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

constants : remove unneeded type annotations

Browse source
This commit is contained in:
Jared Van Bortel 2023-11-09 21:03:05 -05:00
parent 5608cd8d89
commit 795dc0f048
1 changed files with 115 additions and 115 deletions
Download patch file Download diff file Expand all files Collapse all files

230
gguf-py/gguf/constants.py
View file

@ -18,59 +18,59 @@ GGUF_DEFAULT_ALIGNMENT = 32
class GeneralKeys(StrEnum): class GeneralKeys(StrEnum):
ARCHITECTURE: str = "general.architecture" ARCHITECTURE = "general.architecture"
QUANTIZATION_VERSION: str = "general.quantization_version" QUANTIZATION_VERSION = "general.quantization_version"
ALIGNMENT: str = "general.alignment" ALIGNMENT = "general.alignment"
NAME: str = "general.name" NAME = "general.name"
AUTHOR: str = "general.author" AUTHOR = "general.author"
URL: str = "general.url" URL = "general.url"
DESCRIPTION: str = "general.description" DESCRIPTION = "general.description"
LICENSE: str = "general.license" LICENSE = "general.license"
SOURCE_URL: str = "general.source.url" SOURCE_URL = "general.source.url"
SOURCE_HF_REPO: str = "general.source.huggingface.repository" SOURCE_HF_REPO = "general.source.huggingface.repository"
FILE_TYPE: str = "general.file_type" FILE_TYPE = "general.file_type"
class AttentionKeys(StrEnum): class AttentionKeys(StrEnum):
HEAD_COUNT: str = "{arch}.attention.head_count" HEAD_COUNT = "{arch}.attention.head_count"
HEAD_COUNT_KV: str = "{arch}.attention.head_count_kv" HEAD_COUNT_KV = "{arch}.attention.head_count_kv"
MAX_ALIBI_BIAS: str = "{arch}.attention.max_alibi_bias" MAX_ALIBI_BIAS = "{arch}.attention.max_alibi_bias"
CLAMP_KQV: str = "{arch}.attention.clamp_kqv" CLAMP_KQV = "{arch}.attention.clamp_kqv"
LAYERNORM_EPS: str = "{arch}.attention.layer_norm_epsilon" LAYERNORM_EPS = "{arch}.attention.layer_norm_epsilon"
LAYERNORM_RMS_EPS: str = "{arch}.attention.layer_norm_rms_epsilon" LAYERNORM_RMS_EPS = "{arch}.attention.layer_norm_rms_epsilon"
class RopeKeys(StrEnum): class RopeKeys(StrEnum):
DIMENSION_COUNT: str = "{arch}.rope.dimension_count" DIMENSION_COUNT = "{arch}.rope.dimension_count"
FREQ_BASE: str = "{arch}.rope.freq_base" FREQ_BASE = "{arch}.rope.freq_base"
SCALING_TYPE: str = "{arch}.rope.scaling.type" SCALING_TYPE = "{arch}.rope.scaling.type"
SCALING_FACTOR: str = "{arch}.rope.scaling.factor" SCALING_FACTOR = "{arch}.rope.scaling.factor"
SCALING_ORIG_CTX_LEN: str = "{arch}.rope.scaling.original_context_length" SCALING_ORIG_CTX_LEN = "{arch}.rope.scaling.original_context_length"
SCALING_FINETUNED: str = "{arch}.rope.scaling.finetuned" SCALING_FINETUNED = "{arch}.rope.scaling.finetuned"
class TokenizerKeys(StrEnum): class TokenizerKeys(StrEnum):
MODEL: str = "tokenizer.ggml.model" MODEL = "tokenizer.ggml.model"
LIST: str = "tokenizer.ggml.tokens" LIST = "tokenizer.ggml.tokens"
TOKEN_TYPE: str = "tokenizer.ggml.token_type" TOKEN_TYPE = "tokenizer.ggml.token_type"
