File size: 39,130 Bytes
5dadba4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
main: build = 3906 (7eee341b)
main: built with Apple clang version 15.0.0 (clang-1500.3.9.4) for arm64-apple-darwin23.6.0
main: quantizing 'salamandra-2b_bf16.gguf' to './salamandra-2b_Q3_K_S.gguf' as Q3_K_S
llama_model_loader: loaded meta data with 29 key-value pairs and 219 tensors from salamandra-2b_bf16.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                         general.size_label str              = 2.3B
llama_model_loader: - kv   3:                            general.license str              = apache-2.0
llama_model_loader: - kv   4:                               general.tags arr[str,1]       = ["text-generation"]
llama_model_loader: - kv   5:                          general.languages arr[str,36]      = ["bg", "ca", "code", "cs", "cy", "da"...
llama_model_loader: - kv   6:                          llama.block_count u32              = 24
llama_model_loader: - kv   7:                       llama.context_length u32              = 8192
llama_model_loader: - kv   8:                     llama.embedding_length u32              = 2048
llama_model_loader: - kv   9:                  llama.feed_forward_length u32              = 5440
llama_model_loader: - kv  10:                 llama.attention.head_count u32              = 16
llama_model_loader: - kv  11:              llama.attention.head_count_kv u32              = 16
llama_model_loader: - kv  12:                       llama.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  13:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  14:                          general.file_type u32              = 32
llama_model_loader: - kv  15:                           llama.vocab_size u32              = 256000
llama_model_loader: - kv  16:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv  17:            tokenizer.ggml.add_space_prefix bool             = true
llama_model_loader: - kv  18:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  19:                         tokenizer.ggml.pre str              = default
llama_model_loader: - kv  20:                      tokenizer.ggml.tokens arr[str,256000]  = ["<unk>", "<s>", "</s>", "<pad>", "<|...
llama_model_loader: - kv  21:                      tokenizer.ggml.scores arr[f32,256000]  = [-1000.000000, -1000.000000, -1000.00...
llama_model_loader: - kv  22:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv  23:                tokenizer.ggml.bos_token_id u32              = 1
llama_model_loader: - kv  24:                tokenizer.ggml.eos_token_id u32              = 2
llama_model_loader: - kv  25:            tokenizer.ggml.unknown_token_id u32              = 0
llama_model_loader: - kv  26:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  27:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  28:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   49 tensors
llama_model_loader: - type bf16:  170 tensors
================================ Have weights data with 168 entries
[   1/ 219]                        output.weight - [ 2048, 256000,     1,     1], type =   bf16, size = 1000.000 MB
[   2/ 219]                    token_embd.weight - [ 2048, 256000,     1,     1], type =   bf16, 
====== llama_model_quantize_internal: did not find weights for token_embd.weight
converting to q3_K .. load_imatrix: imatrix dataset='./imatrix/oscar/imatrix-dataset.txt'
load_imatrix: loaded 168 importance matrix entries from imatrix/oscar/imatrix.dat computed on 44176 chunks
prepare_imatrix: have 168 importance matrix entries
size =  1000.00 MiB ->   214.84 MiB
[   3/ 219]               blk.0.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[   4/ 219]                blk.0.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[   5/ 219]                blk.0.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[   6/ 219]                  blk.0.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[   7/ 219]                blk.0.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[   8/ 219]                  blk.0.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[   9/ 219]             blk.0.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  10/ 219]                  blk.0.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  11/ 219]                  blk.0.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  12/ 219]               blk.1.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  13/ 219]                blk.1.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  14/ 219]                blk.1.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  15/ 219]                  blk.1.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  16/ 219]                blk.1.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  17/ 219]                  blk.1.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  18/ 219]             blk.1.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  19/ 219]                  blk.1.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  20/ 219]                  blk.1.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  21/ 219]              blk.10.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  22/ 219]               blk.10.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  23/ 219]               blk.10.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  24/ 219]                 blk.10.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  25/ 219]               blk.10.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  26/ 219]                 blk.10.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  27/ 219]            blk.10.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  28/ 219]                 blk.10.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  29/ 219]                 blk.10.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  30/ 219]              blk.11.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  31/ 219]               blk.11.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  32/ 219]               blk.11.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  33/ 219]                 blk.11.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  34/ 219]               blk.11.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  35/ 219]                 blk.11.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  36/ 219]            blk.11.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  37/ 219]                 blk.11.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  38/ 219]                 blk.11.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  39/ 219]              blk.12.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  40/ 219]               blk.12.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  41/ 219]               blk.12.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  42/ 219]                 blk.12.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  43/ 219]               blk.12.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  44/ 219]                 blk.12.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  45/ 219]            blk.12.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  46/ 219]                 blk.12.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  47/ 219]                 blk.12.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  48/ 219]              blk.13.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  49/ 219]               blk.13.