
132
+ ^blk\.13\.attn_qkv\.weight$=iq4_nl
133
+ ^blk\.13\.ffn_down\.weight$=iq4_nl
134
+ ^blk\.13\.ffn_gate\.weight$=iq4_nl
135
+ ^blk\.13\.ffn_up\.weight$=iq4_nl
136
+ ^blk\.13\.ssm_out\.weight$=q6_k
137
+ ^blk\.14\.attn_gate\.weight$=iq4_nl
138
+ ^blk\.14\.attn_qkv\.weight$=iq4_nl
139
+ ^blk\.14\.ffn_down\.weight$=iq4_nl
140
+ ^blk\.14\.ffn_gate\.weight$=iq4_nl
141
+ ^blk\.14\.ffn_up\.weight$=iq4_nl
142
+ ^blk\.14\.ssm_out\.weight$=q6_k
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=iq4_nl
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=iq4_nl
148
+ ^blk\.15\.ffn_gate\.weight$=iq4_nl
149
+ ^blk\.15\.ffn_up\.weight$=iq4_nl
150
+ ^blk\.16\.attn_gate\.weight$=iq4_nl
151
+ ^blk\.16\.attn_qkv\.weight$=iq4_nl
152
+ ^blk\.16\.ffn_down\.weight$=iq4_nl
153
+ ^blk\.16\.ffn_gate\.weight$=iq4_nl
154
+ ^blk\.16\.ffn_up\.weight$=iq4_nl
155
+ ^blk\.16\.ssm_out\.weight$=q6_k
156
+ ^blk\.17\.attn_gate\.weight$=iq4_nl
157
+ ^blk\.17\.attn_qkv\.weight$=iq4_nl
158
+ ^blk\.17\.ffn_down\.weight$=iq4_nl
159
+ ^blk\.17\.ffn_gate\.weight$=iq4_nl
160
+ ^blk\.17\.ffn_up\.weight$=iq4_nl
161
+ ^blk\.17\.ssm_out\.weight$=q6_k
162
+ ^blk\.18\.attn_gate\.weight$=iq4_nl
163
+ ^blk\.18\.attn_qkv\.weight$=iq4_nl
164
+ ^blk\.18\.ffn_down\.weight$=iq4_nl
165
+ ^blk\.18\.ffn_gate\.weight$=iq4_nl
166
+ ^blk\.18\.ffn_up\.weight$=iq4_nl
167
+ ^blk\.18\.ssm_out\.weight$=q6_k
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=iq4_nl
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=iq4_nl
173
+ ^blk\.19\.ffn_gate\.weight$=iq4_nl
174
+ ^blk\.19\.ffn_up\.weight$=iq4_nl
175
+ ^blk\.20\.attn_gate\.weight$=iq4_nl
176
+ ^blk\.20\.attn_qkv\.weight$=iq4_nl
177
+ ^blk\.20\.ffn_down\.weight$=iq4_nl
178
+ ^blk\.20\.ffn_gate\.weight$=iq4_nl
179
+ ^blk\.20\.ffn_up\.weight$=iq4_nl
180
+ ^blk\.20\.ssm_out\.weight$=q6_k
181
+ ^blk\.21\.attn_gate\.weight$=iq4_nl
182
+ ^blk\.21\.attn_qkv\.weight$=iq4_nl
183
+ ^blk\.21\.ffn_down\.weight$=iq4_nl
184
+ ^blk\.21\.ffn_gate\.weight$=iq4_nl
185
+ ^blk\.21\.ffn_up\.weight$=iq4_nl
186
+ ^blk\.21\.ssm_out\.weight$=q6_k
187
+ ^blk\.22\.attn_gate\.weight$=iq4_nl
188
+ ^blk\.22\.attn_qkv\.weight$=iq4_nl
189
+ ^blk\.22\.ffn_down\.weight$=iq4_nl
190
+ ^blk\.22\.ffn_gate\.weight$=iq4_nl
191
+ ^blk\.22\.ffn_up\.weight$=iq4_nl
192
+ ^blk\.22\.ssm_out\.weight$=q6_k
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=iq4_nl
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=iq4_nl
198
+ ^blk\.23\.ffn_gate\.weight$=iq4_nl
199
+ ^blk\.23\.ffn_up\.weight$=iq4_nl
200
+ ^blk\.24\.attn_gate\.weight$=iq4_nl
201
+ ^blk\.24\.attn_qkv\.weight$=iq4_nl
202
+ ^blk\.24\.ffn_down\.weight$=iq4_nl
203
+ ^blk\.24\.ffn_gate\.weight$=iq4_nl
204
+ ^blk\.24\.ffn_up\.weight$=iq4_nl
205
+ ^blk\.24\.ssm_out\.weight$=q6_k
206
+ ^blk\.25\.attn_gate\.weight$=iq4_nl
207
+ ^blk\.25\.attn_qkv\.weight$=iq4_nl
208
+ ^blk\.25\.ffn_down\.weight$=iq4_nl
209
+ ^blk\.25\.ffn_gate\.weight$=iq4_nl
210
+ ^blk\.25\.ffn_up\.weight$=iq4_nl
211
+ ^blk\.25\.ssm_out\.weight$=q6_k
212
+ ^blk\.26\.attn_gate\.weight$=iq4_nl
213
+ ^blk\.26\.attn_qkv\.weight$=iq4_nl
214
+ ^blk\.26\.ffn_down\.weight$=iq4_nl
215
+ ^blk\.26\.ffn_gate\.weight$=iq4_nl
216
+ ^blk\.26\.ffn_up\.weight$=iq4_nl
217
+ ^blk\.26\.ssm_out\.weight$=q6_k
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=iq4_nl
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q6_k
223
+ ^blk\.27\.ffn_gate\.weight$=iq4_nl
224
+ ^blk\.27\.ffn_up\.weight$=iq4_nl
225
+ ^blk\.28\.attn_gate\.weight$=iq4_nl
226
+ ^blk\.28\.attn_qkv\.weight$=iq4_nl
227
+ ^blk\.28\.ffn_down\.weight$=iq4_nl
228
+ ^blk\.28\.ffn_gate\.weight$=iq4_nl
229
+ ^blk\.28\.ffn_up\.weight$=iq4_nl
230
+ ^blk\.28\.ssm_out\.weight$=q6_k
231
+ ^blk\.29\.attn_gate\.weight$=iq4_nl
232
+ ^blk\.29\.attn_qkv\.weight$=iq4_nl
233
+ ^blk\.29\.ffn_down\.weight$=iq4_nl
234
+ ^blk\.29\.ffn_gate\.weight$=iq4_nl
235
+ ^blk\.29\.ffn_up\.weight$=iq4_nl
236
+ ^blk\.29\.ssm_out\.weight$=q6_k
237
+ ^blk\.30\.attn_gate\.weight$=iq4_nl
238
+ ^blk\.30\.attn_qkv\.weight$=iq4_nl
239
+ ^blk\.30\.ffn_down\.weight$=iq4_nl
240
+ ^blk\.30\.ffn_gate\.weight$=iq4_nl
241
+ ^blk\.30\.ffn_up\.weight$=iq4_nl
242
+ ^blk\.30\.ssm_out\.weight$=q6_k
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=iq4_nl
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q6_k
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-IQ4_XS.layout.json ADDED
@@ -0,0 +1,190 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-imatrix.gguf",
4
+ "ftype": "IQ4_XS",
5
+ "base_type": "iq4_xs",
6
+ "family": "iq",
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
+ "token_embd.weight": {
16
+ "share": 0.2059,
17
+ "type": "q5_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q5_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2512619520,
130
+ "map_body+embd": 2408140800,
131
+ "ratio": 0.9584,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.9,
135
+ "body_bpw": 4.423,
136
+ "body_bpw_heuristic": 4.468,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2427715584,
139
+ "file_bpw_pred": 4.618,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "iq4_xs": 23,
143
+ "q8_0": 1
144
+ },
145
+ "attn_k": {
146
+ "q6_k": 7,
147
+ "q8_0": 1
148
+ },
149
+ "attn_output": {
150
+ "q6_k": 7,
151
+ "q8_0": 1
152
+ },
153
+ "attn_q": {
154
+ "iq4_xs": 8
155
+ },
156
+ "attn_qkv": {
157
+ "iq4_xs": 23,
158
+ "q6_k": 1
159
+ },
160
+ "attn_v": {
161
+ "q8_0": 8
162
+ },
163
+ "ffn_down": {
164
+ "iq4_xs": 31,
165
+ "q6_k": 1
166
+ },
167
+ "ffn_gate": {
168
+ "iq4_xs": 32
169
+ },
170
+ "ffn_up": {
171
+ "iq4_xs": 31,
172
+ "q6_k": 1
173
+ },
174
+ "ssm_out": {
175
+ "iq4_xs": 22,
176
+ "q8_0": 1,
177
+ "q5_k": 1
178
+ }
179
+ },
180
+ "prior": "layout/prior.json",
181
+ "pin_rule": "floor",
182
+ "embd_cap": "heur",
183
+ "generator": "auto_quant_v2.layout",
184
+ "generator_key": "3bf8b43e20a68f0b",
185
+ "quant_type": "IQ4_XS",
186
+ "quantize_ftype": "IQ4_XS",
187
+ "llama_cpp_version": "b11627",
188
+ "unknown_kinds": {},
189
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
190
+ }
layouts/UserRepGen-4B-IQ4_XS.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q5_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q6_k
52
+ ^blk\.0\.ffn_down\.weight$=iq4_xs
53
+ ^blk\.0\.ffn_gate\.weight$=iq4_xs
54
+ ^blk\.0\.ffn_up\.weight$=iq4_xs
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=iq4_xs
57
+ ^blk\.1\.attn_qkv\.weight$=iq4_xs
58
+ ^blk\.1\.ffn_down\.weight$=iq4_xs
59
+ ^blk\.1\.ffn_gate\.weight$=iq4_xs
60
+ ^blk\.1\.ffn_up\.weight$=iq4_xs
61
+ ^blk\.1\.ssm_out\.weight$=q5_k
62
+ ^blk\.2\.attn_gate\.weight$=iq4_xs
63
+ ^blk\.2\.attn_qkv\.weight$=iq4_xs
64
+ ^blk\.2\.ffn_down\.weight$=iq4_xs
65
+ ^blk\.2\.ffn_gate\.weight$=iq4_xs
66
+ ^blk\.2\.ffn_up\.weight$=iq4_xs
67
+ ^blk\.2\.ssm_out\.weight$=iq4_xs
68
+ ^blk\.3\.attn_k\.weight$=q6_k
69
+ ^blk\.3\.attn_output\.weight$=q6_k
70
+ ^blk\.3\.attn_q\.weight$=iq4_xs
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=iq4_xs
73
+ ^blk\.3\.ffn_gate\.weight$=iq4_xs
74
+ ^blk\.3\.ffn_up\.weight$=iq4_xs
75
+ ^blk\.4\.attn_gate\.weight$=iq4_xs
76
+ ^blk\.4\.attn_qkv\.weight$=iq4_xs
77
+ ^blk\.4\.ffn_down\.weight$=iq4_xs
78
+ ^blk\.4\.ffn_gate\.weight$=iq4_xs
79
+ ^blk\.4\.ffn_up\.weight$=iq4_xs
80
+ ^blk\.4\.ssm_out\.weight$=iq4_xs
81
+ ^blk\.5\.attn_gate\.weight$=iq4_xs
82
+ ^blk\.5\.attn_qkv\.weight$=iq4_xs
83
+ ^blk\.5\.ffn_down\.weight$=iq4_xs
84
+ ^blk\.5\.ffn_gate\.weight$=iq4_xs
85
+ ^blk\.5\.ffn_up\.weight$=iq4_xs
86
+ ^blk\.5\.ssm_out\.weight$=iq4_xs
87
+ ^blk\.6\.attn_gate\.weight$=iq4_xs
88
+ ^blk\.6\.attn_qkv\.weight$=iq4_xs
89
+ ^blk\.6\.ffn_down\.weight$=iq4_xs
90
+ ^blk\.6\.ffn_gate\.weight$=iq4_xs
91
+ ^blk\.6\.ffn_up\.weight$=iq4_xs
92
+ ^blk\.6\.ssm_out\.weight$=iq4_xs
93
+ ^blk\.7\.attn_k\.weight$=q6_k
94
+ ^blk\.7\.attn_output\.weight$=q6_k
95
+ ^blk\.7\.attn_q\.weight$=iq4_xs
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=iq4_xs
98
+ ^blk\.7\.ffn_gate\.weight$=iq4_xs
99
+ ^blk\.7\.ffn_up\.weight$=iq4_xs
100
+ ^blk\.8\.attn_gate\.weight$=iq4_xs
101
+ ^blk\.8\.attn_qkv\.weight$=iq4_xs
102
+ ^blk\.8\.ffn_down\.weight$=iq4_xs
103
+ ^blk\.8\.ffn_gate\.weight$=iq4_xs
104
+ ^blk\.8\.ffn_up\.weight$=iq4_xs
105
+ ^blk\.8\.ssm_out\.weight$=iq4_xs
106
+ ^blk\.9\.attn_gate\.weight$=iq4_xs
107
+ ^blk\.9\.attn_qkv\.weight$=iq4_xs
108
+ ^blk\.9\.ffn_down\.weight$=iq4_xs
109
+ ^blk\.9\.ffn_gate\.weight$=iq4_xs
110
+ ^blk\.9\.ffn_up\.weight$=iq4_xs
111
+ ^blk\.9\.ssm_out\.weight$=iq4_xs
112
+ ^blk\.10\.attn_gate\.weight$=iq4_xs
113
+ ^blk\.10\.attn_qkv\.weight$=iq4_xs
114
+ ^blk\.10\.ffn_down\.weight$=iq4_xs
115
+ ^blk\.10\.ffn_gate\.weight$=iq4_xs
116
+ ^blk\.10\.ffn_up\.weight$=iq4_xs
117
+ ^blk\.10\.ssm_out\.weight$=iq4_xs
118
+ ^blk\.11\.attn_k\.weight$=q6_k
119
+ ^blk\.11\.attn_output\.weight$=q6_k
120
+ ^blk\.11\.attn_q\.weight$=iq4_xs
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=iq4_xs
123
+ ^blk\.11\.ffn_gate\.weight$=iq4_xs
124
+ ^blk\.11\.ffn_up\.weight$=iq4_xs
125
+ ^blk\.12\.attn_gate\.weight$=iq4_xs
126
+ ^blk\.12\.attn_qkv\.weight$=iq4_xs
127
+ ^blk\.12\.ffn_down\.weight$=iq4_xs
128
+ ^blk\.12\.ffn_gate\.weight$=iq4_xs
129
+ ^blk\.12\.ffn_up\.weight$=iq4_xs
130
+ ^blk\.12\.ssm_out\.weight$=iq4_xs
131
+ ^blk\.13\.attn_gate\.weight$=iq4_xs
132
+ ^blk\.13\.attn_qkv\.weight$=iq4_xs
133
+ ^blk\.13\.ffn_down\.weight$=iq4_xs
134
+ ^blk\.13\.ffn_gate\.weight$=iq4_xs
135
+ ^blk\.13\.ffn_up\.weight$=iq4_xs
136
+ ^blk\.13\.ssm_out\.weight$=iq4_xs
137
+ ^blk\.14\.attn_gate\.weight$=iq4_xs
138
+ ^blk\.14\.attn_qkv\.weight$=iq4_xs
139
+ ^blk\.14\.ffn_down\.weight$=iq4_xs
140
+ ^blk\.14\.ffn_gate\.weight$=iq4_xs
141
+ ^blk\.14\.ffn_up\.weight$=iq4_xs
142
+ ^blk\.14\.ssm_out\.weight$=iq4_xs
143
+ ^blk\.15\.attn_k\.weight$=q6_k
144
+ ^blk\.15\.attn_output\.weight$=q6_k
145
+ ^blk\.15\.attn_q\.weight$=iq4_xs
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=iq4_xs
148
+ ^blk\.15\.ffn_gate\.weight$=iq4_xs
149
+ ^blk\.15\.ffn_up\.weight$=iq4_xs
150
+ ^blk\.16\.attn_gate\.weight$=iq4_xs
151
+ ^blk\.16\.attn_qkv\.weight$=iq4_xs
152
+ ^blk\.16\.ffn_down\.weight$=iq4_xs
153
+ ^blk\.16\.ffn_gate\.weight$=iq4_xs
154
+ ^blk\.16\.ffn_up\.weight$=iq4_xs
155
+ ^blk\.16\.ssm_out\.weight$=iq4_xs
156
+ ^blk\.17\.attn_gate\.weight$=iq4_xs
157
+ ^blk\.17\.attn_qkv\.weight$=iq4_xs
158
+ ^blk\.17\.ffn_down\.weight$=iq4_xs
159
+ ^blk\.17\.ffn_gate\.weight$=iq4_xs
160
+ ^blk\.17\.ffn_up\.weight$=iq4_xs
161
+ ^blk\.17\.ssm_out\.weight$=iq4_xs
162
+ ^blk\.18\.attn_gate\.weight$=iq4_xs
163
+ ^blk\.18\.attn_qkv\.weight$=iq4_xs
164
+ ^blk\.18\.ffn_down\.weight$=iq4_xs
165
+ ^blk\.18\.ffn_gate\.weight$=iq4_xs
166
+ ^blk\.18\.ffn_up\.weight$=iq4_xs
167
+ ^blk\.18\.ssm_out\.weight$=iq4_xs
168
+ ^blk\.19\.attn_k\.weight$=q6_k
169
+ ^blk\.19\.attn_output\.weight$=q6_k
170
+ ^blk\.19\.attn_q\.weight$=iq4_xs
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=iq4_xs
173
+ ^blk\.19\.ffn_gate\.weight$=iq4_xs
174
+ ^blk\.19\.ffn_up\.weight$=iq4_xs
175
+ ^blk\.20\.attn_gate\.weight$=iq4_xs
176
+ ^blk\.20\.attn_qkv\.weight$=iq4_xs
177
+ ^blk\.20\.ffn_down\.weight$=iq4_xs
178
+ ^blk\.20\.ffn_gate\.weight$=iq4_xs
179
+ ^blk\.20\.ffn_up\.weight$=iq4_xs
180
+ ^blk\.20\.ssm_out\.weight$=iq4_xs
181
+ ^blk\.21\.attn_gate\.weight$=iq4_xs
182
+ ^blk\.21\.attn_qkv\.weight$=iq4_xs
183
+ ^blk\.21\.ffn_down\.weight$=iq4_xs
184
+ ^blk\.21\.ffn_gate\.weight$=iq4_xs
185
+ ^blk\.21\.ffn_up\.weight$=iq4_xs
186
+ ^blk\.21\.ssm_out\.weight$=iq4_xs
187
+ ^blk\.22\.attn_gate\.weight$=iq4_xs
188
+ ^blk\.22\.attn_qkv\.weight$=iq4_xs
189
+ ^blk\.22\.ffn_down\.weight$=iq4_xs
190
+ ^blk\.22\.ffn_gate\.weight$=iq4_xs
191
+ ^blk\.22\.ffn_up\.weight$=iq4_xs
192
+ ^blk\.22\.ssm_out\.weight$=iq4_xs
193
+ ^blk\.23\.attn_k\.weight$=q6_k
194
+ ^blk\.23\.attn_output\.weight$=q6_k
195
+ ^blk\.23\.attn_q\.weight$=iq4_xs
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=iq4_xs
198
+ ^blk\.23\.ffn_gate\.weight$=iq4_xs
199
+ ^blk\.23\.ffn_up\.weight$=iq4_xs
200
+ ^blk\.24\.attn_gate\.weight$=iq4_xs
201
+ ^blk\.24\.attn_qkv\.weight$=iq4_xs
202
+ ^blk\.24\.ffn_down\.weight$=iq4_xs
203
+ ^blk\.24\.ffn_gate\.weight$=iq4_xs
204
+ ^blk\.24\.ffn_up\.weight$=iq4_xs
205
+ ^blk\.24\.ssm_out\.weight$=iq4_xs
206
+ ^blk\.25\.attn_gate\.weight$=iq4_xs
207
+ ^blk\.25\.attn_qkv\.weight$=iq4_xs
208
+ ^blk\.25\.ffn_down\.weight$=iq4_xs
209
+ ^blk\.25\.ffn_gate\.weight$=iq4_xs
210
+ ^blk\.25\.ffn_up\.weight$=iq4_xs
211
+ ^blk\.25\.ssm_out\.weight$=iq4_xs
212
+ ^blk\.26\.attn_gate\.weight$=iq4_xs
213
+ ^blk\.26\.attn_qkv\.weight$=iq4_xs
214
+ ^blk\.26\.ffn_down\.weight$=iq4_xs
215
+ ^blk\.26\.ffn_gate\.weight$=iq4_xs
216
+ ^blk\.26\.ffn_up\.weight$=iq4_xs
217
+ ^blk\.26\.ssm_out\.weight$=iq4_xs
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q6_k
220
+ ^blk\.27\.attn_q\.weight$=iq4_xs
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=iq4_xs
223
+ ^blk\.27\.ffn_gate\.weight$=iq4_xs
224
+ ^blk\.27\.ffn_up\.weight$=iq4_xs
225
+ ^blk\.28\.attn_gate\.weight$=iq4_xs
226
+ ^blk\.28\.attn_qkv\.weight$=iq4_xs
227
+ ^blk\.28\.ffn_down\.weight$=iq4_xs
228
+ ^blk\.28\.ffn_gate\.weight$=iq4_xs
229
+ ^blk\.28\.ffn_up\.weight$=iq4_xs
230
+ ^blk\.28\.ssm_out\.weight$=iq4_xs
231
+ ^blk\.29\.attn_gate\.weight$=iq4_xs
232
+ ^blk\.29\.attn_qkv\.weight$=iq4_xs
233
+ ^blk\.29\.ffn_down\.weight$=iq4_xs
234
+ ^blk\.29\.ffn_gate\.weight$=iq4_xs
235
+ ^blk\.29\.ffn_up\.weight$=iq4_xs
236
+ ^blk\.29\.ssm_out\.weight$=iq4_xs
237
+ ^blk\.30\.attn_gate\.weight$=iq4_xs
238
+ ^blk\.30\.attn_qkv\.weight$=iq4_xs
239
+ ^blk\.30\.ffn_down\.weight$=iq4_xs
240
+ ^blk\.30\.ffn_gate\.weight$=iq4_xs
241
+ ^blk\.30\.ffn_up\.weight$=iq4_xs
242
+ ^blk\.30\.ssm_out\.weight$=iq4_xs
243
+ ^blk\.31\.attn_k\.weight$=q6_k
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=iq4_xs
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q6_k
248
+ ^blk\.31\.ffn_gate\.weight$=iq4_xs
249
+ ^blk\.31\.ffn_up\.weight$=q6_k
layouts/UserRepGen-4B-Q2_K.layout.json ADDED
@@ -0,0 +1,191 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.2541,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2033059840,
130
+ "map_body+embd": 1699266560,
131
+ "ratio": 0.8358,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 3.011,
136
+ "body_bpw_heuristic": 3.392,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 1718841344,
139
+ "file_bpw_pred": 3.27,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q2_k": 22,
143
+ "q5_k": 1,
144
+ "q4_k": 1
145
+ },
146
+ "attn_k": {
147
+ "q4_k": 6,
148
+ "q5_k": 2
149
+ },
150
+ "attn_output": {
151
+ "q4_k": 7,
152
+ "q5_k": 1
153
+ },
154
+ "attn_q": {
155
+ "q2_k": 8
156
+ },
157
+ "attn_qkv": {
158
+ "q2_k": 23,
159
+ "q4_k": 1
160
+ },
161
+ "attn_v": {
162
+ "q5_k": 8
163
+ },
164
+ "ffn_down": {
165
+ "q2_k": 30,
166
+ "q4_k": 2
167
+ },
168
+ "ffn_gate": {
169
+ "q2_k": 31,
170
+ "q4_k": 1
171
+ },
172
+ "ffn_up": {
173
+ "q2_k": 23,
174
+ "q4_k": 9
175
+ },
176
+ "ssm_out": {
177
+ "q4_k": 23,
178
+ "q5_k": 1
179
+ }
180
+ },
181
+ "prior": "layout/prior.json",
182
+ "pin_rule": "floor",
183
+ "embd_cap": "heur",
184
+ "generator": "auto_quant_v2.layout",
185
+ "generator_key": "3bf8b43e20a68f0b",
186
+ "quant_type": "Q2_K",
187
+ "quantize_ftype": "Q2_K",
188
+ "llama_cpp_version": "b11627",
189
+ "unknown_kinds": {},
190
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
191
+ }
layouts/UserRepGen-4B-Q2_K.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q5_k
51
+ ^blk\.0\.attn_qkv\.weight$=q4_k
52
+ ^blk\.0\.ffn_down\.weight$=q4_k
53
+ ^blk\.0\.ffn_gate\.weight$=q2_k
54
+ ^blk\.0\.ffn_up\.weight$=q4_k
55
+ ^blk\.0\.ssm_out\.weight$=q5_k
56
+ ^blk\.1\.attn_gate\.weight$=q4_k
57
+ ^blk\.1\.attn_qkv\.weight$=q2_k