SCORES: str = "tokenizer.ggml.scores" SCORES = "tokenizer.ggml.scores"
MERGES: str = "tokenizer.ggml.merges" MERGES = "tokenizer.ggml.merges"
BOS_ID: str = "tokenizer.ggml.bos_token_id" BOS_ID = "tokenizer.ggml.bos_token_id"
EOS_ID: str = "tokenizer.ggml.eos_token_id" EOS_ID = "tokenizer.ggml.eos_token_id"
UNK_ID: str = "tokenizer.ggml.unknown_token_id" UNK_ID = "tokenizer.ggml.unknown_token_id"
SEP_ID: str = "tokenizer.ggml.seperator_token_id" SEP_ID = "tokenizer.ggml.seperator_token_id"
PAD_ID: str = "tokenizer.ggml.padding_token_id" PAD_ID = "tokenizer.ggml.padding_token_id"
HF_JSON: str = "tokenizer.huggingface.json" HF_JSON = "tokenizer.huggingface.json"
RWKV: str = "tokenizer.rwkv.world" RWKV = "tokenizer.rwkv.world"
class LLMKeys(StrEnum): class LLMKeys(StrEnum):
CONTEXT_LENGTH: str = "{arch}.context_length" CONTEXT_LENGTH = "{arch}.context_length"
EMBEDDING_LENGTH: str = "{arch}.embedding_length" EMBEDDING_LENGTH = "{arch}.embedding_length"
BLOCK_COUNT: str = "{arch}.block_count" BLOCK_COUNT = "{arch}.block_count"
FEED_FORWARD_LENGTH: str = "{arch}.feed_forward_length" FEED_FORWARD_LENGTH = "{arch}.feed_forward_length"
USE_PARALLEL_RESIDUAL: str = "{arch}.use_parallel_residual" USE_PARALLEL_RESIDUAL = "{arch}.use_parallel_residual"
TENSOR_DATA_LAYOUT: str = "{arch}.tensor_data_layout" TENSOR_DATA_LAYOUT = "{arch}.tensor_data_layout"
class Keys: class Keys:
@ -87,42 +87,42 @@ class Keys:
class MODEL_ARCH(IntEnum): class MODEL_ARCH(IntEnum):
LLAMA: int = auto() LLAMA = auto()
FALCON: int = auto() FALCON = auto()
BAICHUAN: int = auto() BAICHUAN = auto()
GPT2: int = auto() GPT2 = auto()
GPTJ: int = auto() GPTJ = auto()
GPTNEOX: int = auto() GPTNEOX = auto()
MPT: int = auto() MPT = auto()
STARCODER: int = auto() STARCODER = auto()
PERSIMMON: int = auto() PERSIMMON = auto()
REFACT: int = auto() REFACT = auto()
BERT: int = auto() BERT = auto()
BLOOM: int = auto() BLOOM = auto()
class MODEL_TENSOR(IntEnum): class MODEL_TENSOR(IntEnum):
TOKEN_EMBD: int = auto() TOKEN_EMBD = auto()
TOKEN_EMBD_NORM: int = auto() TOKEN_EMBD_NORM = auto()
TOKEN_TYPES: int = auto() TOKEN_TYPES = auto()
POS_EMBD: int = auto() POS_EMBD = auto()
OUTPUT: int = auto() OUTPUT = auto()
OUTPUT_NORM: int = auto() OUTPUT_NORM = auto()
ROPE_FREQS: int = auto() ROPE_FREQS = auto()
ATTN_Q: int = auto() ATTN_Q = auto()
ATTN_K: int = auto() ATTN_K = auto()
ATTN_V: int = auto() ATTN_V = auto()
ATTN_QKV: int = auto() ATTN_QKV = auto()
ATTN_OUT: int = auto() ATTN_OUT = auto()
ATTN_NORM: int = auto() ATTN_NORM = auto()
ATTN_NORM_2: int = auto() ATTN_NORM_2 = auto()
ATTN_ROT_EMBD: int = auto() ATTN_ROT_EMBD = auto()
FFN_GATE: int = auto() FFN_GATE = auto()
FFN_DOWN: int = auto() FFN_DOWN = auto()
FFN_UP: int = auto() FFN_UP = auto()
FFN_NORM: int = auto() FFN_NORM = auto()
ATTN_Q_NORM: int = auto() ATTN_Q_NORM = auto()
ATTN_K_NORM: int = auto() ATTN_K_NORM = auto()
MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = { MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
