2025-summer/team-10
10
Fork 0
Code Issues Pull requests Projects Releases Activity
team-10/env/Lib/site-packages/transformers/utils/quantization_config.py
Nicolò dbf492898c integrata generazione immagini
2025-08-02 07:34:44 +02:00

2050 lines
91 KiB
Python
Raw Permalink Blame History

#!/usr/bin/env python
# coding=utf-8
# Copyright 2023 The HuggingFace Inc. team. All rights reserved.
# Modifications Copyright (C) 2025, Advanced Micro Devices, Inc. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import copy
import dataclasses
import importlib.metadata
import json
import os
from dataclasses import dataclass, is_dataclass
from enum import Enum
from inspect import Parameter, signature
from typing import Any, Optional, Union
from packaging import version
from ..utils import (
is_auto_awq_available,
is_compressed_tensors_available,
is_gptqmodel_available,
is_hqq_available,
is_quark_available,
is_torch_available,
is_torchao_available,
logging,
)
from .import_utils import is_auto_gptq_available
if is_torch_available():
import torch
logger = logging.get_logger(__name__)
class QuantizationMethod(str, Enum):
BITS_AND_BYTES = "bitsandbytes"
GPTQ = "gptq"
AWQ = "awq"
AQLM = "aqlm"
VPTQ = "vptq"
QUANTO = "quanto"
EETQ = "eetq"
HIGGS = "higgs"
HQQ = "hqq"
COMPRESSED_TENSORS = "compressed-tensors"
FBGEMM_FP8 = "fbgemm_fp8"
TORCHAO = "torchao"
BITNET = "bitnet"
SPQR = "spqr"
FP8 = "fp8"
QUARK = "quark"
FPQUANT = "fp_quant"
AUTOROUND = "auto-round"
class AWQLinearVersion(str, Enum):
GEMM = "gemm"
GEMV = "gemv"
EXLLAMA = "exllama"
IPEX = "ipex"
@staticmethod
def from_str(version: str):
version = version.lower()
if version == "gemm":
return AWQLinearVersion.GEMM
elif version == "gemv":
return AWQLinearVersion.GEMV
elif version == "exllama":
return AWQLinearVersion.EXLLAMA
elif version == "ipex":
return AWQLinearVersion.IPEX
else:
raise ValueError(f"Unknown AWQLinearVersion {version}")
class AwqBackendPackingMethod(str, Enum):
AUTOAWQ = "autoawq"
LLMAWQ = "llm-awq"
@dataclass
class QuantizationConfigMixin:
"""
Mixin class for quantization config
"""
quant_method: QuantizationMethod
@classmethod
def from_dict(cls, config_dict, return_unused_kwargs=False, **kwargs):
"""
Instantiates a [`QuantizationConfigMixin`] from a Python dictionary of parameters.
Args:
config_dict (`dict[str, Any]`):
Dictionary that will be used to instantiate the configuration object.
return_unused_kwargs (`bool`,*optional*, defaults to `False`):
Whether or not to return a list of unused keyword arguments. Used for `from_pretrained` method in
`PreTrainedModel`.
kwargs (`dict[str, Any]`):
Additional parameters from which to initialize the configuration object.
Returns:
[`QuantizationConfigMixin`]: The configuration object instantiated from those parameters.
"""
config = cls(**config_dict)
to_remove = []
for key, value in kwargs.items():
if hasattr(config, key):
setattr(config, key, value)
to_remove.append(key)
for key in to_remove:
kwargs.pop(key, None)
if return_unused_kwargs:
return config, kwargs
else:
return config
def to_json_file(self, json_file_path: Union[str, os.PathLike]):
"""
Save this instance to a JSON file.
Args:
json_file_path (`str` or `os.PathLike`):
Path to the JSON file in which this configuration instance's parameters will be saved.
use_diff (`bool`, *optional*, defaults to `True`):
If set to `True`, only the difference between the config instance and the default
`QuantizationConfig()` is serialized to JSON file.
"""
with open(json_file_path, "w", encoding="utf-8") as writer:
config_dict = self.to_dict()
json_string = json.dumps(config_dict, indent=2, sort_keys=True) + "\n"
writer.write(json_string)
def to_dict(self) -> dict[str, Any]:
"""
Serializes this instance to a Python dictionary. Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance.
"""
return copy.deepcopy(self.__dict__)
def __iter__(self):
"""allows `dict(obj)` for situations where obj may be a dict or QuantizationConfigMixin"""
for attr, value in copy.deepcopy(self.__dict__).items():
yield attr, value
def __repr__(self):
return f"{self.__class__.__name__} {self.to_json_string()}"
def to_json_string(self, use_diff: bool = True) -> str:
"""
Serializes this instance to a JSON string.
Args:
use_diff (`bool`, *optional*, defaults to `True`):
If set to `True`, only the difference between the config instance and the default `PretrainedConfig()`
is serialized to JSON string.
Returns:
`str`: String containing all the attributes that make up this configuration instance in JSON format.
"""
if use_diff is True:
config_dict = self.to_diff_dict()
else:
config_dict = self.to_dict()
return json.dumps(config_dict, indent=2, sort_keys=True) + "\n"
def update(self, **kwargs):
"""
Updates attributes of this class instance with attributes from `kwargs` if they match existing attributes,
returning all the unused kwargs.
Args:
kwargs (`dict[str, Any]`):
Dictionary of attributes to tentatively update this class.
Returns:
`dict[str, Any]`: Dictionary containing all the key-value pairs that were not used to update the instance.
"""
to_remove = []
for key, value in kwargs.items():
if hasattr(self, key):
setattr(self, key, value)
to_remove.append(key)
# Remove all the attributes that were updated, without modifying the input dict
unused_kwargs = {key: value for key, value in kwargs.items() if key not in to_remove}
return unused_kwargs
@dataclass
class AutoRoundConfig(QuantizationConfigMixin):
"""This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded AutoRound quantization.
Args:
bits (`int`, *optional*, defaults to 4):
The number of bits to quantize to, supported numbers are (2, 3, 4, 8).
group_size (`int`, *optional*, defaults to 128): Group-size value
sym (`bool`, *optional*, defaults to `True`): Symmetric quantization or not
backend (`str`, *optional*, defaults to `"auto"`): The kernel to use, e.g., ipex,marlin, exllamav2, triton, etc. Ref. https://github.com/intel/auto-round?tab=readme-ov-file#specify-backend
"""
def __init__(
self,
bits: int = 4,
group_size: int = 128,
sym: bool = True,
backend: str = "auto",
**kwargs,
):
self.bits = bits
self.group_size = group_size
self.sym = sym
self.backend = backend
self.packing_format = "auto_round:gptq"
if kwargs is not None:
for key, value in kwargs.items():
setattr(self, key, value)
self.quant_method = QuantizationMethod.AUTOROUND
self.post_init()
def post_init(self):
r"""Safety checker that arguments are correct."""
if self.bits not in [2, 3, 4, 8]:
raise ValueError(f"Only support quantization to [2,3,4,8] bits but found {self.bits}")
if self.group_size != -1 and self.group_size <= 0:
raise ValueError("group_size must be greater than 0 or equal to -1")
def get_loading_attributes(self):
loading_attibutes_dict = {"backend": self.backend}
return loading_attibutes_dict
def to_dict(self):
config_dict = super().to_dict()
return config_dict
@classmethod
def from_dict(cls, config_dict, return_unused_kwargs=False, **kwargs):
quant_method = config_dict["quant_method"]
if "auto-round" not in quant_method and "gptq" not in quant_method and "awq" not in quant_method:
raise NotImplementedError(
"Failed to convert to auto_round format. Only `gptqv1`, `awq`, and `auto-round` formats are supported."
)
if "gptq" in quant_method and "meta" in config_dict:
raise NotImplementedError("Failed to convert gptq format to auto_round format. Only supports `gptqv1`")
if "awq" in quant_method and config_dict.get("version", "gemm") != "gemm":
raise NotImplementedError(
"Failed to convert awq format to auto_round format. Only supports awq format with gemm version"
)
if "auto-round" not in quant_method:
config_dict["packing_format"] = f"auto_round:{quant_method}"
return super().from_dict(config_dict, return_unused_kwargs=return_unused_kwargs, **kwargs)
@dataclass
class HqqConfig(QuantizationConfigMixin):
"""
This is wrapper around hqq's BaseQuantizeConfig.
Args:
nbits (`int`, *optional*, defaults to 4):
Number of bits. Supported values are (8, 4, 3, 2, 1).
group_size (`int`, *optional*, defaults to 64):
Group-size value. Supported values are any value that is divisible by weight.shape[axis]).
view_as_float (`bool`, *optional*, defaults to `False`):
View the quantized weight as float (used in distributed training) if set to `True`.
axis (`Optional[int]`, *optional*):
Axis along which grouping is performed. Supported values are 0 or 1.
dynamic_config (dict, *optional*):
Parameters for dynamic configuration. The key is the name tag of the layer and the value is a quantization config.
If set, each layer specified by its id will use its dedicated quantization configuration.
skip_modules (`list[str]`, *optional*, defaults to `['lm_head']`):
List of `nn.Linear` layers to skip.
kwargs (`dict[str, Any]`, *optional*):
Additional parameters from which to initialize the configuration object.