58
+ ^blk\.1\.ffn_down\.weight$=q2_k
59
+ ^blk\.1\.ffn_gate\.weight$=q2_k
60
+ ^blk\.1\.ffn_up\.weight$=q4_k
61
+ ^blk\.1\.ssm_out\.weight$=q4_k
62
+ ^blk\.2\.attn_gate\.weight$=q2_k
63
+ ^blk\.2\.attn_qkv\.weight$=q2_k
64
+ ^blk\.2\.ffn_down\.weight$=q2_k
65
+ ^blk\.2\.ffn_gate\.weight$=q2_k
66
+ ^blk\.2\.ffn_up\.weight$=q4_k
67
+ ^blk\.2\.ssm_out\.weight$=q4_k
68
+ ^blk\.3\.attn_k\.weight$=q4_k
69
+ ^blk\.3\.attn_output\.weight$=q4_k
70
+ ^blk\.3\.attn_q\.weight$=q2_k
71
+ ^blk\.3\.attn_v\.weight$=q5_k
72
+ ^blk\.3\.ffn_down\.weight$=q2_k
73
+ ^blk\.3\.ffn_gate\.weight$=q2_k
74
+ ^blk\.3\.ffn_up\.weight$=q4_k
75
+ ^blk\.4\.attn_gate\.weight$=q2_k
76
+ ^blk\.4\.attn_qkv\.weight$=q2_k
77
+ ^blk\.4\.ffn_down\.weight$=q2_k
78
+ ^blk\.4\.ffn_gate\.weight$=q2_k
79
+ ^blk\.4\.ffn_up\.weight$=q4_k
80
+ ^blk\.4\.ssm_out\.weight$=q4_k
81
+ ^blk\.5\.attn_gate\.weight$=q2_k
82
+ ^blk\.5\.attn_qkv\.weight$=q2_k
83
+ ^blk\.5\.ffn_down\.weight$=q2_k
84
+ ^blk\.5\.ffn_gate\.weight$=q2_k
85
+ ^blk\.5\.ffn_up\.weight$=q4_k
86
+ ^blk\.5\.ssm_out\.weight$=q4_k
87
+ ^blk\.6\.attn_gate\.weight$=q2_k
88
+ ^blk\.6\.attn_qkv\.weight$=q2_k
89
+ ^blk\.6\.ffn_down\.weight$=q2_k
90
+ ^blk\.6\.ffn_gate\.weight$=q2_k
91
+ ^blk\.6\.ffn_up\.weight$=q4_k
92
+ ^blk\.6\.ssm_out\.weight$=q4_k
93
+ ^blk\.7\.attn_k\.weight$=q4_k
94
+ ^blk\.7\.attn_output\.weight$=q4_k
95
+ ^blk\.7\.attn_q\.weight$=q2_k
96
+ ^blk\.7\.attn_v\.weight$=q5_k
97
+ ^blk\.7\.ffn_down\.weight$=q2_k
98
+ ^blk\.7\.ffn_gate\.weight$=q2_k
99
+ ^blk\.7\.ffn_up\.weight$=q4_k
100
+ ^blk\.8\.attn_gate\.weight$=q2_k
101
+ ^blk\.8\.attn_qkv\.weight$=q2_k
102
+ ^blk\.8\.ffn_down\.weight$=q2_k
103
+ ^blk\.8\.ffn_gate\.weight$=q2_k
104
+ ^blk\.8\.ffn_up\.weight$=q2_k
105
+ ^blk\.8\.ssm_out\.weight$=q4_k
106
+ ^blk\.9\.attn_gate\.weight$=q2_k
107
+ ^blk\.9\.attn_qkv\.weight$=q2_k
108
+ ^blk\.9\.ffn_down\.weight$=q2_k
109
+ ^blk\.9\.ffn_gate\.weight$=q2_k
110
+ ^blk\.9\.ffn_up\.weight$=q2_k
111
+ ^blk\.9\.ssm_out\.weight$=q4_k
112
+ ^blk\.10\.attn_gate\.weight$=q2_k
113
+ ^blk\.10\.attn_qkv\.weight$=q2_k
114
+ ^blk\.10\.ffn_down\.weight$=q2_k
115
+ ^blk\.10\.ffn_gate\.weight$=q2_k
116
+ ^blk\.10\.ffn_up\.weight$=q2_k
117
+ ^blk\.10\.ssm_out\.weight$=q4_k
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q4_k
120
+ ^blk\.11\.attn_q\.weight$=q2_k
121
+ ^blk\.11\.attn_v\.weight$=q5_k
122
+ ^blk\.11\.ffn_down\.weight$=q2_k
123
+ ^blk\.11\.ffn_gate\.weight$=q2_k
124
+ ^blk\.11\.ffn_up\.weight$=q2_k
125
+ ^blk\.12\.attn_gate\.weight$=q2_k
126
+ ^blk\.12\.attn_qkv\.weight$=q2_k
127
+ ^blk\.12\.ffn_down\.weight$=q2_k
128
+ ^blk\.12\.ffn_gate\.weight$=q2_k
129
+ ^blk\.12\.ffn_up\.weight$=q2_k
130
+ ^blk\.12\.ssm_out\.weight$=q4_k
131
+ ^blk\.13\.attn_gate\.weight$=q2_k
132
+ ^blk\.13\.attn_qkv\.weight$=q2_k
133
+ ^blk\.13\.ffn_down\.weight$=q2_k
134
+ ^blk\.13\.ffn_gate\.weight$=q2_k
135
+ ^blk\.13\.ffn_up\.weight$=q2_k
136
+ ^blk\.13\.ssm_out\.weight$=q4_k
137
+ ^blk\.14\.attn_gate\.weight$=q2_k
138
+ ^blk\.14\.attn_qkv\.weight$=q2_k
139
+ ^blk\.14\.ffn_down\.weight$=q2_k
140
+ ^blk\.14\.ffn_gate\.weight$=q2_k
141
+ ^blk\.14\.ffn_up\.weight$=q2_k
142
+ ^blk\.14\.ssm_out\.weight$=q4_k
143
+ ^blk\.15\.attn_k\.weight$=q4_k
144
+ ^blk\.15\.attn_output\.weight$=q4_k
145
+ ^blk\.15\.attn_q\.weight$=q2_k
146
+ ^blk\.15\.attn_v\.weight$=q5_k
147
+ ^blk\.15\.ffn_down\.weight$=q2_k
148
+ ^blk\.15\.ffn_gate\.weight$=q2_k
149
+ ^blk\.15\.ffn_up\.weight$=q2_k
150
+ ^blk\.16\.attn_gate\.weight$=q2_k
151
+ ^blk\.16\.attn_qkv\.weight$=q2_k
152
+ ^blk\.16\.ffn_down\.weight$=q2_k
153
+ ^blk\.16\.ffn_gate\.weight$=q2_k
154
+ ^blk\.16\.ffn_up\.weight$=q2_k
155
+ ^blk\.16\.ssm_out\.weight$=q4_k
156
+ ^blk\.17\.attn_gate\.weight$=q2_k
157
+ ^blk\.17\.attn_qkv\.weight$=q2_k
158
+ ^blk\.17\.ffn_down\.weight$=q2_k
159
+ ^blk\.17\.ffn_gate\.weight$=q2_k
160
+ ^blk\.17\.ffn_up\.weight$=q2_k
161
+ ^blk\.17\.ssm_out\.weight$=q4_k
162
+ ^blk\.18\.attn_gate\.weight$=q2_k
163
+ ^blk\.18\.attn_qkv\.weight$=q2_k
164
+ ^blk\.18\.ffn_down\.weight$=q2_k
165
+ ^blk\.18\.ffn_gate\.weight$=q2_k
166
+ ^blk\.18\.ffn_up\.weight$=q2_k
167
+ ^blk\.18\.ssm_out\.weight$=q4_k
168
+ ^blk\.19\.attn_k\.weight$=q4_k
169
+ ^blk\.19\.attn_output\.weight$=q4_k
170
+ ^blk\.19\.attn_q\.weight$=q2_k
171
+ ^blk\.19\.attn_v\.weight$=q5_k
172
+ ^blk\.19\.ffn_down\.weight$=q2_k
173
+ ^blk\.19\.ffn_gate\.weight$=q2_k
174
+ ^blk\.19\.ffn_up\.weight$=q2_k
175
+ ^blk\.20\.attn_gate\.weight$=q2_k
176
+ ^blk\.20\.attn_qkv\.weight$=q2_k
177
+ ^blk\.20\.ffn_down\.weight$=q2_k
178
+ ^blk\.20\.ffn_gate\.weight$=q2_k
179
+ ^blk\.20\.ffn_up\.weight$=q2_k
180
+ ^blk\.20\.ssm_out\.weight$=q4_k
181
+ ^blk\.21\.attn_gate\.weight$=q2_k
182
+ ^blk\.21\.attn_qkv\.weight$=q2_k
183
+ ^blk\.21\.ffn_down\.weight$=q2_k
184
+ ^blk\.21\.ffn_gate\.weight$=q2_k
185
+ ^blk\.21\.ffn_up\.weight$=q2_k
186
+ ^blk\.21\.ssm_out\.weight$=q4_k
187
+ ^blk\.22\.attn_gate\.weight$=q2_k
188
+ ^blk\.22\.attn_qkv\.weight$=q2_k
189
+ ^blk\.22\.ffn_down\.weight$=q2_k
190
+ ^blk\.22\.ffn_gate\.weight$=q2_k
191
+ ^blk\.22\.ffn_up\.weight$=q2_k
192
+ ^blk\.22\.ssm_out\.weight$=q4_k
193
+ ^blk\.23\.attn_k\.weight$=q4_k
194
+ ^blk\.23\.attn_output\.weight$=q4_k
195
+ ^blk\.23\.attn_q\.weight$=q2_k
196
+ ^blk\.23\.attn_v\.weight$=q5_k
197
+ ^blk\.23\.ffn_down\.weight$=q2_k
198
+ ^blk\.23\.ffn_gate\.weight$=q2_k
199
+ ^blk\.23\.ffn_up\.weight$=q2_k
200
+ ^blk\.24\.attn_gate\.weight$=q2_k
201
+ ^blk\.24\.attn_qkv\.weight$=q2_k
202
+ ^blk\.24\.ffn_down\.weight$=q2_k
203
+ ^blk\.24\.ffn_gate\.weight$=q2_k
204
+ ^blk\.24\.ffn_up\.weight$=q2_k
205
+ ^blk\.24\.ssm_out\.weight$=q4_k
206
+ ^blk\.25\.attn_gate\.weight$=q2_k
207
+ ^blk\.25\.attn_qkv\.weight$=q2_k
208
+ ^blk\.25\.ffn_down\.weight$=q2_k
209
+ ^blk\.25\.ffn_gate\.weight$=q2_k
210
+ ^blk\.25\.ffn_up\.weight$=q2_k
211
+ ^blk\.25\.ssm_out\.weight$=q4_k
212
+ ^blk\.26\.attn_gate\.weight$=q2_k
213
+ ^blk\.26\.attn_qkv\.weight$=q2_k
214
+ ^blk\.26\.ffn_down\.weight$=q2_k
215
+ ^blk\.26\.ffn_gate\.weight$=q2_k
216
+ ^blk\.26\.ffn_up\.weight$=q2_k
217
+ ^blk\.26\.ssm_out\.weight$=q4_k
218
+ ^blk\.27\.attn_k\.weight$=q5_k
219
+ ^blk\.27\.attn_output\.weight$=q4_k
220
+ ^blk\.27\.attn_q\.weight$=q2_k
221
+ ^blk\.27\.attn_v\.weight$=q5_k
222
+ ^blk\.27\.ffn_down\.weight$=q2_k
223
+ ^blk\.27\.ffn_gate\.weight$=q2_k
224
+ ^blk\.27\.ffn_up\.weight$=q2_k
225
+ ^blk\.28\.attn_gate\.weight$=q2_k
226
+ ^blk\.28\.attn_qkv\.weight$=q2_k
227
+ ^blk\.28\.ffn_down\.weight$=q2_k
228
+ ^blk\.28\.ffn_gate\.weight$=q2_k
229
+ ^blk\.28\.ffn_up\.weight$=q2_k
230
+ ^blk\.28\.ssm_out\.weight$=q4_k
231
+ ^blk\.29\.attn_gate\.weight$=q2_k
232
+ ^blk\.29\.attn_qkv\.weight$=q2_k
233
+ ^blk\.29\.ffn_down\.weight$=q2_k
234
+ ^blk\.29\.ffn_gate\.weight$=q2_k
235
+ ^blk\.29\.ffn_up\.weight$=q2_k
236
+ ^blk\.29\.ssm_out\.weight$=q4_k
237
+ ^blk\.30\.attn_gate\.weight$=q2_k
238
+ ^blk\.30\.attn_qkv\.weight$=q2_k
239
+ ^blk\.30\.ffn_down\.weight$=q2_k
240
+ ^blk\.30\.ffn_gate\.weight$=q2_k
241
+ ^blk\.30\.ffn_up\.weight$=q2_k
242
+ ^blk\.30\.ssm_out\.weight$=q4_k
243
+ ^blk\.31\.attn_k\.weight$=q4_k
244
+ ^blk\.31\.attn_output\.weight$=q5_k
245
+ ^blk\.31\.attn_q\.weight$=q2_k
246
+ ^blk\.31\.attn_v\.weight$=q5_k
247
+ ^blk\.31\.ffn_down\.weight$=q4_k
248
+ ^blk\.31\.ffn_gate\.weight$=q4_k
249
+ ^blk\.31\.ffn_up\.weight$=q4_k
layouts/UserRepGen-4B-Q3_K_L.layout.json ADDED
@@ -0,0 +1,196 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.2064,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2506721280,
130
+ "map_body+embd": 2146222080,
131
+ "ratio": 0.8562,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.5,
135
+ "body_bpw": 4.014,
136
+ "body_bpw_heuristic": 4.455,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2165796864,
139
+ "file_bpw_pred": 4.12,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q3_k": 17,
143
+ "q4_k": 6,
144
+ "q6_k": 1
145
+ },
146
+ "attn_k": {
147
+ "q5_k": 8
148
+ },
149
+ "attn_output": {
150
+ "q5_k": 7,
151
+ "q6_k": 1
152
+ },
153
+ "attn_q": {
154
+ "q3_k": 7,
155
+ "q4_k": 1
156
+ },
157
+ "attn_qkv": {
158
+ "q3_k": 23,
159
+ "q5_k": 1
160
+ },
161
+ "attn_v": {
162
+ "q6_k": 6,
163
+ "q5_k": 2
164
+ },
165
+ "ffn_down": {
166
+ "q4_k": 23,
167
+ "q3_k": 8,
168
+ "q5_k": 1
169
+ },
170
+ "ffn_gate": {
171
+ "q3_k": 30,
172
+ "q4_k": 1,
173
+ "q5_k": 1
174
+ },
175
+ "ffn_up": {
176
+ "q3_k": 16,
177
+ "q4_k": 15,
178
+ "q5_k": 1
179
+ },
180
+ "ssm_out": {
181
+ "q5_k": 18,
182
+ "q4_k": 5,
183
+ "q6_k": 1
184
+ }
185
+ },
186
+ "prior": "layout/prior.json",
187
+ "pin_rule": "floor",
188
+ "embd_cap": "heur",
189
+ "generator": "auto_quant_v2.layout",
190
+ "generator_key": "3bf8b43e20a68f0b",
191
+ "quant_type": "Q3_K_L",
192
+ "quantize_ftype": "Q3_K_L",
193
+ "llama_cpp_version": "b11627",
194
+ "unknown_kinds": {},
195
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
196
+ }
layouts/UserRepGen-4B-Q3_K_L.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q6_k
51
+ ^blk\.0\.attn_qkv\.weight$=q5_k
52
+ ^blk\.0\.ffn_down\.weight$=q4_k
53
+ ^blk\.0\.ffn_gate\.weight$=q4_k
54
+ ^blk\.0\.ffn_up\.weight$=q4_k
55
+ ^blk\.0\.ssm_out\.weight$=q6_k
56
+ ^blk\.1\.attn_gate\.weight$=q4_k
57
+ ^blk\.1\.attn_qkv\.weight$=q3_k
58
+ ^blk\.1\.ffn_down\.weight$=q4_k
59
+ ^blk\.1\.ffn_gate\.weight$=q3_k
60
+ ^blk\.1\.ffn_up\.weight$=q4_k
61
+ ^blk\.1\.ssm_out\.weight$=q5_k
62
+ ^blk\.2\.attn_gate\.weight$=q4_k
63
+ ^blk\.2\.attn_qkv\.weight$=q3_k
64
+ ^blk\.2\.ffn_down\.weight$=q4_k
65
+ ^blk\.2\.ffn_gate\.weight$=q3_k
66
+ ^blk\.2\.ffn_up\.weight$=q4_k
67
+ ^blk\.2\.ssm_out\.weight$=q5_k
68
+ ^blk\.3\.attn_k\.weight$=q5_k
69
+ ^blk\.3\.attn_output\.weight$=q5_k
70
+ ^blk\.3\.attn_q\.weight$=q3_k
71
+ ^blk\.3\.attn_v\.weight$=q5_k
72
+ ^blk\.3\.ffn_down\.weight$=q4_k
73
+ ^blk\.3\.ffn_gate\.weight$=q3_k
74
+ ^blk\.3\.ffn_up\.weight$=q4_k
75
+ ^blk\.4\.attn_gate\.weight$=q4_k
76
+ ^blk\.4\.attn_qkv\.weight$=q3_k
77
+ ^blk\.4\.ffn_down\.weight$=q4_k
78
+ ^blk\.4\.ffn_gate\.weight$=q3_k
79
+ ^blk\.4\.ffn_up\.weight$=q4_k
80
+ ^blk\.4\.ssm_out\.weight$=q5_k
81
+ ^blk\.5\.attn_gate\.weight$=q4_k
82
+ ^blk\.5\.attn_qkv\.weight$=q3_k
83
+ ^blk\.5\.ffn_down\.weight$=q4_k
84
+ ^blk\.5\.ffn_gate\.weight$=q3_k
85
+ ^blk\.5\.ffn_up\.weight$=q4_k
86
+ ^blk\.5\.ssm_out\.weight$=q5_k
87
+ ^blk\.6\.attn_gate\.weight$=q4_k
88
+ ^blk\.6\.attn_qkv\.weight$=q3_k
89
+ ^blk\.6\.ffn_down\.weight$=q4_k
90
+ ^blk\.6\.ffn_gate\.weight$=q3_k
91
+ ^blk\.6\.ffn_up\.weight$=q4_k
92
+ ^blk\.6\.ssm_out\.weight$=q5_k
93
+ ^blk\.7\.attn_k\.weight$=q5_k
94
+ ^blk\.7\.attn_output\.weight$=q5_k
95
+ ^blk\.7\.attn_q\.weight$=q3_k
96
+ ^blk\.7\.attn_v\.weight$=q5_k
97
+ ^blk\.7\.ffn_down\.weight$=q4_k
98
+ ^blk\.7\.ffn_gate\.weight$=q3_k
99
+ ^blk\.7\.ffn_up\.weight$=q4_k
100
+ ^blk\.8\.attn_gate\.weight$=q3_k
101
+ ^blk\.8\.attn_qkv\.weight$=q3_k
102
+ ^blk\.8\.ffn_down\.weight$=q4_k
103
+ ^blk\.8\.ffn_gate\.weight$=q3_k
104
+ ^blk\.8\.ffn_up\.weight$=q3_k
105
+ ^blk\.8\.ssm_out\.weight$=q5_k
106
+ ^blk\.9\.attn_gate\.weight$=q3_k
107
+ ^blk\.9\.attn_qkv\.weight$=q3_k
108
+ ^blk\.9\.ffn_down\.weight$=q4_k
109
+ ^blk\.9\.ffn_gate\.weight$=q3_k
110
+ ^blk\.9\.ffn_up\.weight$=q3_k
111
+ ^blk\.9\.ssm_out\.weight$=q5_k
112
+ ^blk\.10\.attn_gate\.weight$=q3_k
113
+ ^blk\.10\.attn_qkv\.weight$=q3_k
114
+ ^blk\.10\.ffn_down\.weight$=q4_k
115
+ ^blk\.10\.ffn_gate\.weight$=q3_k
116
+ ^blk\.10\.ffn_up\.weight$=q3_k
117
+ ^blk\.10\.ssm_out\.weight$=q5_k
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q5_k
120
+ ^blk\.11\.attn_q\.weight$=q3_k
121
+ ^blk\.11\.attn_v\.weight$=q6_k
122
+ ^blk\.11\.ffn_down\.weight$=q4_k
123
+ ^blk\.11\.ffn_gate\.weight$=q3_k
124
+ ^blk\.11\.ffn_up\.weight$=q3_k
125
+ ^blk\.12\.attn_gate\.weight$=q3_k
126
+ ^blk\.12\.attn_qkv\.weight$=q3_k
127
+ ^blk\.12\.ffn_down\.weight$=q4_k
128
+ ^blk\.12\.ffn_gate\.weight$=q3_k
129
+ ^blk\.12\.ffn_up\.weight$=q3_k
130
+ ^blk\.12\.ssm_out\.weight$=q5_k
131
+ ^blk\.13\.attn_gate\.weight$=q3_k
132
+ ^blk\.13\.attn_qkv\.weight$=q3_k
133
+ ^blk\.13\.ffn_down\.weight$=q3_k
134
+ ^blk\.13\.ffn_gate\.weight$=q3_k
135
+ ^blk\.13\.ffn_up\.weight$=q3_k
136
+ ^blk\.13\.ssm_out\.weight$=q5_k
137
+ ^blk\.14\.attn_gate\.weight$=q3_k
138
+ ^blk\.14\.attn_qkv\.weight$=q3_k
139
+ ^blk\.14\.ffn_down\.weight$=q3_k
140
+ ^blk\.14\.ffn_gate\.weight$=q3_k
141
+ ^blk\.14\.ffn_up\.weight$=q3_k
142
+ ^blk\.14\.ssm_out\.weight$=q5_k
143
+ ^blk\.15\.attn_k\.weight$=q5_k
144
+ ^blk\.15\.attn_output\.weight$=q5_k
145
+ ^blk\.15\.attn_q\.weight$=q3_k
146
+ ^blk\.15\.attn_v\.weight$=q6_k
147
+ ^blk\.15\.ffn_down\.weight$=q4_k
148
+ ^blk\.15\.ffn_gate\.weight$=q3_k
149
+ ^blk\.15\.ffn_up\.weight$=q3_k
150
+ ^blk\.16\.attn_gate\.weight$=q3_k
151
+ ^blk\.16\.attn_qkv\.weight$=q3_k
152
+ ^blk\.16\.ffn_down\.weight$=q3_k
153
+ ^blk\.16\.ffn_gate\.weight$=q3_k
154
+ ^blk\.16\.ffn_up\.weight$=q3_k
155
+ ^blk\.16\.ssm_out\.weight$=q5_k
156
+ ^blk\.17\.attn_gate\.weight$=q3_k
157
+ ^blk\.17\.attn_qkv\.weight$=q3_k
158
+ ^blk\.17\.ffn_down\.weight$=q3_k
159
+ ^blk\.17\.ffn_gate\.weight$=q3_k
160
+ ^blk\.17\.ffn_up\.weight$=q3_k
161
+ ^blk\.17\.ssm_out\.weight$=q4_k
162
+ ^blk\.18\.attn_gate\.weight$=q3_k
163
+ ^blk\.18\.attn_qkv\.weight$=q3_k
164
+ ^blk\.18\.ffn_down\.weight$=q3_k
165
+ ^blk\.18\.ffn_gate\.weight$=q3_k
166
+ ^blk\.18\.ffn_up\.weight$=q3_k
167
+ ^blk\.18\.ssm_out\.weight$=q4_k
168
+ ^blk\.19\.attn_k\.weight$=q5_k
169
+ ^blk\.19\.attn_output\.weight$=q5_k
170
+ ^blk\.19\.attn_q\.weight$=q3_k
171
+ ^blk\.19\.attn_v\.weight$=q6_k
172
+ ^blk\.19\.ffn_down\.weight$=q4_k
173
+ ^blk\.19\.ffn_gate\.weight$=q3_k
174
+ ^blk\.19\.ffn_up\.weight$=q3_k
175
+ ^blk\.20\.attn_gate\.weight$=q3_k
176
+ ^blk\.20\.attn_qkv\.weight$=q3_k
177
+ ^blk\.20\.ffn_down\.weight$=q3_k
178
+ ^blk\.20\.ffn_gate\.weight$=q3_k
179
+ ^blk\.20\.ffn_up\.weight$=q3_k
180
+ ^blk\.20\.ssm_out\.weight$=q4_k
181
+ ^blk\.21\.attn_gate\.weight$=q3_k
182
+ ^blk\.21\.attn_qkv\.weight$=q3_k
183
+ ^blk\.21\.ffn_down\.weight$=q3_k
184
+ ^blk\.21\.ffn_gate\.weight$=q3_k
185
+ ^blk\.21\.ffn_up\.weight$=q3_k
186
+ ^blk\.21\.ssm_out\.weight$=q4_k
187
+ ^blk\.22\.attn_gate\.weight$=q3_k
188
+ ^blk\.22\.attn_qkv\.weight$=q3_k
189
+ ^blk\.22\.ffn_down\.weight$=q3_k
190
+ ^blk\.22\.ffn_gate\.weight$=q3_k
191
+ ^blk\.22\.ffn_up\.weight$=q3_k
192
+ ^blk\.22\.ssm_out\.weight$=q4_k
193
+ ^blk\.23\.attn_k\.weight$=q5_k
194
+ ^blk\.23\.attn_output\.weight$=q5_k
195
+ ^blk\.23\.attn_q\.weight$=q3_k
196
+ ^blk\.23\.attn_v\.weight$=q6_k
197
+ ^blk\.23\.ffn_down\.weight$=q4_k
198
+ ^blk\.23\.ffn_gate\.weight$=q3_k
199
+ ^blk\.23\.ffn_up\.weight$=q3_k
200
+ ^blk\.24\.attn_gate\.weight$=q4_k
201
+ ^blk\.24\.attn_qkv\.weight$=q3_k
202
+ ^blk\.24\.ffn_down\.weight$=q4_k
203
+ ^blk\.24\.ffn_gate\.weight$=q3_k
204
+ ^blk\.24\.ffn_up\.weight$=q4_k
205
+ ^blk\.24\.ssm_out\.weight$=q5_k
206
+ ^blk\.25\.attn_gate\.weight$=q3_k
207
+ ^blk\.25\.attn_qkv\.weight$=q3_k
208
+ ^blk\.25\.ffn_down\.weight$=q4_k
209
+ ^blk\.25\.ffn_gate\.weight$=q3_k
210
+ ^blk\.25\.ffn_up\.weight$=q4_k
211
+ ^blk\.25\.ssm_out\.weight$=q5_k
212
+ ^blk\.26\.attn_gate\.weight$=q3_k
213
+ ^blk\.26\.attn_qkv\.weight$=q3_k
214
+ ^blk\.26\.ffn_down\.weight$=q4_k
215
+ ^blk\.26\.ffn_gate\.weight$=q3_k
216
+ ^blk\.26\.ffn_up\.weight$=q4_k
217
+ ^blk\.26\.ssm_out\.weight$=q5_k
218
+ ^blk\.27\.attn_k\.weight$=q5_k
219
+ ^blk\.27\.attn_output\.weight$=q5_k
220
+ ^blk\.27\.attn_q\.weight$=q3_k
221
+ ^blk\.27\.attn_v\.weight$=q6_k
222
+ ^blk\.27\.ffn_down\.weight$=q4_k
223
+ ^blk\.27\.ffn_gate\.weight$=q3_k
224
+ ^blk\.27\.ffn_up\.weight$=q4_k
225
+ ^blk\.28\.attn_gate\.weight$=q3_k
226
+ ^blk\.28\.attn_qkv\.weight$=q3_k
227
+ ^blk\.28\.ffn_down\.weight$=q4_k