@ -428,52 +428,52 @@ GGML_QUANT_SIZES = {
# Aliases for backward compatibility. # Aliases for backward compatibility.
# general # general
KEY_GENERAL_ARCHITECTURE: str = Keys.GENERAL.ARCHITECTURE KEY_GENERAL_ARCHITECTURE = Keys.GENERAL.ARCHITECTURE
KEY_GENERAL_QUANTIZATION_VERSION: str = Keys.GENERAL.QUANTIZATION_VERSION KEY_GENERAL_QUANTIZATION_VERSION = Keys.GENERAL.QUANTIZATION_VERSION
KEY_GENERAL_ALIGNMENT: str = Keys.GENERAL.ALIGNMENT KEY_GENERAL_ALIGNMENT = Keys.GENERAL.ALIGNMENT
KEY_GENERAL_NAME: str = Keys.GENERAL.NAME KEY_GENERAL_NAME = Keys.GENERAL.NAME
KEY_GENERAL_AUTHOR: str = Keys.GENERAL.AUTHOR KEY_GENERAL_AUTHOR = Keys.GENERAL.AUTHOR
KEY_GENERAL_URL: str = Keys.GENERAL.URL KEY_GENERAL_URL = Keys.GENERAL.URL
KEY_GENERAL_DESCRIPTION: str = Keys.GENERAL.DESCRIPTION KEY_GENERAL_DESCRIPTION = Keys.GENERAL.DESCRIPTION
KEY_GENERAL_LICENSE: str = Keys.GENERAL.LICENSE KEY_GENERAL_LICENSE = Keys.GENERAL.LICENSE
KEY_GENERAL_SOURCE_URL: str = Keys.GENERAL.SOURCE_URL KEY_GENERAL_SOURCE_URL = Keys.GENERAL.SOURCE_URL
KEY_GENERAL_SOURCE_HF_REPO: str = Keys.GENERAL.SOURCE_HF_REPO KEY_GENERAL_SOURCE_HF_REPO = Keys.GENERAL.SOURCE_HF_REPO
KEY_GENERAL_FILE_TYPE: str = Keys.GENERAL.FILE_TYPE KEY_GENERAL_FILE_TYPE = Keys.GENERAL.FILE_TYPE
# LLM # LLM
KEY_CONTEXT_LENGTH: str = Keys.LLM.CONTEXT_LENGTH KEY_CONTEXT_LENGTH = Keys.LLM.CONTEXT_LENGTH
KEY_EMBEDDING_LENGTH: str = Keys.LLM.EMBEDDING_LENGTH KEY_EMBEDDING_LENGTH = Keys.LLM.EMBEDDING_LENGTH
KEY_BLOCK_COUNT: str = Keys.LLM.BLOCK_COUNT KEY_BLOCK_COUNT = Keys.LLM.BLOCK_COUNT
KEY_FEED_FORWARD_LENGTH: str = Keys.LLM.FEED_FORWARD_LENGTH KEY_FEED_FORWARD_LENGTH = Keys.LLM.FEED_FORWARD_LENGTH
KEY_USE_PARALLEL_RESIDUAL: str = Keys.LLM.USE_PARALLEL_RESIDUAL KEY_USE_PARALLEL_RESIDUAL = Keys.LLM.USE_PARALLEL_RESIDUAL
KEY_TENSOR_DATA_LAYOUT: str = Keys.LLM.TENSOR_DATA_LAYOUT KEY_TENSOR_DATA_LAYOUT = Keys.LLM.TENSOR_DATA_LAYOUT
# attention # attention
KEY_ATTENTION_HEAD_COUNT: str = Keys.ATTENTION.HEAD_COUNT KEY_ATTENTION_HEAD_COUNT = Keys.ATTENTION.HEAD_COUNT
KEY_ATTENTION_HEAD_COUNT_KV: str = Keys.ATTENTION.HEAD_COUNT_KV KEY_ATTENTION_HEAD_COUNT_KV = Keys.ATTENTION.HEAD_COUNT_KV
KEY_ATTENTION_MAX_ALIBI_BIAS: str = Keys.ATTENTION.MAX_ALIBI_BIAS KEY_ATTENTION_MAX_ALIBI_BIAS = Keys.ATTENTION.MAX_ALIBI_BIAS