"""
def __init__(
self,
nbits: int = 4,
group_size: int = 64,
view_as_float: bool = False,
axis: Optional[int] = None,
dynamic_config: Optional[dict] = None,
skip_modules: list[str] = ["lm_head"],
**kwargs,
):
if is_hqq_available():
from hqq.core.quantize import BaseQuantizeConfig as HQQBaseQuantizeConfig
else:
raise ImportError(
"A valid HQQ version (>=0.2.1) is not available. Please follow the instructions to install it: `https://github.com/mobiusml/hqq/`."
)
for deprecated_key in ["quant_zero", "quant_scale", "offload_meta"]:
if deprecated_key in kwargs:
logger.info(
deprecated_key + " is deprecated. This parameter will be ignored in quantization settings."
)
if axis is None:
axis = 1
logger.info("Setting axis=1 as faster backends such as TorchAO or BitBlas are only compatible with it.")
if axis not in [0, 1]:
raise ValueError("Invalid axis value. Only 0 and 1 are allowed.")
if dynamic_config is not None:
self.quant_config = {}
for key in dynamic_config:
self.quant_config[key] = HQQBaseQuantizeConfig(**dynamic_config[key])
else:
self.quant_config = HQQBaseQuantizeConfig(
**{
"nbits": nbits,
"group_size": group_size,
"view_as_float": view_as_float,
"axis": axis,
}
)
self.quant_method = QuantizationMethod.HQQ
self.skip_modules = skip_modules
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
"""
pass
@classmethod
def from_dict(cls, config: dict[str, Any]):
"""
Override from_dict, used in AutoQuantizationConfig.from_dict in quantizers/auto.py
"""
instance = cls()
instance.quant_config = config["quant_config"]
instance.skip_modules = config["skip_modules"]
return instance
def to_dict(self) -> dict[str, Any]:
"""
Serializes this instance to a Python dictionary. Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance.
"""
return {
"quant_config": self.quant_config,
"quant_method": self.quant_method,
"skip_modules": self.skip_modules,
}
def __repr__(self):
config_dict = self.to_dict()
return f"{self.__class__.__name__} {json.dumps(config_dict, indent=2, sort_keys=True)}\n"
def to_diff_dict(self) -> dict[str, Any]:
"""
Removes all attributes from config which correspond to the default config attributes for better readability and
serializes to a Python dictionary.
Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance,
"""
config_dict = self.to_dict()
# get the default config dict
default_config_dict = HqqConfig().to_dict()
serializable_config_dict = {}
# only serialize values that differ from the default config
for key, value in config_dict.items():
if value != default_config_dict[key]:
serializable_config_dict[key] = value
return serializable_config_dict
@dataclass
class BitsAndBytesConfig(QuantizationConfigMixin):
"""
This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded using `bitsandbytes`.
This replaces `load_in_8bit` or `load_in_4bit`therefore both options are mutually exclusive.
Currently only supports `LLM.int8()`, `FP4`, and `NF4` quantization. If more methods are added to `bitsandbytes`,
then more arguments will be added to this class.
Args:
load_in_8bit (`bool`, *optional*, defaults to `False`):
This flag is used to enable 8-bit quantization with LLM.int8().
load_in_4bit (`bool`, *optional*, defaults to `False`):
This flag is used to enable 4-bit quantization by replacing the Linear layers with FP4/NF4 layers from
`bitsandbytes`.
llm_int8_threshold (`float`, *optional*, defaults to 6.0):
This corresponds to the outlier threshold for outlier detection as described in `LLM.int8() : 8-bit Matrix
Multiplication for Transformers at Scale` paper: https://huggingface.co/papers/2208.07339 Any hidden states value
that is above this threshold will be considered an outlier and the operation on those values will be done
in fp16. Values are usually normally distributed, that is, most values are in the range [-3.5, 3.5], but
there are some exceptional systematic outliers that are very differently distributed for large models.
These outliers are often in the interval [-60, -6] or [6, 60]. Int8 quantization works well for values of
magnitude ~5, but beyond that, there is a significant performance penalty. A good default threshold is 6,
but a lower threshold might be needed for more unstable models (small models, fine-tuning).
llm_int8_skip_modules (`list[str]`, *optional*):
An explicit list of the modules that we do not want to convert in 8-bit. This is useful for models such as
Jukebox that has several heads in different places and not necessarily at the last position. For example
for `CausalLM` models, the last `lm_head` is kept in its original `dtype`.
llm_int8_enable_fp32_cpu_offload (`bool`, *optional*, defaults to `False`):
This flag is used for advanced use cases and users that are aware of this feature. If you want to split
your model in different parts and run some parts in int8 on GPU and some parts in fp32 on CPU, you can use
this flag. This is useful for offloading large models such as `google/flan-t5-xxl`. Note that the int8
operations will not be run on CPU.
llm_int8_has_fp16_weight (`bool`, *optional*, defaults to `False`):
This flag runs LLM.int8() with 16-bit main weights. This is useful for fine-tuning as the weights do not
have to be converted back and forth for the backward pass.
bnb_4bit_compute_dtype (`torch.dtype` or str, *optional*, defaults to `torch.float32`):
This sets the computational type which might be different than the input type. For example, inputs might be
fp32, but computation can be set to bf16 for speedups.
bnb_4bit_quant_type (`str`, *optional*, defaults to `"fp4"`):
This sets the quantization data type in the bnb.nn.Linear4Bit layers. Options are FP4 and NF4 data types
which are specified by `fp4` or `nf4`.
bnb_4bit_use_double_quant (`bool`, *optional*, defaults to `False`):
This flag is used for nested quantization where the quantization constants from the first quantization are
quantized again.
bnb_4bit_quant_storage (`torch.dtype` or str, *optional*, defaults to `torch.uint8`):
This sets the storage type to pack the quanitzed 4-bit prarams.
kwargs (`dict[str, Any]`, *optional*):
Additional parameters from which to initialize the configuration object.
"""
def __init__(
self,
load_in_8bit=False,
load_in_4bit=False,
llm_int8_threshold=6.0,
llm_int8_skip_modules=None,
llm_int8_enable_fp32_cpu_offload=False,
llm_int8_has_fp16_weight=False,
bnb_4bit_compute_dtype=None,
bnb_4bit_quant_type="fp4",
bnb_4bit_use_double_quant=False,
bnb_4bit_quant_storage=None,
**kwargs,
):
self.quant_method = QuantizationMethod.BITS_AND_BYTES
if load_in_4bit and load_in_8bit:
raise ValueError("load_in_4bit and load_in_8bit are both True, but only one can be used at the same time")
self._load_in_8bit = load_in_8bit
self._load_in_4bit = load_in_4bit
self.llm_int8_threshold = llm_int8_threshold
self.llm_int8_skip_modules = llm_int8_skip_modules
self.llm_int8_enable_fp32_cpu_offload = llm_int8_enable_fp32_cpu_offload
self.llm_int8_has_fp16_weight = llm_int8_has_fp16_weight
self.bnb_4bit_quant_type = bnb_4bit_quant_type
self.bnb_4bit_use_double_quant = bnb_4bit_use_double_quant
if bnb_4bit_compute_dtype is None:
self.bnb_4bit_compute_dtype = torch.float32
elif isinstance(bnb_4bit_compute_dtype, str):
self.bnb_4bit_compute_dtype = getattr(torch, bnb_4bit_compute_dtype)
elif isinstance(bnb_4bit_compute_dtype, torch.dtype):
self.bnb_4bit_compute_dtype = bnb_4bit_compute_dtype
else:
raise ValueError("bnb_4bit_compute_dtype must be a string or a torch.dtype")
if bnb_4bit_quant_storage is None:
self.bnb_4bit_quant_storage = torch.uint8
elif isinstance(bnb_4bit_quant_storage, str):
if bnb_4bit_quant_storage not in ["float16", "float32", "int8", "uint8", "float64", "bfloat16"]:
raise ValueError(
"`bnb_4bit_quant_storage` must be a valid string (one of 'float16', 'float32', 'int8', 'uint8', 'float64', 'bfloat16') "
)
self.bnb_4bit_quant_storage = getattr(torch, bnb_4bit_quant_storage)
elif isinstance(bnb_4bit_quant_storage, torch.dtype):
self.bnb_4bit_quant_storage = bnb_4bit_quant_storage
else:
raise ValueError("bnb_4bit_quant_storage must be a string or a torch.dtype")
if kwargs:
logger.info(f"Unused kwargs: {list(kwargs.keys())}. These kwargs are not used in {self.__class__}.")
self.post_init()
@property
def load_in_4bit(self):
return self._load_in_4bit
@load_in_4bit.setter
def load_in_4bit(self, value: bool):
if not isinstance(value, bool):
raise TypeError("load_in_4bit must be a boolean")
if self.load_in_8bit and value:
raise ValueError("load_in_4bit and load_in_8bit are both True, but only one can be used at the same time")
self._load_in_4bit = value
@property
def load_in_8bit(self):
return self._load_in_8bit
@load_in_8bit.setter
def load_in_8bit(self, value: bool):
if not isinstance(value, bool):
raise TypeError("load_in_8bit must be a boolean")
if self.load_in_4bit and value:
raise ValueError("load_in_4bit and load_in_8bit are both True, but only one can be used at the same time")
self._load_in_8bit = value
def post_init(self):
r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
"""
if not isinstance(self.load_in_4bit, bool):
raise TypeError("load_in_4bit must be a boolean")
if not isinstance(self.load_in_8bit, bool):
raise TypeError("load_in_8bit must be a boolean")
if not isinstance(self.llm_int8_threshold, float):
raise TypeError("llm_int8_threshold must be a float")
if self.llm_int8_skip_modules is not None and not isinstance(self.llm_int8_skip_modules, list):
raise TypeError("llm_int8_skip_modules must be a list of strings")
if not isinstance(self.llm_int8_enable_fp32_cpu_offload, bool):
raise TypeError("llm_int8_enable_fp32_cpu_offload must be a boolean")
if not isinstance(self.llm_int8_has_fp16_weight, bool):
raise TypeError("llm_int8_has_fp16_weight must be a boolean")
if self.bnb_4bit_compute_dtype is not None and not isinstance(self.bnb_4bit_compute_dtype, torch.dtype):
raise TypeError("bnb_4bit_compute_dtype must be torch.dtype")
if not isinstance(self.bnb_4bit_quant_type, str):
raise TypeError("bnb_4bit_quant_type must be a string")
if not isinstance(self.bnb_4bit_use_double_quant, bool):
raise TypeError("bnb_4bit_use_double_quant must be a boolean")
if self.load_in_4bit and not version.parse(importlib.metadata.version("bitsandbytes")) >= version.parse(
"0.39.0"
):
raise ValueError(
"4 bit quantization requires bitsandbytes>=0.39.0 - please upgrade your bitsandbytes version"
)
def is_quantizable(self):
r"""
Returns `True` if the model is quantizable, `False` otherwise.