228
+ ^blk\.28\.ffn_gate\.weight$=q3_k
229
+ ^blk\.28\.ffn_up\.weight$=q4_k
230
+ ^blk\.28\.ssm_out\.weight$=q5_k
231
+ ^blk\.29\.attn_gate\.weight$=q3_k
232
+ ^blk\.29\.attn_qkv\.weight$=q3_k
233
+ ^blk\.29\.ffn_down\.weight$=q4_k
234
+ ^blk\.29\.ffn_gate\.weight$=q3_k
235
+ ^blk\.29\.ffn_up\.weight$=q4_k
236
+ ^blk\.29\.ssm_out\.weight$=q5_k
237
+ ^blk\.30\.attn_gate\.weight$=q3_k
238
+ ^blk\.30\.attn_qkv\.weight$=q3_k
239
+ ^blk\.30\.ffn_down\.weight$=q4_k
240
+ ^blk\.30\.ffn_gate\.weight$=q3_k
241
+ ^blk\.30\.ffn_up\.weight$=q4_k
242
+ ^blk\.30\.ssm_out\.weight$=q5_k
243
+ ^blk\.31\.attn_k\.weight$=q5_k
244
+ ^blk\.31\.attn_output\.weight$=q6_k
245
+ ^blk\.31\.attn_q\.weight$=q4_k
246
+ ^blk\.31\.attn_v\.weight$=q6_k
247
+ ^blk\.31\.ffn_down\.weight$=q5_k
248
+ ^blk\.31\.ffn_gate\.weight$=q5_k
249
+ ^blk\.31\.ffn_up\.weight$=q5_k
layouts/UserRepGen-4B-Q3_K_M.layout.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.214,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2416936960,
130
+ "map_body+embd": 2042327040,
131
+ "ratio": 0.845,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 3.78,
136
+ "body_bpw_heuristic": 4.253,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2061901824,
139
+ "file_bpw_pred": 3.922,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q3_k": 22,
143
+ "q6_k": 1,
144
+ "q4_k": 1
145
+ },
146
+ "attn_k": {
147
+ "q5_k": 8
148
+ },
149
+ "attn_output": {
150
+ "q5_k": 7,
151
+ "q6_k": 1
152
+ },
153
+ "attn_q": {
154
+ "q3_k": 8
155
+ },
156
+ "attn_qkv": {
157
+ "q3_k": 23,
158
+ "q5_k": 1
159
+ },
160
+ "attn_v": {
161
+ "q5_k": 6,
162
+ "q6_k": 2
163
+ },
164
+ "ffn_down": {
165
+ "q3_k": 25,
166
+ "q4_k": 6,
167
+ "q5_k": 1
168
+ },
169
+ "ffn_gate": {
170
+ "q3_k": 31,
171
+ "q5_k": 1
172
+ },
173
+ "ffn_up": {
174
+ "q3_k": 23,
175
+ "q4_k": 8,
176
+ "q5_k": 1
177
+ },
178
+ "ssm_out": {
179
+ "q4_k": 16,
180
+ "q5_k": 7,
181
+ "q6_k": 1
182
+ }
183
+ },
184
+ "prior": "layout/prior.json",
185
+ "pin_rule": "floor",
186
+ "embd_cap": "heur",
187
+ "generator": "auto_quant_v2.layout",
188
+ "generator_key": "3bf8b43e20a68f0b",
189
+ "quant_type": "Q3_K_M",
190
+ "quantize_ftype": "Q3_K_M",
191
+ "llama_cpp_version": "b11627",
192
+ "unknown_kinds": {},
193
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
194
+ }
layouts/UserRepGen-4B-Q3_K_M.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q6_k
51
+ ^blk\.0\.attn_qkv\.weight$=q5_k
52
+ ^blk\.0\.ffn_down\.weight$=q4_k
53
+ ^blk\.0\.ffn_gate\.weight$=q3_k
54
+ ^blk\.0\.ffn_up\.weight$=q4_k
55
+ ^blk\.0\.ssm_out\.weight$=q6_k
56
+ ^blk\.1\.attn_gate\.weight$=q4_k
57
+ ^blk\.1\.attn_qkv\.weight$=q3_k
58
+ ^blk\.1\.ffn_down\.weight$=q4_k
59
+ ^blk\.1\.ffn_gate\.weight$=q3_k
60
+ ^blk\.1\.ffn_up\.weight$=q4_k
61
+ ^blk\.1\.ssm_out\.weight$=q5_k
62
+ ^blk\.2\.attn_gate\.weight$=q3_k
63
+ ^blk\.2\.attn_qkv\.weight$=q3_k
64
+ ^blk\.2\.ffn_down\.weight$=q4_k
65
+ ^blk\.2\.ffn_gate\.weight$=q3_k
66
+ ^blk\.2\.ffn_up\.weight$=q4_k
67
+ ^blk\.2\.ssm_out\.weight$=q5_k
68
+ ^blk\.3\.attn_k\.weight$=q5_k
69
+ ^blk\.3\.attn_output\.weight$=q5_k
70
+ ^blk\.3\.attn_q\.weight$=q3_k
71
+ ^blk\.3\.attn_v\.weight$=q5_k
72
+ ^blk\.3\.ffn_down\.weight$=q4_k
73
+ ^blk\.3\.ffn_gate\.weight$=q3_k
74
+ ^blk\.3\.ffn_up\.weight$=q4_k
75
+ ^blk\.4\.attn_gate\.weight$=q3_k
76
+ ^blk\.4\.attn_qkv\.weight$=q3_k
77
+ ^blk\.4\.ffn_down\.weight$=q3_k
78
+ ^blk\.4\.ffn_gate\.weight$=q3_k
79
+ ^blk\.4\.ffn_up\.weight$=q4_k
80
+ ^blk\.4\.ssm_out\.weight$=q5_k
81
+ ^blk\.5\.attn_gate\.weight$=q3_k
82
+ ^blk\.5\.attn_qkv\.weight$=q3_k
83
+ ^blk\.5\.ffn_down\.weight$=q3_k
84
+ ^blk\.5\.ffn_gate\.weight$=q3_k
85
+ ^blk\.5\.ffn_up\.weight$=q4_k
86
+ ^blk\.5\.ssm_out\.weight$=q5_k
87
+ ^blk\.6\.attn_gate\.weight$=q3_k
88
+ ^blk\.6\.attn_qkv\.weight$=q3_k
89
+ ^blk\.6\.ffn_down\.weight$=q3_k
90
+ ^blk\.6\.ffn_gate\.weight$=q3_k
91
+ ^blk\.6\.ffn_up\.weight$=q4_k
92
+ ^blk\.6\.ssm_out\.weight$=q5_k
93
+ ^blk\.7\.attn_k\.weight$=q5_k
94
+ ^blk\.7\.attn_output\.weight$=q5_k
95
+ ^blk\.7\.attn_q\.weight$=q3_k
96
+ ^blk\.7\.attn_v\.weight$=q5_k
97
+ ^blk\.7\.ffn_down\.weight$=q4_k
98
+ ^blk\.7\.ffn_gate\.weight$=q3_k
99
+ ^blk\.7\.ffn_up\.weight$=q4_k
100
+ ^blk\.8\.attn_gate\.weight$=q3_k
101
+ ^blk\.8\.attn_qkv\.weight$=q3_k
102
+ ^blk\.8\.ffn_down\.weight$=q3_k
103
+ ^blk\.8\.ffn_gate\.weight$=q3_k
104
+ ^blk\.8\.ffn_up\.weight$=q3_k
105
+ ^blk\.8\.ssm_out\.weight$=q4_k
106
+ ^blk\.9\.attn_gate\.weight$=q3_k
107
+ ^blk\.9\.attn_qkv\.weight$=q3_k
108
+ ^blk\.9\.ffn_down\.weight$=q3_k
109
+ ^blk\.9\.ffn_gate\.weight$=q3_k
110
+ ^blk\.9\.ffn_up\.weight$=q3_k
111
+ ^blk\.9\.ssm_out\.weight$=q4_k
112
+ ^blk\.10\.attn_gate\.weight$=q3_k
113
+ ^blk\.10\.attn_qkv\.weight$=q3_k
114
+ ^blk\.10\.ffn_down\.weight$=q3_k
115
+ ^blk\.10\.ffn_gate\.weight$=q3_k
116
+ ^blk\.10\.ffn_up\.weight$=q3_k
117
+ ^blk\.10\.ssm_out\.weight$=q4_k
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q5_k
120
+ ^blk\.11\.attn_q\.weight$=q3_k
121
+ ^blk\.11\.attn_v\.weight$=q5_k
122
+ ^blk\.11\.ffn_down\.weight$=q3_k
123
+ ^blk\.11\.ffn_gate\.weight$=q3_k
124
+ ^blk\.11\.ffn_up\.weight$=q3_k
125
+ ^blk\.12\.attn_gate\.weight$=q3_k
126
+ ^blk\.12\.attn_qkv\.weight$=q3_k
127
+ ^blk\.12\.ffn_down\.weight$=q3_k
128
+ ^blk\.12\.ffn_gate\.weight$=q3_k
129
+ ^blk\.12\.ffn_up\.weight$=q3_k
130
+ ^blk\.12\.ssm_out\.weight$=q4_k
131
+ ^blk\.13\.attn_gate\.weight$=q3_k
132
+ ^blk\.13\.attn_qkv\.weight$=q3_k
133
+ ^blk\.13\.ffn_down\.weight$=q3_k
134
+ ^blk\.13\.ffn_gate\.weight$=q3_k
135
+ ^blk\.13\.ffn_up\.weight$=q3_k
136
+ ^blk\.13\.ssm_out\.weight$=q4_k
137
+ ^blk\.14\.attn_gate\.weight$=q3_k
138
+ ^blk\.14\.attn_qkv\.weight$=q3_k
139
+ ^blk\.14\.ffn_down\.weight$=q3_k
140
+ ^blk\.14\.ffn_gate\.weight$=q3_k
141
+ ^blk\.14\.ffn_up\.weight$=q3_k
142
+ ^blk\.14\.ssm_out\.weight$=q4_k
143
+ ^blk\.15\.attn_k\.weight$=q5_k
144
+ ^blk\.15\.attn_output\.weight$=q5_k
145
+ ^blk\.15\.attn_q\.weight$=q3_k
146
+ ^blk\.15\.attn_v\.weight$=q5_k
147
+ ^blk\.15\.ffn_down\.weight$=q3_k
148
+ ^blk\.15\.ffn_gate\.weight$=q3_k
149
+ ^blk\.15\.ffn_up\.weight$=q3_k
150
+ ^blk\.16\.attn_gate\.weight$=q3_k
151
+ ^blk\.16\.attn_qkv\.weight$=q3_k
152
+ ^blk\.16\.ffn_down\.weight$=q3_k
153
+ ^blk\.16\.ffn_gate\.weight$=q3_k
154
+ ^blk\.16\.ffn_up\.weight$=q3_k
155
+ ^blk\.16\.ssm_out\.weight$=q4_k
156
+ ^blk\.17\.attn_gate\.weight$=q3_k
157
+ ^blk\.17\.attn_qkv\.weight$=q3_k
158
+ ^blk\.17\.ffn_down\.weight$=q3_k
159
+ ^blk\.17\.ffn_gate\.weight$=q3_k
160
+ ^blk\.17\.ffn_up\.weight$=q3_k
161
+ ^blk\.17\.ssm_out\.weight$=q4_k
162
+ ^blk\.18\.attn_gate\.weight$=q3_k
163
+ ^blk\.18\.attn_qkv\.weight$=q3_k
164
+ ^blk\.18\.ffn_down\.weight$=q3_k
165
+ ^blk\.18\.ffn_gate\.weight$=q3_k
166
+ ^blk\.18\.ffn_up\.weight$=q3_k
167
+ ^blk\.18\.ssm_out\.weight$=q4_k
168
+ ^blk\.19\.attn_k\.weight$=q5_k
169
+ ^blk\.19\.attn_output\.weight$=q5_k
170
+ ^blk\.19\.attn_q\.weight$=q3_k
171
+ ^blk\.19\.attn_v\.weight$=q5_k
172
+ ^blk\.19\.ffn_down\.weight$=q3_k
173
+ ^blk\.19\.ffn_gate\.weight$=q3_k
174
+ ^blk\.19\.ffn_up\.weight$=q3_k
175
+ ^blk\.20\.attn_gate\.weight$=q3_k
176
+ ^blk\.20\.attn_qkv\.weight$=q3_k
177
+ ^blk\.20\.ffn_down\.weight$=q3_k
178
+ ^blk\.20\.ffn_gate\.weight$=q3_k
179
+ ^blk\.20\.ffn_up\.weight$=q3_k
180
+ ^blk\.20\.ssm_out\.weight$=q4_k
181
+ ^blk\.21\.attn_gate\.weight$=q3_k
182
+ ^blk\.21\.attn_qkv\.weight$=q3_k
183
+ ^blk\.21\.ffn_down\.weight$=q3_k
184
+ ^blk\.21\.ffn_gate\.weight$=q3_k
185
+ ^blk\.21\.ffn_up\.weight$=q3_k
186
+ ^blk\.21\.ssm_out\.weight$=q4_k
187
+ ^blk\.22\.attn_gate\.weight$=q3_k
188
+ ^blk\.22\.attn_qkv\.weight$=q3_k
189
+ ^blk\.22\.ffn_down\.weight$=q3_k
190
+ ^blk\.22\.ffn_gate\.weight$=q3_k
191
+ ^blk\.22\.ffn_up\.weight$=q3_k
192
+ ^blk\.22\.ssm_out\.weight$=q4_k
193
+ ^blk\.23\.attn_k\.weight$=q5_k
194
+ ^blk\.23\.attn_output\.weight$=q5_k
195
+ ^blk\.23\.attn_q\.weight$=q3_k
196
+ ^blk\.23\.attn_v\.weight$=q5_k
197
+ ^blk\.23\.ffn_down\.weight$=q3_k
198
+ ^blk\.23\.ffn_gate\.weight$=q3_k
199
+ ^blk\.23\.ffn_up\.weight$=q3_k
200
+ ^blk\.24\.attn_gate\.weight$=q3_k
201
+ ^blk\.24\.attn_qkv\.weight$=q3_k
202
+ ^blk\.24\.ffn_down\.weight$=q3_k
203
+ ^blk\.24\.ffn_gate\.weight$=q3_k
204
+ ^blk\.24\.ffn_up\.weight$=q3_k
205
+ ^blk\.24\.ssm_out\.weight$=q5_k
206
+ ^blk\.25\.attn_gate\.weight$=q3_k
207
+ ^blk\.25\.attn_qkv\.weight$=q3_k
208
+ ^blk\.25\.ffn_down\.weight$=q3_k
209
+ ^blk\.25\.ffn_gate\.weight$=q3_k
210
+ ^blk\.25\.ffn_up\.weight$=q3_k
211
+ ^blk\.25\.ssm_out\.weight$=q5_k
212
+ ^blk\.26\.attn_gate\.weight$=q3_k
213
+ ^blk\.26\.attn_qkv\.weight$=q3_k
214
+ ^blk\.26\.ffn_down\.weight$=q3_k
215
+ ^blk\.26\.ffn_gate\.weight$=q3_k
216
+ ^blk\.26\.ffn_up\.weight$=q3_k
217
+ ^blk\.26\.ssm_out\.weight$=q4_k
218
+ ^blk\.27\.attn_k\.weight$=q5_k
219
+ ^blk\.27\.attn_output\.weight$=q5_k
220
+ ^blk\.27\.attn_q\.weight$=q3_k
221
+ ^blk\.27\.attn_v\.weight$=q6_k
222
+ ^blk\.27\.ffn_down\.weight$=q4_k
223
+ ^blk\.27\.ffn_gate\.weight$=q3_k
224
+ ^blk\.27\.ffn_up\.weight$=q3_k
225
+ ^blk\.28\.attn_gate\.weight$=q3_k
226
+ ^blk\.28\.attn_qkv\.weight$=q3_k
227
+ ^blk\.28\.ffn_down\.weight$=q3_k
228
+ ^blk\.28\.ffn_gate\.weight$=q3_k
229
+ ^blk\.28\.ffn_up\.weight$=q3_k
230
+ ^blk\.28\.ssm_out\.weight$=q4_k
231
+ ^blk\.29\.attn_gate\.weight$=q3_k
232
+ ^blk\.29\.attn_qkv\.weight$=q3_k
233
+ ^blk\.29\.ffn_down\.weight$=q3_k
234
+ ^blk\.29\.ffn_gate\.weight$=q3_k
235
+ ^blk\.29\.ffn_up\.weight$=q3_k
236
+ ^blk\.29\.ssm_out\.weight$=q4_k
237
+ ^blk\.30\.attn_gate\.weight$=q3_k
238
+ ^blk\.30\.attn_qkv\.weight$=q3_k
239
+ ^blk\.30\.ffn_down\.weight$=q3_k
240
+ ^blk\.30\.ffn_gate\.weight$=q3_k
241
+ ^blk\.30\.ffn_up\.weight$=q3_k
242
+ ^blk\.30\.ssm_out\.weight$=q4_k
243
+ ^blk\.31\.attn_k\.weight$=q5_k
244
+ ^blk\.31\.attn_output\.weight$=q6_k
245
+ ^blk\.31\.attn_q\.weight$=q3_k
246
+ ^blk\.31\.attn_v\.weight$=q6_k
247
+ ^blk\.31\.ffn_down\.weight$=q5_k
248
+ ^blk\.31\.ffn_gate\.weight$=q5_k
249
+ ^blk\.31\.ffn_up\.weight$=q5_k
layouts/UserRepGen-4B-Q3_K_S.layout.json ADDED
@@ -0,0 +1,190 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.2252,
17
+ "type": "q4_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q4_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2295531520,
130
+ "map_body+embd": 1937039360,
131
+ "ratio": 0.8438,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.9,
135
+ "body_bpw": 3.544,
136
+ "body_bpw_heuristic": 3.981,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 1956614144,
139
+ "file_bpw_pred": 3.722,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q3_k": 23,
143
+ "q5_k": 1
144
+ },
145
+ "attn_k": {
146
+ "q4_k": 5,
147
+ "q5_k": 3
148
+ },
149
+ "attn_output": {
150
+ "q4_k": 7,
151
+ "q5_k": 1
152
+ },
153
+ "attn_q": {
154
+ "q3_k": 8
155
+ },
156
+ "attn_qkv": {
157
+ "q3_k": 23,
158
+ "q4_k": 1
159
+ },
160
+ "attn_v": {
161
+ "q5_k": 8
162
+ },
163
+ "ffn_down": {
164
+ "q3_k": 31,
165
+ "q5_k": 1
166
+ },
167
+ "ffn_gate": {
168
+ "q3_k": 32
169
+ },
170
+ "ffn_up": {
171
+ "q3_k": 31,
172
+ "q5_k": 1
173
+ },
174
+ "ssm_out": {
175
+ "q3_k": 19,
176
+ "q4_k": 4,
177
+ "q5_k": 1
178
+ }
179
+ },
180
+ "prior": "layout/prior.json",
181
+ "pin_rule": "floor",
182
+ "embd_cap": "heur",
183
+ "generator": "auto_quant_v2.layout",
184
+ "generator_key": "3bf8b43e20a68f0b",
185
+ "quant_type": "Q3_K_S",
186
+ "quantize_ftype": "Q3_K_S",
187
+ "llama_cpp_version": "b11627",
188
+ "unknown_kinds": {},
189
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
190
+ }
layouts/UserRepGen-4B-Q3_K_S.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q4_k
50
+ ^blk\.0\.attn_gate\.weight$=q5_k
51
+ ^blk\.0\.attn_qkv\.weight$=q4_k
52
+ ^blk\.0\.ffn_down\.weight$=q3_k
53
+ ^blk\.0\.ffn_gate\.weight$=q3_k
54
+ ^blk\.0\.ffn_up\.weight$=q3_k
55
+ ^blk\.0\.ssm_out\.weight$=q5_k
56
+ ^blk\.1\.attn_gate\.weight$=q3_k
57
+ ^blk\.1\.attn_qkv\.weight$=q3_k
58
+ ^blk\.1\.ffn_down\.weight$=q3_k
59
+ ^blk\.1\.ffn_gate\.weight$=q3_k
60
+ ^blk\.1\.ffn_up\.weight$=q3_k
61
+ ^blk\.1\.ssm_out\.weight$=q4_k
62
+ ^blk\.2\.attn_gate\.weight$=q3_k
63
+ ^blk\.2\.attn_qkv\.weight$=q3_k
64
+ ^blk\.2\.ffn_down\.weight$=q3_k
65
+ ^blk\.2\.ffn_gate\.weight$=q3_k
66
+ ^blk\.2\.ffn_up\.weight$=q3_k
67
+ ^blk\.2\.ssm_out\.weight$=q4_k
68
+ ^blk\.3\.attn_k\.weight$=q4_k
69
+ ^blk\.3\.attn_output\.weight$=q4_k
70
+ ^blk\.3\.attn_q\.weight$=q3_k
71
+ ^blk\.3\.attn_v\.weight$=q5_k
72
+ ^blk\.3\.ffn_down\.weight$=q3_k
73
+ ^blk\.3\.ffn_gate\.weight$=q3_k
74
+ ^blk\.3\.ffn_up\.weight$=q3_k
75
+ ^blk\.4\.attn_gate\.weight$=q3_k
76
+ ^blk\.4\.attn_qkv\.weight$=q3_k
77
+ ^blk\.4\.ffn_down\.weight$=q3_k
78
+ ^blk\.4\.ffn_gate\.weight$=q3_k
79
+ ^blk\.4\.ffn_up\.weight$=q3_k
80
+ ^blk\.4\.ssm_out\.weight$=q4_k
81
+ ^blk\.5\.attn_gate\.weight$=q3_k
82
+ ^blk\.5\.attn_qkv\.weight$=q3_k
83
+ ^blk\.5\.ffn_down\.weight$=q3_k
84
+ ^blk\.5\.ffn_gate\.weight$=q3_k
85
+ ^blk\.5\.ffn_up\.weight$=q3_k
86
+ ^blk\.5\.ssm_out\.weight$=q4_k
87
+ ^blk\.6\.attn_gate\.weight$=q3_k
88
+ ^blk\.6\.attn_qkv\.weight$=q3_k
89
+ ^blk\.6\.ffn_down\.weight$=q3_k
90
+ ^blk\.6\.ffn_gate\.weight$=q3_k
91
+ ^blk\.6\.ffn_up\.weight$=q3_k
92
+ ^blk\.6\.ssm_out\.weight$=q3_k
93
+ ^blk\.7\.attn_k\.weight$=q4_k
94
+ ^blk\.7\.attn_output\.weight$=q4_k
95
+ ^blk\.7\.attn_q\.weight$=q3_k
96
+ ^blk\.7\.attn_v\.weight$=q5_k
97
+ ^blk\.7\.ffn_down\.weight$=q3_k
98
+ ^blk\.7\.ffn_gate\.weight$=q3_k
99
+ ^blk\.7\.ffn_up\.weight$=q3_k
100
+ ^blk\.8\.attn_gate\.weight$=q3_k
101
+ ^blk\.8\.attn_qkv\.weight$=q3_k
102
+ ^blk\.8\.ffn_down\.weight$=q3_k
103
+ ^blk\.8\.ffn_gate\.weight$=q3_k
104
+ ^blk\.8\.ffn_up\.weight$=q3_k
105
+ ^blk\.8\.ssm_out\.weight$=q3_k
106
+ ^blk\.9\.attn_gate\.weight$=q3_k
107
+ ^blk\.9\.attn_qkv\.weight$=q3_k
108
+ ^blk\.9\.ffn_down\.weight$=q3_k
109
+ ^blk\.9\.ffn_gate\.weight$=q3_k
110
+ ^blk\.9\.ffn_up\.weight$=q3_k
111
+ ^blk\.9\.ssm_out\.weight$=q3_k
112
+ ^blk\.10\.attn_gate\.weight$=q3_k
113
+ ^blk\.10\.attn_qkv\.weight$=q3_k
114
+ ^blk\.10\.ffn_down\.weight$=q3_k
115
+ ^blk\.10\.ffn_gate\.weight$=q3_k
116
+ ^blk\.10\.ffn_up\.weight$=q3_k
117
+ ^blk\.10\.ssm_out\.weight$=q3_k
118
+ ^blk\.11\.attn_k\.weight$=q5_k
119
+ ^blk\.11\.attn_output\.weight$=q4_k
120
+ ^blk\.11\.attn_q\.weight$=q3_k
121
+ ^blk\.11\.attn_v\.weight$=q5_k
122
+ ^blk\.11\.ffn_down\.weight$=q3_k
123
+ ^blk\.11\.ffn_gate\.weight$=q3_k
124
+ ^blk\.11\.ffn_up\.weight$=q3_k
125
+ ^blk\.12\.attn_gate\.weight$=q3_k
126
+ ^blk\.12\.attn_qkv\.weight$=q3_k
127
+ ^blk\.12\.ffn_down\.weight$=q3_k
128
+ ^blk\.12\.ffn_gate\.weight$=q3_k
129
+ ^blk\.12\.ffn_up\.weight$=q3_k
130
+ ^blk\.12\.ssm_out\.weight$=q3_k
131
+ ^blk\.13\.attn_gate\.weight$=q3_k
132
+ ^blk\.13\.attn_qkv\.weight$=q3_k
133
+ ^blk\.13\.ffn_down\.weight$=q3_k
134
+ ^blk\.13\.ffn_gate\.weight$=q3_k
135
+ ^blk\.13\.ffn_up\.weight$=q3_k
136
+ ^blk\.13\.ssm_out\.weight$=q3_k
137
+ ^blk\.14\.attn_gate\.weight$=q3_k
138
+ ^blk\.14\.attn_qkv\.weight$=q3_k
139
+ ^blk\.14\.ffn_down\.weight$=q3_k
140
+ ^blk\.14\.ffn_gate\.weight$=q3_k
141
+ ^blk\.14\.ffn_up\.weight$=q3_k
142
+ ^blk\.14\.ssm_out\.weight$=q3_k
143
+ ^blk\.15\.attn_k\.weight$=q5_k
144
+ ^blk\.15\.attn_output\.weight$=q4_k
145
+ ^blk\.15\.attn_q\.weight$=q3_k
146
+ ^blk\.15\.attn_v\.weight$=q5_k
147
+ ^blk\.15\.ffn_down\.weight$=q3_k
148
+ ^blk\.15\.ffn_gate\.weight$=q3_k
149
+ ^blk\.15\.ffn_up\.weight$=q3_k
150
+ ^blk\.16\.attn_gate\.weight$=q3_k
151
+ ^blk\.16\.attn_qkv\.weight$=q3_k
152
+ ^blk\.16\.ffn_down\.weight$=q3_k
153
+ ^blk\.16\.ffn_gate\.weight$=q3_k
154
+ ^blk\.16\.ffn_up\.weight$=q3_k
155
+ ^blk\.16\.ssm_out\.weight$=q3_k
156
+ ^blk\.17\.attn_gate\.weight$=q3_k
157
+ ^blk\.17\.attn_qkv\.weight$=q3_k
158
