Add files using upload-large-folder tool

.gitattributes

@@ -801,3 +801,5 @@ evalkit_eagle/lib/python3.10/site-packages/pandas/core/__pycache__/frame.cpython
 evalkit_eagle/lib/python3.10/site-packages/pandas/io/formats/__pycache__/style.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text
 evalkit_eagle/lib/python3.10/site-packages/pandas/core/indexes/__pycache__/base.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text
 evalkit_eagle/lib/python3.10/site-packages/pandas/io/__pycache__/stata.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text
+evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda121.so filter=lfs diff=lfs merge=lfs -text
+evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda124.so filter=lfs diff=lfs merge=lfs -text

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/diagnostics/cuda.py

@@ -0,0 +1,176 @@
+import logging
+import os
+from pathlib import Path
+from typing import Dict, Iterable, Iterator
+
+import torch
+
+from bitsandbytes.cextension import get_cuda_bnb_library_path
+from bitsandbytes.consts import NONPYTORCH_DOC_URL
+from bitsandbytes.cuda_specs import CUDASpecs
+from bitsandbytes.diagnostics.utils import print_dedented
+
+CUDART_PATH_PREFERRED_ENVVARS = ("CONDA_PREFIX", "LD_LIBRARY_PATH")
+
+CUDART_PATH_IGNORED_ENVVARS = {
+    "DBUS_SESSION_BUS_ADDRESS",  # hardware related
+    "GOOGLE_VM_CONFIG_LOCK_FILE",  # GCP: requires elevated permissions, causing problems in VMs and Jupyter notebooks
+    "HOME",  # Linux shell default
+    "LESSCLOSE",
+    "LESSOPEN",  # related to the `less` command
+    "MAIL",  # something related to emails
+    "OLDPWD",
+    "PATH",  # this is for finding binaries, not libraries
+    "PWD",  # PWD: this is how the shell keeps track of the current working dir
+    "SHELL",  # binary for currently invoked shell
+    "SSH_AUTH_SOCK",  # SSH stuff, therefore unrelated
+    "SSH_TTY",
+    "TMUX",  # Terminal Multiplexer
+    "XDG_DATA_DIRS",  # XDG: Desktop environment stuff
+    "XDG_GREETER_DATA_DIR",  # XDG: Desktop environment stuff
+    "XDG_RUNTIME_DIR",
+    "_",  # current Python interpreter
+}
+
+CUDA_RUNTIME_LIB_PATTERNS = (
+    "cudart64*.dll",  # Windows
+    "libcudart*.so*",  # libcudart.so, libcudart.so.11.0, libcudart.so.12.0, libcudart.so.12.1, libcudart.so.12.2 etc.
+    "nvcuda*.dll",  # Windows
+)
+
+logger = logging.getLogger(__name__)
+
+
+def find_cuda_libraries_in_path_list(paths_list_candidate: str) -> Iterable[Path]:
+    for dir_string in paths_list_candidate.split(os.pathsep):
+        if not dir_string:
+            continue
+        if os.sep not in dir_string:
+            continue
+        try:
+            dir = Path(dir_string)
+            try:
+                if not dir.exists():
+                    logger.warning(f"The directory listed in your path is found to be non-existent: {dir}")
+                    continue
+            except OSError:  # Assume an esoteric error trying to poke at the directory
+                pass
+            for lib_pattern in CUDA_RUNTIME_LIB_PATTERNS:
+                for pth in dir.glob(lib_pattern):
+                    if pth.is_file():
+                        yield pth
+        except (OSError, PermissionError):
+            pass
+
+
+def is_relevant_candidate_env_var(env_var: str, value: str) -> bool:
+    return (
+        env_var in CUDART_PATH_PREFERRED_ENVVARS  # is a preferred location
+        or (
+            os.sep in value  # might contain a path
+            and env_var not in CUDART_PATH_IGNORED_ENVVARS  # not ignored
+            and "CONDA" not in env_var  # not another conda envvar
+            and "BASH_FUNC" not in env_var  # not a bash function defined via envvar
+            and "\n" not in value  # likely e.g. a script or something?
+        )
+    )
+
+
+def get_potentially_lib_path_containing_env_vars() -> Dict[str, str]:
+    return {env_var: value for env_var, value in os.environ.items() if is_relevant_candidate_env_var(env_var, value)}
+
+
+def find_cudart_libraries() -> Iterator[Path]:
+    """
+    Searches for a cuda installations, in the following order of priority:
+        1. active conda env
+        2. LD_LIBRARY_PATH
+        3. any other env vars, while ignoring those that
+            - are known to be unrelated
+            - don't contain the path separator `/`
+
+    If multiple libraries are found in part 3, we optimistically try one,
+    while giving a warning message.
+    """
+    candidate_env_vars = get_potentially_lib_path_containing_env_vars()
+
+    for envvar in CUDART_PATH_PREFERRED_ENVVARS:
+        if envvar in candidate_env_vars:
+            directory = candidate_env_vars[envvar]
+            yield from find_cuda_libraries_in_path_list(directory)
+            candidate_env_vars.pop(envvar)
+
+    for env_var, value in candidate_env_vars.items():
+        yield from find_cuda_libraries_in_path_list(value)
+
+
+def print_cuda_diagnostics(cuda_specs: CUDASpecs) -> None:
+    print(
+        f"PyTorch settings found: CUDA_VERSION={cuda_specs.cuda_version_string}, "
+        f"Highest Compute Capability: {cuda_specs.highest_compute_capability}.",
+    )
+
+    binary_path = get_cuda_bnb_library_path(cuda_specs)
+    if not binary_path.exists():
+        print_dedented(
+            f"""
+        Library not found: {binary_path}. Maybe you need to compile it from source?
+        If you compiled from source, try again with `make CUDA_VERSION=DETECTED_CUDA_VERSION`,
+        for example, `make CUDA_VERSION=113`.
+
+        The CUDA version for the compile might depend on your conda install, if using conda.
+        Inspect CUDA version via `conda list | grep cuda`.
+        """,
+        )
+
+    cuda_major, cuda_minor = cuda_specs.cuda_version_tuple
+    if cuda_major < 11:
+        print_dedented(
+            """
+            WARNING: CUDA versions lower than 11 are currently not supported for LLM.int8().
+            You will be only to use 8-bit optimizers and quantization routines!
+            """,
+        )
+
+    print(f"To manually override the PyTorch CUDA version please see: {NONPYTORCH_DOC_URL}")
+
+    # 7.5 is the minimum CC for int8 tensor cores
+    if not cuda_specs.has_imma:
+        print_dedented(
+            """
+            WARNING: Compute capability < 7.5 detected! Only slow 8-bit matmul is supported for your GPU!
+            If you run into issues with 8-bit matmul, you can try 4-bit quantization:
+            https://huggingface.co/blog/4bit-transformers-bitsandbytes
+            """,
+        )
+
+    # TODO:
+    # (1) CUDA missing cases (no CUDA installed by CUDA driver (nvidia-smi accessible)
+    # (2) Multiple CUDA versions installed
+
+
+def print_cuda_runtime_diagnostics() -> None:
+    cudart_paths = list(find_cudart_libraries())
+    if not cudart_paths:
+        print("CUDA SETUP: WARNING! CUDA runtime files not found in any environmental path.")
+    elif len(cudart_paths) > 1:
+        print_dedented(
+            f"""
+            Found duplicate CUDA runtime files (see below).
+
+            We select the PyTorch default CUDA runtime, which is {torch.version.cuda},
+            but this might mismatch with the CUDA version that is needed for bitsandbytes.
+            To override this behavior set the `BNB_CUDA_VERSION=<version string, e.g. 122>` environmental variable.
+
+            For example, if you want to use the CUDA version 122,
+                BNB_CUDA_VERSION=122 python ...
+
+            OR set the environmental variable in your .bashrc:
+                export BNB_CUDA_VERSION=122
+
+            In the case of a manual override, make sure you set LD_LIBRARY_PATH, e.g.
+            export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.2,
+            """,
+        )
+        for pth in cudart_paths:
+            print(f"* Found CUDA runtime at: {pth}")