"""
return self.load_in_8bit or self.load_in_4bit
def quantization_method(self):
r"""
This method returns the quantization method used for the model. If the model is not quantizable, it returns
`None`.
"""
if self.load_in_8bit:
return "llm_int8"
elif self.load_in_4bit and self.bnb_4bit_quant_type == "fp4":
return "fp4"
elif self.load_in_4bit and self.bnb_4bit_quant_type == "nf4":
return "nf4"
else:
return None
def to_dict(self) -> dict[str, Any]:
"""
Serializes this instance to a Python dictionary. Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance.
"""
output = copy.deepcopy(self.__dict__)
output["bnb_4bit_compute_dtype"] = str(output["bnb_4bit_compute_dtype"]).split(".")[1]
output["bnb_4bit_quant_storage"] = str(output["bnb_4bit_quant_storage"]).split(".")[1]
output["load_in_4bit"] = self.load_in_4bit
output["load_in_8bit"] = self.load_in_8bit
return output
def __repr__(self):
config_dict = self.to_dict()
return f"{self.__class__.__name__} {json.dumps(config_dict, indent=2, sort_keys=True)}\n"
def to_diff_dict(self) -> dict[str, Any]:
"""
Removes all attributes from config which correspond to the default config attributes for better readability and
serializes to a Python dictionary.
Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance,
"""
config_dict = self.to_dict()
# get the default config dict
default_config_dict = BitsAndBytesConfig().to_dict()
serializable_config_dict = {}
# only serialize values that differ from the default config
for key, value in config_dict.items():
if value != default_config_dict[key]:
serializable_config_dict[key] = value
return serializable_config_dict
class ExllamaVersion(int, Enum):
ONE = 1
TWO = 2
@dataclass
class GPTQConfig(QuantizationConfigMixin):
"""
This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded using `optimum` api for gptq quantization relying on auto_gptq backend.
Args:
bits (`int`):
The number of bits to quantize to, supported numbers are (2, 3, 4, 8).
tokenizer (`str` or `PreTrainedTokenizerBase`, *optional*):
The tokenizer used to process the dataset. You can pass either:
- A custom tokenizer object.
- A string, the *model id* of a predefined tokenizer hosted inside a model repo on huggingface.co.
- A path to a *directory* containing vocabulary files required by the tokenizer, for instance saved
using the [`~PreTrainedTokenizer.save_pretrained`] method, e.g., `./my_model_directory/`.
dataset (`Union[list[str]]`, *optional*):
The dataset used for quantization. You can provide your own dataset in a list of string or just use the
original datasets used in GPTQ paper ['wikitext2','c4','c4-new']
group_size (`int`, *optional*, defaults to 128):
The group size to use for quantization. Recommended value is 128 and -1 uses per-column quantization.
damp_percent (`float`, *optional*, defaults to 0.1):
The percent of the average Hessian diagonal to use for dampening. Recommended value is 0.1.
desc_act (`bool`, *optional*, defaults to `False`):
Whether to quantize columns in order of decreasing activation size. Setting it to False can significantly
speed up inference but the perplexity may become slightly worse. Also known as act-order.
sym (`bool`, *optional*, defaults to `True`):
Whether to use symmetric quantization.
true_sequential (`bool`, *optional*, defaults to `True`):
Whether to perform sequential quantization even within a single Transformer block. Instead of quantizing
the entire block at once, we perform layer-wise quantization. As a result, each layer undergoes
quantization using inputs that have passed through the previously quantized layers.
checkpoint_format (`str`, *optional*, defaults to `"gptq"`):
GPTQ weight format. `gptq`(v1) is supported by both gptqmodel and auto-gptq. `gptq_v2` is gptqmodel only.
meta (`dict[str, any]`, *optional*):
Properties, such as tooling:version, that do not directly contributes to quantization or quant inference are stored in meta.
i.e. `meta.quantizer`: ["optimum:_version_", "gptqmodel:_version_"]
backend (`str`, *optional*):
Controls which gptq kernel to be used. Valid values for gptqmodel are `auto`, `auto_trainable` and more. For auto-gptq, only
valid value is None and `auto_trainable`. Ref gptqmodel backends: https://github.com/ModelCloud/GPTQModel/blob/main/gptqmodel/utils/backend.py
use_cuda_fp16 (`bool`, *optional*, defaults to `False`):
Whether or not to use optimized cuda kernel for fp16 model. Need to have model in fp16. Auto-gptq only.
model_seqlen (`int`, *optional*):
The maximum sequence length that the model can take.
block_name_to_quantize (`str`, *optional*):
The transformers block name to quantize. If None, we will infer the block name using common patterns (e.g. model.layers)
module_name_preceding_first_block (`list[str]`, *optional*):
The layers that are preceding the first Transformer block.
batch_size (`int`, *optional*, defaults to 1):
The batch size used when processing the dataset
pad_token_id (`int`, *optional*):
The pad token id. Needed to prepare the dataset when `batch_size` > 1.
use_exllama (`bool`, *optional*):
Whether to use exllama backend. Defaults to `True` if unset. Only works with `bits` = 4.
max_input_length (`int`, *optional*):
The maximum input length. This is needed to initialize a buffer that depends on the maximum expected input
length. It is specific to the exllama backend with act-order.
exllama_config (`dict[str, Any]`, *optional*):
The exllama config. You can specify the version of the exllama kernel through the `version` key. Defaults
to `{"version": 1}` if unset.
cache_block_outputs (`bool`, *optional*, defaults to `True`):
Whether to cache block outputs to reuse as inputs for the succeeding block.
modules_in_block_to_quantize (`list[list[str]]`, *optional*):
List of list of module names to quantize in the specified block. This argument is useful to exclude certain linear modules from being quantized.
The block to quantize can be specified by setting `block_name_to_quantize`. We will quantize each list sequentially. If not set, we will quantize all linear layers.
Example: `modules_in_block_to_quantize =[["self_attn.k_proj", "self_attn.v_proj", "self_attn.q_proj"], ["self_attn.o_proj"]]`.
In this example, we will first quantize the q,k,v layers simultaneously since they are independent.
Then, we will quantize `self_attn.o_proj` layer with the q,k,v layers quantized. This way, we will get
better results since it reflects the real input `self_attn.o_proj` will get when the model is quantized.
"""
def __init__(
self,
bits: int,
tokenizer: Any = None,
dataset: Optional[Union[list[str], str]] = None,
group_size: int = 128,
damp_percent: float = 0.1,
desc_act: bool = False,
sym: bool = True,
true_sequential: bool = True,
checkpoint_format: str = "gptq",
meta: Optional[dict[str, Any]] = None,
backend: Optional[str] = None,
use_cuda_fp16: bool = False,
model_seqlen: Optional[int] = None,
block_name_to_quantize: Optional[str] = None,
module_name_preceding_first_block: Optional[list[str]] = None,
batch_size: int = 1,
pad_token_id: Optional[int] = None,
use_exllama: Optional[bool] = None,
max_input_length: Optional[int] = None,
exllama_config: Optional[dict[str, Any]] = None,
cache_block_outputs: bool = True,
modules_in_block_to_quantize: Optional[list[list[str]]] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.GPTQ
self.bits = bits
self.tokenizer = tokenizer
self.dataset = dataset
self.group_size = group_size
self.damp_percent = damp_percent
self.desc_act = desc_act
self.sym = sym
self.true_sequential = true_sequential
self.checkpoint_format = checkpoint_format.lower()
self.meta = meta
self.backend = backend.lower() if isinstance(backend, str) else backend
self.use_cuda_fp16 = use_cuda_fp16
self.model_seqlen = model_seqlen
self.block_name_to_quantize = block_name_to_quantize
self.module_name_preceding_first_block = module_name_preceding_first_block
self.batch_size = batch_size
self.pad_token_id = pad_token_id
self.use_exllama = use_exllama
self.max_input_length = max_input_length
self.exllama_config = exllama_config
self.cache_block_outputs = cache_block_outputs
self.modules_in_block_to_quantize = modules_in_block_to_quantize
self.post_init()
def get_loading_attributes(self):
attibutes_dict = copy.deepcopy(self.__dict__)
loading_attibutes = [
"use_exllama",
"exllama_config",
"use_cuda_fp16",
"max_input_length",
"backend",
]
loading_attibutes_dict = {i: j for i, j in attibutes_dict.items() if i in loading_attibutes}
return loading_attibutes_dict
def post_init(self):
r"""
Safety checker that arguments are correct
"""
if self.bits not in [2, 3, 4, 8]:
raise ValueError(f"Only support quantization to [2,3,4,8] bits but found {self.bits}")
if self.group_size != -1 and self.group_size <= 0:
raise ValueError("group_size must be greater than 0 or equal to -1")
if not (0 < self.damp_percent < 1):
raise ValueError("damp_percent must between 0 and 1.")