+ ^blk\.17\.ffn_down\.weight$=q3_k
159
+ ^blk\.17\.ffn_gate\.weight$=q3_k
160
+ ^blk\.17\.ffn_up\.weight$=q3_k
161
+ ^blk\.17\.ssm_out\.weight$=q3_k
162
+ ^blk\.18\.attn_gate\.weight$=q3_k
163
+ ^blk\.18\.attn_qkv\.weight$=q3_k
164
+ ^blk\.18\.ffn_down\.weight$=q3_k
165
+ ^blk\.18\.ffn_gate\.weight$=q3_k
166
+ ^blk\.18\.ffn_up\.weight$=q3_k
167
+ ^blk\.18\.ssm_out\.weight$=q3_k
168
+ ^blk\.19\.attn_k\.weight$=q4_k
169
+ ^blk\.19\.attn_output\.weight$=q4_k
170
+ ^blk\.19\.attn_q\.weight$=q3_k
171
+ ^blk\.19\.attn_v\.weight$=q5_k
172
+ ^blk\.19\.ffn_down\.weight$=q3_k
173
+ ^blk\.19\.ffn_gate\.weight$=q3_k
174
+ ^blk\.19\.ffn_up\.weight$=q3_k
175
+ ^blk\.20\.attn_gate\.weight$=q3_k
176
+ ^blk\.20\.attn_qkv\.weight$=q3_k
177
+ ^blk\.20\.ffn_down\.weight$=q3_k
178
+ ^blk\.20\.ffn_gate\.weight$=q3_k
179
+ ^blk\.20\.ffn_up\.weight$=q3_k
180
+ ^blk\.20\.ssm_out\.weight$=q3_k
181
+ ^blk\.21\.attn_gate\.weight$=q3_k
182
+ ^blk\.21\.attn_qkv\.weight$=q3_k
183
+ ^blk\.21\.ffn_down\.weight$=q3_k
184
+ ^blk\.21\.ffn_gate\.weight$=q3_k
185
+ ^blk\.21\.ffn_up\.weight$=q3_k
186
+ ^blk\.21\.ssm_out\.weight$=q3_k
187
+ ^blk\.22\.attn_gate\.weight$=q3_k
188
+ ^blk\.22\.attn_qkv\.weight$=q3_k
189
+ ^blk\.22\.ffn_down\.weight$=q3_k
190
+ ^blk\.22\.ffn_gate\.weight$=q3_k
191
+ ^blk\.22\.ffn_up\.weight$=q3_k
192
+ ^blk\.22\.ssm_out\.weight$=q3_k
193
+ ^blk\.23\.attn_k\.weight$=q4_k
194
+ ^blk\.23\.attn_output\.weight$=q4_k
195
+ ^blk\.23\.attn_q\.weight$=q3_k
196
+ ^blk\.23\.attn_v\.weight$=q5_k
197
+ ^blk\.23\.ffn_down\.weight$=q3_k
198
+ ^blk\.23\.ffn_gate\.weight$=q3_k
199
+ ^blk\.23\.ffn_up\.weight$=q3_k
200
+ ^blk\.24\.attn_gate\.weight$=q3_k
201
+ ^blk\.24\.attn_qkv\.weight$=q3_k
202
+ ^blk\.24\.ffn_down\.weight$=q3_k
203
+ ^blk\.24\.ffn_gate\.weight$=q3_k
204
+ ^blk\.24\.ffn_up\.weight$=q3_k
205
+ ^blk\.24\.ssm_out\.weight$=q3_k
206
+ ^blk\.25\.attn_gate\.weight$=q3_k
207
+ ^blk\.25\.attn_qkv\.weight$=q3_k
208
+ ^blk\.25\.ffn_down\.weight$=q3_k
209
+ ^blk\.25\.ffn_gate\.weight$=q3_k
210
+ ^blk\.25\.ffn_up\.weight$=q3_k
211
+ ^blk\.25\.ssm_out\.weight$=q3_k
212
+ ^blk\.26\.attn_gate\.weight$=q3_k
213
+ ^blk\.26\.attn_qkv\.weight$=q3_k
214
+ ^blk\.26\.ffn_down\.weight$=q3_k
215
+ ^blk\.26\.ffn_gate\.weight$=q3_k
216
+ ^blk\.26\.ffn_up\.weight$=q3_k
217
+ ^blk\.26\.ssm_out\.weight$=q3_k
218
+ ^blk\.27\.attn_k\.weight$=q5_k
219
+ ^blk\.27\.attn_output\.weight$=q4_k
220
+ ^blk\.27\.attn_q\.weight$=q3_k
221
+ ^blk\.27\.attn_v\.weight$=q5_k
222
+ ^blk\.27\.ffn_down\.weight$=q3_k
223
+ ^blk\.27\.ffn_gate\.weight$=q3_k
224
+ ^blk\.27\.ffn_up\.weight$=q3_k
225
+ ^blk\.28\.attn_gate\.weight$=q3_k
226
+ ^blk\.28\.attn_qkv\.weight$=q3_k
227
+ ^blk\.28\.ffn_down\.weight$=q3_k
228
+ ^blk\.28\.ffn_gate\.weight$=q3_k
229
+ ^blk\.28\.ffn_up\.weight$=q3_k
230
+ ^blk\.28\.ssm_out\.weight$=q3_k
231
+ ^blk\.29\.attn_gate\.weight$=q3_k
232
+ ^blk\.29\.attn_qkv\.weight$=q3_k
233
+ ^blk\.29\.ffn_down\.weight$=q3_k
234
+ ^blk\.29\.ffn_gate\.weight$=q3_k
235
+ ^blk\.29\.ffn_up\.weight$=q3_k
236
+ ^blk\.29\.ssm_out\.weight$=q3_k
237
+ ^blk\.30\.attn_gate\.weight$=q3_k
238
+ ^blk\.30\.attn_qkv\.weight$=q3_k
239
+ ^blk\.30\.ffn_down\.weight$=q3_k
240
+ ^blk\.30\.ffn_gate\.weight$=q3_k
241
+ ^blk\.30\.ffn_up\.weight$=q3_k
242
+ ^blk\.30\.ssm_out\.weight$=q3_k
243
+ ^blk\.31\.attn_k\.weight$=q4_k
244
+ ^blk\.31\.attn_output\.weight$=q5_k
245
+ ^blk\.31\.attn_q\.weight$=q3_k
246
+ ^blk\.31\.attn_v\.weight$=q5_k
247
+ ^blk\.31\.ffn_down\.weight$=q5_k
248
+ ^blk\.31\.ffn_gate\.weight$=q3_k
249
+ ^blk\.31\.ffn_up\.weight$=q5_k
layouts/UserRepGen-4B-Q4_K_L.layout.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1815,
17
+ "type": "q5_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q5_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2854307840,
130
+ "map_body+embd": 2871029760,
131
+ "ratio": 1.0059,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.5,
135
+ "body_bpw": 5.462,
136
+ "body_bpw_heuristic": 5.235,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2890604544,
139
+ "file_bpw_pred": 5.498,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q4_k": 18,
143
+ "q6_k": 5,
144
+ "q8_0": 1
145
+ },
146
+ "attn_k": {
147
+ "q8_0": 8
148
+ },
149
+ "attn_output": {
150
+ "q8_0": 8
151
+ },
152
+ "attn_q": {
153
+ "q4_k": 7,
154
+ "q6_k": 1
155
+ },
156
+ "attn_qkv": {
157
+ "q4_k": 22,
158
+ "q8_0": 1,
159
+ "q5_k": 1
160
+ },
161
+ "attn_v": {
162
+ "q8_0": 8
163
+ },
164
+ "ffn_down": {
165
+ "q6_k": 19,
166
+ "q4_k": 12,
167
+ "q8_0": 1
168
+ },
169
+ "ffn_gate": {
170
+ "q4_k": 30,
171
+ "q6_k": 1,
172
+ "q8_0": 1
173
+ },
174
+ "ffn_up": {
175
+ "q4_k": 16,
176
+ "q6_k": 15,
177
+ "q8_0": 1
178
+ },
179
+ "ssm_out": {
180
+ "q6_k": 18,
181
+ "q8_0": 6
182
+ }
183
+ },
184
+ "prior": "layout/prior.json",
185
+ "pin_rule": "floor",
186
+ "embd_cap": "heur",
187
+ "generator": "auto_quant_v2.layout",
188
+ "generator_key": "3bf8b43e20a68f0b",
189
+ "quant_type": "Q4_K_L",
190
+ "quantize_ftype": "Q4_K_M",
191
+ "llama_cpp_version": "b11627",
192
+ "unknown_kinds": {},
193
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
194
+ }
layouts/UserRepGen-4B-Q4_K_L.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q5_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q6_k
53
+ ^blk\.0\.ffn_gate\.weight$=q6_k
54
+ ^blk\.0\.ffn_up\.weight$=q6_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q6_k
57
+ ^blk\.1\.attn_qkv\.weight$=q5_k
58
+ ^blk\.1\.ffn_down\.weight$=q6_k
59
+ ^blk\.1\.ffn_gate\.weight$=q4_k
60
+ ^blk\.1\.ffn_up\.weight$=q6_k
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=q6_k
63
+ ^blk\.2\.attn_qkv\.weight$=q4_k
64
+ ^blk\.2\.ffn_down\.weight$=q6_k
65
+ ^blk\.2\.ffn_gate\.weight$=q4_k
66
+ ^blk\.2\.ffn_up\.weight$=q6_k
67
+ ^blk\.2\.ssm_out\.weight$=q8_0
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q4_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q6_k
73
+ ^blk\.3\.ffn_gate\.weight$=q4_k
74
+ ^blk\.3\.ffn_up\.weight$=q6_k
75
+ ^blk\.4\.attn_gate\.weight$=q6_k
76
+ ^blk\.4\.attn_qkv\.weight$=q4_k
77
+ ^blk\.4\.ffn_down\.weight$=q6_k
78
+ ^blk\.4\.ffn_gate\.weight$=q4_k
79
+ ^blk\.4\.ffn_up\.weight$=q6_k
80
+ ^blk\.4\.ssm_out\.weight$=q8_0
81
+ ^blk\.5\.attn_gate\.weight$=q6_k
82
+ ^blk\.5\.attn_qkv\.weight$=q4_k
83
+ ^blk\.5\.ffn_down\.weight$=q6_k
84
+ ^blk\.5\.ffn_gate\.weight$=q4_k
85
+ ^blk\.5\.ffn_up\.weight$=q6_k
86
+ ^blk\.5\.ssm_out\.weight$=q8_0
87
+ ^blk\.6\.attn_gate\.weight$=q6_k
88
+ ^blk\.6\.attn_qkv\.weight$=q4_k
89
+ ^blk\.6\.ffn_down\.weight$=q6_k
90
+ ^blk\.6\.ffn_gate\.weight$=q4_k
91
+ ^blk\.6\.ffn_up\.weight$=q6_k
92
+ ^blk\.6\.ssm_out\.weight$=q8_0
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q4_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q6_k
98
+ ^blk\.7\.ffn_gate\.weight$=q4_k
99
+ ^blk\.7\.ffn_up\.weight$=q6_k
100
+ ^blk\.8\.attn_gate\.weight$=q4_k
101
+ ^blk\.8\.attn_qkv\.weight$=q4_k
102
+ ^blk\.8\.ffn_down\.weight$=q4_k
103
+ ^blk\.8\.ffn_gate\.weight$=q4_k
104
+ ^blk\.8\.ffn_up\.weight$=q4_k
105
+ ^blk\.8\.ssm_out\.weight$=q6_k
106
+ ^blk\.9\.attn_gate\.weight$=q4_k
107
+ ^blk\.9\.attn_qkv\.weight$=q4_k
108
+ ^blk\.9\.ffn_down\.weight$=q4_k
109
+ ^blk\.9\.ffn_gate\.weight$=q4_k
110
+ ^blk\.9\.ffn_up\.weight$=q4_k
111
+ ^blk\.9\.ssm_out\.weight$=q6_k
112
+ ^blk\.10\.attn_gate\.weight$=q4_k
113
+ ^blk\.10\.attn_qkv\.weight$=q4_k
114
+ ^blk\.10\.ffn_down\.weight$=q4_k
115
+ ^blk\.10\.ffn_gate\.weight$=q4_k
116
+ ^blk\.10\.ffn_up\.weight$=q4_k
117
+ ^blk\.10\.ssm_out\.weight$=q6_k
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q4_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q6_k
123
+ ^blk\.11\.ffn_gate\.weight$=q4_k
124
+ ^blk\.11\.ffn_up\.weight$=q4_k
125
+ ^blk\.12\.attn_gate\.weight$=q4_k
126
+ ^blk\.12\.attn_qkv\.weight$=q4_k
127
+ ^blk\.12\.ffn_down\.weight$=q4_k
128
+ ^blk\.12\.ffn_gate\.weight$=q4_k
129
+ ^blk\.12\.ffn_up\.weight$=q4_k
130
+ ^blk\.12\.ssm_out\.weight$=q6_k
131
+ ^blk\.13\.attn_gate\.weight$=q4_k
132
+ ^blk\.13\.attn_qkv\.weight$=q4_k
133
+ ^blk\.13\.ffn_down\.weight$=q4_k
134
+ ^blk\.13\.ffn_gate\.weight$=q4_k
135
+ ^blk\.13\.ffn_up\.weight$=q4_k
136
+ ^blk\.13\.ssm_out\.weight$=q6_k
137
+ ^blk\.14\.attn_gate\.weight$=q4_k
138
+ ^blk\.14\.attn_qkv\.weight$=q4_k
139
+ ^blk\.14\.ffn_down\.weight$=q4_k
140
+ ^blk\.14\.ffn_gate\.weight$=q4_k
141
+ ^blk\.14\.ffn_up\.weight$=q4_k
142
+ ^blk\.14\.ssm_out\.weight$=q6_k
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q4_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q6_k
148
+ ^blk\.15\.ffn_gate\.weight$=q4_k
149
+ ^blk\.15\.ffn_up\.weight$=q4_k
150
+ ^blk\.16\.attn_gate\.weight$=q4_k
151
+ ^blk\.16\.attn_qkv\.weight$=q4_k
152
+ ^blk\.16\.ffn_down\.weight$=q4_k
153
+ ^blk\.16\.ffn_gate\.weight$=q4_k
154
+ ^blk\.16\.ffn_up\.weight$=q4_k
155
+ ^blk\.16\.ssm_out\.weight$=q6_k
156
+ ^blk\.17\.attn_gate\.weight$=q4_k
157
+ ^blk\.17\.attn_qkv\.weight$=q4_k
158
+ ^blk\.17\.ffn_down\.weight$=q4_k
159
+ ^blk\.17\.ffn_gate\.weight$=q4_k
160
+ ^blk\.17\.ffn_up\.weight$=q4_k
161
+ ^blk\.17\.ssm_out\.weight$=q6_k
162
+ ^blk\.18\.attn_gate\.weight$=q4_k
163
+ ^blk\.18\.attn_qkv\.weight$=q4_k
164
+ ^blk\.18\.ffn_down\.weight$=q4_k
165
+ ^blk\.18\.ffn_gate\.weight$=q4_k
166
+ ^blk\.18\.ffn_up\.weight$=q4_k
167
+ ^blk\.18\.ssm_out\.weight$=q6_k
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q4_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q6_k
173
+ ^blk\.19\.ffn_gate\.weight$=q4_k
174
+ ^blk\.19\.ffn_up\.weight$=q4_k
175
+ ^blk\.20\.attn_gate\.weight$=q4_k
176
+ ^blk\.20\.attn_qkv\.weight$=q4_k
177
+ ^blk\.20\.ffn_down\.weight$=q4_k
178
+ ^blk\.20\.ffn_gate\.weight$=q4_k
179
+ ^blk\.20\.ffn_up\.weight$=q4_k
180
+ ^blk\.20\.ssm_out\.weight$=q6_k
181
+ ^blk\.21\.attn_gate\.weight$=q4_k
182
+ ^blk\.21\.attn_qkv\.weight$=q4_k
183
+ ^blk\.21\.ffn_down\.weight$=q4_k
184
+ ^blk\.21\.ffn_gate\.weight$=q4_k
185
+ ^blk\.21\.ffn_up\.weight$=q4_k
186
+ ^blk\.21\.ssm_out\.weight$=q6_k
187
+ ^blk\.22\.attn_gate\.weight$=q4_k
188
+ ^blk\.22\.attn_qkv\.weight$=q4_k
189
+ ^blk\.22\.ffn_down\.weight$=q4_k
190
+ ^blk\.22\.ffn_gate\.weight$=q4_k
191
+ ^blk\.22\.ffn_up\.weight$=q4_k
192
+ ^blk\.22\.ssm_out\.weight$=q6_k
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q4_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q6_k
198
+ ^blk\.23\.ffn_gate\.weight$=q4_k
199
+ ^blk\.23\.ffn_up\.weight$=q4_k
200
+ ^blk\.24\.attn_gate\.weight$=q4_k
201
+ ^blk\.24\.attn_qkv\.weight$=q4_k
202
+ ^blk\.24\.ffn_down\.weight$=q6_k
203
+ ^blk\.24\.ffn_gate\.weight$=q4_k
204
+ ^blk\.24\.ffn_up\.weight$=q6_k
205
+ ^blk\.24\.ssm_out\.weight$=q6_k
206
+ ^blk\.25\.attn_gate\.weight$=q4_k
207
+ ^blk\.25\.attn_qkv\.weight$=q4_k
208
+ ^blk\.25\.ffn_down\.weight$=q6_k
209
+ ^blk\.25\.ffn_gate\.weight$=q4_k
210
+ ^blk\.25\.ffn_up\.weight$=q6_k
211
+ ^blk\.25\.ssm_out\.weight$=q6_k
212
+ ^blk\.26\.attn_gate\.weight$=q4_k
213
+ ^blk\.26\.attn_qkv\.weight$=q4_k
214
+ ^blk\.26\.ffn_down\.weight$=q6_k
215
+ ^blk\.26\.ffn_gate\.weight$=q4_k
216
+ ^blk\.26\.ffn_up\.weight$=q6_k
217
+ ^blk\.26\.ssm_out\.weight$=q6_k
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q4_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q6_k
223
+ ^blk\.27\.ffn_gate\.weight$=q4_k
224
+ ^blk\.27\.ffn_up\.weight$=q6_k
225
+ ^blk\.28\.attn_gate\.weight$=q4_k
226
+ ^blk\.28\.attn_qkv\.weight$=q4_k
227
+ ^blk\.28\.ffn_down\.weight$=q6_k
228
+ ^blk\.28\.ffn_gate\.weight$=q4_k
229
+ ^blk\.28\.ffn_up\.weight$=q6_k
230
+ ^blk\.28\.ssm_out\.weight$=q6_k
231
+ ^blk\.29\.attn_gate\.weight$=q4_k
232
+ ^blk\.29\.attn_qkv\.weight$=q4_k
233
+ ^blk\.29\.ffn_down\.weight$=q6_k
234
+ ^blk\.29\.ffn_gate\.weight$=q4_k
235
+ ^blk\.29\.ffn_up\.weight$=q6_k
236
+ ^blk\.29\.ssm_out\.weight$=q6_k
237
+ ^blk\.30\.attn_gate\.weight$=q4_k
238
+ ^blk\.30\.attn_qkv\.weight$=q4_k
239
+ ^blk\.30\.ffn_down\.weight$=q6_k
240
+ ^blk\.30\.ffn_gate\.weight$=q4_k
241
+ ^blk\.30\.ffn_up\.weight$=q6_k
242
+ ^blk\.30\.ssm_out\.weight$=q6_k
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q6_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q8_0
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-Q4_K_M.layout.json ADDED
@@ -0,0 +1,191 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1815,
17
+ "type": "q5_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q5_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2854307840,
130
+ "map_body+embd": 2699100160,
131
+ "ratio": 0.9456,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 5.076,
136
+ "body_bpw_heuristic": 5.235,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2718674944,
139
+ "file_bpw_pred": 5.171,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q4_k": 22,
143
+ "q8_0": 1,
144
+ "q5_k": 1
145
+ },
146
+ "attn_k": {
147
+ "q8_0": 8
148
+ },
149
+ "attn_output": {
150
+ "q8_0": 8
151
+ },
152
+ "attn_q": {
153
+ "q4_k": 8
154
+ },
155
+ "attn_qkv": {
156
+ "q4_k": 23,
157
+ "q8_0": 1
158
+ },
159
+ "attn_v": {
160
+ "q8_0": 8
161
+ },
162
+ "ffn_down": {
163
+ "q4_k": 28,
164
+ "q6_k": 3,
165
+ "q8_0": 1
166
+ },
167
+ "ffn_gate": {
168
+ "q4_k": 31,
169
+ "q6_k": 1
170
+ },
171
+ "ffn_up": {
172
+ "q4_k": 23,
173
+ "q6_k": 8,
174
+ "q8_0": 1
175
+ },
176
+ "ssm_out": {
177
+ "q6_k": 18,
178
+ "q8_0": 6
179
+ }
180
+ },
181
+ "prior": "layout/prior.json",
182
+ "pin_rule": "floor",
183
+ "embd_cap": "heur",
184
+ "generator": "auto_quant_v2.layout",
185
+ "generator_key": "3bf8b43e20a68f0b",
186
+ "quant_type": "Q4_K_M",
187
+ "quantize_ftype": "Q4_K_M",
188
+ "llama_cpp_version": "b11627",
189
+ "unknown_kinds": {},
190
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
191
+ }
layouts/UserRepGen-4B-Q4_K_M.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q5_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q6_k
53
+ ^blk\.0\.ffn_gate\.weight$=q4_k
54
+ ^blk\.0\.ffn_up\.weight$=q6_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q5_k
57
+ ^blk\.1\.attn_qkv\.weight$=q4_k
58
+ ^blk\.1\.ffn_down\.weight$=q4_k
59
+ ^blk\.1\.ffn_gate\.weight$=q4_k
60
+ ^blk\.1\.ffn_up\.weight$=q6_k
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=q4_k
63
+ ^blk\.2\.attn_qkv\.weight$=q4_k
64
+ ^blk\.2\.ffn_down\.weight$=q4_k
65
+ ^blk\.2\.ffn_gate\.weight$=q4_k
66
+ ^blk\.2\.ffn_up\.weight$=q6_k
67
+ ^blk\.2\.ssm_out\.weight$=q8_0
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q4_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q6_k
73
+ ^blk\.3\.ffn_gate\.weight$=q4_k
74
+ ^blk\.3\.ffn_up\.weight$=q6_k
75
+ ^blk\.4\.attn_gate\.weight$=q4_k
76
+ ^blk\.4\.attn_qkv\.weight$=q4_k
77
+ ^blk\.4\.ffn_down\.weight$=q4_k
78
+ ^blk\.4\.ffn_gate\.weight$=q4_k
79
+ ^blk\.4\.ffn_up\.weight$=q6_k
80
+ ^blk\.4\.ssm_out\.weight$=q8_0
81
+ ^blk\.5\.attn_gate\.weight$=q4_k
82
+ ^blk\.5\.attn_qkv\.weight$=q4_k
83
+ ^blk\.5\.ffn_down\.weight$=q4_k
84
+ ^blk\.5\.ffn_gate\.weight$=q4_k
85
+ ^blk\.5\.ffn_up\.weight$=q6_k
86
+ ^blk\.5\.ssm_out\.weight$=q8_0
87
+ ^blk\.6\.attn_gate\.weight$=q4_k
88
+ ^blk\.6\.attn_qkv\.weight$=q4_k
89
+ ^blk\.6\.ffn_down\.weight$=q4_k
90
+ ^blk\.6\.ffn_gate\.weight$=q4_k
91
+ ^blk\.6\.ffn_up\.weight$=q6_k
92
+ ^blk\.6\.ssm_out\.weight$=q8_0
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q4_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q4_k
98
+ ^blk\.7\.ffn_gate\.weight$=q4_k
99
+ ^blk\.7\.ffn_up\.weight$=q6_k
100
+ ^blk\.8\.attn_gate\.weight$=q4_k
101
+ ^blk\.8\.attn_qkv\.weight$=q4_k
102
+ ^blk\.8\.ffn_down\.weight$=q4_k
103
+ ^blk\.8\.ffn_gate\.weight$=q4_k
104
+ ^blk\.8\.ffn_up\.weight$=q4_k
105
+ ^blk\.8\.ssm_out\.weight$=q6_k
106
+ ^blk\.9\.attn_gate\.weight$=q4_k
107
+ ^blk\.9\.attn_qkv\.weight$=q4_k
108
+ ^blk\.9\.ffn_down\.weight$=q4_k
109
+ ^blk\.9\.ffn_gate\.weight$=q4_k
