CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

add max_position_embeddings

Browse Source
This commit is contained in:
Meng Zhang
2023-09-15 12:35:17 +08:00
parent 57f064d7c2
commit a17ef39792
1 changed files with 12 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
gguf-py/gguf/gguf.py
View File

@@ -42,6 +42,7 @@ KEY_BLOCK_COUNT = "{arch}.block_count"
KEY_FEED_FORWARD_LENGTH = "{arch}.feed_forward_length" KEY_FEED_FORWARD_LENGTH = "{arch}.feed_forward_length"
KEY_USE_PARALLEL_RESIDUAL = "{arch}.use_parallel_residual" KEY_USE_PARALLEL_RESIDUAL = "{arch}.use_parallel_residual"
KEY_TENSOR_DATA_LAYOUT = "{arch}.tensor_data_layout" KEY_TENSOR_DATA_LAYOUT = "{arch}.tensor_data_layout"
KEY_MAX_POSITION_EMBEDDINGS = "{arch}.max_position_embeddings"
# attention # attention
KEY_ATTENTION_HEAD_COUNT = "{arch}.attention.head_count" KEY_ATTENTION_HEAD_COUNT = "{arch}.attention.head_count"
@@ -182,8 +183,8 @@ MODEL_TENSOR_NAMES: dict[MODEL_ARCH, dict[MODEL_TENSOR, str]] = {
MODEL_TENSOR.ATTN_QKV: "blk.{bid}.attn_qkv", MODEL_TENSOR.ATTN_QKV: "blk.{bid}.attn_qkv",
MODEL_TENSOR.ATTN_OUT: "blk.{bid}.attn_output", MODEL_TENSOR.ATTN_OUT: "blk.{bid}.attn_output",
MODEL_TENSOR.FFN_NORM: "blk.{bid}.ffn_norm", MODEL_TENSOR.FFN_NORM: "blk.{bid}.ffn_norm",
MODEL_TENSOR.FFN_UP: "blk.{bid}.ffn_up",
MODEL_TENSOR.FFN_DOWN: "blk.{bid}.ffn_down", MODEL_TENSOR.FFN_DOWN: "blk.{bid}.ffn_down",
MODEL_TENSOR.FFN_UP: "blk.{bid}.ffn_up",
}, },
MODEL_ARCH.GPT2: { MODEL_ARCH.GPT2: {
# TODO # TODO
@@ -717,6 +718,10 @@ class GGUFWriter:
self.add_uint32( self.add_uint32(
KEY_EMBEDDING_LENGTH.format(arch=self.arch), length) KEY_EMBEDDING_LENGTH.format(arch=self.arch), length)
def add_max_position_embeddings(self, length: int):
self.add_uint32(
KEY_MAX_POSITION_EMBEDDINGS.format(arch=self.arch), length)
def add_block_count(self, length: int): def add_block_count(self, length: int):
self.add_uint32( self.add_uint32(
KEY_BLOCK_COUNT.format(arch=self.arch), length) KEY_BLOCK_COUNT.format(arch=self.arch), length)