if self.dataset is not None:
if isinstance(self.dataset, str):
if self.dataset in ["ptb", "ptb-new"]:
raise ValueError(
f"""{self.dataset} dataset was deprecated. You can only choose between
['wikitext2','c4','c4-new']"""
)
if self.dataset not in ["wikitext2", "c4", "c4-new"]:
raise ValueError(
f"""You have entered a string value for dataset. You can only choose between
['wikitext2','c4','c4-new'], but we found {self.dataset}"""
)
elif not isinstance(self.dataset, list):
raise ValueError(
f"""dataset needs to be either a list of string or a value in
['wikitext2','c4','c4-new'], but we found {self.dataset}"""
)
# make sure backend is back/forward compatible with both gptqmodel (full) and auto-gptq (partial)
if is_gptqmodel_available():
# convert auto-gptq control into gptqmodel backend
if self.backend is None:
self.backend = "auto_trainable" if self.use_exllama is not None and not self.use_exllama else "auto"
else:
# convert gptqmodel backend `auto_trainable` into auto-gptq control
if self.backend == "auto_trainable":
self.use_exllama = False
# auto-gptq specific kernel control logic
if self.use_exllama is None:
# New default behaviour
self.use_exllama = True
if self.exllama_config is None:
self.exllama_config = {"version": ExllamaVersion.ONE}
else:
if "version" not in self.exllama_config:
raise ValueError("`exllama_config` needs to have a `version` key.")
elif self.exllama_config["version"] not in [ExllamaVersion.ONE, ExllamaVersion.TWO]:
exllama_version = self.exllama_config["version"]
raise ValueError(
f"Only supported versions are in [ExllamaVersion.ONE, ExllamaVersion.TWO] - not recognized version {exllama_version}"
)
if self.bits == 4 and self.use_exllama:
if self.exllama_config["version"] == ExllamaVersion.ONE:
logger.info(
"You have activated exllama backend. Note that you can get better inference "
"speed using exllamav2 kernel by setting `exllama_config`."
)
elif self.exllama_config["version"] == ExllamaVersion.TWO:
if is_auto_gptq_available():
optimum_version = version.parse(importlib.metadata.version("optimum"))
autogptq_version = version.parse(importlib.metadata.version("auto_gptq"))
if optimum_version <= version.parse("1.13.2") or autogptq_version <= version.parse("0.4.2"):
raise ValueError(
f"You need optimum > 1.13.2 and auto-gptq > 0.4.2 . Make sure to have that version installed - detected version : optimum {optimum_version} and autogptq {autogptq_version}"
)
if self.modules_in_block_to_quantize is not None:
optimum_version = version.parse(importlib.metadata.version("optimum"))
if optimum_version < version.parse("1.15.0"):
raise ValueError(
"You current version of `optimum` does not support `modules_in_block_to_quantize` quantization argument, please upgrade `optimum` package to a version superior than 1.15.0 ."
)
def to_dict(self):
config_dict = super().to_dict()
config_dict.pop("disable_exllama", None)
return config_dict
def to_dict_optimum(self):
"""
Get compatible dict for optimum gptq config
"""
quant_dict = self.to_dict()
# make it compatible with optimum config
quant_dict["disable_exllama"] = not self.use_exllama
return quant_dict
@classmethod
def from_dict_optimum(cls, config_dict):
"""
Get compatible class with optimum gptq config dict
"""
if "disable_exllama" in config_dict:
config_dict["use_exllama"] = not config_dict["disable_exllama"]
# switch to None to not trigger the warning
config_dict.pop("disable_exllama")
config = cls(**config_dict)
return config
@dataclass
class AwqConfig(QuantizationConfigMixin):
"""
This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded using `auto-awq` library awq quantization relying on auto_awq backend.
Args:
bits (`int`, *optional*, defaults to 4):
The number of bits to quantize to.
group_size (`int`, *optional*, defaults to 128):
The group size to use for quantization. Recommended value is 128 and -1 uses per-column quantization.
zero_point (`bool`, *optional*, defaults to `True`):
Whether to use zero point quantization.
version (`AWQLinearVersion`, *optional*, defaults to `AWQLinearVersion.GEMM`):
The version of the quantization algorithm to use. GEMM is better for big batch_size (e.g. >= 8) otherwise,
GEMV is better (e.g. < 8 ). GEMM models are compatible with Exllama kernels.
backend (`AwqBackendPackingMethod`, *optional*, defaults to `AwqBackendPackingMethod.AUTOAWQ`):
The quantization backend. Some models might be quantized using `llm-awq` backend. This is useful for users
that quantize their own models using `llm-awq` library.
do_fuse (`bool`, *optional*, defaults to `False`):
Whether to fuse attention and mlp layers together for faster inference
fuse_max_seq_len (`int`, *optional*):
The Maximum sequence length to generate when using fusing.
modules_to_fuse (`dict`, *optional*, default to `None`):
Overwrite the natively supported fusing scheme with the one specified by the users.
modules_to_not_convert (`list`, *optional*, default to `None`):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision (e.g. Whisper encoder, Llava encoder, Mixtral gate layers).
Note you cannot quantize directly with transformers, please refer to `AutoAWQ` documentation for quantizing HF models.
exllama_config (`dict[str, Any]`, *optional*):
You can specify the version of the exllama kernel through the `version` key, the maximum sequence
length through the `max_input_len` key, and the maximum batch size through the `max_batch_size` key.
Defaults to `{"version": 2, "max_input_len": 2048, "max_batch_size": 8}` if unset.
"""
def __init__(
self,
bits: int = 4,
group_size: int = 128,
zero_point: bool = True,
version: AWQLinearVersion = AWQLinearVersion.GEMM,
backend: AwqBackendPackingMethod = AwqBackendPackingMethod.AUTOAWQ,
do_fuse: Optional[bool] = None,
fuse_max_seq_len: Optional[int] = None,
modules_to_fuse: Optional[dict] = None,
modules_to_not_convert: Optional[list] = None,
exllama_config: Optional[dict[str, int]] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.AWQ
self.bits = bits
self.group_size = group_size
self.zero_point = zero_point
self.version = version
self.backend = backend
self.fuse_max_seq_len = fuse_max_seq_len
self.modules_to_not_convert = modules_to_not_convert
self.exllama_config = exllama_config
self.modules_to_fuse = modules_to_fuse
if do_fuse is None:
self.do_fuse = modules_to_fuse is not None and len(modules_to_fuse) > 0
else:
self.do_fuse = do_fuse
self.fuse_max_seq_len = fuse_max_seq_len
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
if self.backend not in [AwqBackendPackingMethod.AUTOAWQ, AwqBackendPackingMethod.LLMAWQ]:
raise ValueError(
f"Only supported quantization backends in {AwqBackendPackingMethod.AUTOAWQ} and {AwqBackendPackingMethod.LLMAWQ} - not recognized backend {self.backend}"
)
self.version = AWQLinearVersion.from_str(self.version)
if self.version not in [
AWQLinearVersion.GEMM,
AWQLinearVersion.GEMV,
AWQLinearVersion.EXLLAMA,
AWQLinearVersion.IPEX,
]:
raise ValueError(
f"Only supported versions are in [AWQLinearVersion.GEMM, AWQLinearVersion.GEMV, AWQLinearVersion.EXLLAMA, AWQLinearVersion.IPEX] - not recognized version {self.version}"
)
if self.backend == AwqBackendPackingMethod.LLMAWQ:
# Only cuda device can run this function
if not (torch.cuda.is_available() or torch.xpu.is_available()):
raise ValueError("LLM-AWQ backend is only supported on CUDA and XPU")
if torch.cuda.is_available():
compute_capability = torch.cuda.get_device_capability()
major, minor = compute_capability
if major < 8:
raise ValueError("LLM-AWQ backend is only supported on CUDA GPUs with compute capability >= 8.0")
if self.do_fuse and self.fuse_max_seq_len is None:
raise ValueError(
"You cannot enable fused modules without specifying a `fuse_max_seq_len`, make sure to pass a valid `fuse_max_seq_len` for your usecase"
)
if self.do_fuse:
awq_version_supports_fusing = False
MIN_AWQ_VERSION = "0.1.7"
if is_auto_awq_available():
awq_version_supports_fusing = version.parse(importlib.metadata.version("autoawq")) >= version.parse(
MIN_AWQ_VERSION
)
if not awq_version_supports_fusing:
raise ValueError(
f"You current version of `autoawq` does not support module fusing, please upgrade `autoawq` package to at least {MIN_AWQ_VERSION}."