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/diagnostics/main.py

@@ -0,0 +1,85 @@
+import sys
+import traceback
+
+import torch
+
+from bitsandbytes.consts import PACKAGE_GITHUB_URL
+from bitsandbytes.cuda_specs import get_cuda_specs
+from bitsandbytes.diagnostics.cuda import (
+    print_cuda_diagnostics,
+    print_cuda_runtime_diagnostics,
+)
+from bitsandbytes.diagnostics.utils import print_dedented, print_header
+
+
+def sanity_check():
+    from bitsandbytes.cextension import lib
+
+    if lib is None:
+        print_dedented(
+            """
+            Couldn't load the bitsandbytes library, likely due to missing binaries.
+            Please ensure bitsandbytes is properly installed.
+
+            For source installations, compile the binaries with `cmake -DCOMPUTE_BACKEND=cuda -S .`.
+            See the documentation for more details if needed.
+
+            Trying a simple check anyway, but this will likely fail...
+            """,
+        )
+
+    from bitsandbytes.optim import Adam
+
+    p = torch.nn.Parameter(torch.rand(10, 10).cuda())
+    a = torch.rand(10, 10).cuda()
+    p1 = p.data.sum().item()
+    adam = Adam([p])
+    out = a * p
+    loss = out.sum()
+    loss.backward()
+    adam.step()
+    p2 = p.data.sum().item()
+    assert p1 != p2
+
+
+def main():
+    print_header("")
+    print_header("BUG REPORT INFORMATION")
+    print_header("")
+
+    print_header("OTHER")
+    cuda_specs = get_cuda_specs()
+    print("CUDA specs:", cuda_specs)
+    if not torch.cuda.is_available():
+        print("Torch says CUDA is not available. Possible reasons:")
+        print("1. CUDA driver not installed")
+        print("2. CUDA not installed")
+        print("3. You have multiple conflicting CUDA libraries")
+    if cuda_specs:
+        print_cuda_diagnostics(cuda_specs)
+    print_cuda_runtime_diagnostics()
+    print_header("")
+    print_header("DEBUG INFO END")
+    print_header("")
+    print("Checking that the library is importable and CUDA is callable...")
+    try:
+        sanity_check()
+        print("SUCCESS!")
+        print("Installation was successful!")
+        return
+    except ImportError:
+        print(
+            f"WARNING: {__package__} is currently running as CPU-only!\n"
+            "Therefore, 8-bit optimizers and GPU quantization are unavailable.\n\n"
+            f"If you think that this is so erroneously,\nplease report an issue!",
+        )
+    except Exception:
+        traceback.print_exc()
+    print_dedented(
+        f"""
+        Above we output some debug information.
+        Please provide this info when creating an issue via {PACKAGE_GITHUB_URL}/issues/new/choose
+        WARNING: Please be sure to sanitize sensitive info from the output before posting it.
+        """,
+    )
+    sys.exit(1)

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/diagnostics/utils.py

@@ -0,0 +1,12 @@
+import textwrap
+
+HEADER_WIDTH = 60
+
+
+def print_header(txt: str, width: int = HEADER_WIDTH, filler: str = "+") -> None:
+    txt = f" {txt} " if txt else ""
+    print(txt.center(width, filler))
+
+
+def print_dedented(text):
+    print("\n".join(textwrap.dedent(text).strip().split("\n")))

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda121.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:906a547b09630095e0355cc3780c84169e1b222a00571b0f738760da5f3df665
+size 25822992

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda124.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06689764b9601734d0e8f5bec701255fd38d4480de704d637800c8874348717a
+size 25123696

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/nn/__init__.py

@@ -0,0 +1,26 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+#
+# This source code is licensed under the MIT license found in the
+# LICENSE file in the root directory of this source tree.
+from .modules import (
+    Embedding,
+    Embedding4bit,
+    Embedding8bit,
+    EmbeddingFP4,
+    EmbeddingNF4,
+    Int8Params,
+    Linear4bit,
+    Linear8bitLt,
+    LinearFP4,
+    LinearNF4,
+    OutlierAwareLinear,
+    Params4bit,
+    StableEmbedding,
+    SwitchBackLinearBnb,
+)
+from .triton_based_modules import (
+    StandardLinear,
+    SwitchBackLinear,
+    SwitchBackLinearGlobal,
+    SwitchBackLinearVectorwise,
+)

