@@ -1974,12 +1974,15 @@ enum e_model {
1974
1974
MODEL_17M,
1975
1975
MODEL_22M,
1976
1976
MODEL_33M,
1977
+ MODEL_60M,
1977
1978
MODEL_70M,
1978
1979
MODEL_109M,
1979
1980
MODEL_137M,
1980
1981
MODEL_160M,
1982
+ MODEL_220M,
1981
1983
MODEL_335M,
1982
1984
MODEL_410M,
1985
+ MODEL_770M,
1983
1986
MODEL_0_5B,
1984
1987
MODEL_1B,
1985
1988
MODEL_1_4B,
@@ -1990,6 +1993,7 @@ enum e_model {
1990
1993
MODEL_6_9B,
1991
1994
MODEL_7B,
1992
1995
MODEL_8B,
1996
+ MODEL_11B,
1993
1997
MODEL_12B,
1994
1998
MODEL_13B,
1995
1999
MODEL_14B,
@@ -4213,12 +4217,15 @@ static const char * llama_model_type_name(e_model type) {
4213
4217
case MODEL_17M: return "17M";
4214
4218
case MODEL_22M: return "22M";
4215
4219
case MODEL_33M: return "33M";
4220
+ case MODEL_60M: return "60M";
4216
4221
case MODEL_70M: return "70M";
4217
4222
case MODEL_109M: return "109M";
4218
4223
case MODEL_137M: return "137M";
4219
4224
case MODEL_160M: return "160M";
4225
+ case MODEL_220M: return "220M";
4220
4226
case MODEL_335M: return "335M";
4221
4227
case MODEL_410M: return "410M";
4228
+ case MODEL_770M: return "770M";
4222
4229
case MODEL_0_5B: return "0.5B";
4223
4230
case MODEL_1B: return "1B";
4224
4231
case MODEL_1_4B: return "1.4B";
@@ -4229,6 +4236,7 @@ static const char * llama_model_type_name(e_model type) {
4229
4236
case MODEL_6_9B: return "6.9B";
4230
4237
case MODEL_7B: return "7B";
4231
4238
case MODEL_8B: return "8B";
4239
+ case MODEL_11B: return "11B";
4232
4240
case MODEL_12B: return "12B";
4233
4241
case MODEL_13B: return "13B";
4234
4242
case MODEL_14B: return "14B";
@@ -4823,7 +4831,19 @@ static void llm_load_hparams(
4823
4831
if (ml.get_key(LLM_KV_DECODER_START_TOKEN_ID, decoder_start_token_id, false)) {
4824
4832
hparams.dec_start_token_id = decoder_start_token_id;
4825
4833
}
4826
- model.type = e_model::MODEL_UNKNOWN;
4834
+
4835
+ switch (hparams.n_layer) {
4836
+ case 6: model.type = e_model::MODEL_60M; break;
4837
+ case 12: model.type = e_model::MODEL_220M; break;
4838
+ case 24:
4839
+ switch (hparams.n_ff) {
4840
+ case 4096: model.type = e_model::MODEL_770M; break;
4841
+ case 16384: model.type = e_model::MODEL_3B; break;
4842
+ case 65536: model.type = e_model::MODEL_11B; break;
4843
+ default: model.type = e_model::MODEL_UNKNOWN;
4844
+ } break;
4845
+ default: model.type = e_model::MODEL_UNKNOWN;
4846
+ }
4827
4847
} break;
4828
4848
default: (void)0;
4829
4849
}
0 commit comments