110
+ ^blk\.9\.ffn_up\.weight$=q4_k
111
+ ^blk\.9\.ssm_out\.weight$=q6_k
112
+ ^blk\.10\.attn_gate\.weight$=q4_k
113
+ ^blk\.10\.attn_qkv\.weight$=q4_k
114
+ ^blk\.10\.ffn_down\.weight$=q4_k
115
+ ^blk\.10\.ffn_gate\.weight$=q4_k
116
+ ^blk\.10\.ffn_up\.weight$=q4_k
117
+ ^blk\.10\.ssm_out\.weight$=q6_k
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q4_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q4_k
123
+ ^blk\.11\.ffn_gate\.weight$=q4_k
124
+ ^blk\.11\.ffn_up\.weight$=q4_k
125
+ ^blk\.12\.attn_gate\.weight$=q4_k
126
+ ^blk\.12\.attn_qkv\.weight$=q4_k
127
+ ^blk\.12\.ffn_down\.weight$=q4_k
128
+ ^blk\.12\.ffn_gate\.weight$=q4_k
129
+ ^blk\.12\.ffn_up\.weight$=q4_k
130
+ ^blk\.12\.ssm_out\.weight$=q6_k
131
+ ^blk\.13\.attn_gate\.weight$=q4_k
132
+ ^blk\.13\.attn_qkv\.weight$=q4_k
133
+ ^blk\.13\.ffn_down\.weight$=q4_k
134
+ ^blk\.13\.ffn_gate\.weight$=q4_k
135
+ ^blk\.13\.ffn_up\.weight$=q4_k
136
+ ^blk\.13\.ssm_out\.weight$=q6_k
137
+ ^blk\.14\.attn_gate\.weight$=q4_k
138
+ ^blk\.14\.attn_qkv\.weight$=q4_k
139
+ ^blk\.14\.ffn_down\.weight$=q4_k
140
+ ^blk\.14\.ffn_gate\.weight$=q4_k
141
+ ^blk\.14\.ffn_up\.weight$=q4_k
142
+ ^blk\.14\.ssm_out\.weight$=q6_k
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q4_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q4_k
148
+ ^blk\.15\.ffn_gate\.weight$=q4_k
149
+ ^blk\.15\.ffn_up\.weight$=q4_k
150
+ ^blk\.16\.attn_gate\.weight$=q4_k
151
+ ^blk\.16\.attn_qkv\.weight$=q4_k
152
+ ^blk\.16\.ffn_down\.weight$=q4_k
153
+ ^blk\.16\.ffn_gate\.weight$=q4_k
154
+ ^blk\.16\.ffn_up\.weight$=q4_k
155
+ ^blk\.16\.ssm_out\.weight$=q6_k
156
+ ^blk\.17\.attn_gate\.weight$=q4_k
157
+ ^blk\.17\.attn_qkv\.weight$=q4_k
158
+ ^blk\.17\.ffn_down\.weight$=q4_k
159
+ ^blk\.17\.ffn_gate\.weight$=q4_k
160
+ ^blk\.17\.ffn_up\.weight$=q4_k
161
+ ^blk\.17\.ssm_out\.weight$=q6_k
162
+ ^blk\.18\.attn_gate\.weight$=q4_k
163
+ ^blk\.18\.attn_qkv\.weight$=q4_k
164
+ ^blk\.18\.ffn_down\.weight$=q4_k
165
+ ^blk\.18\.ffn_gate\.weight$=q4_k
166
+ ^blk\.18\.ffn_up\.weight$=q4_k
167
+ ^blk\.18\.ssm_out\.weight$=q6_k
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q4_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q4_k
173
+ ^blk\.19\.ffn_gate\.weight$=q4_k
174
+ ^blk\.19\.ffn_up\.weight$=q4_k
175
+ ^blk\.20\.attn_gate\.weight$=q4_k
176
+ ^blk\.20\.attn_qkv\.weight$=q4_k
177
+ ^blk\.20\.ffn_down\.weight$=q4_k
178
+ ^blk\.20\.ffn_gate\.weight$=q4_k
179
+ ^blk\.20\.ffn_up\.weight$=q4_k
180
+ ^blk\.20\.ssm_out\.weight$=q6_k
181
+ ^blk\.21\.attn_gate\.weight$=q4_k
182
+ ^blk\.21\.attn_qkv\.weight$=q4_k
183
+ ^blk\.21\.ffn_down\.weight$=q4_k
184
+ ^blk\.21\.ffn_gate\.weight$=q4_k
185
+ ^blk\.21\.ffn_up\.weight$=q4_k
186
+ ^blk\.21\.ssm_out\.weight$=q6_k
187
+ ^blk\.22\.attn_gate\.weight$=q4_k
188
+ ^blk\.22\.attn_qkv\.weight$=q4_k
189
+ ^blk\.22\.ffn_down\.weight$=q4_k
190
+ ^blk\.22\.ffn_gate\.weight$=q4_k
191
+ ^blk\.22\.ffn_up\.weight$=q4_k
192
+ ^blk\.22\.ssm_out\.weight$=q6_k
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q4_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q4_k
198
+ ^blk\.23\.ffn_gate\.weight$=q4_k
199
+ ^blk\.23\.ffn_up\.weight$=q4_k
200
+ ^blk\.24\.attn_gate\.weight$=q4_k
201
+ ^blk\.24\.attn_qkv\.weight$=q4_k
202
+ ^blk\.24\.ffn_down\.weight$=q4_k
203
+ ^blk\.24\.ffn_gate\.weight$=q4_k
204
+ ^blk\.24\.ffn_up\.weight$=q4_k
205
+ ^blk\.24\.ssm_out\.weight$=q6_k
206
+ ^blk\.25\.attn_gate\.weight$=q4_k
207
+ ^blk\.25\.attn_qkv\.weight$=q4_k
208
+ ^blk\.25\.ffn_down\.weight$=q4_k
209
+ ^blk\.25\.ffn_gate\.weight$=q4_k
210
+ ^blk\.25\.ffn_up\.weight$=q4_k
211
+ ^blk\.25\.ssm_out\.weight$=q6_k
212
+ ^blk\.26\.attn_gate\.weight$=q4_k
213
+ ^blk\.26\.attn_qkv\.weight$=q4_k
214
+ ^blk\.26\.ffn_down\.weight$=q4_k
215
+ ^blk\.26\.ffn_gate\.weight$=q4_k
216
+ ^blk\.26\.ffn_up\.weight$=q4_k
217
+ ^blk\.26\.ssm_out\.weight$=q6_k
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q4_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q6_k
223
+ ^blk\.27\.ffn_gate\.weight$=q4_k
224
+ ^blk\.27\.ffn_up\.weight$=q4_k
225
+ ^blk\.28\.attn_gate\.weight$=q4_k
226
+ ^blk\.28\.attn_qkv\.weight$=q4_k
227
+ ^blk\.28\.ffn_down\.weight$=q4_k
228
+ ^blk\.28\.ffn_gate\.weight$=q4_k
229
+ ^blk\.28\.ffn_up\.weight$=q4_k
230
+ ^blk\.28\.ssm_out\.weight$=q6_k
231
+ ^blk\.29\.attn_gate\.weight$=q4_k
232
+ ^blk\.29\.attn_qkv\.weight$=q4_k
233
+ ^blk\.29\.ffn_down\.weight$=q4_k
234
+ ^blk\.29\.ffn_gate\.weight$=q4_k
235
+ ^blk\.29\.ffn_up\.weight$=q4_k
236
+ ^blk\.29\.ssm_out\.weight$=q6_k
237
+ ^blk\.30\.attn_gate\.weight$=q4_k
238
+ ^blk\.30\.attn_qkv\.weight$=q4_k
239
+ ^blk\.30\.ffn_down\.weight$=q4_k
240
+ ^blk\.30\.ffn_gate\.weight$=q4_k
241
+ ^blk\.30\.ffn_up\.weight$=q4_k
242
+ ^blk\.30\.ssm_out\.weight$=q6_k
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q4_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q6_k
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-Q4_K_S.layout.json ADDED
@@ -0,0 +1,190 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1926,
17
+ "type": "q5_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q5_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 2687436800,
130
+ "map_body+embd": 2517237760,
131
+ "ratio": 0.9367,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.9,
135
+ "body_bpw": 4.668,
136
+ "body_bpw_heuristic": 4.86,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 2536812544,
139
+ "file_bpw_pred": 4.825,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q4_k": 23,
143
+ "q8_0": 1
144
+ },
145
+ "attn_k": {
146
+ "q6_k": 6,
147
+ "q8_0": 2
148
+ },
149
+ "attn_output": {
150
+ "q6_k": 7,
151
+ "q8_0": 1
152
+ },
153
+ "attn_q": {
154
+ "q4_k": 8
155
+ },
156
+ "attn_qkv": {
157
+ "q4_k": 23,
158
+ "q6_k": 1
159
+ },
160
+ "attn_v": {
161
+ "q8_0": 8
162
+ },
163
+ "ffn_down": {
164
+ "q4_k": 31,
165
+ "q6_k": 1
166
+ },
167
+ "ffn_gate": {
168
+ "q4_k": 32
169
+ },
170
+ "ffn_up": {
171
+ "q4_k": 31,
172
+ "q6_k": 1
173
+ },
174
+ "ssm_out": {
175
+ "q4_k": 21,
176
+ "q6_k": 2,
177
+ "q8_0": 1
178
+ }
179
+ },
180
+ "prior": "layout/prior.json",
181
+ "pin_rule": "floor",
182
+ "embd_cap": "heur",
183
+ "generator": "auto_quant_v2.layout",
184
+ "generator_key": "3bf8b43e20a68f0b",
185
+ "quant_type": "Q4_K_S",
186
+ "quantize_ftype": "Q4_K_S",
187
+ "llama_cpp_version": "b11627",
188
+ "unknown_kinds": {},
189
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
190
+ }
layouts/UserRepGen-4B-Q4_K_S.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q5_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q6_k
52
+ ^blk\.0\.ffn_down\.weight$=q4_k
53
+ ^blk\.0\.ffn_gate\.weight$=q4_k
54
+ ^blk\.0\.ffn_up\.weight$=q4_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q4_k
57
+ ^blk\.1\.attn_qkv\.weight$=q4_k
58
+ ^blk\.1\.ffn_down\.weight$=q4_k
59
+ ^blk\.1\.ffn_gate\.weight$=q4_k
60
+ ^blk\.1\.ffn_up\.weight$=q4_k
61
+ ^blk\.1\.ssm_out\.weight$=q6_k
62
+ ^blk\.2\.attn_gate\.weight$=q4_k
63
+ ^blk\.2\.attn_qkv\.weight$=q4_k
64
+ ^blk\.2\.ffn_down\.weight$=q4_k
65
+ ^blk\.2\.ffn_gate\.weight$=q4_k
66
+ ^blk\.2\.ffn_up\.weight$=q4_k
67
+ ^blk\.2\.ssm_out\.weight$=q6_k
68
+ ^blk\.3\.attn_k\.weight$=q6_k
69
+ ^blk\.3\.attn_output\.weight$=q6_k
70
+ ^blk\.3\.attn_q\.weight$=q4_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q4_k
73
+ ^blk\.3\.ffn_gate\.weight$=q4_k
74
+ ^blk\.3\.ffn_up\.weight$=q4_k
75
+ ^blk\.4\.attn_gate\.weight$=q4_k
76
+ ^blk\.4\.attn_qkv\.weight$=q4_k
77
+ ^blk\.4\.ffn_down\.weight$=q4_k
78
+ ^blk\.4\.ffn_gate\.weight$=q4_k
79
+ ^blk\.4\.ffn_up\.weight$=q4_k
80
+ ^blk\.4\.ssm_out\.weight$=q4_k
81
+ ^blk\.5\.attn_gate\.weight$=q4_k
82
+ ^blk\.5\.attn_qkv\.weight$=q4_k
83
+ ^blk\.5\.ffn_down\.weight$=q4_k
84
+ ^blk\.5\.ffn_gate\.weight$=q4_k
85
+ ^blk\.5\.ffn_up\.weight$=q4_k
86
+ ^blk\.5\.ssm_out\.weight$=q4_k
87
+ ^blk\.6\.attn_gate\.weight$=q4_k
88
+ ^blk\.6\.attn_qkv\.weight$=q4_k
89
+ ^blk\.6\.ffn_down\.weight$=q4_k
90
+ ^blk\.6\.ffn_gate\.weight$=q4_k
91
+ ^blk\.6\.ffn_up\.weight$=q4_k
92
+ ^blk\.6\.ssm_out\.weight$=q4_k
93
+ ^blk\.7\.attn_k\.weight$=q6_k
94
+ ^blk\.7\.attn_output\.weight$=q6_k
95
+ ^blk\.7\.attn_q\.weight$=q4_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q4_k
98
+ ^blk\.7\.ffn_gate\.weight$=q4_k
99
+ ^blk\.7\.ffn_up\.weight$=q4_k
100
+ ^blk\.8\.attn_gate\.weight$=q4_k
101
+ ^blk\.8\.attn_qkv\.weight$=q4_k
102
+ ^blk\.8\.ffn_down\.weight$=q4_k
103
+ ^blk\.8\.ffn_gate\.weight$=q4_k
104
+ ^blk\.8\.ffn_up\.weight$=q4_k
105
+ ^blk\.8\.ssm_out\.weight$=q4_k
106
+ ^blk\.9\.attn_gate\.weight$=q4_k
107
+ ^blk\.9\.attn_qkv\.weight$=q4_k
108
+ ^blk\.9\.ffn_down\.weight$=q4_k
109
+ ^blk\.9\.ffn_gate\.weight$=q4_k
110
+ ^blk\.9\.ffn_up\.weight$=q4_k
111
+ ^blk\.9\.ssm_out\.weight$=q4_k
112
+ ^blk\.10\.attn_gate\.weight$=q4_k
113
+ ^blk\.10\.attn_qkv\.weight$=q4_k
114
+ ^blk\.10\.ffn_down\.weight$=q4_k
115
+ ^blk\.10\.ffn_gate\.weight$=q4_k
116
+ ^blk\.10\.ffn_up\.weight$=q4_k
117
+ ^blk\.10\.ssm_out\.weight$=q4_k
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q6_k
120
+ ^blk\.11\.attn_q\.weight$=q4_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q4_k
123
+ ^blk\.11\.ffn_gate\.weight$=q4_k
124
+ ^blk\.11\.ffn_up\.weight$=q4_k
125
+ ^blk\.12\.attn_gate\.weight$=q4_k
126
+ ^blk\.12\.attn_qkv\.weight$=q4_k
127
+ ^blk\.12\.ffn_down\.weight$=q4_k
128
+ ^blk\.12\.ffn_gate\.weight$=q4_k
129
+ ^blk\.12\.ffn_up\.weight$=q4_k
130
+ ^blk\.12\.ssm_out\.weight$=q4_k
131
+ ^blk\.13\.attn_gate\.weight$=q4_k
132
+ ^blk\.13\.attn_qkv\.weight$=q4_k
133
+ ^blk\.13\.ffn_down\.weight$=q4_k
134
+ ^blk\.13\.ffn_gate\.weight$=q4_k
135
+ ^blk\.13\.ffn_up\.weight$=q4_k
136
+ ^blk\.13\.ssm_out\.weight$=q4_k
137
+ ^blk\.14\.attn_gate\.weight$=q4_k
138
+ ^blk\.14\.attn_qkv\.weight$=q4_k
139
+ ^blk\.14\.ffn_down\.weight$=q4_k
140
+ ^blk\.14\.ffn_gate\.weight$=q4_k
141
+ ^blk\.14\.ffn_up\.weight$=q4_k
142
+ ^blk\.14\.ssm_out\.weight$=q4_k
143
+ ^blk\.15\.attn_k\.weight$=q6_k
144
+ ^blk\.15\.attn_output\.weight$=q6_k
145
+ ^blk\.15\.attn_q\.weight$=q4_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q4_k
148
+ ^blk\.15\.ffn_gate\.weight$=q4_k
149
+ ^blk\.15\.ffn_up\.weight$=q4_k
150
+ ^blk\.16\.attn_gate\.weight$=q4_k
151
+ ^blk\.16\.attn_qkv\.weight$=q4_k
152
+ ^blk\.16\.ffn_down\.weight$=q4_k
153
+ ^blk\.16\.ffn_gate\.weight$=q4_k
154
+ ^blk\.16\.ffn_up\.weight$=q4_k
155
+ ^blk\.16\.ssm_out\.weight$=q4_k
156
+ ^blk\.17\.attn_gate\.weight$=q4_k
157
+ ^blk\.17\.attn_qkv\.weight$=q4_k
158
+ ^blk\.17\.ffn_down\.weight$=q4_k
159
+ ^blk\.17\.ffn_gate\.weight$=q4_k
160
+ ^blk\.17\.ffn_up\.weight$=q4_k
161
+ ^blk\.17\.ssm_out\.weight$=q4_k
162
+ ^blk\.18\.attn_gate\.weight$=q4_k
163
+ ^blk\.18\.attn_qkv\.weight$=q4_k
164
+ ^blk\.18\.ffn_down\.weight$=q4_k
165
+ ^blk\.18\.ffn_gate\.weight$=q4_k
166
+ ^blk\.18\.ffn_up\.weight$=q4_k
167
+ ^blk\.18\.ssm_out\.weight$=q4_k
168
+ ^blk\.19\.attn_k\.weight$=q6_k
169
+ ^blk\.19\.attn_output\.weight$=q6_k
170
+ ^blk\.19\.attn_q\.weight$=q4_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q4_k
173
+ ^blk\.19\.ffn_gate\.weight$=q4_k
174
+ ^blk\.19\.ffn_up\.weight$=q4_k
175
+ ^blk\.20\.attn_gate\.weight$=q4_k
176
+ ^blk\.20\.attn_qkv\.weight$=q4_k
177
+ ^blk\.20\.ffn_down\.weight$=q4_k
178
+ ^blk\.20\.ffn_gate\.weight$=q4_k
179
+ ^blk\.20\.ffn_up\.weight$=q4_k
180
+ ^blk\.20\.ssm_out\.weight$=q4_k
181
+ ^blk\.21\.attn_gate\.weight$=q4_k
182
+ ^blk\.21\.attn_qkv\.weight$=q4_k
183
+ ^blk\.21\.ffn_down\.weight$=q4_k
184
+ ^blk\.21\.ffn_gate\.weight$=q4_k
185
+ ^blk\.21\.ffn_up\.weight$=q4_k
186
+ ^blk\.21\.ssm_out\.weight$=q4_k
187
+ ^blk\.22\.attn_gate\.weight$=q4_k
188
+ ^blk\.22\.attn_qkv\.weight$=q4_k
189
+ ^blk\.22\.ffn_down\.weight$=q4_k
190
+ ^blk\.22\.ffn_gate\.weight$=q4_k
191
+ ^blk\.22\.ffn_up\.weight$=q4_k
192
+ ^blk\.22\.ssm_out\.weight$=q4_k
193
+ ^blk\.23\.attn_k\.weight$=q6_k
194
+ ^blk\.23\.attn_output\.weight$=q6_k
195
+ ^blk\.23\.attn_q\.weight$=q4_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q4_k
198
+ ^blk\.23\.ffn_gate\.weight$=q4_k
199
+ ^blk\.23\.ffn_up\.weight$=q4_k
200
+ ^blk\.24\.attn_gate\.weight$=q4_k
201
+ ^blk\.24\.attn_qkv\.weight$=q4_k
202
+ ^blk\.24\.ffn_down\.weight$=q4_k
203
+ ^blk\.24\.ffn_gate\.weight$=q4_k
204
+ ^blk\.24\.ffn_up\.weight$=q4_k
205
+ ^blk\.24\.ssm_out\.weight$=q4_k
206
+ ^blk\.25\.attn_gate\.weight$=q4_k
207
+ ^blk\.25\.attn_qkv\.weight$=q4_k
208
+ ^blk\.25\.ffn_down\.weight$=q4_k
209
+ ^blk\.25\.ffn_gate\.weight$=q4_k
210
+ ^blk\.25\.ffn_up\.weight$=q4_k
211
+ ^blk\.25\.ssm_out\.weight$=q4_k
212
+ ^blk\.26\.attn_gate\.weight$=q4_k
213
+ ^blk\.26\.attn_qkv\.weight$=q4_k
214
+ ^blk\.26\.ffn_down\.weight$=q4_k
215
+ ^blk\.26\.ffn_gate\.weight$=q4_k
216
+ ^blk\.26\.ffn_up\.weight$=q4_k
217
+ ^blk\.26\.ssm_out\.weight$=q4_k
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q6_k
220
+ ^blk\.27\.attn_q\.weight$=q4_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q4_k
223
+ ^blk\.27\.ffn_gate\.weight$=q4_k
224
+ ^blk\.27\.ffn_up\.weight$=q4_k
225
+ ^blk\.28\.attn_gate\.weight$=q4_k
226
+ ^blk\.28\.attn_qkv\.weight$=q4_k
227
+ ^blk\.28\.ffn_down\.weight$=q4_k
228
+ ^blk\.28\.ffn_gate\.weight$=q4_k
229
+ ^blk\.28\.ffn_up\.weight$=q4_k
230
+ ^blk\.28\.ssm_out\.weight$=q4_k
231
+ ^blk\.29\.attn_gate\.weight$=q4_k
232
+ ^blk\.29\.attn_qkv\.weight$=q4_k
233
+ ^blk\.29\.ffn_down\.weight$=q4_k
234
+ ^blk\.29\.ffn_gate\.weight$=q4_k
235
+ ^blk\.29\.ffn_up\.weight$=q4_k
236
+ ^blk\.29\.ssm_out\.weight$=q4_k
237
+ ^blk\.30\.attn_gate\.weight$=q4_k
238
+ ^blk\.30\.attn_qkv\.weight$=q4_k
239
+ ^blk\.30\.ffn_down\.weight$=q4_k
240
+ ^blk\.30\.ffn_gate\.weight$=q4_k
241
+ ^blk\.30\.ffn_up\.weight$=q4_k
242
+ ^blk\.30\.ssm_out\.weight$=q4_k
243
+ ^blk\.31\.attn_k\.weight$=q6_k
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q4_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q6_k
248
+ ^blk\.31\.ffn_gate\.weight$=q4_k
249
+ ^blk\.31\.ffn_up\.weight$=q6_k
layouts/UserRepGen-4B-Q5_K_M.layout.json ADDED
@@ -0,0 +1,187 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1578,
17
+ "type": "q6_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q6_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 3285043200,
130
+ "map_body+embd": 3262515200,
131
+ "ratio": 0.9931,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.7,
135
+ "body_bpw": 6.151,
136
+ "body_bpw_heuristic": 6.201,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 3282089984,
139
+ "file_bpw_pred": 6.243,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q5_k": 22,
143
+ "q8_0": 2
144
+ },
145
+ "attn_k": {
146
+ "q8_0": 8
147
+ },
148
+ "attn_output": {
149
+ "q8_0": 8
150
+ },
151
+ "attn_q": {
152
+ "q5_k": 8
153
+ },
154
+ "attn_qkv": {
155
+ "q5_k": 23,
156
+ "q8_0": 1
157
+ },
158
+ "attn_v": {
159
+ "q8_0": 8
160
+ },
161
+ "ffn_down": {
162
+ "q5_k": 27,
163
+ "q8_0": 5
164
+ },
165
+ "ffn_gate": {
166
+ "q5_k": 31,
167
+ "q8_0": 1
168
+ },
169
+ "ffn_up": {
170
+ "q5_k": 23,
171
+ "q8_0": 9
172
+ },
173
+ "ssm_out": {
174
+ "q8_0": 24
175
+ }
176
+ },
177
+ "prior": "layout/prior.json",
178
+ "pin_rule": "floor",
179
+ "embd_cap": "heur",
180
+ "generator": "auto_quant_v2.layout",
181
+ "generator_key": "3bf8b43e20a68f0b",
182
+ "quant_type": "Q5_K_M",
183
+ "quantize_ftype": "Q5_K_M",
184
+ "llama_cpp_version": "b11627",
185
+ "unknown_kinds": {},
186
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
187
+ }
layouts/UserRepGen-4B-Q5_K_M.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q6_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q8_0
53
+ ^blk\.0\.ffn_gate\.weight$=q5_k
54
+ ^blk\.0\.ffn_up\.weight$=q8_0
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q8_0
57
+ ^blk\.1\.attn_qkv\.weight$=q5_k