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/nn/modules.py

@@ -0,0 +1,1061 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+#
+# This source code is licensed under the MIT license found in the
+# LICENSE file in the root directory of this source tree.
+import copy
+from typing import Any, Dict, Optional, TypeVar, Union, overload
+import warnings
+
+import torch
+from torch import Tensor, device, dtype, nn
+import torch.nn.functional as F
+
+import bitsandbytes as bnb
+from bitsandbytes.autograd._functions import get_tile_inds, undo_layout
+from bitsandbytes.functional import QuantState
+from bitsandbytes.optim import GlobalOptimManager
+from bitsandbytes.utils import (
+    INVERSE_LINEAR_8BIT_WEIGHTS_FORMAT_MAPPING,
+    OutlierTracer,
+)
+
+T = TypeVar("T", bound="torch.nn.Module")
+
+
+class StableEmbedding(torch.nn.Embedding):
+    """
+    Custom embedding layer designed to improve stability during training for NLP tasks by using 32-bit optimizer states. It is designed to reduce gradient variations that can result from quantization. This embedding layer is initialized with Xavier uniform initialization followed by layer normalization.
+
+    Example:
+
+    ```
+    # Initialize StableEmbedding layer with vocabulary size 1000, embedding dimension 300
+    embedding_layer = StableEmbedding(num_embeddings=1000, embedding_dim=300)
+
+    # Reset embedding parameters
+    embedding_layer.reset_parameters()
+
+    # Perform a forward pass with input tensor
+    input_tensor = torch.tensor([1, 2, 3])
+    output_embedding = embedding_layer(input_tensor)
+    ```
+
+    Attributes:
+        norm (`torch.nn.LayerNorm`): Layer normalization applied after the embedding.
+
+    Methods:
+        reset_parameters(): Reset embedding parameters using Xavier uniform initialization.
+        forward(input: Tensor) -> Tensor: Forward pass through the stable embedding layer.
+    """
+
+    def __init__(
+        self,
+        num_embeddings: int,
+        embedding_dim: int,
+        padding_idx: Optional[int] = None,
+        max_norm: Optional[float] = None,
+        norm_type: float = 2.0,
+        scale_grad_by_freq: bool = False,
+        sparse: bool = False,
+        _weight: Optional[Tensor] = None,
+        device=None,
+        dtype=None,
+    ) -> None:
+        """
+        Args:
+            num_embeddings (`int`):
+                The number of unique embeddings (vocabulary size).
+            embedding_dim (`int`):
+                The dimensionality of the embedding.
+            padding_idx (`Optional[int]`):
+                Pads the output with zeros at the given index.
+            max_norm (`Optional[float]`):
+                Renormalizes embeddings to have a maximum L2 norm.
+            norm_type (`float`, defaults to `2.0`):
+                The p-norm to compute for the `max_norm` option.
+            scale_grad_by_freq (`bool`, defaults to `False`):
+                Scale gradient by frequency during backpropagation.
+            sparse (`bool`, defaults to `False`):
+                Computes dense gradients. Set to `True` to compute sparse gradients instead.
+            _weight (`Optional[Tensor]`):
+                Pretrained embeddings.
+        """
+        super().__init__(
+            num_embeddings,
+            embedding_dim,
+            padding_idx,
+            max_norm,
+            norm_type,
+            scale_grad_by_freq,
+            sparse,
+            _weight,
+            device,
+            dtype,
+        )
+        self.norm = torch.nn.LayerNorm(embedding_dim, device=device)
+        GlobalOptimManager.get_instance().register_module_override(self, "weight", {"optim_bits": 32})
+
+    def reset_parameters(self) -> None:
+        torch.nn.init.xavier_uniform_(self.weight)
+        self._fill_padding_idx_with_zero()
+
+    """ !!! This is a redefinition of _fill_padding_idx_with_zero in torch.nn.Embedding
+        to make the Layer compatible with Pytorch < 1.9.
+        This means that if this changes in future PyTorch releases this need to change too
+        which is cumbersome. However, with this we can ensure compatibility with previous
+        PyTorch releases.
+    """
+
+    def _fill_padding_idx_with_zero(self) -> None:
+        if self.padding_idx is not None:
+            with torch.no_grad():
+                self.weight[self.padding_idx].fill_(0)
+
+    def forward(self, input: Tensor) -> Tensor:
+        emb = F.embedding(
+            input,
+            self.weight,
+            self.padding_idx,
+            self.max_norm,
+            self.norm_type,
+            self.scale_grad_by_freq,
+            self.sparse,
+        )
+
+        # always apply layer norm in full precision
+        emb = emb.to(torch.get_default_dtype())
+
+        return self.norm(emb).to(self.weight.dtype)
+
+
+class Embedding(torch.nn.Embedding):
+    """
+    Embedding class to store and retrieve word embeddings from their indices.
+    """
+
+    def __init__(
+        self,
+        num_embeddings: int,
+        embedding_dim: int,
+        padding_idx: Optional[int] = None,
+        max_norm: Optional[float] = None,
+        norm_type: float = 2.0,
+        scale_grad_by_freq: bool = False,
+        sparse: bool = False,
+        _weight: Optional[Tensor] = None,
+        device: Optional[device] = None,
+    ) -> None:
+        """
+        Args:
+            num_embeddings (`int`):
+                The number of unique embeddings (vocabulary size).
+            embedding_dim (`int`):
+                The dimensionality of the embedding.
+            padding_idx (`Optional[int]`):
+                Pads the output with zeros at the given index.
+            max_norm (`Optional[float]`):
+                Renormalizes embeddings to have a maximum L2 norm.
+            norm_type (`float`, defaults to `2.0`):
+                The p-norm to compute for the `max_norm` option.
+            scale_grad_by_freq (`bool`, defaults to `False`):
+                Scale gradient by frequency during backpropagation.
+            sparse (`bool`, defaults to `False`):
+                Computes dense gradients. Set to `True` to compute sparse gradients instead.
+            _weight (`Optional[Tensor]`):
+                Pretrained embeddings.
+        """
+        super().__init__(
+            num_embeddings,
+            embedding_dim,
+            padding_idx,
+            max_norm,
+            norm_type,
+            scale_grad_by_freq,
+            sparse,
+            _weight,
+            device=device,
+        )
+        GlobalOptimManager.get_instance().register_module_override(self, "weight", {"optim_bits": 32})
+
+    def reset_parameters(self) -> None:
+        torch.nn.init.xavier_uniform_(self.weight)
+        self._fill_padding_idx_with_zero()
+
+    """ !!! This is a redefinition of _fill_padding_idx_with_zero in torch.nn.Embedding
+        to make the Layer compatible with Pytorch < 1.9.
+        This means that if this changes in future PyTorch releases this need to change too
+        which is cumbersome. However, with this we can ensure compatibility with previous
+        PyTorch releases.
+    """
+
+    def _fill_padding_idx_with_zero(self) -> None:
+        if self.padding_idx is not None:
+            with torch.no_grad():
+                self.weight[self.padding_idx].fill_(0)
+
+    def forward(self, input: Tensor) -> Tensor:
+        emb = F.embedding(
+            input,
+            self.weight,
+            self.padding_idx,
+            self.max_norm,
+            self.norm_type,
+            self.scale_grad_by_freq,
+            self.sparse,
+        )
+
+        return emb
+
+
+class Params4bit(torch.nn.Parameter):
+    def __new__(
+        cls,
+        data: Optional[torch.Tensor] = None,
+        requires_grad=False,  # quantized weights should be frozen by default
+        quant_state: Optional[QuantState] = None,
+        blocksize: int = 64,
+        compress_statistics: bool = True,
+        quant_type: str = "fp4",
+        quant_storage: torch.dtype = torch.uint8,
+        module: Optional["Linear4bit"] = None,
+        bnb_quantized: bool = False,
+    ) -> "Params4bit":
+        if data is None:
+            data = torch.empty(0)
+
+        self = torch.Tensor._make_subclass(cls, data, requires_grad)
+        self.blocksize = blocksize
+        self.compress_statistics = compress_statistics
+        self.quant_type = quant_type
+        self.quant_state = quant_state
+        self.quant_storage = quant_storage
+        self.bnb_quantized = bnb_quantized
+        self.data = data
+        self.module = module
+        return self
+
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        state["data"] = self.data
+        state["requires_grad"] = self.requires_grad
+        return state
+
+    def __setstate__(self, state):
+        self.requires_grad = state["requires_grad"]
+        self.blocksize = state["blocksize"]
+        self.compress_statistics = state["compress_statistics"]
+        self.quant_type = state["quant_type"]
+        self.quant_state = state["quant_state"]
+        self.data = state["data"]
+        self.quant_storage = state["quant_storage"]
+        self.bnb_quantized = state["bnb_quantized"]
+        self.module = state["module"]
+
+    def __deepcopy__(self, memo):
+        new_instance = type(self).__new__(type(self))
+        state = self.__getstate__()
+        new_instance.__setstate__(state)
+        new_instance.quant_state = copy.deepcopy(state["quant_state"])
