Skip to content

Commit 2245d19

Browse files
committed
llama : add model types for various T5 variants
1 parent de1338e commit 2245d19

File tree

1 file changed

+21
-1
lines changed

1 file changed

+21
-1
lines changed

Diff for: llama.cpp

+21-1
Original file line numberDiff line numberDiff line change
@@ -1974,12 +1974,15 @@ enum e_model {
19741974
MODEL_17M,
19751975
MODEL_22M,
19761976
MODEL_33M,
1977+
MODEL_60M,
19771978
MODEL_70M,
19781979
MODEL_109M,
19791980
MODEL_137M,
19801981
MODEL_160M,
1982+
MODEL_220M,
19811983
MODEL_335M,
19821984
MODEL_410M,
1985+
MODEL_770M,
19831986
MODEL_0_5B,
19841987
MODEL_1B,
19851988
MODEL_1_4B,
@@ -1990,6 +1993,7 @@ enum e_model {
19901993
MODEL_6_9B,
19911994
MODEL_7B,
19921995
MODEL_8B,
1996+
MODEL_11B,
19931997
MODEL_12B,
19941998
MODEL_13B,
19951999
MODEL_14B,
@@ -4213,12 +4217,15 @@ static const char * llama_model_type_name(e_model type) {
42134217
case MODEL_17M: return "17M";
42144218
case MODEL_22M: return "22M";
42154219
case MODEL_33M: return "33M";
4220+
case MODEL_60M: return "60M";
42164221
case MODEL_70M: return "70M";
42174222
case MODEL_109M: return "109M";
42184223
case MODEL_137M: return "137M";
42194224
case MODEL_160M: return "160M";
4225+
case MODEL_220M: return "220M";
42204226
case MODEL_335M: return "335M";
42214227
case MODEL_410M: return "410M";
4228+
case MODEL_770M: return "770M";
42224229
case MODEL_0_5B: return "0.5B";
42234230
case MODEL_1B: return "1B";
42244231
case MODEL_1_4B: return "1.4B";
@@ -4229,6 +4236,7 @@ static const char * llama_model_type_name(e_model type) {
42294236
case MODEL_6_9B: return "6.9B";
42304237
case MODEL_7B: return "7B";
42314238
case MODEL_8B: return "8B";
4239+
case MODEL_11B: return "11B";
42324240
case MODEL_12B: return "12B";
42334241
case MODEL_13B: return "13B";
42344242
case MODEL_14B: return "14B";
@@ -4823,7 +4831,19 @@ static void llm_load_hparams(
48234831
if (ml.get_key(LLM_KV_DECODER_START_TOKEN_ID, decoder_start_token_id, false)) {
48244832
hparams.dec_start_token_id = decoder_start_token_id;
48254833
}
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+
}
48274847
} break;
48284848
default: (void)0;
48294849
}

0 commit comments

Comments
 (0)