)
if self.modules_to_not_convert is not None:
awq_version_supports_non_conversion = False
MIN_AWQ_VERSION = "0.1.8"
if is_auto_awq_available():
awq_version_supports_non_conversion = version.parse(
importlib.metadata.version("autoawq")
) >= version.parse(MIN_AWQ_VERSION)
if not awq_version_supports_non_conversion:
raise ValueError(
f"You current version of `autoawq` does not support module quantization skipping, please upgrade `autoawq` package to at least {MIN_AWQ_VERSION}."
)
if self.do_fuse and self.modules_to_fuse is not None:
required_keys = [
"hidden_size",
"num_attention_heads",
"num_key_value_heads",
"mlp",
"attention",
"layernorm",
"use_alibi",
]
if not all(key in self.modules_to_fuse for key in required_keys):
raise ValueError(
f"Required fields are missing in the fusing mapping, required fields are {required_keys}"
)
if self.version == AWQLinearVersion.EXLLAMA:
awq_version_supports_exllama = False
MIN_AWQ_VERSION = "0.2.0"
if is_auto_awq_available():
awq_version_supports_exllama = version.parse(importlib.metadata.version("autoawq")) >= version.parse(
MIN_AWQ_VERSION
)
if not awq_version_supports_exllama:
raise ValueError(
f"You current version of `autoawq` does not support exllama backend, "
f"please upgrade `autoawq` package to at least {MIN_AWQ_VERSION}."
)
if self.exllama_config is None:
self.exllama_config = {"version": ExllamaVersion.TWO, "max_input_len": 2048, "max_batch_size": 8}
else:
if "version" not in self.exllama_config:
raise ValueError("`exllama_config` needs to have a `version` key.")
elif self.exllama_config["version"] not in [ExllamaVersion.ONE, ExllamaVersion.TWO]:
exllama_version = self.exllama_config["version"]
raise ValueError(
f"Only supported versions are in [ExllamaVersion.ONE, ExllamaVersion.TWO] - not recognized version {exllama_version}"
)
def get_loading_attributes(self):
attibutes_dict = copy.deepcopy(self.__dict__)
loading_attibutes = ["version", "do_fuse", "modules_to_fuse", "fuse_max_seq_len", "exllama_config"]
loading_attibutes_dict = {i: j for i, j in attibutes_dict.items() if i in loading_attibutes}
return loading_attibutes_dict
@dataclass
class AqlmConfig(QuantizationConfigMixin):
"""
This is a wrapper class about `aqlm` parameters.
Args:
in_group_size (`int`, *optional*, defaults to 8):
The group size along the input dimension.
out_group_size (`int`, *optional*, defaults to 1):
The group size along the output dimension. It's recommended to always use 1.
num_codebooks (`int`, *optional*, defaults to 1):
Number of codebooks for the Additive Quantization procedure.
nbits_per_codebook (`int`, *optional*, defaults to 16):
Number of bits encoding a single codebook vector. Codebooks size is 2**nbits_per_codebook.
linear_weights_not_to_quantize (`Optional[list[str]]`, *optional*):
List of full paths of `nn.Linear` weight parameters that shall not be quantized.
kwargs (`dict[str, Any]`, *optional*):
Additional parameters from which to initialize the configuration object.
"""
def __init__(
self,
in_group_size: int = 8,
out_group_size: int = 1,
num_codebooks: int = 1,
nbits_per_codebook: int = 16,
linear_weights_not_to_quantize: Optional[list[str]] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.AQLM
self.in_group_size = in_group_size
self.out_group_size = out_group_size
self.num_codebooks = num_codebooks
self.nbits_per_codebook = nbits_per_codebook
self.linear_weights_not_to_quantize = linear_weights_not_to_quantize
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
"""
if not isinstance(self.in_group_size, int):
raise TypeError("in_group_size must be a float")
if not isinstance(self.out_group_size, int):
raise TypeError("out_group_size must be a float")
if not isinstance(self.num_codebooks, int):
raise TypeError("num_codebooks must be a float")
if not isinstance(self.nbits_per_codebook, int):
raise TypeError("nbits_per_codebook must be a float")
if self.linear_weights_not_to_quantize is not None and not isinstance(
self.linear_weights_not_to_quantize, list
):
raise ValueError("linear_weights_not_to_quantize must be a list of strings")
if self.linear_weights_not_to_quantize is None:
self.linear_weights_not_to_quantize = []
@dataclass
class VptqLayerConfig(QuantizationConfigMixin):
"""
This is used to explain vptq config params for each layer
Args:
enable_norm (`bool`, *optional*, defaults to `True`): to control if we have scale/bias for fp-weight
enable_perm (`bool`, *optional*, defaults to `True`): to perm input_channel or not
group_num (`int`, *optional*, defaults to `1`): how many single groups for vector-quantization
group_size (`int`, *optional*, defaults to `-1`): depends on out-features
indices_as_float (`bool`, *optional*, defaults to `False`): for Finetuning
is_indice_packed (`bool`, *optional*, defaults to `True`): should always be True
num_centroids (`list`, *optional*, defaults to `[-1, -1]`): centroid numbers of clusters
num_res_centroids (`list`, *optional*, defaults to `[-1, -1]`): ditto for residual
outlier_size (`int`, *optional*, defaults to `1`): outliers
vector_lens (`list`, *optional*, defaults to `[-1, -1]`): centroid vector length in quantization
"""
def __init__(
self,
enable_norm: bool = True,
enable_perm: bool = True,
group_num: int = 1,
group_size: int = -1,
in_features: int = -1,
indices_as_float: bool = False,
is_indice_packed: bool = True,
num_centroids: tuple = [-1, -1],
num_res_centroids: tuple = [-1, -1],
out_features: int = -1,
outlier_size: int = 0,
vector_lens: tuple = [-1, -1],
**kwargs,
):
self.enable_norm = enable_norm
self.enable_perm = enable_perm
self.group_num = group_num
self.group_size = group_size
self.in_features = in_features
self.indices_as_float = indices_as_float
self.is_indice_packed = is_indice_packed
self.num_centroids = num_centroids
self.num_res_centroids = num_res_centroids
self.out_features = out_features
self.outlier_size = outlier_size
self.vector_lens = vector_lens
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
if self.is_indice_packed is False:
raise ValueError("is_indice_packed should always be True")
@dataclass
class VptqConfig(QuantizationConfigMixin):
"""
This is a wrapper class about `vptq` parameters.
Args:
enable_proxy_error (`bool`, *optional*, defaults to `False`): calculate proxy error for each layer
config_for_layers (`Dict`, *optional*, defaults to `{}`): quantization params for each layer
shared_layer_config (`Dict`, *optional*, defaults to `{}`): shared quantization params among layers
modules_to_not_convert (`list`, *optional*, default to `None`):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision (e.g. Whisper encoder, Llava encoder, Mixtral gate layers).
kwargs (`dict[str, Any]`, *optional*):
Additional parameters from which to initialize the configuration object.
"""
def __init__(
self,
enable_proxy_error: bool = False,
config_for_layers: dict[str, Any] = {},
shared_layer_config: dict[str, Any] = {},
modules_to_not_convert: Optional[list] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.VPTQ
self.enable_proxy_error = enable_proxy_error
self.config_for_layers: dict[str, Any] = config_for_layers
self.shared_layer_config: dict[str, Any] = shared_layer_config
self.modules_to_not_convert = modules_to_not_convert
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
for layer_param in self.config_for_layers.values():
VptqLayerConfig(**layer_param)
if self.enable_proxy_error is True:
raise ValueError("enable_proxy_error should always be False until we support training")
@dataclass
class QuantoConfig(QuantizationConfigMixin):
"""
This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded using `quanto`.
Args:
weights (`str`, *optional*, defaults to `"int8"`):
The target dtype for the weights after quantization. Supported values are ("float8","int8","int4","int2")
activations (`str`, *optional*):
The target dtype for the activations after quantization. Supported values are (None,"int8","float8")
modules_to_not_convert (`list`, *optional*, default to `None`):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision (e.g. Whisper encoder, Llava encoder, Mixtral gate layers).
"""
def __init__(
self,
weights="int8",
activations=None,
modules_to_not_convert: Optional[list] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.QUANTO
self.weights = weights
self.activations = activations
self.modules_to_not_convert = modules_to_not_convert
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
accepted_weights = ["float8", "int8", "int4", "int2"]
accepted_activations = [None, "int8", "float8"]
if self.weights not in accepted_weights:
raise ValueError(f"Only support weights in {accepted_weights} but found {self.weights}")
if self.activations not in accepted_activations:
raise ValueError(f"Only support weights in {accepted_activations} but found {self.activations}")
@dataclass
class EetqConfig(QuantizationConfigMixin):
"""
This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded using `eetq`.
Args:
weights (`str`, *optional*, defaults to `"int8"`):
The target dtype for the weights. Supported value is only "int8"
modules_to_not_convert (`list`, *optional*, default to `None`):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision.
"""
def __init__(
self,
weights: str = "int8",
modules_to_not_convert: Optional[list] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.EETQ
self.weights = weights
self.modules_to_not_convert = modules_to_not_convert
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
accepted_weights = ["int8"]
if self.weights not in accepted_weights:
raise ValueError(f"Only support weights in {accepted_weights} but found {self.weights}")
class CompressedTensorsConfig(QuantizationConfigMixin):
"""
This is a wrapper class that handles compressed-tensors quantization config options.