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  50/ 219]               blk.13.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  51/ 219]                 blk.13.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  52/ 219]               blk.13.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  53/ 219]                 blk.13.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  54/ 219]            blk.13.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  55/ 219]                 blk.13.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  56/ 219]                 blk.13.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  57/ 219]              blk.14.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  58/ 219]               blk.14.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  59/ 219]               blk.14.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  60/ 219]                 blk.14.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  61/ 219]               blk.14.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  62/ 219]                 blk.14.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  63/ 219]            blk.14.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  64/ 219]                 blk.14.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  65/ 219]                 blk.14.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  66/ 219]              blk.15.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  67/ 219]               blk.15.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  68/ 219]               blk.15.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  69/ 219]                 blk.15.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  70/ 219]               blk.15.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  71/ 219]                 blk.15.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  72/ 219]            blk.15.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  73/ 219]                 blk.15.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  74/ 219]                 blk.15.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  75/ 219]              blk.16.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  76/ 219]               blk.16.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  77/ 219]               blk.16.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  78/ 219]                 blk.16.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  79/ 219]               blk.16.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  80/ 219]                 blk.16.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  81/ 219]            blk.16.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  82/ 219]                 blk.16.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  83/ 219]                 blk.16.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  84/ 219]              blk.17.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  85/ 219]               blk.17.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  86/ 219]               blk.17.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  87/ 219]                 blk.17.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  88/ 219]               blk.17.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  89/ 219]                 blk.17.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  90/ 219]            blk.17.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  91/ 219]                 blk.17.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  92/ 219]                 blk.17.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  93/ 219]              blk.18.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  94/ 219]               blk.18.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[  95/ 219]               blk.18.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  96/ 219]                 blk.18.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[  97/ 219]               blk.18.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[  98/ 219]                 blk.18.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[  99/ 219]            blk.18.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 100/ 219]                 blk.18.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 101/ 219]                 blk.18.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 102/ 219]              blk.19.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 103/ 219]               blk.19.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 104/ 219]               blk.19.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 105/ 219]                 blk.19.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 106/ 219]               blk.19.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 107/ 219]                 blk.19.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 108/ 219]            blk.19.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 109/ 219]                 blk.19.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 110/ 219]                 blk.19.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 111/ 219]               blk.2.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 112/ 219]                blk.2.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 113/ 219]                blk.2.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 114/ 219]                  blk.2.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 115/ 219]                blk.2.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 116/ 219]                  blk.2.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 117/ 219]             blk.2.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 118/ 219]                  blk.2.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 119/ 219]                  blk.2.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 120/ 219]              blk.20.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 121/ 219]               blk.20.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 122/ 219]               blk.20.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 123/ 219]                 blk.20.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 124/ 219]               blk.20.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 125/ 219]                 blk.20.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 126/ 219]            blk.20.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 127/ 219]                 blk.20.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 128/ 219]                 blk.20.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 129/ 219]              blk.21.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 130/ 219]               blk.21.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 131/ 219]               blk.21.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 132/ 219]                 blk.21.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 133/ 219]               blk.21.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 134/ 219]                 blk.21.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 135/ 219]            blk.21.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 136/ 219]                 blk.21.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 137/ 219]                 blk.21.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 138/ 219]              blk.22.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 139/ 219]               blk.22.