+        new_instance.data = copy.deepcopy(state["data"])
+        return new_instance
+
+    def __copy__(self):
+        new_instance = type(self).__new__(type(self))
+        state = self.__getstate__()
+        new_instance.__setstate__(state)
+        return new_instance
+
+    @classmethod
+    def from_prequantized(
+        cls,
+        data: torch.Tensor,
+        quantized_stats: Dict[str, Any],
+        requires_grad: bool = False,
+        device="cuda",
+        module: Optional["Linear4bit"] = None,
+        **kwargs,
+    ) -> "Params4bit":
+        self = torch.Tensor._make_subclass(cls, data.to(device))
+        self.requires_grad = requires_grad
+        self.quant_state = QuantState.from_dict(qs_dict=quantized_stats, device=device)
+        self.blocksize = self.quant_state.blocksize
+        self.compress_statistics = self.quant_state.nested
+        self.quant_type = self.quant_state.quant_type
+        self.bnb_quantized = True
+
+        self.quant_storage = data.dtype
+        self.module = module
+
+        if self.module is not None:
+            self.module.quant_state = self.quant_state
+
+        return self
+
+    def _quantize(self, device):
+        w = self.data.contiguous().to(device)
+        w_4bit, quant_state = bnb.functional.quantize_4bit(
+            w,
+            blocksize=self.blocksize,
+            compress_statistics=self.compress_statistics,
+            quant_type=self.quant_type,
+            quant_storage=self.quant_storage,
+        )
+        self.data = w_4bit
+        self.quant_state = quant_state
+        if self.module is not None:
+            self.module.quant_state = quant_state
+        self.bnb_quantized = True
+        return self
+
+    def cuda(self, device: Optional[Union[int, device, str]] = None, non_blocking: bool = False):
+        return self.to(device="cuda" if device is None else device, non_blocking=non_blocking)
+
+    @overload
+    def to(
+        self: T,
+        device: Optional[Union[int, device]] = ...,
+        dtype: Optional[Union[dtype, str]] = ...,
+        non_blocking: bool = ...,
+    ) -> T: ...
+
+    @overload
+    def to(self: T, dtype: Union[dtype, str], non_blocking: bool = ...) -> T: ...
+
+    @overload
+    def to(self: T, tensor: Tensor, non_blocking: bool = ...) -> T: ...
+
+    def to(self, *args, **kwargs):
+        device, dtype, non_blocking, convert_to_format = torch._C._nn._parse_to(*args, **kwargs)
+
+        if device is not None and device.type == "cuda" and not self.bnb_quantized:
+            return self._quantize(device)
+        else:
+            if self.quant_state is not None:
+                self.quant_state.to(device)
+
+            new_param = Params4bit(
+                super().to(device=device, dtype=dtype, non_blocking=non_blocking),
+                requires_grad=self.requires_grad,
+                quant_state=self.quant_state,
+                blocksize=self.blocksize,
+                compress_statistics=self.compress_statistics,
+                quant_type=self.quant_type,
+                quant_storage=self.quant_storage,
+            )
+
+            return new_param
+
+
+def fix_4bit_weight_quant_state_from_module(module: Union["Embedding4bit", "Linear4bit"]):
+    if getattr(module.weight, "quant_state", None) is not None:
+        return
+
+    if getattr(module, "quant_state", None) is None:
+        warnings.warn(
+            "FP4 quantization state not initialized. Please call .cuda() or .to(device) on the LinearFP4 layer first.",
+        )
+
+    # the quant state got lost when the parameter got converted. This happens for example for fsdp
+    # since we registered the module, we can recover the state here
+    assert module.weight.shape[1] == 1
+    if not isinstance(module.weight, Params4bit):
+        module.weight = Params4bit(module.weight, quant_storage=module.quant_storage, bnb_quantized=True)
+    module.weight.quant_state = module.quant_state
+
+
+class Linear4bit(nn.Linear):
+    """
+    This class is the base module for the 4-bit quantization algorithm presented in [QLoRA](https://arxiv.org/abs/2305.14314).
+    QLoRA 4-bit linear layers uses blockwise k-bit quantization under the hood, with the possibility of selecting various
+    compute datatypes such as FP4 and NF4.
+
+    In order to quantize a linear layer one should first load the original fp16 / bf16 weights into
+    the Linear4bit module, then call `quantized_module.to("cuda")` to quantize the fp16 / bf16 weights.
+
+    Example:
+
+    ```python
+    import torch
+    import torch.nn as nn
+
+    import bitsandbytes as bnb
+    from bnb.nn import Linear4bit
+
+    fp16_model = nn.Sequential(
+        nn.Linear(64, 64),
+        nn.Linear(64, 64)
+    )
+
+    quantized_model = nn.Sequential(
+        Linear4bit(64, 64),
+        Linear4bit(64, 64)
+    )
+
+    quantized_model.load_state_dict(fp16_model.state_dict())
+    quantized_model = quantized_model.to(0) # Quantization happens here
+    ```
+    """
+
+    def __init__(
+        self,
+        input_features,
+        output_features,
+        bias=True,
+        compute_dtype=None,
+        compress_statistics=True,
+        quant_type="fp4",
+        quant_storage=torch.uint8,
+        device=None,
+    ):
+        """
+        Initialize Linear4bit class.
+
+        Args:
+            input_features (`str`):
+                Number of input features of the linear layer.
+            output_features (`str`):
+                Number of output features of the linear layer.
+            bias (`bool`, defaults to `True`):
+                Whether the linear class uses the bias term as well.
+        """
+        super().__init__(input_features, output_features, bias, device)
+        self.weight = Params4bit(
+            self.weight.data,
+            requires_grad=False,
+            compress_statistics=compress_statistics,
+            quant_type=quant_type,
+            quant_storage=quant_storage,
+            module=self,
+        )
+        # self.persistent_buffers = []  # TODO consider as way to save quant state
+        self.compute_dtype = compute_dtype
+        self.compute_type_is_set = False
+        self.quant_state = None
+        self.quant_storage = quant_storage
+
+    def set_compute_type(self, x):
+        if x.dtype in [torch.float32, torch.bfloat16]:
+            # the input is in a dtype that is safe to compute in, we switch
+            # to this type for speed and stability
+            self.compute_dtype = x.dtype
+        elif x.dtype == torch.float16:
+            # we take the compoute dtype passed into the layer
+            if self.compute_dtype == torch.float32 and (x.numel() == x.shape[-1]):
+                # single batch inference with input torch.float16 and compute_dtype float32 -> slow inference when it could be fast
+                # warn the user about this
+                warnings.warn(
+                    "Input type into Linear4bit is torch.float16, but bnb_4bit_compute_dtype=torch.float32 (default). This will lead to slow inference.",
+                )
+                warnings.filterwarnings("ignore", message=".*inference.")
+            if self.compute_dtype == torch.float32 and (x.numel() != x.shape[-1]):
+                warnings.warn(
+                    "Input type into Linear4bit is torch.float16, but bnb_4bit_compute_dtype=torch.float32 (default). This will lead to slow inference or training speed.",
+                )
+                warnings.filterwarnings("ignore", message=".*inference or training")
+
+    def _save_to_state_dict(self, destination, prefix, keep_vars):
+        """
+        save weight and bias,
+        then fill state_dict with components of quant_state
+        """
+        super()._save_to_state_dict(destination, prefix, keep_vars)  # saving weight and bias
+
+        if getattr(self.weight, "quant_state", None) is not None:
+            for k, v in self.weight.quant_state.as_dict(packed=True).items():
+                destination[prefix + "weight." + k] = v if keep_vars else v.detach()
+
+    def forward(self, x: torch.Tensor):
+        fix_4bit_weight_quant_state_from_module(self)
+
+        # weights are cast automatically as Int8Params, but the bias has to be cast manually
+        if self.bias is not None and self.bias.dtype != x.dtype:
+            self.bias.data = self.bias.data.to(x.dtype)
+
+        if not self.compute_type_is_set:
+            self.set_compute_type(x)
+            self.compute_type_is_set = True
+
+        inp_dtype = x.dtype
+        if self.compute_dtype is not None:
+            x = x.to(self.compute_dtype)
+
+        bias = None if self.bias is None else self.bias.to(self.compute_dtype)
+
+        return bnb.matmul_4bit(x, self.weight.t(), bias=bias, quant_state=self.weight.quant_state).to(inp_dtype)
+
+
+class LinearFP4(Linear4bit):
+    """
+    Implements the FP4 data type.
+    """
+
+    def __init__(
+        self,
+        input_features,
+        output_features,
+        bias=True,
+        compute_dtype=None,
+        compress_statistics=True,
+        quant_storage=torch.uint8,
+        device=None,
+    ):
+        """
+        Args:
+            input_features (`str`):
+                Number of input features of the linear layer.
+            output_features (`str`):
+                Number of output features of the linear layer.
+            bias (`bool`, defaults to `True`):
+                Whether the linear class uses the bias term as well.
+        """
+        super().__init__(
+            input_features,