It is a wrapper around `compressed_tensors.QuantizationConfig`
Args:
config_groups (`typing.dict[str, typing.Union[ForwardRef('QuantizationScheme'), typing.list[str]]]`, *optional*):
dictionary mapping group name to a quantization scheme definition
format (`str`, *optional*, defaults to `"dense"`):
format the model is represented as. Set `run_compressed` True to execute model as the
compressed format if not `dense`
quantization_status (`QuantizationStatus`, *optional*, defaults to `"initialized"`):
status of model in the quantization lifecycle, ie 'initialized', 'calibration', 'frozen'
kv_cache_scheme (`typing.Union[QuantizationArgs, NoneType]`, *optional*):
specifies quantization of the kv cache. If None, kv cache is not quantized.
global_compression_ratio (`typing.Union[float, NoneType]`, *optional*):
0-1 float percentage of model compression
ignore (`typing.Union[typing.list[str], NoneType]`, *optional*):
layer names or types to not quantize, supports regex prefixed by 're:'
sparsity_config (`typing.dict[str, typing.Any]`, *optional*):
configuration for sparsity compression
quant_method (`str`, *optional*, defaults to `"compressed-tensors"`):
do not override, should be compressed-tensors
run_compressed (`bool`, *optional*, defaults to `True`): alter submodules (usually linear) in order to
emulate compressed model execution if True, otherwise use default submodule
"""
def __init__(
self,
config_groups: Optional[dict[str, Union["QuantizationScheme", list[str]]]] = None, # noqa: F821
format: str = "dense",
quantization_status: "QuantizationStatus" = "initialized", # noqa: F821
kv_cache_scheme: Optional["QuantizationArgs"] = None, # noqa: F821
global_compression_ratio: Optional[float] = None,
ignore: Optional[list[str]] = None,
sparsity_config: Optional[dict[str, Any]] = None,
quant_method: str = "compressed-tensors",
run_compressed: bool = True,
**kwargs,
):
if is_compressed_tensors_available():
from compressed_tensors.config import SparsityCompressionConfig
from compressed_tensors.quantization import QuantizationConfig
else:
raise ImportError(
"compressed_tensors is not installed and is required for compressed-tensors quantization. Please install it with `pip install compressed-tensors`."
)
self.quantization_config = None
self.sparsity_config = None
self.run_compressed = run_compressed
# parse from dict to load nested QuantizationScheme objects
if config_groups or kv_cache_scheme:
self.quantization_config = QuantizationConfig.model_validate(
{
"config_groups": config_groups,
"quant_method": quant_method,
"format": format,
"quantization_status": quantization_status,
"kv_cache_scheme": kv_cache_scheme,
"global_compression_ratio": global_compression_ratio,
"ignore": ignore,
**kwargs,
}
)
if sparsity_config:
self.sparsity_config = SparsityCompressionConfig.load_from_registry(
sparsity_config.get("format"), **sparsity_config
)
self.quant_method = QuantizationMethod.COMPRESSED_TENSORS
def post_init(self):
if self.run_compressed:
if self.is_sparsification_compressed:
logger.warning(
"`run_compressed` is only supported for quantized_compressed models"
" and not for sparsified models. Setting `run_compressed=False`"
)
self.run_compressed = False
elif not self.is_quantization_compressed:
logger.warning(
"`run_compressed` is only supported for compressed models. Setting `run_compressed=False`"
)
self.run_compressed = False
@classmethod
def from_dict(cls, config_dict, return_unused_kwargs=False, **kwargs):
"""
Instantiates a [`CompressedTensorsConfig`] from a Python dictionary of parameters.
Optionally unwraps any args from the nested quantization_config
Args:
config_dict (`dict[str, Any]`):
Dictionary that will be used to instantiate the configuration object.
return_unused_kwargs (`bool`,*optional*, defaults to `False`):
Whether or not to return a list of unused keyword arguments. Used for `from_pretrained` method in
`PreTrainedModel`.
kwargs (`dict[str, Any]`):
Additional parameters from which to initialize the configuration object.
Returns:
[`QuantizationConfigMixin`]: The configuration object instantiated from those parameters.
"""
if "quantization_config" in config_dict:
config_dict = dict(
sparsity_config=config_dict.get("sparsity_config"),
**config_dict["quantization_config"],
)
return super().from_dict(config_dict, return_unused_kwargs=return_unused_kwargs, **kwargs)
def to_dict(self) -> dict[str, Any]:
"""
Quantization config to be added to config.json
Serializes this instance to a Python dictionary. Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance.
"""
quantization_config = {}
if self.quantization_config is not None:
quantization_config = self.quantization_config.model_dump()
else:
quantization_config["quant_method"] = QuantizationMethod.COMPRESSED_TENSORS
if self.sparsity_config is not None:
quantization_config["sparsity_config"] = self.sparsity_config.model_dump()
else:
quantization_config["sparsity_config"] = {}
return quantization_config
def to_diff_dict(self) -> dict[str, Any]:
"""
Removes all attributes from config which correspond to the default config attributes for better readability and
serializes to a Python dictionary.
Returns:
`dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance,
"""
config_dict = self.to_dict()
# get the default config dict
default_config_dict = CompressedTensorsConfig().to_dict()
serializable_config_dict = {}
# only serialize values that differ from the default config
for key, value in config_dict.items():
if key not in default_config_dict or value != default_config_dict[key]:
serializable_config_dict[key] = value
return serializable_config_dict
def get_loading_attributes(self):
return {"run_compressed": self.run_compressed}
@property
def is_quantized(self):
return bool(self.quantization_config) and bool(self.quantization_config.config_groups)
@property
def is_quantization_compressed(self):
from compressed_tensors.quantization import QuantizationStatus
return self.is_quantized and self.quantization_config.quantization_status == QuantizationStatus.COMPRESSED
@property
def is_sparsification_compressed(self):
from compressed_tensors.config import (
CompressionFormat,
SparsityCompressionConfig,
)
return (
isinstance(self.sparsity_config, SparsityCompressionConfig)
and self.sparsity_config.format != CompressionFormat.dense.value
)
@dataclass
class FbgemmFp8Config(QuantizationConfigMixin):
"""
This is a wrapper class about all possible attributes and features that you can play with a model that has been
loaded using fbgemm fp8 quantization.
Args:
activation_scale_ub (`float`, *optional*, defaults to 1200.0):
The activation scale upper bound. This is used when quantizing the input activation.
modules_to_not_convert (`list`, *optional*, default to `None`):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision.
"""
def __init__(
self,
activation_scale_ub: float = 1200.0,
modules_to_not_convert: Optional[list] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.FBGEMM_FP8
self.activation_scale_ub = activation_scale_ub
self.modules_to_not_convert = modules_to_not_convert
def get_loading_attributes(self):
attibutes_dict = copy.deepcopy(self.__dict__)
loading_attibutes = ["activation_scale_ub"]
loading_attibutes_dict = {i: j for i, j in attibutes_dict.items() if i in loading_attibutes}
return loading_attibutes_dict
@dataclass
class HiggsConfig(QuantizationConfigMixin):
"""
HiggsConfig is a configuration class for quantization using the HIGGS method.
Args:
bits (int, *optional*, defaults to 4):
Number of bits to use for quantization. Can be 2, 3 or 4. Default is 4.
p (int, *optional*, defaults to 2):
Quantization grid dimension. 1 and 2 are supported. 2 is always better in practice. Default is 2.
modules_to_not_convert (`list`, *optional*, default to ["lm_head"]):
List of linear layers that should not be quantized.
hadamard_size (int, *optional*, defaults to 512):
Hadamard size for the HIGGS method. Default is 512. Input dimension of matrices is padded to this value. Decreasing this below 512 will reduce the quality of the quantization.
group_size (int, *optional*, defaults to 256):
Group size for the HIGGS method. Can be 64, 128 or 256. Decreasing it barely affects the performance. Default is 256. Must be a divisor of hadamard_size.
tune_metadata ('dict', *optional*, defaults to {}):
Module-wise metadata (gemm block shapes, GPU metadata, etc.) for saving the kernel tuning results. Default is an empty dictionary. Is set automatically during tuning.
"""
def __init__(
self,
bits: int = 4,
p: int = 2,
modules_to_not_convert: Optional[list[str]] = None,
hadamard_size: int = 512,
group_size: int = 256,
tune_metadata: Optional[dict[str, Any]] = None,
**kwargs,
):
if tune_metadata is None:
tune_metadata = {}
self.quant_method = QuantizationMethod.HIGGS
self.bits = bits
self.p = p
self.modules_to_not_convert = modules_to_not_convert
self.hadamard_size = hadamard_size
self.group_size = group_size
self.tune_metadata = tune_metadata
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
"""
if self.bits not in [2, 3, 4]:
raise ValueError("bits must be 2, 3, or 4")
if self.p not in [1, 2]:
raise ValueError("p must be 1 or 2. 2 is always better in practice")
if self.group_size not in [64, 128, 256]:
raise ValueError("group_size must be 64, 128, or 256")
if self.hadamard_size % self.group_size != 0:
raise ValueError("hadamard_size must be divisible by group_size")
@dataclass
class FPQuantConfig(QuantizationConfigMixin):
"""
FPQuantConfig is a configuration class for quantization using the FPQuant method.