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 140/ 219]               blk.22.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 141/ 219]                 blk.22.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 142/ 219]               blk.22.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 143/ 219]                 blk.22.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 144/ 219]            blk.22.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 145/ 219]                 blk.22.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 146/ 219]                 blk.22.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 147/ 219]              blk.23.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 148/ 219]               blk.23.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 149/ 219]               blk.23.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 150/ 219]                 blk.23.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 151/ 219]               blk.23.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 152/ 219]                 blk.23.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 153/ 219]            blk.23.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 154/ 219]                 blk.23.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 155/ 219]                 blk.23.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 156/ 219]               blk.3.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 157/ 219]                blk.3.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 158/ 219]                blk.3.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 159/ 219]                  blk.3.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 160/ 219]                blk.3.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 161/ 219]                  blk.3.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 162/ 219]             blk.3.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 163/ 219]                  blk.3.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 164/ 219]                  blk.3.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 165/ 219]               blk.4.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 166/ 219]                blk.4.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 167/ 219]                blk.4.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 168/ 219]                  blk.4.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 169/ 219]                blk.4.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 170/ 219]                  blk.4.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 171/ 219]             blk.4.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 172/ 219]                  blk.4.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 173/ 219]                  blk.4.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 174/ 219]               blk.5.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 175/ 219]                blk.5.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 176/ 219]                blk.5.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 177/ 219]                  blk.5.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 178/ 219]                blk.5.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 179/ 219]                  blk.5.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 180/ 219]             blk.5.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 181/ 219]                  blk.5.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 182/ 219]                  blk.5.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 183/ 219]               blk.6.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 184/ 219]                blk.6.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 185/ 219]                blk.6.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 186/ 219]                  blk.6.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 187/ 219]                blk.6.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 188/ 219]                  blk.6.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 189/ 219]             blk.6.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 190/ 219]                  blk.6.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 191/ 219]                  blk.6.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 192/ 219]               blk.7.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 193/ 219]                blk.7.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 194/ 219]                blk.7.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 195/ 219]                  blk.7.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 196/ 219]                blk.7.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 197/ 219]                  blk.7.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 198/ 219]             blk.7.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 199/ 219]                  blk.7.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 200/ 219]                  blk.7.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 201/ 219]               blk.8.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 202/ 219]                blk.8.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 203/ 219]                blk.8.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 204/ 219]                  blk.8.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 205/ 219]                blk.8.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 206/ 219]                  blk.8.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 207/ 219]             blk.8.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 208/ 219]                  blk.8.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 209/ 219]                  blk.8.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 210/ 219]               blk.9.attn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 211/ 219]                blk.9.ffn_down.weight - [ 5440,  2048,     1,     1], type =   bf16, 

llama_tensor_get_type : tensor cols 5440 x 2048 are not divisible by 256, required for q3_K - using fallback quantization iq4_nl
converting to iq4_nl .. size =    21.25 MiB ->     5.98 MiB
[ 212/ 219]                blk.9.ffn_gate.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 213/ 219]                  blk.9.ffn_up.weight - [ 2048,  5440,     1,     1], type =   bf16, converting to q3_K .. size =    21.25 MiB ->     4.57 MiB
[ 214/ 219]                blk.9.ffn_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
[ 215/ 219]                  blk.9.attn_k.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 216/ 219]             blk.9.attn_output.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 217/ 219]                  blk.9.attn_q.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 218/ 219]                  blk.9.attn_v.weight - [ 2048,  2048,     1,     1], type =   bf16, converting to q3_K .. size =     8.00 MiB ->     1.72 MiB
[ 219/ 219]                   output_norm.weight - [ 2048,     1,     1,     1], type =    f32, size =    0.008 MB
llama_model_quantize_internal: model size  =  4298.38 MB
llama_model_quantize_internal: quant size  =  1742.80 MB
llama_model_quantize_internal: WARNING: 24 of 169 tensor(s) required fallback quantization

main: quantize time =  6724.72 ms
main:    total time =  6724.72 ms