+            output_features,
+            bias,
+            compute_dtype,
+            compress_statistics,
+            "fp4",
+            quant_storage,
+            device,
+        )
+
+
+class LinearNF4(Linear4bit):
+    """Implements the NF4 data type.
+
+    Constructs a quantization data type where each bin has equal area under a standard normal distribution N(0, 1) that
+    is normalized into the range [-1, 1].
+
+    For more information read the paper: QLoRA: Efficient Finetuning of Quantized LLMs (https://arxiv.org/abs/2305.14314)
+
+    Implementation of the NF4 data type in bitsandbytes can be found in the `create_normal_map` function in
+    the `functional.py` file: https://github.com/TimDettmers/bitsandbytes/blob/main/bitsandbytes/functional.py#L236.
+    """
+
+    def __init__(
+        self,
+        input_features,
+        output_features,
+        bias=True,
+        compute_dtype=None,
+        compress_statistics=True,
+        quant_storage=torch.uint8,
+        device=None,
+    ):
+        """
+        Args:
+            input_features (`str`):
+                Number of input features of the linear layer.
+            output_features (`str`):
+                Number of output features of the linear layer.
+            bias (`bool`, defaults to `True`):
+                Whether the linear class uses the bias term as well.
+        """
+        super().__init__(
+            input_features,
+            output_features,
+            bias,
+            compute_dtype,
+            compress_statistics,
+            "nf4",
+            quant_storage,
+            device,
+        )
+
+
+class Int8Params(torch.nn.Parameter):
+    def __new__(
+        cls,
+        data: Optional[torch.Tensor] = None,
+        requires_grad=True,
+        has_fp16_weights=False,
+        CB: Optional[torch.Tensor] = None,
+        SCB: Optional[torch.Tensor] = None,
+    ):
+        if data is None:
+            data = torch.empty(0)
+        obj = torch.Tensor._make_subclass(cls, data, requires_grad)
+        obj.CB = CB
+        obj.SCB = SCB
+        obj.has_fp16_weights = has_fp16_weights
+        return obj
+
+    def cuda(self, device):
+        if self.has_fp16_weights:
+            return super().cuda(device)
+        else:
+            # We quantize the weight and store in 8bit row-major
+            B = self.data.contiguous().half().cuda(device)
+            CB, SCB, _ = bnb.functional.int8_vectorwise_quant(B)
+            self.data = CB
+            self.CB = CB
+            self.SCB = SCB
+
+        return self
+
+    def __deepcopy__(self, memo):
+        # adjust this if new arguments are added to the constructor
+        new_instance = type(self).__new__(
+            type(self),
+            data=copy.deepcopy(self.data, memo),
+            requires_grad=self.requires_grad,
+            has_fp16_weights=self.has_fp16_weights,
+            CB=copy.deepcopy(self.CB, memo),
+            SCB=copy.deepcopy(self.SCB, memo),
+        )
+        return new_instance
+
+    @overload
+    def to(
+        self: T,
+        device: Optional[Union[int, device]] = ...,
+        dtype: Optional[Union[dtype, str]] = ...,
+        non_blocking: bool = ...,
+    ) -> T: ...
+
+    @overload
+    def to(self: T, dtype: Union[dtype, str], non_blocking: bool = ...) -> T: ...
+
+    @overload
+    def to(self: T, tensor: Tensor, non_blocking: bool = ...) -> T: ...
+
+    def to(self, *args, **kwargs):
+        device, dtype, non_blocking, convert_to_format = torch._C._nn._parse_to(*args, **kwargs)
+
+        if device is not None and device.type == "cuda" and self.data.device.type == "cpu":
+            return self.cuda(device)
+        else:
+            new_param = Int8Params(
+                super().to(device=device, dtype=dtype, non_blocking=non_blocking),
+                requires_grad=self.requires_grad,
+                has_fp16_weights=self.has_fp16_weights,
+            )
+            new_param.CB = self.CB
+            new_param.SCB = self.SCB
+
+            return new_param
+
+
+def maybe_rearrange_weight(state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs):
+    weight = state_dict.get(f"{prefix}weight")
+    if weight is None:
+        # if the state dict has no weights for this layer (e.g., LoRA finetuning), do nothing
+        return
+    weight_format = state_dict.pop(f"{prefix}weight_format", "row")
+
+    if isinstance(weight_format, torch.Tensor):
+        weight_format = weight_format.item()
+
+    # For new weights format storage type, we explicitly check
+    # if weights_format is on the mapping
+    if isinstance(weight_format, int) and weight_format not in INVERSE_LINEAR_8BIT_WEIGHTS_FORMAT_MAPPING:
+        raise ValueError(f"Expected supported weight format - got {weight_format}")
+    elif isinstance(weight_format, int) and weight_format in INVERSE_LINEAR_8BIT_WEIGHTS_FORMAT_MAPPING:
+        weight_format = INVERSE_LINEAR_8BIT_WEIGHTS_FORMAT_MAPPING[weight_format]
+
+    if weight_format != "row":
+        tile_indices = get_tile_inds(weight_format, weight.device)
+        state_dict[f"{prefix}weight"] = undo_layout(weight, tile_indices)
+
+
+class Embedding8bit(nn.Embedding):
+    """
+    This class implements [LLM.int8()](https://arxiv.org/abs/2208.07339) algorithm for embedding layer
+
+    Quantization API is similar to Linear8bitLt:
+    ```python
+    import torch
+    import torch.nn as nn
+
+    from bitsandbytes.nn import Embedding8bit
+
+    fp16_module = nn.Embedding(128, 64)
+    int8_module = Embedding8bit(128, 64)
+
+    int8_module.load_state_dict(fp16_module.state_dict())
+
+    int8_module = int8_module.to(0) # Quantization happens here
+    ```
+    """
+
+    def __init__(self, num_embeddings, embedding_dim, device=None, dtype=None):
+        super().__init__(num_embeddings, embedding_dim, device=device, dtype=dtype)
+        self.dtype = self.weight.data.dtype
+
+        self.weight = Int8Params(self.weight.data, has_fp16_weights=False, requires_grad=False)
+
+    def _save_to_state_dict(self, destination, prefix, keep_vars):
+        raise NotImplementedError("Saving Embedding8bit module is not implemented")
+
+    def forward(self, input: Tensor) -> Tensor:
+        if not hasattr(self.weight, "SCB"):
+            raise RuntimeError("Embedding layer is not quantized. Please call .cuda() or .to(device) first.")
+
+        rows = self.weight.data
+        row_stats = self.weight.SCB
+
+        assert rows.shape == (self.num_embeddings, self.embedding_dim)
+        assert row_stats.shape == (self.num_embeddings,)
+
+        compressed_output = F.embedding(input, rows)
+        compressed_output_stats = F.embedding(input, row_stats.view(self.num_embeddings, 1))
+
+        output = compressed_output * (compressed_output_stats / 127.0)
+
+        return output.to(self.dtype)
+
+
+class Embedding4bit(nn.Embedding):
+    """
+    This is the base class similar to Linear4bit. It implements the 4-bit quantization algorithm presented in
+    [QLoRA](https://arxiv.org/abs/2305.14314) for embeddings.
+
+    Quantization API is similar to Linear4bit:
+    ```python
+    import torch
+    import torch.nn as nn
+
+    from bitsandbytes.nn import Embedding4bit
+
+    fp16_module = nn.Embedding(128, 64)
+    quantized_module = Embedding4bit(128, 64)
+
+    quantized_module.load_state_dict(fp16_module.state_dict())
+
+    quantized_module = quantized_module.to(0) # Quantization happens here
+    ```
+    """
+
+    def __init__(
+        self,
+        num_embeddings,
+        embedding_dim,
+        dtype=None,
+        quant_type="fp4",
+        quant_storage=torch.uint8,
+        device=None,
+    ):
+        super().__init__(num_embeddings, embedding_dim, device=device, dtype=dtype)
+        self.dtype = self.weight.data.dtype
+
+        self.weight = Params4bit(
+            self.weight.data,
+            requires_grad=False,
+            compress_statistics=None,
+            quant_type=quant_type,
+            quant_storage=quant_storage,
+            module=self,
+        )
+
+        blocksize = self.weight.blocksize
+
+        if embedding_dim % blocksize != 0:
+            warnings.warn(
+                f"Embedding size {embedding_dim} is not divisible by block size {blocksize}. "
+                "This will lead to slow inference.",
+            )
+
+    def _forward_with_partial_dequantize(self, input: Tensor):
+        assert self.embedding_dim % self.weight.quant_state.blocksize == 0
+
+        w_4bit_uint8 = self.weight.data.view(torch.uint8).view(self.num_embeddings * self.embedding_dim // 2, 1)
+
+        output_4bit = torch.nn.functional.embedding(
+            weight=w_4bit_uint8.view(self.num_embeddings, self.embedding_dim // 2),
+            input=input,
+        ).view(-1, 1)
+        assert output_4bit.shape == (input.numel() * self.embedding_dim // 2, 1)
+
+        blocks_per_emb = self.embedding_dim // self.weight.blocksize
+
+        absmax = self.weight.quant_state.absmax
+        assert absmax.shape == (self.num_embeddings * blocks_per_emb,)
+
+        output_absmax = torch.nn.functional.embedding(
+            weight=absmax.view(self.num_embeddings, blocks_per_emb),
+            input=input,
+        ).view(
+            -1,
+        )
+        assert output_absmax.shape == (input.numel() * blocks_per_emb,)
+
+        output_quant_state = copy.deepcopy(self.weight.quant_state)
+        output_quant_state.absmax = output_absmax
+        output_quant_state.shape = torch.Size((*input.shape, self.embedding_dim))
+