Args:
forward_dtype (`str`, *optional*, defaults to `"mxfp4"`):
The dtype to use for the forward pass.
forward_method (`str`, *optional*, defaults to `"abs_max"`):
The scaling to use for the forward pass. Can be `"abs_max"` or `"quest"`. `"abs_max"` is better for PTQ, `"quest"` is better for QAT.
backward_dtype (`str`, *optional*, defaults to `"bf16"`):
The dtype to use for the backward pass.
store_master_weights (`bool`, *optional*, defaults to `False`):
Whether to store the master weights. Needed for QAT over layer weights.
hadamard_group_size (`int`, *optional*, defaults to 32):
The group size for the hadamard transform before quantization for `"quest"` it matches the MXFP4 group size (32).
pseudoquantization (`bool`, *optional*, defaults to `False`):
Whether to use Triton-based pseudo-quantization. Is mandatory for non-Blackwell GPUs. Doesn't provide any speedup. For debugging purposes.
modules_to_not_convert (`list`, *optional*):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision.
"""
def __init__(
self,
forward_dtype: str = "mxfp4",
forward_method: str = "abs_max",
backward_dtype: str = "bf16",
store_master_weights: bool = False,
hadamard_group_size: int = 32,
pseudoquantization: bool = False,
modules_to_not_convert: Optional[list[str]] = None,
**kwargs,
):
self.forward_dtype = forward_dtype
self.forward_method = forward_method
self.backward_dtype = backward_dtype
self.store_master_weights = store_master_weights
self.hadamard_group_size = hadamard_group_size
self.pseudoquantization = pseudoquantization
self.modules_to_not_convert = modules_to_not_convert
self.quant_method = QuantizationMethod.FPQUANT
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
"""
if self.forward_dtype not in ["mxfp4"]:
raise ValueError("Only 'mxfp4' is supported for forward_dtype for now.")
if self.forward_method not in ["abs_max", "quest"]:
raise ValueError("Only 'abs_max' and 'quest' are supported for forward_method for now.")
if self.backward_dtype not in ["bf16"]:
raise ValueError("Only 'bf16' is supported for backward_dtype for now.")
if self.hadamard_group_size not in [32]:
raise ValueError("Only a hadamard_group_size of 32 is supported for now.")
if self.modules_to_not_convert is None:
self.modules_to_not_convert = ["lm_head"]
@dataclass
class TorchAoConfig(QuantizationConfigMixin):
quant_method: QuantizationMethod
quant_type: Union[str, "AOBaseConfig"] # noqa: F821
modules_to_not_convert: Optional[list]
quant_type_kwargs: dict[str, Any]
include_input_output_embeddings: bool
untie_embedding_weights: bool
"""This is a config class for torchao quantization/sparsity techniques.
Args:
quant_type (`Union[str, AOBaseConfig]`):
The type of quantization we want to use. Can be either:
- A string: currently supporting: `int4_weight_only`, `int8_weight_only` and `int8_dynamic_activation_int8_weight`.
- An AOBaseConfig instance: for more advanced configuration options.
modules_to_not_convert (`list`, *optional*, default to `None`):
The list of modules to not quantize, useful for quantizing models that explicitly require to have
some modules left in their original precision.
inlcude_embedding (`bool`, default to `False`):
Whether to include embedding in quantization or not, input embedding will be removed from
the module_not_to_convert list as well if this flag is set.
untie_embedding_weights (`bool`, default to `False`):
Whether to untie the weights when we are quantizing input embedding weights that is tied
to other weights.
kwargs (`dict[str, Any]`, *optional*):
The keyword arguments for the chosen type of quantization, for example, int4_weight_only quantization supports two keyword arguments
`group_size` and `inner_k_tiles` currently. More API examples and documentation of arguments can be found in
https://github.com/pytorch/ao/tree/main/torchao/quantization#other-available-quantization-techniques
Example:
```python
# AOBaseConfig-based configuration
config = Int4WeightOnlyConfig(group_size=32)
quantization_config = TorchAoConfig(config)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda", torch_dtype=torch.bfloat16, quantization_config=quantization_config)
# String-based configuration
quantization_config = TorchAoConfig("int4_weight_only", group_size=32)
# int4_weight_only quant is only working with *torch.bfloat16* dtype right now
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda", torch_dtype=torch.bfloat16, quantization_config=quantization_config)
# autoquant
# `autoquant` is a convenient way for users to search for the best quantization for each layer
# `min_sqnr` is an option to control the accuracy of the model, higher value means the model is more
# accurate, we can start with 30 and adjust it to larger or smaller (e.g. 40, 20)
# defaults to None, which means we'll try to get the best performing quantized model without
# considering accuracy
quantization_config = TorchAoConfig("autoquant", min_sqnr=30)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda", torch_dtype=torch.bfloat16, quantization_config=quantization_config)
# run through example inputs, quantization methods will be selected based on the shape of example input
tokenizer = AutoTokenizer.from_pretrained(model_name)
input_text = "What are we having for dinner?"
input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")
MAX_NEW_TOKENS = 1000
model.generate(**input_ids, max_new_tokens=MAX_NEW_TOKENS, cache_implementation="static")
# manually ran finalize_autoquant if needed
if hasattr(quantized_model, "finalize_autoquant"):
print("finalizing autoquant")
quantized_model.finalize_autoquant()
```
"""
def __init__(
self,
quant_type: Union[str, "AOBaseConfig"], # noqa: F821
modules_to_not_convert: Optional[list] = None,
include_input_output_embeddings: bool = False,
untie_embedding_weights: bool = False,
**kwargs,
):
self.quant_method = QuantizationMethod.TORCHAO
self.quant_type = quant_type
self.modules_to_not_convert = modules_to_not_convert
self.quant_type_kwargs = kwargs.get("quant_type_kwargs", kwargs)
self.include_input_output_embeddings = include_input_output_embeddings
self.untie_embedding_weights = untie_embedding_weights
self.post_init()
@staticmethod
def _get_ao_version() -> version.Version:
"""Centralized check for TorchAO availability and version requirements."""
if not is_torchao_available():
raise ValueError("TorchAoConfig requires torchao to be installed. Install with `pip install torchao`")
return version.parse(importlib.metadata.version("torchao"))
def post_init(self):
"""Validate configuration and set defaults."""
ao_version = self._get_ao_version()
# Handle quant_type based on type and version
if isinstance(self.quant_type, str):
self._validate_string_quant_type()
elif ao_version > version.parse("0.9.0"):
from torchao.quantization.quant_api import AOBaseConfig
if not isinstance(self.quant_type, AOBaseConfig):
raise TypeError(
f"quant_type must be either a string or an AOBaseConfig instance, got {type(self.quant_type)}"
)
else:
raise ValueError(
f"In torchao <= 0.9.0, quant_type must be a string. Got {type(self.quant_type)}. "
f"Please upgrade to torchao > 0.9.0 to use AOBaseConfig instances."
)
def _validate_string_quant_type(self):
"""Validate string quant_type and its kwargs."""
methods = self._get_torchao_quant_type_to_method()
if self.quant_type not in methods:
raise ValueError(
f"Unsupported string quantization type: {self.quant_type}. "
f"Supported types: {', '.join(methods.keys())}"
)
# Validate kwargs against method signature
method = methods[self.quant_type]
sig = signature(method)
valid_kwargs = {
param.name
for param in sig.parameters.values()
if param.kind in [Parameter.KEYWORD_ONLY, Parameter.POSITIONAL_OR_KEYWORD]
}
invalid_kwargs = set(self.quant_type_kwargs) - valid_kwargs
if invalid_kwargs:
raise ValueError(
f"Unexpected keyword arg for {self.quant_type}: {', '.join(invalid_kwargs)}. "
f"Valid kwargs: {', '.join(valid_kwargs)}"
)
def _get_torchao_quant_type_to_method(self):
"""Get mapping of quant_type strings to their corresponding methods."""
from torchao.quantization import (
autoquant,
int4_weight_only,
int8_dynamic_activation_int8_weight,
int8_weight_only,
)
return {
"int4_weight_only": int4_weight_only,
"int8_weight_only": int8_weight_only,
"int8_dynamic_activation_int8_weight": int8_dynamic_activation_int8_weight,
"autoquant": autoquant,
}
def get_apply_tensor_subclass(self):
"""Create the appropriate quantization method based on configuration."""
if isinstance(self.quant_type, str):
methods = self._get_torchao_quant_type_to_method()
quant_type_kwargs = self.quant_type_kwargs.copy()
if (
not torch.cuda.is_available()
and is_torchao_available()
and self.quant_type == "int4_weight_only"
and version.parse(importlib.metadata.version("torchao")) >= version.parse("0.8.0")
and quant_type_kwargs.get("layout", None) is None
):
if torch.xpu.is_available():
if version.parse(importlib.metadata.version("torchao")) >= version.parse(
"0.11.0"
) and version.parse(importlib.metadata.version("torch")) > version.parse("2.7.9"):
from torchao.dtypes import Int4XPULayout
from torchao.quantization.quant_primitives import ZeroPointDomain
quant_type_kwargs["layout"] = Int4XPULayout()
quant_type_kwargs["zero_point_domain"] = ZeroPointDomain.INT
else:
raise ValueError(
"TorchAoConfig requires torchao >= 0.11.0 and torch >= 2.8.0 for XPU support. Please upgrade the version or use run on CPU with the cpu version pytorch."