58
+ ^blk\.1\.ffn_down\.weight$=q5_k
59
+ ^blk\.1\.ffn_gate\.weight$=q5_k
60
+ ^blk\.1\.ffn_up\.weight$=q8_0
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=q5_k
63
+ ^blk\.2\.attn_qkv\.weight$=q5_k
64
+ ^blk\.2\.ffn_down\.weight$=q5_k
65
+ ^blk\.2\.ffn_gate\.weight$=q5_k
66
+ ^blk\.2\.ffn_up\.weight$=q8_0
67
+ ^blk\.2\.ssm_out\.weight$=q8_0
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q5_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q8_0
73
+ ^blk\.3\.ffn_gate\.weight$=q5_k
74
+ ^blk\.3\.ffn_up\.weight$=q8_0
75
+ ^blk\.4\.attn_gate\.weight$=q5_k
76
+ ^blk\.4\.attn_qkv\.weight$=q5_k
77
+ ^blk\.4\.ffn_down\.weight$=q5_k
78
+ ^blk\.4\.ffn_gate\.weight$=q5_k
79
+ ^blk\.4\.ffn_up\.weight$=q8_0
80
+ ^blk\.4\.ssm_out\.weight$=q8_0
81
+ ^blk\.5\.attn_gate\.weight$=q5_k
82
+ ^blk\.5\.attn_qkv\.weight$=q5_k
83
+ ^blk\.5\.ffn_down\.weight$=q5_k
84
+ ^blk\.5\.ffn_gate\.weight$=q5_k
85
+ ^blk\.5\.ffn_up\.weight$=q8_0
86
+ ^blk\.5\.ssm_out\.weight$=q8_0
87
+ ^blk\.6\.attn_gate\.weight$=q5_k
88
+ ^blk\.6\.attn_qkv\.weight$=q5_k
89
+ ^blk\.6\.ffn_down\.weight$=q5_k
90
+ ^blk\.6\.ffn_gate\.weight$=q5_k
91
+ ^blk\.6\.ffn_up\.weight$=q8_0
92
+ ^blk\.6\.ssm_out\.weight$=q8_0
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q5_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q8_0
98
+ ^blk\.7\.ffn_gate\.weight$=q5_k
99
+ ^blk\.7\.ffn_up\.weight$=q8_0
100
+ ^blk\.8\.attn_gate\.weight$=q5_k
101
+ ^blk\.8\.attn_qkv\.weight$=q5_k
102
+ ^blk\.8\.ffn_down\.weight$=q5_k
103
+ ^blk\.8\.ffn_gate\.weight$=q5_k
104
+ ^blk\.8\.ffn_up\.weight$=q5_k
105
+ ^blk\.8\.ssm_out\.weight$=q8_0
106
+ ^blk\.9\.attn_gate\.weight$=q5_k
107
+ ^blk\.9\.attn_qkv\.weight$=q5_k
108
+ ^blk\.9\.ffn_down\.weight$=q5_k
109
+ ^blk\.9\.ffn_gate\.weight$=q5_k
110
+ ^blk\.9\.ffn_up\.weight$=q5_k
111
+ ^blk\.9\.ssm_out\.weight$=q8_0
112
+ ^blk\.10\.attn_gate\.weight$=q5_k
113
+ ^blk\.10\.attn_qkv\.weight$=q5_k
114
+ ^blk\.10\.ffn_down\.weight$=q5_k
115
+ ^blk\.10\.ffn_gate\.weight$=q5_k
116
+ ^blk\.10\.ffn_up\.weight$=q5_k
117
+ ^blk\.10\.ssm_out\.weight$=q8_0
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q5_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q5_k
123
+ ^blk\.11\.ffn_gate\.weight$=q5_k
124
+ ^blk\.11\.ffn_up\.weight$=q5_k
125
+ ^blk\.12\.attn_gate\.weight$=q5_k
126
+ ^blk\.12\.attn_qkv\.weight$=q5_k
127
+ ^blk\.12\.ffn_down\.weight$=q5_k
128
+ ^blk\.12\.ffn_gate\.weight$=q5_k
129
+ ^blk\.12\.ffn_up\.weight$=q5_k
130
+ ^blk\.12\.ssm_out\.weight$=q8_0
131
+ ^blk\.13\.attn_gate\.weight$=q5_k
132
+ ^blk\.13\.attn_qkv\.weight$=q5_k
133
+ ^blk\.13\.ffn_down\.weight$=q5_k
134
+ ^blk\.13\.ffn_gate\.weight$=q5_k
135
+ ^blk\.13\.ffn_up\.weight$=q5_k
136
+ ^blk\.13\.ssm_out\.weight$=q8_0
137
+ ^blk\.14\.attn_gate\.weight$=q5_k
138
+ ^blk\.14\.attn_qkv\.weight$=q5_k
139
+ ^blk\.14\.ffn_down\.weight$=q5_k
140
+ ^blk\.14\.ffn_gate\.weight$=q5_k
141
+ ^blk\.14\.ffn_up\.weight$=q5_k
142
+ ^blk\.14\.ssm_out\.weight$=q8_0
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q5_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q5_k
148
+ ^blk\.15\.ffn_gate\.weight$=q5_k
149
+ ^blk\.15\.ffn_up\.weight$=q5_k
150
+ ^blk\.16\.attn_gate\.weight$=q5_k
151
+ ^blk\.16\.attn_qkv\.weight$=q5_k
152
+ ^blk\.16\.ffn_down\.weight$=q5_k
153
+ ^blk\.16\.ffn_gate\.weight$=q5_k
154
+ ^blk\.16\.ffn_up\.weight$=q5_k
155
+ ^blk\.16\.ssm_out\.weight$=q8_0
156
+ ^blk\.17\.attn_gate\.weight$=q5_k
157
+ ^blk\.17\.attn_qkv\.weight$=q5_k
158
+ ^blk\.17\.ffn_down\.weight$=q5_k
159
+ ^blk\.17\.ffn_gate\.weight$=q5_k
160
+ ^blk\.17\.ffn_up\.weight$=q5_k
161
+ ^blk\.17\.ssm_out\.weight$=q8_0
162
+ ^blk\.18\.attn_gate\.weight$=q5_k
163
+ ^blk\.18\.attn_qkv\.weight$=q5_k
164
+ ^blk\.18\.ffn_down\.weight$=q5_k
165
+ ^blk\.18\.ffn_gate\.weight$=q5_k
166
+ ^blk\.18\.ffn_up\.weight$=q5_k
167
+ ^blk\.18\.ssm_out\.weight$=q8_0
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q5_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q5_k
173
+ ^blk\.19\.ffn_gate\.weight$=q5_k
174
+ ^blk\.19\.ffn_up\.weight$=q5_k
175
+ ^blk\.20\.attn_gate\.weight$=q5_k
176
+ ^blk\.20\.attn_qkv\.weight$=q5_k
177
+ ^blk\.20\.ffn_down\.weight$=q5_k
178
+ ^blk\.20\.ffn_gate\.weight$=q5_k
179
+ ^blk\.20\.ffn_up\.weight$=q5_k
180
+ ^blk\.20\.ssm_out\.weight$=q8_0
181
+ ^blk\.21\.attn_gate\.weight$=q5_k
182
+ ^blk\.21\.attn_qkv\.weight$=q5_k
183
+ ^blk\.21\.ffn_down\.weight$=q5_k
184
+ ^blk\.21\.ffn_gate\.weight$=q5_k
185
+ ^blk\.21\.ffn_up\.weight$=q5_k
186
+ ^blk\.21\.ssm_out\.weight$=q8_0
187
+ ^blk\.22\.attn_gate\.weight$=q5_k
188
+ ^blk\.22\.attn_qkv\.weight$=q5_k
189
+ ^blk\.22\.ffn_down\.weight$=q5_k
190
+ ^blk\.22\.ffn_gate\.weight$=q5_k
191
+ ^blk\.22\.ffn_up\.weight$=q5_k
192
+ ^blk\.22\.ssm_out\.weight$=q8_0
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q5_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q5_k
198
+ ^blk\.23\.ffn_gate\.weight$=q5_k
199
+ ^blk\.23\.ffn_up\.weight$=q5_k
200
+ ^blk\.24\.attn_gate\.weight$=q5_k
201
+ ^blk\.24\.attn_qkv\.weight$=q5_k
202
+ ^blk\.24\.ffn_down\.weight$=q5_k
203
+ ^blk\.24\.ffn_gate\.weight$=q5_k
204
+ ^blk\.24\.ffn_up\.weight$=q5_k
205
+ ^blk\.24\.ssm_out\.weight$=q8_0
206
+ ^blk\.25\.attn_gate\.weight$=q5_k
207
+ ^blk\.25\.attn_qkv\.weight$=q5_k
208
+ ^blk\.25\.ffn_down\.weight$=q5_k
209
+ ^blk\.25\.ffn_gate\.weight$=q5_k
210
+ ^blk\.25\.ffn_up\.weight$=q5_k
211
+ ^blk\.25\.ssm_out\.weight$=q8_0
212
+ ^blk\.26\.attn_gate\.weight$=q5_k
213
+ ^blk\.26\.attn_qkv\.weight$=q5_k
214
+ ^blk\.26\.ffn_down\.weight$=q5_k
215
+ ^blk\.26\.ffn_gate\.weight$=q5_k
216
+ ^blk\.26\.ffn_up\.weight$=q5_k
217
+ ^blk\.26\.ssm_out\.weight$=q8_0
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q5_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q8_0
223
+ ^blk\.27\.ffn_gate\.weight$=q5_k
224
+ ^blk\.27\.ffn_up\.weight$=q5_k
225
+ ^blk\.28\.attn_gate\.weight$=q5_k
226
+ ^blk\.28\.attn_qkv\.weight$=q5_k
227
+ ^blk\.28\.ffn_down\.weight$=q5_k
228
+ ^blk\.28\.ffn_gate\.weight$=q5_k
229
+ ^blk\.28\.ffn_up\.weight$=q5_k
230
+ ^blk\.28\.ssm_out\.weight$=q8_0
231
+ ^blk\.29\.attn_gate\.weight$=q5_k
232
+ ^blk\.29\.attn_qkv\.weight$=q5_k
233
+ ^blk\.29\.ffn_down\.weight$=q5_k
234
+ ^blk\.29\.ffn_gate\.weight$=q5_k
235
+ ^blk\.29\.ffn_up\.weight$=q5_k
236
+ ^blk\.29\.ssm_out\.weight$=q8_0
237
+ ^blk\.30\.attn_gate\.weight$=q5_k
238
+ ^blk\.30\.attn_qkv\.weight$=q5_k
239
+ ^blk\.30\.ffn_down\.weight$=q5_k
240
+ ^blk\.30\.ffn_gate\.weight$=q5_k
241
+ ^blk\.30\.ffn_up\.weight$=q5_k
242
+ ^blk\.30\.ssm_out\.weight$=q8_0
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q5_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q8_0
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-Q5_K_S.layout.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1673,
17
+ "type": "q6_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q6_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 3097692160,
130
+ "map_body+embd": 3056322560,
131
+ "ratio": 0.9866,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.901,
135
+ "body_bpw": 5.688,
136
+ "body_bpw_heuristic": 5.781,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 3075897344,
139
+ "file_bpw_pred": 5.851,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q5_k": 23,
143
+ "q8_0": 1
144
+ },
145
+ "attn_k": {
146
+ "q8_0": 7,
147
+ "q5_k": 1
148
+ },
149
+ "attn_output": {
150
+ "q8_0": 8
151
+ },
152
+ "attn_q": {
153
+ "q5_k": 8
154
+ },
155
+ "attn_qkv": {
156
+ "q5_k": 23,
157
+ "q8_0": 1
158
+ },
159
+ "attn_v": {
160
+ "q8_0": 8
161
+ },
162
+ "ffn_down": {
163
+ "q5_k": 31,
164
+ "q8_0": 1
165
+ },
166
+ "ffn_gate": {
167
+ "q5_k": 32
168
+ },
169
+ "ffn_up": {
170
+ "q5_k": 31,
171
+ "q8_0": 1
172
+ },
173
+ "ssm_out": {
174
+ "q5_k": 20,
175
+ "q6_k": 3,
176
+ "q8_0": 1
177
+ }
178
+ },
179
+ "prior": "layout/prior.json",
180
+ "pin_rule": "floor",
181
+ "embd_cap": "heur",
182
+ "generator": "auto_quant_v2.layout",
183
+ "generator_key": "3bf8b43e20a68f0b",
184
+ "quant_type": "Q5_K_S",
185
+ "quantize_ftype": "Q5_K_S",
186
+ "llama_cpp_version": "b11627",
187
+ "unknown_kinds": {},
188
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
189
+ }
layouts/UserRepGen-4B-Q5_K_S.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q6_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q5_k
53
+ ^blk\.0\.ffn_gate\.weight$=q5_k
54
+ ^blk\.0\.ffn_up\.weight$=q5_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q5_k
57
+ ^blk\.1\.attn_qkv\.weight$=q5_k
58
+ ^blk\.1\.ffn_down\.weight$=q5_k
59
+ ^blk\.1\.ffn_gate\.weight$=q5_k
60
+ ^blk\.1\.ffn_up\.weight$=q5_k
61
+ ^blk\.1\.ssm_out\.weight$=q6_k
62
+ ^blk\.2\.attn_gate\.weight$=q5_k
63
+ ^blk\.2\.attn_qkv\.weight$=q5_k
64
+ ^blk\.2\.ffn_down\.weight$=q5_k
65
+ ^blk\.2\.ffn_gate\.weight$=q5_k
66
+ ^blk\.2\.ffn_up\.weight$=q5_k
67
+ ^blk\.2\.ssm_out\.weight$=q6_k
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q5_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q5_k
73
+ ^blk\.3\.ffn_gate\.weight$=q5_k
74
+ ^blk\.3\.ffn_up\.weight$=q5_k
75
+ ^blk\.4\.attn_gate\.weight$=q5_k
76
+ ^blk\.4\.attn_qkv\.weight$=q5_k
77
+ ^blk\.4\.ffn_down\.weight$=q5_k
78
+ ^blk\.4\.ffn_gate\.weight$=q5_k
79
+ ^blk\.4\.ffn_up\.weight$=q5_k
80
+ ^blk\.4\.ssm_out\.weight$=q6_k
81
+ ^blk\.5\.attn_gate\.weight$=q5_k
82
+ ^blk\.5\.attn_qkv\.weight$=q5_k
83
+ ^blk\.5\.ffn_down\.weight$=q5_k
84
+ ^blk\.5\.ffn_gate\.weight$=q5_k
85
+ ^blk\.5\.ffn_up\.weight$=q5_k
86
+ ^blk\.5\.ssm_out\.weight$=q5_k
87
+ ^blk\.6\.attn_gate\.weight$=q5_k
88
+ ^blk\.6\.attn_qkv\.weight$=q5_k
89
+ ^blk\.6\.ffn_down\.weight$=q5_k
90
+ ^blk\.6\.ffn_gate\.weight$=q5_k
91
+ ^blk\.6\.ffn_up\.weight$=q5_k
92
+ ^blk\.6\.ssm_out\.weight$=q5_k
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q5_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q5_k
98
+ ^blk\.7\.ffn_gate\.weight$=q5_k
99
+ ^blk\.7\.ffn_up\.weight$=q5_k
100
+ ^blk\.8\.attn_gate\.weight$=q5_k
101
+ ^blk\.8\.attn_qkv\.weight$=q5_k
102
+ ^blk\.8\.ffn_down\.weight$=q5_k
103
+ ^blk\.8\.ffn_gate\.weight$=q5_k
104
+ ^blk\.8\.ffn_up\.weight$=q5_k
105
+ ^blk\.8\.ssm_out\.weight$=q5_k
106
+ ^blk\.9\.attn_gate\.weight$=q5_k
107
+ ^blk\.9\.attn_qkv\.weight$=q5_k
108
+ ^blk\.9\.ffn_down\.weight$=q5_k
109
+ ^blk\.9\.ffn_gate\.weight$=q5_k
110
+ ^blk\.9\.ffn_up\.weight$=q5_k
111
+ ^blk\.9\.ssm_out\.weight$=q5_k
112
+ ^blk\.10\.attn_gate\.weight$=q5_k
113
+ ^blk\.10\.attn_qkv\.weight$=q5_k
114
+ ^blk\.10\.ffn_down\.weight$=q5_k
115
+ ^blk\.10\.ffn_gate\.weight$=q5_k
116
+ ^blk\.10\.ffn_up\.weight$=q5_k
117
+ ^blk\.10\.ssm_out\.weight$=q5_k
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q5_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q5_k
123
+ ^blk\.11\.ffn_gate\.weight$=q5_k
124
+ ^blk\.11\.ffn_up\.weight$=q5_k
125
+ ^blk\.12\.attn_gate\.weight$=q5_k
126
+ ^blk\.12\.attn_qkv\.weight$=q5_k
127
+ ^blk\.12\.ffn_down\.weight$=q5_k
128
+ ^blk\.12\.ffn_gate\.weight$=q5_k
129
+ ^blk\.12\.ffn_up\.weight$=q5_k
130
+ ^blk\.12\.ssm_out\.weight$=q5_k
131
+ ^blk\.13\.attn_gate\.weight$=q5_k
132
+ ^blk\.13\.attn_qkv\.weight$=q5_k
133
+ ^blk\.13\.ffn_down\.weight$=q5_k
134
+ ^blk\.13\.ffn_gate\.weight$=q5_k
135
+ ^blk\.13\.ffn_up\.weight$=q5_k
136
+ ^blk\.13\.ssm_out\.weight$=q5_k
137
+ ^blk\.14\.attn_gate\.weight$=q5_k
138
+ ^blk\.14\.attn_qkv\.weight$=q5_k
139
+ ^blk\.14\.ffn_down\.weight$=q5_k
140
+ ^blk\.14\.ffn_gate\.weight$=q5_k
141
+ ^blk\.14\.ffn_up\.weight$=q5_k
142
+ ^blk\.14\.ssm_out\.weight$=q5_k
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q5_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q5_k
148
+ ^blk\.15\.ffn_gate\.weight$=q5_k
149
+ ^blk\.15\.ffn_up\.weight$=q5_k
150
+ ^blk\.16\.attn_gate\.weight$=q5_k
151
+ ^blk\.16\.attn_qkv\.weight$=q5_k
152
+ ^blk\.16\.ffn_down\.weight$=q5_k
153
+ ^blk\.16\.ffn_gate\.weight$=q5_k
154
+ ^blk\.16\.ffn_up\.weight$=q5_k
155
+ ^blk\.16\.ssm_out\.weight$=q5_k
156
+ ^blk\.17\.attn_gate\.weight$=q5_k
157
+ ^blk\.17\.attn_qkv\.weight$=q5_k
158
+ ^blk\.17\.ffn_down\.weight$=q5_k
159
+ ^blk\.17\.ffn_gate\.weight$=q5_k
160
+ ^blk\.17\.ffn_up\.weight$=q5_k
161
+ ^blk\.17\.ssm_out\.weight$=q5_k
162
+ ^blk\.18\.attn_gate\.weight$=q5_k
163
+ ^blk\.18\.attn_qkv\.weight$=q5_k
164
+ ^blk\.18\.ffn_down\.weight$=q5_k
165
+ ^blk\.18\.ffn_gate\.weight$=q5_k
166
+ ^blk\.18\.ffn_up\.weight$=q5_k
167
+ ^blk\.18\.ssm_out\.weight$=q5_k
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q5_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q5_k
173
+ ^blk\.19\.ffn_gate\.weight$=q5_k
174
+ ^blk\.19\.ffn_up\.weight$=q5_k
175
+ ^blk\.20\.attn_gate\.weight$=q5_k
176
+ ^blk\.20\.attn_qkv\.weight$=q5_k
177
+ ^blk\.20\.ffn_down\.weight$=q5_k
178
+ ^blk\.20\.ffn_gate\.weight$=q5_k
179
+ ^blk\.20\.ffn_up\.weight$=q5_k
180
+ ^blk\.20\.ssm_out\.weight$=q5_k
181
+ ^blk\.21\.attn_gate\.weight$=q5_k
182
+ ^blk\.21\.attn_qkv\.weight$=q5_k
183
+ ^blk\.21\.ffn_down\.weight$=q5_k
184
+ ^blk\.21\.ffn_gate\.weight$=q5_k
185
+ ^blk\.21\.ffn_up\.weight$=q5_k
186
+ ^blk\.21\.ssm_out\.weight$=q5_k
187
+ ^blk\.22\.attn_gate\.weight$=q5_k
188
+ ^blk\.22\.attn_qkv\.weight$=q5_k
189
+ ^blk\.22\.ffn_down\.weight$=q5_k
190
+ ^blk\.22\.ffn_gate\.weight$=q5_k
191
+ ^blk\.22\.ffn_up\.weight$=q5_k
192
+ ^blk\.22\.ssm_out\.weight$=q5_k
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q5_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q5_k
198
+ ^blk\.23\.ffn_gate\.weight$=q5_k
199
+ ^blk\.23\.ffn_up\.weight$=q5_k
200
+ ^blk\.24\.attn_gate\.weight$=q5_k
201
+ ^blk\.24\.attn_qkv\.weight$=q5_k
202
+ ^blk\.24\.ffn_down\.weight$=q5_k
203
+ ^blk\.24\.ffn_gate\.weight$=q5_k
204
+ ^blk\.24\.ffn_up\.weight$=q5_k
205
+ ^blk\.24\.ssm_out\.weight$=q5_k
206
+ ^blk\.25\.attn_gate\.weight$=q5_k
207
+ ^blk\.25\.attn_qkv\.weight$=q5_k
208
+ ^blk\.25\.ffn_down\.weight$=q5_k
209
+ ^blk\.25\.ffn_gate\.weight$=q5_k
210
+ ^blk\.25\.ffn_up\.weight$=q5_k
211
+ ^blk\.25\.ssm_out\.weight$=q5_k
212
+ ^blk\.26\.attn_gate\.weight$=q5_k
213
+ ^blk\.26\.attn_qkv\.weight$=q5_k
214
+ ^blk\.26\.ffn_down\.weight$=q5_k
215
+ ^blk\.26\.ffn_gate\.weight$=q5_k
216
+ ^blk\.26\.ffn_up\.weight$=q5_k
217
+ ^blk\.26\.ssm_out\.weight$=q5_k
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q5_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q5_k
223
+ ^blk\.27\.ffn_gate\.weight$=q5_k
224
+ ^blk\.27\.ffn_up\.weight$=q5_k
225
+ ^blk\.28\.attn_gate\.weight$=q5_k
226
+ ^blk\.28\.attn_qkv\.weight$=q5_k
227
+ ^blk\.28\.ffn_down\.weight$=q5_k
228
+ ^blk\.28\.ffn_gate\.weight$=q5_k
229
+ ^blk\.28\.ffn_up\.weight$=q5_k
230
+ ^blk\.28\.ssm_out\.weight$=q5_k
231
+ ^blk\.29\.attn_gate\.weight$=q5_k
232
+ ^blk\.29\.attn_qkv\.weight$=q5_k
233
+ ^blk\.29\.ffn_down\.weight$=q5_k
234
+ ^blk\.29\.ffn_gate\.weight$=q5_k
235
+ ^blk\.29\.ffn_up\.weight$=q5_k
236
+ ^blk\.29\.ssm_out\.weight$=q5_k
237
+ ^blk\.30\.attn_gate\.weight$=q5_k
238
+ ^blk\.30\.attn_qkv\.weight$=q5_k
239
+ ^blk\.30\.ffn_down\.weight$=q5_k
240
+ ^blk\.30\.ffn_gate\.weight$=q5_k
241
+ ^blk\.30\.ffn_up\.weight$=q5_k
242
+ ^blk\.30\.ssm_out\.weight$=q5_k
243
+ ^blk\.31\.attn_k\.weight$=q5_k
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q5_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q5_k
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-Q6_K.layout.json ADDED
@@ -0,0 +1,187 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1422,
17
+ "type": "q6_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q6_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 3646638080,
130
+ "map_body+embd": 3659335680,
131
+ "ratio": 1.0035,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.702,
135
+ "body_bpw": 7.041,
136
+ "body_bpw_heuristic": 7.013,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 3678910464,
