tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-20 06:36:48 -04:00
Code Issues Packages Projects Releases Wiki Activity

Add support for encoder-only T5 models (#8900)

Browse Source 
* gguf-py : add T5ENCODER model architecture

* common : call llama_decode() during warmup only if the model has decoder

* convert-hf : add T5EncoderModel

* llama : add llama_model_has_decoder() API function

* llama : split build_t5() into build_t5_encoder() and build_t5_decoder()

* llama : add support for LLM_ARCH_T5ENCODER

* llama-embedding : add support for LLAMA_POOLING_TYPE_NONE

* llama-embedding : add support for encoder-only models

---------

Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com>
This commit is contained in:
fairydreaming
2024-08-10 11:43:26 +02:00
committed by GitHub
parent 911b437f22
commit 7c3f55c100
6 changed files with 702 additions and 335 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
gguf-py/gguf/constants.py
View File

@@ -217,6 +217,7 @@ class MODEL_ARCH(IntEnum):
CHATGLM = auto()
BITNET = auto()
T5 = auto()
T5ENCODER = auto()
JAIS = auto()
@@ -344,6 +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",
}
@@ -1036,6 +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,