)
else:
from torchao.dtypes import Int4CPULayout
quant_type_kwargs["layout"] = Int4CPULayout()
return methods[self.quant_type](**quant_type_kwargs)
else:
return self.quant_type
def to_dict(self):
"""Convert configuration to a dictionary."""
d = super().to_dict()
if isinstance(self.quant_type, str):
# Handle layout serialization if present
if "quant_type_kwargs" in d and "layout" in d["quant_type_kwargs"]:
if is_dataclass(d["quant_type_kwargs"]["layout"]):
d["quant_type_kwargs"]["layout"] = [
d["quant_type_kwargs"]["layout"].__class__.__name__,
dataclasses.asdict(d["quant_type_kwargs"]["layout"]),
]
if isinstance(d["quant_type_kwargs"]["layout"], list):
assert len(d["quant_type_kwargs"]["layout"]) == 2, "layout saves layout name and layout kwargs"
assert isinstance(d["quant_type_kwargs"]["layout"][0], str), "layout name must be a string"
assert isinstance(d["quant_type_kwargs"]["layout"][1], dict), "layout kwargs must be a dict"
else:
raise ValueError("layout must be a list")
else:
# Handle AOBaseConfig serialization
from torchao.core.config import config_to_dict
# For now we assume there is 1 config per Transformer, however in the future
# We may want to support a config per fqn.
d["quant_type"] = {"default": config_to_dict(self.quant_type)}
return d
@classmethod
def from_dict(cls, config_dict, return_unused_kwargs=False, **kwargs):
"""Create configuration from a dictionary."""
ao_verison = cls._get_ao_version()
assert ao_verison > version.parse("0.9.0"), "TorchAoConfig requires torchao > 0.9.0 for construction from dict"
config_dict = config_dict.copy()
quant_type = config_dict.pop("quant_type")
if isinstance(quant_type, str):
return cls(quant_type=quant_type, **config_dict)
# Check if we only have one key which is "default"
# In the future we may update this
assert len(quant_type) == 1 and "default" in quant_type, (
"Expected only one key 'default' in quant_type dictionary"
)
quant_type = quant_type["default"]
# Deserialize quant_type if needed
from torchao.core.config import config_from_dict
quant_type = config_from_dict(quant_type)
return cls(quant_type=quant_type, **config_dict)
@dataclass
class BitNetQuantConfig(QuantizationConfigMixin):
"""
Configuration class for applying BitNet quantization.
Args:
modules_to_not_convert (`Optional[List]`, *optional*):
Optionally, provides a list of full paths of `nn.Linear` weight parameters
that shall not be quantized. Defaults to None.
linear_class (`str`, *optional*, defaults to `"bitlinear"`):
The type of linear class to use. Can be either `bitlinear` or `autobitlinear`.
quantization_mode (`str`, *optional*, defaults to `"offline"`):
The quantization mode to use. Can be either `online` or `offline`.
In `online` mode, the weight quantization parameters are calculated dynamically
during each forward pass (e.g., based on the current weight values). This can
adapt to weight changes during training (Quantization-Aware Training - QAT).
In `offline` mode, quantization parameters are pre-calculated *before* inference.
These parameters are then fixed and loaded into the quantized model. This
generally results in lower runtime overhead compared to online quantization.
use_rms_norm (`bool`, *optional*, defaults to `False`):
Whether to apply RMSNorm on the activations before quantization. This matches the original BitNet paper's approach
of normalizing activations before quantization/packing.
rms_norm_eps (`float`, *optional*, defaults to 1e-06):
The epsilon value used in the RMSNorm layer for numerical stability.
kwargs (`dict[str, Any]`, *optional*):
Additional keyword arguments that may be used by specific quantization
backends or future versions.
"""
def __init__(
self,
modules_to_not_convert: Optional[list] = None,
linear_class: Optional[str] = "bitlinear",
quantization_mode: Optional[str] = "offline",
use_rms_norm: Optional[bool] = False,
rms_norm_eps: Optional[float] = 1e-6,
**kwargs,
):
if linear_class not in ["bitlinear", "autobitlinear"]:
raise ValueError(f"linear_class must be either 'bitlinear' or 'autobitlinear', but got {linear_class}")
if quantization_mode not in ["online", "offline"]:
raise ValueError(f"quantization_mode must be either 'online' or 'offline', but got {quantization_mode}")
self.quant_method = QuantizationMethod.BITNET
self.modules_to_not_convert = modules_to_not_convert
self.linear_class = linear_class
self.quantization_mode = quantization_mode
self.use_rms_norm = use_rms_norm
self.rms_norm_eps = rms_norm_eps
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
pass
@dataclass
class SpQRConfig(QuantizationConfigMixin):
"""
This is a wrapper class about `spqr` parameters. Refer to the original publication for more details.
Args:
bits (`int`, *optional*, defaults to 3):
Specifies the bit count for the weights and first order zero-points and scales.
Currently only bits = 3 is supported.
beta1 (`int`, *optional*, defaults to 16):
SpQR tile width. Currently only beta1 = 16 is supported.
beta2 (`int`, *optional*, defaults to 16):
SpQR tile height. Currently only beta2 = 16 is supported.
shapes (`Optional`, *optional*):
A dictionary holding the shape of each object. We need this because it's impossible
to deduce the exact size of the parameters just from bits, beta1, beta2.
modules_to_not_convert (`Optional[list[str]]`, *optional*):
Optionally, provides a list of full paths of `nn.Linear` weight parameters that shall not be quantized.
Defaults to None.
kwargs (`dict[str, Any]`, *optional*):
Additional parameters from which to initialize the configuration object.
"""
def __init__(
self,
bits: int = 3,
beta1: int = 16,
beta2: int = 16,
shapes: Optional[dict[str, int]] = None,
modules_to_not_convert: Optional[list[str]] = None,
**kwargs,
):
if shapes is None:
shapes = {}
self.shapes = shapes
self.quant_method = QuantizationMethod.SPQR
self.bits = bits
self.beta1 = beta1
self.beta2 = beta2
self.modules_to_not_convert = modules_to_not_convert
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct - also replaces some NoneType arguments with their default values.
"""
if not isinstance(self.bits, int):
raise TypeError("bits must be an int")
if not isinstance(self.beta1, int):
raise TypeError("beta1 must be an int")
if not isinstance(self.beta2, int):
raise TypeError("beta2 must be an int")
if self.bits != 3:
raise ValueError("SpQR currently only supports bits = 3")
if self.beta1 != 16:
raise ValueError("SpQR currently only supports beta1 = 16")
if self.beta2 != 16:
raise ValueError("SpQR currently only supports beta2 = 16")
if not isinstance(self.shapes, dict):
raise TypeError("shapes must be a dict")
@dataclass
class FineGrainedFP8Config(QuantizationConfigMixin):
"""
FineGrainedFP8Config is a configuration class for fine-grained FP8 quantization used mainly for deepseek models.
Args:
activation_scheme (`str`, *optional*, defaults to `"dynamic"`):
The scheme used for activation, the defaults and only support scheme for now is "dynamic".
weight_block_size (`typing.tuple[int, int]`, *optional*, defaults to `(128, 128)`):
The size of the weight blocks for quantization, default is (128, 128).
modules_to_not_convert (`list`, *optional*):
A list of module names that should not be converted during quantization.
"""
def __init__(
self,
activation_scheme: str = "dynamic",
weight_block_size: tuple[int, int] = (128, 128),
modules_to_not_convert: Optional[list] = None,
**kwargs,
):
self.quant_method = QuantizationMethod.FP8
self.modules_to_not_convert = modules_to_not_convert
self.activation_scheme = activation_scheme
self.weight_block_size = weight_block_size
self.post_init()
def post_init(self):
r"""
Safety checker that arguments are correct
"""
self.activation_scheme = self.activation_scheme.lower()
if self.activation_scheme not in ["dynamic"]:
raise ValueError(f"Activation scheme {self.activation_scheme} not supported")
if len(self.weight_block_size) != 2:
raise ValueError("weight_block_size must be a tuple of two integers")
if self.weight_block_size[0] <= 0 or self.weight_block_size[1] <= 0:
raise ValueError("weight_block_size must be a tuple of two positive integers")
class QuarkConfig(QuantizationConfigMixin):
def __init__(
self,
**kwargs,
):
if is_torch_available() and is_quark_available():
from quark import __version__ as quark_version
from quark.torch.export.config.config import JsonExporterConfig
from quark.torch.export.main_export.quant_config_parser import QuantConfigParser
from quark.torch.quantization.config.config import Config
else:
raise ImportError(
"Quark is not installed. Please refer to https://quark.docs.amd.com/latest/install.html."
)
# This might be e.g. `"fp8"` or `"awq"`.
self.custom_mode = kwargs["quant_method"]
self.legacy = "export" not in kwargs
if self.custom_mode in ["awq", "fp8"]:
# Legacy (quark<1.0) or custom export.
self.quant_config = QuantConfigParser.from_custom_config(kwargs, is_bias_quantized=False)
self.json_export_config = JsonExporterConfig()
else:
self.quant_config = Config.from_dict(kwargs)
if "export" in kwargs:
# TODO: Remove this check once configuration version is handled natively by Quark.
if "min_kv_scale" in kwargs["export"] and version.parse(quark_version) < version.parse("0.8"):
min_kv_scale = kwargs["export"].pop("min_kv_scale")
logger.warning(
f"The parameter `min_kv_scale={min_kv_scale}` was found in the model config.json's `quantization_config.export` configuration, but this parameter is supported only for quark>=0.8. Ignoring this configuration parameter. Please update the `amd-quark` package."
)
self.json_export_config = JsonExporterConfig(**kwargs["export"])
else:
# Legacy (quark<1.0) or custom export.
self.json_export_config = JsonExporterConfig()
self.quant_method = QuantizationMethod.QUARK
Reference in a new issue View git blame Copy permalink