KEY_ATTENTION_CLAMP_KQV: str = Keys.ATTENTION.CLAMP_KQV KEY_ATTENTION_CLAMP_KQV = Keys.ATTENTION.CLAMP_KQV
KEY_ATTENTION_LAYERNORM_EPS: str = Keys.ATTENTION.LAYERNORM_EPS KEY_ATTENTION_LAYERNORM_EPS = Keys.ATTENTION.LAYERNORM_EPS
KEY_ATTENTION_LAYERNORM_RMS_EPS: str = Keys.ATTENTION.LAYERNORM_RMS_EPS KEY_ATTENTION_LAYERNORM_RMS_EPS = Keys.ATTENTION.LAYERNORM_RMS_EPS
# RoPE # RoPE
KEY_ROPE_DIMENSION_COUNT: str = Keys.ROPE.DIMENSION_COUNT KEY_ROPE_DIMENSION_COUNT = Keys.ROPE.DIMENSION_COUNT
KEY_ROPE_FREQ_BASE: str = Keys.ROPE.FREQ_BASE KEY_ROPE_FREQ_BASE = Keys.ROPE.FREQ_BASE
KEY_ROPE_SCALING_TYPE: str = Keys.ROPE.SCALING_TYPE KEY_ROPE_SCALING_TYPE = Keys.ROPE.SCALING_TYPE
KEY_ROPE_SCALING_FACTOR: str = Keys.ROPE.SCALING_FACTOR KEY_ROPE_SCALING_FACTOR = Keys.ROPE.SCALING_FACTOR
KEY_ROPE_SCALING_ORIG_CTX_LEN: str = Keys.ROPE.SCALING_ORIG_CTX_LEN KEY_ROPE_SCALING_ORIG_CTX_LEN = Keys.ROPE.SCALING_ORIG_CTX_LEN
KEY_ROPE_SCALING_FINETUNED: str = Keys.ROPE.SCALING_FINETUNED KEY_ROPE_SCALING_FINETUNED = Keys.ROPE.SCALING_FINETUNED
# tokenization # tokenization
KEY_TOKENIZER_MODEL: str = Keys.TOKENIZER.MODEL KEY_TOKENIZER_MODEL = Keys.TOKENIZER.MODEL
KEY_TOKENIZER_LIST: str = Keys.TOKENIZER.LIST KEY_TOKENIZER_LIST = Keys.TOKENIZER.LIST
KEY_TOKENIZER_TOKEN_TYPE: str = Keys.TOKENIZER.TOKEN_TYPE KEY_TOKENIZER_TOKEN_TYPE = Keys.TOKENIZER.TOKEN_TYPE
KEY_TOKENIZER_SCORES: str = Keys.TOKENIZER.SCORES KEY_TOKENIZER_SCORES = Keys.TOKENIZER.SCORES
KEY_TOKENIZER_MERGES: str = Keys.TOKENIZER.MERGES KEY_TOKENIZER_MERGES = Keys.TOKENIZER.MERGES
KEY_TOKENIZER_BOS_ID: str = Keys.TOKENIZER.BOS_ID KEY_TOKENIZER_BOS_ID = Keys.TOKENIZER.BOS_ID
KEY_TOKENIZER_EOS_ID: str = Keys.TOKENIZER.EOS_ID KEY_TOKENIZER_EOS_ID = Keys.TOKENIZER.EOS_ID
KEY_TOKENIZER_UNK_ID: str = Keys.TOKENIZER.UNK_ID KEY_TOKENIZER_UNK_ID = Keys.TOKENIZER.UNK_ID
KEY_TOKENIZER_SEP_ID: str = Keys.TOKENIZER.SEP_ID KEY_TOKENIZER_SEP_ID = Keys.TOKENIZER.SEP_ID
KEY_TOKENIZER_PAD_ID: str = Keys.TOKENIZER.PAD_ID KEY_TOKENIZER_PAD_ID = Keys.TOKENIZER.PAD_ID
KEY_TOKENIZER_HF_JSON: str = Keys.TOKENIZER.HF_JSON KEY_TOKENIZER_HF_JSON = Keys.TOKENIZER.HF_JSON
KEY_TOKENIZER_RWKV: str = Keys.TOKENIZER.RWKV KEY_TOKENIZER_RWKV = Keys.TOKENIZER.RWKV