vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into compilade/bitnet-ternary

Browse source
This commit is contained in:
Francis Couture-Harpin 2024-08-11 15:52:29 -04:00
commit d911cd1f13
138 changed files with 7065 additions and 1937 deletions
Download patch file Download diff file Expand all files Collapse all files

21
gguf-py/gguf/constants.py
View file

@ -161,6 +161,7 @@ class Keys:
SUFFIX_ID = "tokenizer.ggml.suffix_token_id"
MIDDLE_ID = "tokenizer.ggml.middle_token_id"
EOT_ID = "tokenizer.ggml.eot_token_id"
EOM_ID = "tokenizer.ggml.eom_token_id"
class Adapter:
TYPE = "adapter.type"
@ -216,6 +217,7 @@ class MODEL_ARCH(IntEnum):
CHATGLM = auto()
BITNET = auto()
T5 = auto()
T5ENCODER = auto()
JAIS = auto()
@ -343,6 +345,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.CHATGLM: "chatglm",
MODEL_ARCH.BITNET: "bitnet",
MODEL_ARCH.T5: "t5",
MODEL_ARCH.T5ENCODER: "t5encoder",
MODEL_ARCH.JAIS: "jais",
}
@ -1035,6 +1038,21 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.ENC_FFN_UP,
MODEL_TENSOR.ENC_OUTPUT_NORM,
],
MODEL_ARCH.T5ENCODER: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ENC_ATTN_NORM,
MODEL_TENSOR.ENC_ATTN_Q,
MODEL_TENSOR.ENC_ATTN_K,
MODEL_TENSOR.ENC_ATTN_V,
MODEL_TENSOR.ENC_ATTN_OUT,
MODEL_TENSOR.ENC_ATTN_REL_B,
MODEL_TENSOR.ENC_FFN_NORM,
MODEL_TENSOR.ENC_FFN_GATE,
MODEL_TENSOR.ENC_FFN_DOWN,
MODEL_TENSOR.ENC_FFN_UP,
MODEL_TENSOR.ENC_OUTPUT_NORM,
],
MODEL_ARCH.JAIS: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
@ -1162,7 +1180,7 @@ class LlamaFileType(IntEnum):
MOSTLY_F16 = 1 # except 1d tensors
MOSTLY_Q4_0 = 2 # except 1d tensors
MOSTLY_Q4_1 = 3 # except 1d tensors
MOSTLY_Q4_1_SOME_F16 = 4 # tok_embeddings.weight and output.weight are F16
# MOSTLY_Q4_1_SOME_F16 = 4 # tok_embeddings.weight and output.weight are F16
# MOSTLY_Q4_2 = 5 # support has been removed
# MOSTLY_Q4_3 = 6 # support has been removed
MOSTLY_Q8_0 = 7 # except 1d tensors
@ -1342,3 +1360,4 @@ KEY_TOKENIZER_PRIFIX_ID = Keys.Tokenizer.PREFIX_ID
KEY_TOKENIZER_SUFFIX_ID = Keys.Tokenizer.SUFFIX_ID
KEY_TOKENIZER_MIDDLE_ID = Keys.Tokenizer.MIDDLE_ID
KEY_TOKENIZER_EOT_ID = Keys.Tokenizer.EOT_ID
KEY_TOKENIZER_EOM_ID = Keys.Tokenizer.EOM_ID