+        output = bnb.functional.dequantize_4bit(output_4bit, output_quant_state)
+        assert output.shape == (*input.shape, self.embedding_dim)
+
+        return output.to(self.dtype)
+
+    def _save_to_state_dict(self, destination, prefix, keep_vars):
+        raise NotImplementedError("Saving Embedding4bit module is not implemented")
+
+    def forward(self, input: Tensor) -> Tensor:
+        fix_4bit_weight_quant_state_from_module(self)
+
+        if self.embedding_dim % self.weight.quant_state.blocksize == 0:
+            return self._forward_with_partial_dequantize(input)
+
+        dequantized_weight = bnb.functional.dequantize_4bit(self.weight.data, self.weight.quant_state)
+
+        return torch.nn.functional.embedding(
+            weight=dequantized_weight,
+            input=input,
+        ).to(self.dtype)
+
+
+class EmbeddingFP4(Embedding4bit):
+    def __init__(
+        self,
+        num_embeddings,
+        embedding_dim,
+        dtype=None,
+        quant_storage=torch.uint8,
+        device=None,
+    ):
+        super().__init__(
+            num_embeddings,
+            embedding_dim,
+            dtype=dtype,
+            quant_type="fp4",
+            quant_storage=quant_storage,
+            device=device,
+        )
+
+
+class EmbeddingNF4(Embedding4bit):
+    def __init__(
+        self,
+        num_embeddings,
+        embedding_dim,
+        dtype=None,
+        quant_storage=torch.uint8,
+        device=None,
+    ):
+        super().__init__(
+            num_embeddings,
+            embedding_dim,
+            dtype=dtype,
+            quant_type="nf4",
+            quant_storage=quant_storage,
+            device=device,
+        )
+
+
+class Linear8bitLt(nn.Linear):
+    """
+    This class is the base module for the [LLM.int8()](https://arxiv.org/abs/2208.07339) algorithm.
+    To read more about it, have a look at the paper.
+
+    In order to quantize a linear layer one should first load the original fp16 / bf16 weights into
+    the Linear8bitLt module, then call `int8_module.to("cuda")` to quantize the fp16 weights.
+
+    Example:
+
+    ```python
+    import torch
+    import torch.nn as nn
+
+    import bitsandbytes as bnb
+    from bnb.nn import Linear8bitLt
+
+    fp16_model = nn.Sequential(
+        nn.Linear(64, 64),
+        nn.Linear(64, 64)
+    )
+
+    int8_model = nn.Sequential(
+        Linear8bitLt(64, 64, has_fp16_weights=False),
+        Linear8bitLt(64, 64, has_fp16_weights=False)
+    )
+
+    int8_model.load_state_dict(fp16_model.state_dict())
+    int8_model = int8_model.to(0) # Quantization happens here
+    ```
+    """
+
+    def __init__(
+        self,
+        input_features: int,
+        output_features: int,
+        bias=True,
+        has_fp16_weights=True,
+        threshold=0.0,
+        index=None,
+        device=None,
+    ):
+        """
+        Initialize Linear8bitLt class.
+
+        Args:
+            input_features (`int`):
+                Number of input features of the linear layer.
+            output_features (`int`):
+                Number of output features of the linear layer.
+            bias (`bool`, defaults to `True`):
+                Whether the linear class uses the bias term as well.
+        """
+        super().__init__(input_features, output_features, bias, device)
+        self.state = bnb.MatmulLtState()
+        self.index = index
+
+        self.state.threshold = threshold
+        self.state.has_fp16_weights = has_fp16_weights
+
+        if threshold > 0.0 and not has_fp16_weights:
+            self.state.use_pool = True
+
+        self.weight = Int8Params(self.weight.data, has_fp16_weights=has_fp16_weights, requires_grad=has_fp16_weights)
+        self._register_load_state_dict_pre_hook(maybe_rearrange_weight)
+
+    def _save_to_state_dict(self, destination, prefix, keep_vars):
+        super()._save_to_state_dict(destination, prefix, keep_vars)
+
+        # we only need to save SCB as extra data, because CB for quantized weights is already stored in weight.data
+        scb_name = "SCB"
+
+        # case 1: .cuda was called, SCB is in self.weight
+        param_from_weight = getattr(self.weight, scb_name)
+        # case 2: self.init_8bit_state was called, SCB is in self.state
+        param_from_state = getattr(self.state, scb_name)
+
+        key_name = prefix + f"{scb_name}"
+
+        # We now only save in row-major. This format information is stored for backwards compatibility.
+        format_name = prefix + "weight_format"
+
+        if not self.state.has_fp16_weights:
+            if param_from_weight is not None:
+                destination[key_name] = param_from_weight if keep_vars else param_from_weight.detach()
+                destination[format_name] = torch.tensor(0, dtype=torch.uint8)
+            elif param_from_state is not None:
+                destination[key_name] = param_from_state if keep_vars else param_from_state.detach()
+                destination[format_name] = torch.tensor(0, dtype=torch.uint8)
+
+    def _load_from_state_dict(
+        self,
+        state_dict,
+        prefix,
+        local_metadata,
+        strict,
+        missing_keys,
+        unexpected_keys,
+        error_msgs,
+    ):
+        super()._load_from_state_dict(
+            state_dict,
+            prefix,
+            local_metadata,
+            strict,
+            missing_keys,
+            unexpected_keys,
+            error_msgs,
+        )
+        unexpected_copy = list(unexpected_keys)
+
+        for key in unexpected_copy:
+            input_name = key[len(prefix) :]
+            if input_name == "SCB":
+                if self.weight.SCB is None:
+                    # buffers not yet initialized, can't access them directly without quantizing first
+                    raise RuntimeError(
+                        "Loading a quantized checkpoint into non-quantized Linear8bitLt is "
+                        "not supported. Please call module.cuda() before module.load_state_dict()",
+                    )
+
+                input_param = state_dict[key]
+                self.weight.SCB.copy_(input_param)
+
+                if self.state.SCB is not None:
+                    self.state.SCB = self.weight.SCB
+
+                unexpected_keys.remove(key)
+
+    def init_8bit_state(self):
+        self.state.CB = self.weight.CB
+        self.state.SCB = self.weight.SCB
+        self.weight.CB = None
+        self.weight.SCB = None
+
+    def forward(self, x: torch.Tensor):
+        self.state.is_training = self.training
+        if self.weight.CB is not None:
+            self.init_8bit_state()
+
+        # weights are cast automatically as Int8Params, but the bias has to be cast manually
+        if self.bias is not None and self.bias.dtype != x.dtype:
+            self.bias.data = self.bias.data.to(x.dtype)
+
+        out = bnb.matmul(x, self.weight, bias=self.bias, state=self.state)
+
+        if not self.state.has_fp16_weights and self.state.CB is not None:
+            self.weight.data = self.state.CB
+
+        return out
+
+
+class OutlierAwareLinear(nn.Linear):
+    def __init__(self, input_features, output_features, bias=True, device=None):
+        super().__init__(input_features, output_features, bias, device)
+        self.outlier_dim = None
+        self.is_quantized = False
+
+    def forward_with_outliers(self, x, outlier_idx):
+        raise NotImplementedError("Please override the `forward_with_outliers(self, x, outlier_idx)` function")
+
+    def quantize_weight(self, w, outlier_idx):
+        raise NotImplementedError("Please override the `quantize_weights(self, w, outlier_idx)` function")
+
+    def forward(self, x):
+        if self.outlier_dim is None:
+            tracer = OutlierTracer.get_instance()
+            if not tracer.is_initialized():
+                print("Please use OutlierTracer.initialize(model) before using the OutlierAwareLinear layer")
+            outlier_idx = tracer.get_outliers(self.weight)
+            # print(outlier_idx, tracer.get_hvalue(self.weight))
+            self.outlier_dim = outlier_idx
+
+        if not self.is_quantized:
+            w = self.quantize_weight(self.weight, self.outlier_dim)
+            self.weight.data.copy_(w)
+            self.is_quantized = True
+
+
+class SwitchBackLinearBnb(nn.Linear):
+    def __init__(
+        self,
+        input_features,
+        output_features,
+        bias=True,
+        has_fp16_weights=True,
+        memory_efficient_backward=False,
+        threshold=0.0,
+        index=None,
+        device=None,
+    ):
+        super().__init__(input_features, output_features, bias, device)
+        self.state = bnb.MatmulLtState()
+        self.index = index
+
+        self.state.threshold = threshold
+        self.state.has_fp16_weights = has_fp16_weights
+        self.state.memory_efficient_backward = memory_efficient_backward
+        if threshold > 0.0 and not has_fp16_weights:
+            self.state.use_pool = True
+
+        self.weight = Int8Params(self.weight.data, has_fp16_weights=has_fp16_weights, requires_grad=has_fp16_weights)
+
+    def init_8bit_state(self):
+        self.state.CB = self.weight.CB
+        self.state.SCB = self.weight.SCB
+        self.weight.CB = None
+        self.weight.SCB = None
+
+    def forward(self, x):
+        self.state.is_training = self.training
+
+        if self.weight.CB is not None:
+            self.init_8bit_state()
+
+        out = bnb.matmul_mixed(x.half(), self.weight.half(), bias=None, state=self.state) + self.bias