139
+ "file_bpw_pred": 6.998,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q6_k": 23,
143
+ "q8_0": 1
144
+ },
145
+ "attn_k": {
146
+ "q8_0": 8
147
+ },
148
+ "attn_output": {
149
+ "q8_0": 8
150
+ },
151
+ "attn_q": {
152
+ "q6_k": 8
153
+ },
154
+ "attn_qkv": {
155
+ "q6_k": 23,
156
+ "q8_0": 1
157
+ },
158
+ "attn_v": {
159
+ "q8_0": 8
160
+ },
161
+ "ffn_down": {
162
+ "q6_k": 22,
163
+ "q8_0": 10
164
+ },
165
+ "ffn_gate": {
166
+ "q6_k": 31,
167
+ "q8_0": 1
168
+ },
169
+ "ffn_up": {
170
+ "q6_k": 23,
171
+ "q8_0": 9
172
+ },
173
+ "ssm_out": {
174
+ "q8_0": 24
175
+ }
176
+ },
177
+ "prior": "layout/prior.json",
178
+ "pin_rule": "floor",
179
+ "embd_cap": "heur",
180
+ "generator": "auto_quant_v2.layout",
181
+ "generator_key": "3bf8b43e20a68f0b",
182
+ "quant_type": "Q6_K",
183
+ "quantize_ftype": "Q6_K",
184
+ "llama_cpp_version": "b11627",
185
+ "unknown_kinds": {},
186
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
187
+ }
layouts/UserRepGen-4B-Q6_K.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q6_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q8_0
53
+ ^blk\.0\.ffn_gate\.weight$=q6_k
54
+ ^blk\.0\.ffn_up\.weight$=q8_0
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q6_k
57
+ ^blk\.1\.attn_qkv\.weight$=q6_k
58
+ ^blk\.1\.ffn_down\.weight$=q8_0
59
+ ^blk\.1\.ffn_gate\.weight$=q6_k
60
+ ^blk\.1\.ffn_up\.weight$=q8_0
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=q6_k
63
+ ^blk\.2\.attn_qkv\.weight$=q6_k
64
+ ^blk\.2\.ffn_down\.weight$=q8_0
65
+ ^blk\.2\.ffn_gate\.weight$=q6_k
66
+ ^blk\.2\.ffn_up\.weight$=q8_0
67
+ ^blk\.2\.ssm_out\.weight$=q8_0
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q6_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q8_0
73
+ ^blk\.3\.ffn_gate\.weight$=q6_k
74
+ ^blk\.3\.ffn_up\.weight$=q8_0
75
+ ^blk\.4\.attn_gate\.weight$=q6_k
76
+ ^blk\.4\.attn_qkv\.weight$=q6_k
77
+ ^blk\.4\.ffn_down\.weight$=q8_0
78
+ ^blk\.4\.ffn_gate\.weight$=q6_k
79
+ ^blk\.4\.ffn_up\.weight$=q8_0
80
+ ^blk\.4\.ssm_out\.weight$=q8_0
81
+ ^blk\.5\.attn_gate\.weight$=q6_k
82
+ ^blk\.5\.attn_qkv\.weight$=q6_k
83
+ ^blk\.5\.ffn_down\.weight$=q8_0
84
+ ^blk\.5\.ffn_gate\.weight$=q6_k
85
+ ^blk\.5\.ffn_up\.weight$=q8_0
86
+ ^blk\.5\.ssm_out\.weight$=q8_0
87
+ ^blk\.6\.attn_gate\.weight$=q6_k
88
+ ^blk\.6\.attn_qkv\.weight$=q6_k
89
+ ^blk\.6\.ffn_down\.weight$=q8_0
90
+ ^blk\.6\.ffn_gate\.weight$=q6_k
91
+ ^blk\.6\.ffn_up\.weight$=q8_0
92
+ ^blk\.6\.ssm_out\.weight$=q8_0
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q6_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q8_0
98
+ ^blk\.7\.ffn_gate\.weight$=q6_k
99
+ ^blk\.7\.ffn_up\.weight$=q8_0
100
+ ^blk\.8\.attn_gate\.weight$=q6_k
101
+ ^blk\.8\.attn_qkv\.weight$=q6_k
102
+ ^blk\.8\.ffn_down\.weight$=q6_k
103
+ ^blk\.8\.ffn_gate\.weight$=q6_k
104
+ ^blk\.8\.ffn_up\.weight$=q6_k
105
+ ^blk\.8\.ssm_out\.weight$=q8_0
106
+ ^blk\.9\.attn_gate\.weight$=q6_k
107
+ ^blk\.9\.attn_qkv\.weight$=q6_k
108
+ ^blk\.9\.ffn_down\.weight$=q6_k
109
+ ^blk\.9\.ffn_gate\.weight$=q6_k
110
+ ^blk\.9\.ffn_up\.weight$=q6_k
111
+ ^blk\.9\.ssm_out\.weight$=q8_0
112
+ ^blk\.10\.attn_gate\.weight$=q6_k
113
+ ^blk\.10\.attn_qkv\.weight$=q6_k
114
+ ^blk\.10\.ffn_down\.weight$=q6_k
115
+ ^blk\.10\.ffn_gate\.weight$=q6_k
116
+ ^blk\.10\.ffn_up\.weight$=q6_k
117
+ ^blk\.10\.ssm_out\.weight$=q8_0
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q6_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q6_k
123
+ ^blk\.11\.ffn_gate\.weight$=q6_k
124
+ ^blk\.11\.ffn_up\.weight$=q6_k
125
+ ^blk\.12\.attn_gate\.weight$=q6_k
126
+ ^blk\.12\.attn_qkv\.weight$=q6_k
127
+ ^blk\.12\.ffn_down\.weight$=q6_k
128
+ ^blk\.12\.ffn_gate\.weight$=q6_k
129
+ ^blk\.12\.ffn_up\.weight$=q6_k
130
+ ^blk\.12\.ssm_out\.weight$=q8_0
131
+ ^blk\.13\.attn_gate\.weight$=q6_k
132
+ ^blk\.13\.attn_qkv\.weight$=q6_k
133
+ ^blk\.13\.ffn_down\.weight$=q6_k
134
+ ^blk\.13\.ffn_gate\.weight$=q6_k
135
+ ^blk\.13\.ffn_up\.weight$=q6_k
136
+ ^blk\.13\.ssm_out\.weight$=q8_0
137
+ ^blk\.14\.attn_gate\.weight$=q6_k
138
+ ^blk\.14\.attn_qkv\.weight$=q6_k
139
+ ^blk\.14\.ffn_down\.weight$=q6_k
140
+ ^blk\.14\.ffn_gate\.weight$=q6_k
141
+ ^blk\.14\.ffn_up\.weight$=q6_k
142
+ ^blk\.14\.ssm_out\.weight$=q8_0
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q6_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q6_k
148
+ ^blk\.15\.ffn_gate\.weight$=q6_k
149
+ ^blk\.15\.ffn_up\.weight$=q6_k
150
+ ^blk\.16\.attn_gate\.weight$=q6_k
151
+ ^blk\.16\.attn_qkv\.weight$=q6_k
152
+ ^blk\.16\.ffn_down\.weight$=q6_k
153
+ ^blk\.16\.ffn_gate\.weight$=q6_k
154
+ ^blk\.16\.ffn_up\.weight$=q6_k
155
+ ^blk\.16\.ssm_out\.weight$=q8_0
156
+ ^blk\.17\.attn_gate\.weight$=q6_k
157
+ ^blk\.17\.attn_qkv\.weight$=q6_k
158
+ ^blk\.17\.ffn_down\.weight$=q6_k
159
+ ^blk\.17\.ffn_gate\.weight$=q6_k
160
+ ^blk\.17\.ffn_up\.weight$=q6_k
161
+ ^blk\.17\.ssm_out\.weight$=q8_0
162
+ ^blk\.18\.attn_gate\.weight$=q6_k
163
+ ^blk\.18\.attn_qkv\.weight$=q6_k
164
+ ^blk\.18\.ffn_down\.weight$=q6_k
165
+ ^blk\.18\.ffn_gate\.weight$=q6_k
166
+ ^blk\.18\.ffn_up\.weight$=q6_k
167
+ ^blk\.18\.ssm_out\.weight$=q8_0
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q6_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q6_k
173
+ ^blk\.19\.ffn_gate\.weight$=q6_k
174
+ ^blk\.19\.ffn_up\.weight$=q6_k
175
+ ^blk\.20\.attn_gate\.weight$=q6_k
176
+ ^blk\.20\.attn_qkv\.weight$=q6_k
177
+ ^blk\.20\.ffn_down\.weight$=q6_k
178
+ ^blk\.20\.ffn_gate\.weight$=q6_k
179
+ ^blk\.20\.ffn_up\.weight$=q6_k
180
+ ^blk\.20\.ssm_out\.weight$=q8_0
181
+ ^blk\.21\.attn_gate\.weight$=q6_k
182
+ ^blk\.21\.attn_qkv\.weight$=q6_k
183
+ ^blk\.21\.ffn_down\.weight$=q6_k
184
+ ^blk\.21\.ffn_gate\.weight$=q6_k
185
+ ^blk\.21\.ffn_up\.weight$=q6_k
186
+ ^blk\.21\.ssm_out\.weight$=q8_0
187
+ ^blk\.22\.attn_gate\.weight$=q6_k
188
+ ^blk\.22\.attn_qkv\.weight$=q6_k
189
+ ^blk\.22\.ffn_down\.weight$=q6_k
190
+ ^blk\.22\.ffn_gate\.weight$=q6_k
191
+ ^blk\.22\.ffn_up\.weight$=q6_k
192
+ ^blk\.22\.ssm_out\.weight$=q8_0
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q6_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q6_k
198
+ ^blk\.23\.ffn_gate\.weight$=q6_k
199
+ ^blk\.23\.ffn_up\.weight$=q6_k
200
+ ^blk\.24\.attn_gate\.weight$=q6_k
201
+ ^blk\.24\.attn_qkv\.weight$=q6_k
202
+ ^blk\.24\.ffn_down\.weight$=q6_k
203
+ ^blk\.24\.ffn_gate\.weight$=q6_k
204
+ ^blk\.24\.ffn_up\.weight$=q6_k
205
+ ^blk\.24\.ssm_out\.weight$=q8_0
206
+ ^blk\.25\.attn_gate\.weight$=q6_k
207
+ ^blk\.25\.attn_qkv\.weight$=q6_k
208
+ ^blk\.25\.ffn_down\.weight$=q6_k
209
+ ^blk\.25\.ffn_gate\.weight$=q6_k
210
+ ^blk\.25\.ffn_up\.weight$=q6_k
211
+ ^blk\.25\.ssm_out\.weight$=q8_0
212
+ ^blk\.26\.attn_gate\.weight$=q6_k
213
+ ^blk\.26\.attn_qkv\.weight$=q6_k
214
+ ^blk\.26\.ffn_down\.weight$=q6_k
215
+ ^blk\.26\.ffn_gate\.weight$=q6_k
216
+ ^blk\.26\.ffn_up\.weight$=q6_k
217
+ ^blk\.26\.ssm_out\.weight$=q8_0
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q6_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q8_0
223
+ ^blk\.27\.ffn_gate\.weight$=q6_k
224
+ ^blk\.27\.ffn_up\.weight$=q6_k
225
+ ^blk\.28\.attn_gate\.weight$=q6_k
226
+ ^blk\.28\.attn_qkv\.weight$=q6_k
227
+ ^blk\.28\.ffn_down\.weight$=q6_k
228
+ ^blk\.28\.ffn_gate\.weight$=q6_k
229
+ ^blk\.28\.ffn_up\.weight$=q6_k
230
+ ^blk\.28\.ssm_out\.weight$=q8_0
231
+ ^blk\.29\.attn_gate\.weight$=q6_k
232
+ ^blk\.29\.attn_qkv\.weight$=q6_k
233
+ ^blk\.29\.ffn_down\.weight$=q6_k
234
+ ^blk\.29\.ffn_gate\.weight$=q6_k
235
+ ^blk\.29\.ffn_up\.weight$=q6_k
236
+ ^blk\.29\.ssm_out\.weight$=q8_0
237
+ ^blk\.30\.attn_gate\.weight$=q6_k
238
+ ^blk\.30\.attn_qkv\.weight$=q6_k
239
+ ^blk\.30\.ffn_down\.weight$=q6_k
240
+ ^blk\.30\.ffn_gate\.weight$=q6_k
241
+ ^blk\.30\.ffn_up\.weight$=q6_k
242
+ ^blk\.30\.ssm_out\.weight$=q8_0
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q6_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q8_0
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-Q6_K_L.layout.json ADDED
@@ -0,0 +1,188 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1422,
17
+ "type": "q6_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q6_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 3646638080,
130
+ "map_body+embd": 3819960320,
131
+ "ratio": 1.0475,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.503,
135
+ "body_bpw": 7.402,
136
+ "body_bpw_heuristic": 7.013,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 3839535104,
139
+ "file_bpw_pred": 7.303,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q6_k": 18,
143
+ "q8_0": 6
144
+ },
145
+ "attn_k": {
146
+ "q8_0": 8
147
+ },
148
+ "attn_output": {
149
+ "q8_0": 8
150
+ },
151
+ "attn_q": {
152
+ "q6_k": 7,
153
+ "q8_0": 1
154
+ },
155
+ "attn_qkv": {
156
+ "q6_k": 23,
157
+ "q8_0": 1
158
+ },
159
+ "attn_v": {
160
+ "q8_0": 8
161
+ },
162
+ "ffn_down": {
163
+ "q8_0": 27,
164
+ "q6_k": 5
165
+ },
166
+ "ffn_gate": {
167
+ "q6_k": 30,
168
+ "q8_0": 2
169
+ },
170
+ "ffn_up": {
171
+ "q8_0": 16,
172
+ "q6_k": 16
173
+ },
174
+ "ssm_out": {
175
+ "q8_0": 24
176
+ }
177
+ },
178
+ "prior": "layout/prior.json",
179
+ "pin_rule": "floor",
180
+ "embd_cap": "heur",
181
+ "generator": "auto_quant_v2.layout",
182
+ "generator_key": "3bf8b43e20a68f0b",
183
+ "quant_type": "Q6_K_L",
184
+ "quantize_ftype": "Q6_K",
185
+ "llama_cpp_version": "b11627",
186
+ "unknown_kinds": {},
187
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
188
+ }
layouts/UserRepGen-4B-Q6_K_L.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q6_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q8_0
53
+ ^blk\.0\.ffn_gate\.weight$=q8_0
54
+ ^blk\.0\.ffn_up\.weight$=q8_0
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q8_0
57
+ ^blk\.1\.attn_qkv\.weight$=q6_k
58
+ ^blk\.1\.ffn_down\.weight$=q8_0
59
+ ^blk\.1\.ffn_gate\.weight$=q6_k
60
+ ^blk\.1\.ffn_up\.weight$=q8_0
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=q8_0
63
+ ^blk\.2\.attn_qkv\.weight$=q6_k
64
+ ^blk\.2\.ffn_down\.weight$=q8_0
65
+ ^blk\.2\.ffn_gate\.weight$=q6_k
66
+ ^blk\.2\.ffn_up\.weight$=q8_0
67
+ ^blk\.2\.ssm_out\.weight$=q8_0
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q6_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q8_0
73
+ ^blk\.3\.ffn_gate\.weight$=q6_k
74
+ ^blk\.3\.ffn_up\.weight$=q8_0
75
+ ^blk\.4\.attn_gate\.weight$=q8_0
76
+ ^blk\.4\.attn_qkv\.weight$=q6_k
77
+ ^blk\.4\.ffn_down\.weight$=q8_0
78
+ ^blk\.4\.ffn_gate\.weight$=q6_k
79
+ ^blk\.4\.ffn_up\.weight$=q8_0
80
+ ^blk\.4\.ssm_out\.weight$=q8_0
81
+ ^blk\.5\.attn_gate\.weight$=q8_0
82
+ ^blk\.5\.attn_qkv\.weight$=q6_k
83
+ ^blk\.5\.ffn_down\.weight$=q8_0
84
+ ^blk\.5\.ffn_gate\.weight$=q6_k
85
+ ^blk\.5\.ffn_up\.weight$=q8_0
86
+ ^blk\.5\.ssm_out\.weight$=q8_0
87
+ ^blk\.6\.attn_gate\.weight$=q8_0
88
+ ^blk\.6\.attn_qkv\.weight$=q6_k
89
+ ^blk\.6\.ffn_down\.weight$=q8_0
90
+ ^blk\.6\.ffn_gate\.weight$=q6_k
91
+ ^blk\.6\.ffn_up\.weight$=q8_0
92
+ ^blk\.6\.ssm_out\.weight$=q8_0
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q6_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q8_0
98
+ ^blk\.7\.ffn_gate\.weight$=q6_k
99
+ ^blk\.7\.ffn_up\.weight$=q8_0
100
+ ^blk\.8\.attn_gate\.weight$=q6_k
101
+ ^blk\.8\.attn_qkv\.weight$=q6_k
102
+ ^blk\.8\.ffn_down\.weight$=q8_0
103
+ ^blk\.8\.ffn_gate\.weight$=q6_k
104
+ ^blk\.8\.ffn_up\.weight$=q6_k
105
+ ^blk\.8\.ssm_out\.weight$=q8_0
106
+ ^blk\.9\.attn_gate\.weight$=q6_k
107
+ ^blk\.9\.attn_qkv\.weight$=q6_k
108
+ ^blk\.9\.ffn_down\.weight$=q8_0
109
+ ^blk\.9\.ffn_gate\.weight$=q6_k
110
+ ^blk\.9\.ffn_up\.weight$=q6_k
111
+ ^blk\.9\.ssm_out\.weight$=q8_0
112
+ ^blk\.10\.attn_gate\.weight$=q6_k
113
+ ^blk\.10\.attn_qkv\.weight$=q6_k
114
+ ^blk\.10\.ffn_down\.weight$=q8_0
115
+ ^blk\.10\.ffn_gate\.weight$=q6_k
116
+ ^blk\.10\.ffn_up\.weight$=q6_k
117
+ ^blk\.10\.ssm_out\.weight$=q8_0
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q6_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q8_0
123
+ ^blk\.11\.ffn_gate\.weight$=q6_k
124
+ ^blk\.11\.ffn_up\.weight$=q6_k
125
+ ^blk\.12\.attn_gate\.weight$=q6_k
126
+ ^blk\.12\.attn_qkv\.weight$=q6_k
127
+ ^blk\.12\.ffn_down\.weight$=q8_0
128
+ ^blk\.12\.ffn_gate\.weight$=q6_k
129
+ ^blk\.12\.ffn_up\.weight$=q6_k
130
+ ^blk\.12\.ssm_out\.weight$=q8_0
131
+ ^blk\.13\.attn_gate\.weight$=q6_k
132
+ ^blk\.13\.attn_qkv\.weight$=q6_k
133
+ ^blk\.13\.ffn_down\.weight$=q8_0
134
+ ^blk\.13\.ffn_gate\.weight$=q6_k
135
+ ^blk\.13\.ffn_up\.weight$=q6_k
136
+ ^blk\.13\.ssm_out\.weight$=q8_0
137
+ ^blk\.14\.attn_gate\.weight$=q6_k
138
+ ^blk\.14\.attn_qkv\.weight$=q6_k
139
+ ^blk\.14\.ffn_down\.weight$=q8_0
140
+ ^blk\.14\.ffn_gate\.weight$=q6_k
141
+ ^blk\.14\.ffn_up\.weight$=q6_k
142
+ ^blk\.14\.ssm_out\.weight$=q8_0
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q6_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q8_0
148
+ ^blk\.15\.ffn_gate\.weight$=q6_k
149
+ ^blk\.15\.ffn_up\.weight$=q6_k
150
+ ^blk\.16\.attn_gate\.weight$=q6_k
151
+ ^blk\.16\.attn_qkv\.weight$=q6_k
152
+ ^blk\.16\.ffn_down\.weight$=q8_0
153
+ ^blk\.16\.ffn_gate\.weight$=q6_k
154
+ ^blk\.16\.ffn_up\.weight$=q6_k
155
+ ^blk\.16\.ssm_out\.weight$=q8_0
156
+ ^blk\.17\.attn_gate\.weight$=q6_k
157
+ ^blk\.17\.attn_qkv\.weight$=q6_k
158
+ ^blk\.17\.ffn_down\.weight$=q6_k
159
+ ^blk\.17\.ffn_gate\.weight$=q6_k
160
+ ^blk\.17\.ffn_up\.weight$=q6_k
161
+ ^blk\.17\.ssm_out\.weight$=q8_0
162
+ ^blk\.18\.attn_gate\.weight$=q6_k
163
+ ^blk\.18\.attn_qkv\.weight$=q6_k
164
+ ^blk\.18\.ffn_down\.weight$=q6_k
165
+ ^blk\.18\.ffn_gate\.weight$=q6_k
166
+ ^blk\.18\.ffn_up\.weight$=q6_k
167
+ ^blk\.18\.ssm_out\.weight$=q8_0
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q6_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q8_0
173
+ ^blk\.19\.ffn_gate\.weight$=q6_k
174
+ ^blk\.19\.ffn_up\.weight$=q6_k
175
+ ^blk\.20\.attn_gate\.weight$=q6_k
176
+ ^blk\.20\.attn_qkv\.weight$=q6_k
177
+ ^blk\.20\.ffn_down\.weight$=q6_k
178
+ ^blk\.20\.ffn_gate\.weight$=q6_k
179
+ ^blk\.20\.ffn_up\.weight$=q6_k
180
+ ^blk\.20\.ssm_out\.weight$=q8_0
181
+ ^blk\.21\.attn_gate\.weight$=q6_k
182
+ ^blk\.21\.attn_qkv\.weight$=q6_k
183
+ ^blk\.21\.ffn_down\.weight$=q6_k
184
+ ^blk\.21\.ffn_gate\.weight$=q6_k
185
+ ^blk\.21\.ffn_up\.weight$=q6_k
186
+ ^blk\.21\.ssm_out\.weight$=q8_0
187
+ ^blk\.22\.attn_gate\.weight$=q6_k
188
+ ^blk\.22\.attn_qkv\.weight$=q6_k
189
+ ^blk\.22\.ffn_down\.weight$=q6_k
190
+ ^blk\.22\.ffn_gate\.weight$=q6_k
191
+ ^blk\.22\.ffn_up\.weight$=q6_k
192
+ ^blk\.22\.ssm_out\.weight$=q8_0
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q6_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q8_0
198
+ ^blk\.23\.ffn_gate\.weight$=q6_k
199
+ ^blk\.23\.ffn_up\.weight$=q6_k
200
+ ^blk\.24\.attn_gate\.weight$=q6_k
201
+ ^blk\.24\.attn_qkv\.weight$=q6_k
202
+ ^blk\.24\.ffn_down\.weight$=q8_0
203
+ ^blk\.24\.ffn_gate\.weight$=q6_k
204
+ ^blk\.24\.ffn_up\.weight$=q8_0
205
+ ^blk\.24\.ssm_out\.weight$=q8_0
206
+ ^blk\.25\.attn_gate\.weight$=q6_k
207
+ ^blk\.25\.attn_qkv\.weight$=q6_k
208
+ ^blk\.25\.ffn_down\.weight$=q8_0
209
+ ^blk\.25\.ffn_gate\.weight$=q6_k
210
+ ^blk\.25\.ffn_up\.weight$=q8_0
211
+ ^blk\.25\.ssm_out\.weight$=q8_0
212
+ ^blk\.26\.attn_gate\.weight$=q6_k
213
+ ^blk\.26\.attn_qkv\.weight$=q6_k
214
+ ^blk\.26\.ffn_down\.weight$=q8_0
215
+ ^blk\.26\.ffn_gate\.weight$=q6_k
216
+ ^blk\.26\.ffn_up\.weight$=q8_0
217
+ ^blk\.26\.ssm_out\.weight$=q8_0
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q6_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q8_0
223
+ ^blk\.27\.ffn_gate\.weight$=q6_k
224
+ ^blk\.27\.ffn_up\.weight$=q8_0
225
+ ^blk\.28\.attn_gate\.weight$=q6_k
226
+ ^blk\.28\.attn_qkv\.weight$=q6_k
227
