Update GGMLQuantizationType

This commit is contained in:
Saood Karim
2025-03-31 01:15:45 -05:00
parent 4819257ce6
commit e98daabcf1

View File

@@ -1147,47 +1147,86 @@ class PoolingType(IntEnum):
class GGMLQuantizationType(IntEnum):
F32 = 0
F16 = 1
Q4_0 = 2
Q4_1 = 3
Q5_0 = 6
Q5_1 = 7
Q8_0 = 8
Q8_1 = 9
Q2_K = 10
Q3_K = 11
Q4_K = 12
Q5_K = 13
Q6_K = 14
Q8_K = 15
IQ2_XXS = 16
IQ2_XS = 17
IQ3_XXS = 18
IQ1_S = 19
IQ4_NL = 20
IQ3_S = 21
IQ2_S = 22
IQ4_XS = 23
I8 = 24
I16 = 25
I32 = 26
I64 = 27
F64 = 28
IQ1_M = 29
BF16 = 30
Q4_0_4_4 = 31
Q4_0_4_8 = 32
Q4_0_8_8 = 33
IQ1_BN = 34,
IQ2_BN = 35,
Q8_K64 = 36,
IQ2_K = 37,
IQ3_K = 38,
IQ4_K = 39,
IQ5_K = 40,
IQ6_K = 41,
IQ2_TN = 42,
F32 = 0
F16 = 1
Q4_0 = 2
Q4_1 = 3
Q5_0 = 6
Q5_1 = 7
Q8_0 = 8
Q8_1 = 9
Q2_K = 10
Q3_K = 11
Q4_K = 12
Q5_K = 13
Q6_K = 14
Q8_K = 15
IQ2_XXS = 16
IQ2_XS = 17
IQ3_XXS = 18
IQ1_S = 19
IQ4_NL = 20
IQ3_S = 21
IQ2_S = 22
IQ4_XS = 23
I8 = 24
I16 = 25
I32 = 26
I64 = 27
F64 = 28
IQ1_M = 29
BF16 = 30
Q4_0_4_4 = 31
Q4_0_4_8 = 32
Q4_0_8_8 = 33
I2_S = 36
Q8_0_X4 = 97
Q8_1_X4 = 98
Q8_2_X4 = 99
Q6_0 = 133
IQ1_BN = 134
IQ2_BN = 135
Q8_K64 = 136
IQ2_K = 137
IQ3_K = 138
IQ4_K = 139
IQ5_K = 140
IQ6_K = 141
IQ4_KS = 144
IQ2_KS = 145
IQ4_KSS = 146
Q8_K16 = 147
Q8_K32 = 148
Q8_KR8 = 149
Q8_K128 = 150
Q8_KV = 151
Q4_0_R8 = 202
Q5_0_R4 = 206
Q8_0_R8 = 208
Q2_K_R4 = 210
Q3_K_R4 = 211
Q4_K_R4 = 212
Q5_K_R4 = 213
Q6_K_R4 = 214
IQ2_XXS_R4= 216
IQ2_XS_R4 = 217
IQ3_XXS_R4= 218
IQ1_S_R4 = 219
IQ4_NL_R4 = 220
IQ3_S_R4 = 221
IQ2_S_R4 = 222
IQ4_XS_R8 = 223
IQ1_M_R4 = 229
BF16_R16 = 230
Q6_0_R4 = 233
IQ2_BN_R4 = 335
IQ2_K_R4 = 337
IQ3_K_R4 = 338
IQ4_K_R4 = 339
IQ5_K_R4 = 340
IQ4_KS_R4 = 344
Q8_KV_R8 = 398
Q8_K_R8 = 399
class ExpertGatingFuncType(IntEnum):