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/nn/triton_based_modules.py

@@ -0,0 +1,264 @@
+from functools import partial
+
+import torch
+import torch.nn as nn
+
+from bitsandbytes.triton.dequantize_rowwise import dequantize_rowwise
+from bitsandbytes.triton.int8_matmul_mixed_dequantize import (
+    int8_matmul_mixed_dequantize,
+)
+from bitsandbytes.triton.int8_matmul_rowwise_dequantize import (
+    int8_matmul_rowwise_dequantize,
+)
+from bitsandbytes.triton.quantize_columnwise_and_transpose import (
+    quantize_columnwise_and_transpose,
+)
+from bitsandbytes.triton.quantize_global import (
+    quantize_global,
+    quantize_global_transpose,
+)
+from bitsandbytes.triton.quantize_rowwise import quantize_rowwise
+from bitsandbytes.triton.triton_utils import is_triton_available
+
+
+class _switchback_global(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, X_3D, W, bias):
+        # reshape input to [N * L, D]
+        X = X_3D.view(-1, X_3D.size(-1))
+
+        # rowwise quantize for X, global quantize for W
+        X_int8, state_X = quantize_rowwise(X)
+        W_int8, state_W = quantize_global(W)
+
+        # save for backward.
+        ctx.save_for_backward = X, W
+
+        # matmult, fused dequant and add bias
+        # call "mixed" because we are mixing rowwise quantized and global quantized
+        return int8_matmul_mixed_dequantize(X_int8, W_int8.t(), state_X, state_W, bias).view(*X_3D.size()[:-1], -1)
+
+    @staticmethod
+    def backward(ctx, G_3D):
+        # reshape input to [N_out * L, D]
+        G = G_3D.reshape(-1, G_3D.size(-1))
+
+        grad_X = grad_W = grad_bias = None
+
+        X, W = ctx.save_for_backward
+        if ctx.needs_input_grad[0]:
+            # rowwise quantize for G, global quantize for W
+            # for W, we also fuse the transpose operation because only A @ B^T is supported
+            # so we transpose once then call .t() in the matmul
+            G_int8, state_G = quantize_rowwise(G)
+            W_int8, state_W = quantize_global_transpose(W)
+            grad_X = int8_matmul_mixed_dequantize(G_int8, W_int8.t(), state_G, state_W, None).view(
+                *G_3D.size()[:-1],
+                -1,
+            )
+        if ctx.needs_input_grad[1]:
+            # backward pass uses standard weight grad
+            grad_W = torch.matmul(G.t(), X.to(G.dtype))
+        if ctx.needs_input_grad[2]:
+            grad_bias = G.sum(dim=0)
+
+        return grad_X, grad_W, grad_bias
+
+
+class _switchback_vectorrize(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, X_3D, W, bias):
+        # reshape input to [N * L, D]
+        X = X_3D.view(-1, X_3D.size(-1))
+
+        ctx.save_for_backward = X, W
+        # rowwise quantize for X
+        # columnwise quantize for W (first rowwise, transpose later)
+        X_int8, state_X = quantize_rowwise(X)
+        W_int8, state_W = quantize_rowwise(W)
+
+        # matmult, fused dequant and add bias
+        # call kernel which expects rowwise quantized X and W
+        return int8_matmul_rowwise_dequantize(X_int8, W_int8.t(), state_X, state_W, bias).view(*X_3D.size()[:-1], -1)
+
+    @staticmethod
+    def backward(ctx, G_3D):
+        X, W = ctx.save_for_backward
+
+        G = G_3D.reshape(-1, G_3D.size(-1))
+
+        grad_X = grad_W = grad_bias = None
+
+        if ctx.needs_input_grad[0]:
+            # rowwise quantize for G, columnwise quantize for W and fused transpose
+            # we call .t() for weight later because only A @ B^T is supported
+            G_int8, state_G = quantize_rowwise(G)
+            W_int8, state_W = quantize_columnwise_and_transpose(W)
+            grad_X = int8_matmul_rowwise_dequantize(G_int8, W_int8.t(), state_G, state_W, None).view(
+                *G_3D.size()[:-1],
+                -1,
+            )
+        if ctx.needs_input_grad[1]:
+            # backward pass uses standard weight grad
+            grad_W = torch.matmul(G.t(), X.to(G.dtype))
+        if ctx.needs_input_grad[2]:
+            grad_bias = G.sum(dim=0)
+
+        return grad_X, grad_W, grad_bias
+
+
+class _switchback_global_mem_efficient(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, X_3D, W, bias):
+        # reshape input to [N * L, D]
+        X = X_3D.view(-1, X_3D.size(-1))
+        X_3D_sz = X_3D.size()
+
+        # rowwise quantize for X, global quantize for W
+        X_int8, state_X = quantize_rowwise(X)
+        del X
+        W_int8, state_W = quantize_global(W)
+
+        # save for backward.
+        ctx.save_for_backward = X_int8, state_X, W_int8, state_W
+
+        # matmult, fused dequant and add bias
+        # call "mixed" because we are mixing rowwise quantized and global quantized
+        return int8_matmul_mixed_dequantize(X_int8, W_int8.t(), state_X, state_W, bias).view(*X_3D_sz[:-1], -1)
+
+    @staticmethod
+    def backward(ctx, G_3D):
+        # reshape input to [N_out * L, D]
+        G = G_3D.reshape(-1, G_3D.size(-1))
+        G_3D_sz = G_3D.size()
+
+        grad_X = grad_W = grad_bias = None
+
+        X_int8, state_X, W_int8, state_W = ctx.save_for_backward
+        if ctx.needs_input_grad[1]:
+            real_X = dequantize_rowwise(X_int8, state_X)
+            del X_int8
+            grad_W = torch.matmul(G.t(), real_X.to(G.dtype))
+            del real_X
+        if ctx.needs_input_grad[2]:
+            grad_bias = G.sum(dim=0)
+        if ctx.needs_input_grad[0]:
+            G_int8, state_G = quantize_rowwise(G)
+            del G
+            W_int8 = W_int8.t().contiguous()
+            grad_X = int8_matmul_mixed_dequantize(G_int8, W_int8.t(), state_G, state_W, None).view(*G_3D_sz[:-1], -1)
+
+        return grad_X, grad_W, grad_bias
+
+
+class SwitchBackLinear(nn.Linear):
+    def __init__(
+        self,
+        in_features: int,
+        out_features: int,
+        bias: bool = True,
+        device=None,
+        dtype=None,
+        vector_wise_quantization: bool = False,
+        mem_efficient: bool = False,
+    ):
+        super().__init__(in_features, out_features, bias, device, dtype)
+
+        if not is_triton_available():
+            raise ImportError("""Could not import triton. Please install triton to use SwitchBackLinear.
+                               Alternatively, you can use bnb.nn.SwitchBackLinearBnb, but it will be slower""")
+
+        # By default, we use the global quantization.
+        self.vector_wise_quantization = vector_wise_quantization
+        if self.vector_wise_quantization:
+            self._fn = _switchback_vectorrize
+            if mem_efficient:
+                print("mem efficient is not supported for vector-wise quantization.")
+                exit(1)
+        else:
+            if mem_efficient:
+                self._fn = _switchback_global_mem_efficient
+            else:
+                self._fn = _switchback_global
+
+    def prepare_for_eval(self):
+        # If we just want to do eval, we can pre-quantize the weights instead of doing it on the forward pass.
+        # Note this is experimental and not tested thoroughly.
+        # Note this needs to be explicitly called with something like
+        # def cond_prepare(m):
+        #     if hasattr(m, "prepare_for_eval"):
+        #         m.prepare_for_eval()
+        # model.apply(cond_prepare)
+        print("=> preparing for eval.")
+        if self.vector_wise_quantization:
+            W_int8, state_W = quantize_rowwise(self.weight)
+        else:
+            W_int8, state_W = quantize_global(self.weight)
+
+        self.register_buffer("W_int8", W_int8)
+        self.register_buffer("state_W", state_W)
+
+        del self.weight
+
+    def forward(self, x):
+        if self.training:
+            return self._fn.apply(x, self.weight, self.bias)
+        else:
+            # If it hasn't been "prepared for eval", run the standard forward pass.
+            if not hasattr(self, "W_int8"):
+                return self._fn.apply(x, self.weight, self.bias)
+
+            # Otherwise, use pre-computed weights.
+            X = x.view(-1, x.size(-1))
+            X_int8, state_X = quantize_rowwise(X)
+
+            if self.vector_wise_quantization:
+                return int8_matmul_rowwise_dequantize(X_int8, self.W_int8.t(), state_X, self.state_W, self.bias).view(
+                    *x.size()[:-1],
+                    -1,
+                )
+            else:
+                return int8_matmul_mixed_dequantize(X_int8, self.W_int8.t(), state_X, self.state_W, self.bias).view(
+                    *x.size()[:-1],
+                    -1,
+                )
+
+
+SwitchBackLinearGlobal = partial(SwitchBackLinear, vector_wise_quantization=False)
+SwitchBackLinearGlobalMemEfficient = partial(SwitchBackLinear, vector_wise_quantization=False, mem_efficient=True)
+SwitchBackLinearVectorwise = partial(SwitchBackLinear, vector_wise_quantization=True)
+
+
+# This is just the standard linear function.
+class StandardLinearFunction(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, input, weight, bias=None):
+        X = input.view(-1, input.size(-1))
+
+        ctx.save_for_backward(X, weight, bias)
+        output = input.matmul(weight.t())
+        if bias is not None:
+            output += bias.unsqueeze(0).expand_as(output)
+        return output.view(*input.size()[:-1], -1)
+
+    @staticmethod
+    def backward(ctx, grad_output_3D):
+        input, weight, bias = ctx.saved_tensors
+
+        grad_output = grad_output_3D.reshape(-1, grad_output_3D.size(-1))
+
+        grad_input = grad_weight = grad_bias = None
+
+        if ctx.needs_input_grad[0]:
+            grad_input = grad_output.matmul(weight.to(grad_output.dtype)).view(*grad_output_3D.size()[:-1], -1)
+        if ctx.needs_input_grad[1]:
+            grad_weight = grad_output.t().matmul(input.to(grad_output.dtype))
+        if bias is not None and ctx.needs_input_grad[2]:
+            grad_bias = grad_output.sum(0)
+
+        return grad_input, grad_weight, grad_bias
+
+
+class StandardLinear(nn.Linear):
+    def forward(self, x):
+        return StandardLinearFunction.apply(x, self.weight, self.bias)