+ ^blk\.28\.ffn_down\.weight$=q8_0
228
+ ^blk\.28\.ffn_gate\.weight$=q6_k
229
+ ^blk\.28\.ffn_up\.weight$=q8_0
230
+ ^blk\.28\.ssm_out\.weight$=q8_0
231
+ ^blk\.29\.attn_gate\.weight$=q6_k
232
+ ^blk\.29\.attn_qkv\.weight$=q6_k
233
+ ^blk\.29\.ffn_down\.weight$=q8_0
234
+ ^blk\.29\.ffn_gate\.weight$=q6_k
235
+ ^blk\.29\.ffn_up\.weight$=q8_0
236
+ ^blk\.29\.ssm_out\.weight$=q8_0
237
+ ^blk\.30\.attn_gate\.weight$=q6_k
238
+ ^blk\.30\.attn_qkv\.weight$=q6_k
239
+ ^blk\.30\.ffn_down\.weight$=q8_0
240
+ ^blk\.30\.ffn_gate\.weight$=q6_k
241
+ ^blk\.30\.ffn_up\.weight$=q8_0
242
+ ^blk\.30\.ssm_out\.weight$=q8_0
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q8_0
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q8_0
249
+ ^blk\.31\.ffn_up\.weight$=q8_0
layouts/UserRepGen-4B-Q6_K_S.layout.json ADDED
@@ -0,0 +1,187 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-bf16.gguf",
3
+ "imatrix": "/media/m2teamgroup/models/UserRepGen-4B-GGUF/UserRepGen-4B-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
+ "token_embd.weight": {
16
+ "share": 0.1422,
17
+ "type": "q6_k"
18
+ }
19
+ },
20
+ "arch": "hybrid",
21
+ "tied": true,
22
+ "n_layer": 32,
23
+ "body_rule": true,
24
+ "pins": {
25
+ "^blk\\.0\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
26
+ "^blk\\.0\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
27
+ "^blk\\.1\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
28
+ "^blk\\.1\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
29
+ "^blk\\.2\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
30
+ "^blk\\.2\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
31
+ "^blk\\.4\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
32
+ "^blk\\.4\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
33
+ "^blk\\.5\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
34
+ "^blk\\.5\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
35
+ "^blk\\.6\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
36
+ "^blk\\.6\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
37
+ "^blk\\.8\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
38
+ "^blk\\.8\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
39
+ "^blk\\.9\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
40
+ "^blk\\.9\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
41
+ "^blk\\.10\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
42
+ "^blk\\.10\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
43
+ "^blk\\.12\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
44
+ "^blk\\.12\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
45
+ "^blk\\.13\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
46
+ "^blk\\.13\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
47
+ "^blk\\.14\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
48
+ "^blk\\.14\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
49
+ "^blk\\.16\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
50
+ "^blk\\.16\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
51
+ "^blk\\.17\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
52
+ "^blk\\.17\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
53
+ "^blk\\.18\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
54
+ "^blk\\.18\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
55
+ "^blk\\.20\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
56
+ "^blk\\.20\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
57
+ "^blk\\.21\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
58
+ "^blk\\.21\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
59
+ "^blk\\.22\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
60
+ "^blk\\.22\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
61
+ "^blk\\.24\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
62
+ "^blk\\.24\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
63
+ "^blk\\.25\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
64
+ "^blk\\.25\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
65
+ "^blk\\.26\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
66
+ "^blk\\.26\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
67
+ "^blk\\.28\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
68
+ "^blk\\.28\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
69
+ "^blk\\.29\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
70
+ "^blk\\.29\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32",
71
+ "^blk\\.30\\.ssm_alpha\\.weight$": "tiny: 81,920 params, kept at f32",
72
+ "^blk\\.30\\.ssm_beta\\.weight$": "tiny: 81,920 params, kept at f32"
73
+ },
74
+ "shape_pins": {},
75
+ "fixed_pins": {
76
+ "blk.0.ssm_alpha.weight": "f32",
77
+ "blk.0.ssm_beta.weight": "f32",
78
+ "blk.1.ssm_alpha.weight": "f32",
79
+ "blk.1.ssm_beta.weight": "f32",
80
+ "blk.2.ssm_alpha.weight": "f32",
81
+ "blk.2.ssm_beta.weight": "f32",
82
+ "blk.4.ssm_alpha.weight": "f32",
83
+ "blk.4.ssm_beta.weight": "f32",
84
+ "blk.5.ssm_alpha.weight": "f32",
85
+ "blk.5.ssm_beta.weight": "f32",
86
+ "blk.6.ssm_alpha.weight": "f32",
87
+ "blk.6.ssm_beta.weight": "f32",
88
+ "blk.8.ssm_alpha.weight": "f32",
89
+ "blk.8.ssm_beta.weight": "f32",
90
+ "blk.9.ssm_alpha.weight": "f32",
91
+ "blk.9.ssm_beta.weight": "f32",
92
+ "blk.10.ssm_alpha.weight": "f32",
93
+ "blk.10.ssm_beta.weight": "f32",
94
+ "blk.12.ssm_alpha.weight": "f32",
95
+ "blk.12.ssm_beta.weight": "f32",
96
+ "blk.13.ssm_alpha.weight": "f32",
97
+ "blk.13.ssm_beta.weight": "f32",
98
+ "blk.14.ssm_alpha.weight": "f32",
99
+ "blk.14.ssm_beta.weight": "f32",
100
+ "blk.16.ssm_alpha.weight": "f32",
101
+ "blk.16.ssm_beta.weight": "f32",
102
+ "blk.17.ssm_alpha.weight": "f32",
103
+ "blk.17.ssm_beta.weight": "f32",
104
+ "blk.18.ssm_alpha.weight": "f32",
105
+ "blk.18.ssm_beta.weight": "f32",
106
+ "blk.20.ssm_alpha.weight": "f32",
107
+ "blk.20.ssm_beta.weight": "f32",
108
+ "blk.21.ssm_alpha.weight": "f32",
109
+ "blk.21.ssm_beta.weight": "f32",
110
+ "blk.22.ssm_alpha.weight": "f32",
111
+ "blk.22.ssm_beta.weight": "f32",
112
+ "blk.24.ssm_alpha.weight": "f32",
113
+ "blk.24.ssm_beta.weight": "f32",
114
+ "blk.25.ssm_alpha.weight": "f32",
115
+ "blk.25.ssm_beta.weight": "f32",
116
+ "blk.26.ssm_alpha.weight": "f32",
117
+ "blk.26.ssm_beta.weight": "f32",
118
+ "blk.28.ssm_alpha.weight": "f32",
119
+ "blk.28.ssm_beta.weight": "f32",
120
+ "blk.29.ssm_alpha.weight": "f32",
121
+ "blk.29.ssm_beta.weight": "f32",
122
+ "blk.30.ssm_alpha.weight": "f32",
123
+ "blk.30.ssm_beta.weight": "f32"
124
+ },
125
+ "embeddings": {
126
+ "token_embd.weight": "q6_k"
127
+ },
128
+ "bytes": {
129
+ "heuristic_body+embd": 3646638080,
130
+ "map_body+embd": 3513313280,
131
+ "ratio": 0.9634,
132
+ "note": "quantizable, unpinned tensors only; pinned/unquantized tensors are identical in both"
133
+ },
134
+ "body_base_share": 0.901,
135
+ "body_bpw": 6.714,
136
+ "body_bpw_heuristic": 7.013,
137
+ "total_params": 4205751296,
138
+ "file_bytes_pred": 3532888064,
139
+ "file_bpw_pred": 6.72,
140
+ "histogram": {
141
+ "attn_gate": {
142
+ "q6_k": 23,
143
+ "q8_0": 1
144
+ },
145
+ "attn_k": {
146
+ "q8_0": 8
147
+ },
148
+ "attn_output": {
149
+ "q8_0": 8
150
+ },
151
+ "attn_q": {
152
+ "q6_k": 8
153
+ },
154
+ "attn_qkv": {
155
+ "q6_k": 23,
156
+ "q8_0": 1
157
+ },
158
+ "attn_v": {
159
+ "q8_0": 8
160
+ },
161
+ "ffn_down": {
162
+ "q6_k": 31,
163
+ "q8_0": 1
164
+ },
165
+ "ffn_gate": {
166
+ "q6_k": 32
167
+ },
168
+ "ffn_up": {
169
+ "q6_k": 31,
170
+ "q8_0": 1
171
+ },
172
+ "ssm_out": {
173
+ "q6_k": 17,
174
+ "q8_0": 7
175
+ }
176
+ },
177
+ "prior": "layout/prior.json",
178
+ "pin_rule": "floor",
179
+ "embd_cap": "heur",
180
+ "generator": "auto_quant_v2.layout",
181
+ "generator_key": "3bf8b43e20a68f0b",
182
+ "quant_type": "Q6_K_S",
183
+ "quantize_ftype": "Q6_K",
184
+ "llama_cpp_version": "b11627",
185
+ "unknown_kinds": {},
186
+ "commit": "2d948bec8623bf6a14e732adb3400ec09ebb5e20"
187
+ }
layouts/UserRepGen-4B-Q6_K_S.tensor-types.txt ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ^token_embd\.weight$=q6_k
50
+ ^blk\.0\.attn_gate\.weight$=q8_0
51
+ ^blk\.0\.attn_qkv\.weight$=q8_0
52
+ ^blk\.0\.ffn_down\.weight$=q6_k
53
+ ^blk\.0\.ffn_gate\.weight$=q6_k
54
+ ^blk\.0\.ffn_up\.weight$=q6_k
55
+ ^blk\.0\.ssm_out\.weight$=q8_0
56
+ ^blk\.1\.attn_gate\.weight$=q6_k
57
+ ^blk\.1\.attn_qkv\.weight$=q6_k
58
+ ^blk\.1\.ffn_down\.weight$=q6_k
59
+ ^blk\.1\.ffn_gate\.weight$=q6_k
60
+ ^blk\.1\.ffn_up\.weight$=q6_k
61
+ ^blk\.1\.ssm_out\.weight$=q8_0
62
+ ^blk\.2\.attn_gate\.weight$=q6_k
63
+ ^blk\.2\.attn_qkv\.weight$=q6_k
64
+ ^blk\.2\.ffn_down\.weight$=q6_k
65
+ ^blk\.2\.ffn_gate\.weight$=q6_k
66
+ ^blk\.2\.ffn_up\.weight$=q6_k
67
+ ^blk\.2\.ssm_out\.weight$=q8_0
68
+ ^blk\.3\.attn_k\.weight$=q8_0
69
+ ^blk\.3\.attn_output\.weight$=q8_0
70
+ ^blk\.3\.attn_q\.weight$=q6_k
71
+ ^blk\.3\.attn_v\.weight$=q8_0
72
+ ^blk\.3\.ffn_down\.weight$=q6_k
73
+ ^blk\.3\.ffn_gate\.weight$=q6_k
74
+ ^blk\.3\.ffn_up\.weight$=q6_k
75
+ ^blk\.4\.attn_gate\.weight$=q6_k
76
+ ^blk\.4\.attn_qkv\.weight$=q6_k
77
+ ^blk\.4\.ffn_down\.weight$=q6_k
78
+ ^blk\.4\.ffn_gate\.weight$=q6_k
79
+ ^blk\.4\.ffn_up\.weight$=q6_k
80
+ ^blk\.4\.ssm_out\.weight$=q8_0
81
+ ^blk\.5\.attn_gate\.weight$=q6_k
82
+ ^blk\.5\.attn_qkv\.weight$=q6_k
83
+ ^blk\.5\.ffn_down\.weight$=q6_k
84
+ ^blk\.5\.ffn_gate\.weight$=q6_k
85
+ ^blk\.5\.ffn_up\.weight$=q6_k
86
+ ^blk\.5\.ssm_out\.weight$=q8_0
87
+ ^blk\.6\.attn_gate\.weight$=q6_k
88
+ ^blk\.6\.attn_qkv\.weight$=q6_k
89
+ ^blk\.6\.ffn_down\.weight$=q6_k
90
+ ^blk\.6\.ffn_gate\.weight$=q6_k
91
+ ^blk\.6\.ffn_up\.weight$=q6_k
92
+ ^blk\.6\.ssm_out\.weight$=q8_0
93
+ ^blk\.7\.attn_k\.weight$=q8_0
94
+ ^blk\.7\.attn_output\.weight$=q8_0
95
+ ^blk\.7\.attn_q\.weight$=q6_k
96
+ ^blk\.7\.attn_v\.weight$=q8_0
97
+ ^blk\.7\.ffn_down\.weight$=q6_k
98
+ ^blk\.7\.ffn_gate\.weight$=q6_k
99
+ ^blk\.7\.ffn_up\.weight$=q6_k
100
+ ^blk\.8\.attn_gate\.weight$=q6_k
101
+ ^blk\.8\.attn_qkv\.weight$=q6_k
102
+ ^blk\.8\.ffn_down\.weight$=q6_k
103
+ ^blk\.8\.ffn_gate\.weight$=q6_k
104
+ ^blk\.8\.ffn_up\.weight$=q6_k
105
+ ^blk\.8\.ssm_out\.weight$=q6_k
106
+ ^blk\.9\.attn_gate\.weight$=q6_k
107
+ ^blk\.9\.attn_qkv\.weight$=q6_k
108
+ ^blk\.9\.ffn_down\.weight$=q6_k
109
+ ^blk\.9\.ffn_gate\.weight$=q6_k
110
+ ^blk\.9\.ffn_up\.weight$=q6_k
111
+ ^blk\.9\.ssm_out\.weight$=q6_k
112
+ ^blk\.10\.attn_gate\.weight$=q6_k
113
+ ^blk\.10\.attn_qkv\.weight$=q6_k
114
+ ^blk\.10\.ffn_down\.weight$=q6_k
115
+ ^blk\.10\.ffn_gate\.weight$=q6_k
116
+ ^blk\.10\.ffn_up\.weight$=q6_k
117
+ ^blk\.10\.ssm_out\.weight$=q6_k
118
+ ^blk\.11\.attn_k\.weight$=q8_0
119
+ ^blk\.11\.attn_output\.weight$=q8_0
120
+ ^blk\.11\.attn_q\.weight$=q6_k
121
+ ^blk\.11\.attn_v\.weight$=q8_0
122
+ ^blk\.11\.ffn_down\.weight$=q6_k
123
+ ^blk\.11\.ffn_gate\.weight$=q6_k
124
+ ^blk\.11\.ffn_up\.weight$=q6_k
125
+ ^blk\.12\.attn_gate\.weight$=q6_k
126
+ ^blk\.12\.attn_qkv\.weight$=q6_k
127
+ ^blk\.12\.ffn_down\.weight$=q6_k
128
+ ^blk\.12\.ffn_gate\.weight$=q6_k
129
+ ^blk\.12\.ffn_up\.weight$=q6_k
130
+ ^blk\.12\.ssm_out\.weight$=q6_k
131
+ ^blk\.13\.attn_gate\.weight$=q6_k
132
+ ^blk\.13\.attn_qkv\.weight$=q6_k
133
+ ^blk\.13\.ffn_down\.weight$=q6_k
134
+ ^blk\.13\.ffn_gate\.weight$=q6_k
135
+ ^blk\.13\.ffn_up\.weight$=q6_k
136
+ ^blk\.13\.ssm_out\.weight$=q6_k
137
+ ^blk\.14\.attn_gate\.weight$=q6_k
138
+ ^blk\.14\.attn_qkv\.weight$=q6_k
139
+ ^blk\.14\.ffn_down\.weight$=q6_k
140
+ ^blk\.14\.ffn_gate\.weight$=q6_k
141
+ ^blk\.14\.ffn_up\.weight$=q6_k
142
+ ^blk\.14\.ssm_out\.weight$=q6_k
143
+ ^blk\.15\.attn_k\.weight$=q8_0
144
+ ^blk\.15\.attn_output\.weight$=q8_0
145
+ ^blk\.15\.attn_q\.weight$=q6_k
146
+ ^blk\.15\.attn_v\.weight$=q8_0
147
+ ^blk\.15\.ffn_down\.weight$=q6_k
148
+ ^blk\.15\.ffn_gate\.weight$=q6_k
149
+ ^blk\.15\.ffn_up\.weight$=q6_k
150
+ ^blk\.16\.attn_gate\.weight$=q6_k
151
+ ^blk\.16\.attn_qkv\.weight$=q6_k
152
+ ^blk\.16\.ffn_down\.weight$=q6_k
153
+ ^blk\.16\.ffn_gate\.weight$=q6_k
154
+ ^blk\.16\.ffn_up\.weight$=q6_k
155
+ ^blk\.16\.ssm_out\.weight$=q6_k
156
+ ^blk\.17\.attn_gate\.weight$=q6_k
157
+ ^blk\.17\.attn_qkv\.weight$=q6_k
158
+ ^blk\.17\.ffn_down\.weight$=q6_k
159
+ ^blk\.17\.ffn_gate\.weight$=q6_k
160
+ ^blk\.17\.ffn_up\.weight$=q6_k
161
+ ^blk\.17\.ssm_out\.weight$=q6_k
162
+ ^blk\.18\.attn_gate\.weight$=q6_k
163
+ ^blk\.18\.attn_qkv\.weight$=q6_k
164
+ ^blk\.18\.ffn_down\.weight$=q6_k
165
+ ^blk\.18\.ffn_gate\.weight$=q6_k
166
+ ^blk\.18\.ffn_up\.weight$=q6_k
167
+ ^blk\.18\.ssm_out\.weight$=q6_k
168
+ ^blk\.19\.attn_k\.weight$=q8_0
169
+ ^blk\.19\.attn_output\.weight$=q8_0
170
+ ^blk\.19\.attn_q\.weight$=q6_k
171
+ ^blk\.19\.attn_v\.weight$=q8_0
172
+ ^blk\.19\.ffn_down\.weight$=q6_k
173
+ ^blk\.19\.ffn_gate\.weight$=q6_k
174
+ ^blk\.19\.ffn_up\.weight$=q6_k
175
+ ^blk\.20\.attn_gate\.weight$=q6_k
176
+ ^blk\.20\.attn_qkv\.weight$=q6_k
177
+ ^blk\.20\.ffn_down\.weight$=q6_k
178
+ ^blk\.20\.ffn_gate\.weight$=q6_k
179
+ ^blk\.20\.ffn_up\.weight$=q6_k
180
+ ^blk\.20\.ssm_out\.weight$=q6_k
181
+ ^blk\.21\.attn_gate\.weight$=q6_k
182
+ ^blk\.21\.attn_qkv\.weight$=q6_k
183
+ ^blk\.21\.ffn_down\.weight$=q6_k
184
+ ^blk\.21\.ffn_gate\.weight$=q6_k
185
+ ^blk\.21\.ffn_up\.weight$=q6_k
186
+ ^blk\.21\.ssm_out\.weight$=q6_k
187
+ ^blk\.22\.attn_gate\.weight$=q6_k
188
+ ^blk\.22\.attn_qkv\.weight$=q6_k
189
+ ^blk\.22\.ffn_down\.weight$=q6_k
190
+ ^blk\.22\.ffn_gate\.weight$=q6_k
191
+ ^blk\.22\.ffn_up\.weight$=q6_k
192
+ ^blk\.22\.ssm_out\.weight$=q6_k
193
+ ^blk\.23\.attn_k\.weight$=q8_0
194
+ ^blk\.23\.attn_output\.weight$=q8_0
195
+ ^blk\.23\.attn_q\.weight$=q6_k
196
+ ^blk\.23\.attn_v\.weight$=q8_0
197
+ ^blk\.23\.ffn_down\.weight$=q6_k
198
+ ^blk\.23\.ffn_gate\.weight$=q6_k
199
+ ^blk\.23\.ffn_up\.weight$=q6_k
200
+ ^blk\.24\.attn_gate\.weight$=q6_k
201
+ ^blk\.24\.attn_qkv\.weight$=q6_k
202
+ ^blk\.24\.ffn_down\.weight$=q6_k
203
+ ^blk\.24\.ffn_gate\.weight$=q6_k
204
+ ^blk\.24\.ffn_up\.weight$=q6_k
205
+ ^blk\.24\.ssm_out\.weight$=q8_0
206
+ ^blk\.25\.attn_gate\.weight$=q6_k
207
+ ^blk\.25\.attn_qkv\.weight$=q6_k
208
+ ^blk\.25\.ffn_down\.weight$=q6_k
209
+ ^blk\.25\.ffn_gate\.weight$=q6_k
210
+ ^blk\.25\.ffn_up\.weight$=q6_k
211
+ ^blk\.25\.ssm_out\.weight$=q6_k
212
+ ^blk\.26\.attn_gate\.weight$=q6_k
213
+ ^blk\.26\.attn_qkv\.weight$=q6_k
214
+ ^blk\.26\.ffn_down\.weight$=q6_k
215
+ ^blk\.26\.ffn_gate\.weight$=q6_k
216
+ ^blk\.26\.ffn_up\.weight$=q6_k
217
+ ^blk\.26\.ssm_out\.weight$=q6_k
218
+ ^blk\.27\.attn_k\.weight$=q8_0
219
+ ^blk\.27\.attn_output\.weight$=q8_0
220
+ ^blk\.27\.attn_q\.weight$=q6_k
221
+ ^blk\.27\.attn_v\.weight$=q8_0
222
+ ^blk\.27\.ffn_down\.weight$=q6_k
223
+ ^blk\.27\.ffn_gate\.weight$=q6_k
224
+ ^blk\.27\.ffn_up\.weight$=q6_k
225
+ ^blk\.28\.attn_gate\.weight$=q6_k
226
+ ^blk\.28\.attn_qkv\.weight$=q6_k
227
+ ^blk\.28\.ffn_down\.weight$=q6_k
228
+ ^blk\.28\.ffn_gate\.weight$=q6_k
229
+ ^blk\.28\.ffn_up\.weight$=q6_k
230
+ ^blk\.28\.ssm_out\.weight$=q6_k
231
+ ^blk\.29\.attn_gate\.weight$=q6_k
232
+ ^blk\.29\.attn_qkv\.weight$=q6_k
233
+ ^blk\.29\.ffn_down\.weight$=q6_k
234
+ ^blk\.29\.ffn_gate\.weight$=q6_k
235
+ ^blk\.29\.ffn_up\.weight$=q6_k
236
+ ^blk\.29\.ssm_out\.weight$=q6_k
237
+ ^blk\.30\.attn_gate\.weight$=q6_k
238
+ ^blk\.30\.attn_qkv\.weight$=q6_k
239
+ ^blk\.30\.ffn_down\.weight$=q6_k
240
+ ^blk\.30\.ffn_gate\.weight$=q6_k
241
+ ^blk\.30\.ffn_up\.weight$=q6_k
242
+ ^blk\.30\.ssm_out\.weight$=q6_k
243
+ ^blk\.31\.attn_k\.weight$=q8_0
244
+ ^blk\.31\.attn_output\.weight$=q8_0
245
+ ^blk\.31\.attn_q\.weight$=q6_k
246
+ ^blk\.31\.attn_v\.weight$=q8_0
247
+ ^blk\.31\.ffn_down\.weight$=q8_0
248
+ ^blk\.31\.ffn_gate\.weight$=q6_k
249
+ ^blk\.31\.ffn_up\.weight$=q8_0