tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-11 11:05:39 -04:00
Code Issues Packages Projects Releases Wiki Activity

gguf-py : simplify support for quant types (#8838)

Browse Source 
* gguf-py : use classes for quants

* convert_hf : simplify internal quantization type selection

* gguf-py : fix flake8 lint

* gguf-py : fix BF16 numpy view type

* gguf-py : remove LlamaFileTypeMap

Too specific to 'llama.cpp', and would be a maintenance burden
to keep up to date.

* gguf-py : add generic quantize and dequantize functions

The quant classes no longer need to be known,
only the target or the source type,
for 'quantize' and 'dequantize', respectively.
This commit is contained in:
compilade
2024-08-08 13:33:09 -04:00
committed by GitHub
parent afd27f01fe
commit 3a14e00366
4 changed files with 226 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
gguf-py/gguf/constants.py
View File

@@ -1146,6 +1146,9 @@ class GGMLQuantizationType(IntEnum):
F64 = 28
IQ1_M = 29
BF16 = 30
Q4_0_4_4 = 31
Q4_0_4_8 = 32
Q4_0_8_8 = 33
# TODO: add GGMLFileType from ggml_ftype in ggml.h
@@ -1158,7 +1161,7 @@ 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
@@ -1187,6 +1190,9 @@ class LlamaFileType(IntEnum):
MOSTLY_IQ4_XS = 30 # except 1d tensors
MOSTLY_IQ1_M = 31 # except 1d tensors
MOSTLY_BF16 = 32 # except 1d tensors
MOSTLY_Q4_0_4_4 = 33 # except 1d tensors
MOSTLY_Q4_0_4_8 = 34 # except 1d tensors
MOSTLY_Q4_0_8_8 = 35 # except 1d tensors
GUESSED = 1024 # not specified in the model file
@@ -1260,6 +1266,9 @@ GGML_QUANT_SIZES: dict[GGMLQuantizationType, tuple[int, int]] = {
GGMLQuantizationType.F64: (1, 8),
GGMLQuantizationType.IQ1_M: (256, QK_K // 8 + QK_K // 16 + QK_K // 32),
GGMLQuantizationType.BF16: (1, 2),
GGMLQuantizationType.Q4_0_4_4:(32, 2 + 16),
GGMLQuantizationType.Q4_0_4_8:(32, 2 + 16),
GGMLQuantizationType.Q4_0_8_8:(32, 2 + 16),
}