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/research/__init__.py

@@ -0,0 +1,6 @@
+from . import nn
+from .autograd._functions import (
+    matmul_fp8_global,
+    matmul_fp8_mixed,
+    switchback_bnb,
+)

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/research/autograd/_functions.py

@@ -0,0 +1,396 @@
+from functools import reduce  # Required in Python 3
+import operator
+from typing import Optional
+import warnings
+
+import torch
+
+from bitsandbytes.autograd._functions import GlobalOutlierPooler, MatmulLtState
+import bitsandbytes.functional as F
+
+
+# math.prod not compatible with python < 3.8
+def prod(iterable):
+    return reduce(operator.mul, iterable, 1)
+
+
+class MatMulFP8Mixed(torch.autograd.Function):
+    # forward is the same, but we added the fallback for pre-turing GPUs
+    # backward is mostly the same, but adds one extra clause (see "elif state.CxB is not None")
+
+    @staticmethod
+    def forward(ctx, A, B, out=None, fw_code=None, bw_code=None, bsz=1024, bsz2=1024):
+        # default of pytorch behavior if inputs are empty
+        ctx.is_empty = False
+        if prod(A.shape) == 0:
+            ctx.is_empty = True
+            ctx.A = A
+            ctx.B = B
+
+            B_shape = B.shape
+            if A.shape[-1] == B_shape[0]:
+                return torch.empty(A.shape[:-1] + B_shape[1:], dtype=A.dtype, device=A.device)
+            else:
+                return torch.empty(A.shape[:-1] + B_shape[:1], dtype=A.dtype, device=A.device)
+
+        # 1. Dequantize
+        # 2. MatmulnN
+        cA, state = F.quantize_blockwise(A, code=fw_code, blocksize=bsz)
+        fp8A = F.dequantize_blockwise(cA, state, blocksize=bsz).to(A.dtype)
+
+        cB, state = F.quantize(B.float(), code=fw_code)
+        fp8B = F.dequantize(cB, state).to(B.dtype)
+
+        output = torch.matmul(fp8A, fp8B)
+
+        # output is half
+
+        # 3. Save state
+        ctx.fw_code = fw_code
+        ctx.bw_code = bw_code
+        ctx.bsz = bsz
+        ctx.bsz2 = bsz2
+        ctx.dtype_A, ctx.dtype_B = A.dtype, B.dtype
+
+        if any(ctx.needs_input_grad[:2]):
+            # NOTE: we send back A, and re-quant.
+            ctx.tensors = (A, fp8B)
+        else:
+            ctx.tensors = (None, None)
+
+        return output
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        if ctx.is_empty:
+            return torch.zeros_like(ctx.A), torch.zeros_like(ctx.B), None, None, None, None, None
+
+        req_gradA, req_gradB, _, _, _, _, _ = ctx.needs_input_grad
+        A, B = ctx.tensors
+
+        grad_A, grad_B = None, None
+
+        # TODO: Fix blocksize to be output_dim
+        cgrad_out, state = F.quantize_blockwise(grad_output, code=ctx.bw_code, blocksize=ctx.bsz2)
+        fp8out = F.dequantize_blockwise(cgrad_out, state, blocksize=ctx.bsz2).to(grad_output.dtype)
+
+        # cgrad_output_2, state_2 = F.quantize(grad_output.float(), code=ctx.bw_code)
+        # fp8out_2 = F.dequantize(cgrad_output_2, state_2).to(grad_output.dtype)
+
+        # grad_output_reshape = grad_output.reshape(-1, grad_output.shape[-1]).contiguous()
+        # fp8grad_transpose, stategrad_transpose = F.vectorwise_quant(grad_output_reshape, dim=0, quant_type='vector')
+        # fp8out_transpose = (fp8grad_transpose / 7) * stategrad_transpose
+        # fp8out_transpose = fp8out_transpose.view(grad_output.shape[0], grad_output.shape[1], grad_output.shape[2])
+
+        # not supported by PyTorch. TODO: create work-around
+        if req_gradA:
+            grad_A = torch.matmul(fp8out, B.t().to(fp8out.dtype)).to(A.dtype)
+
+        if req_gradB:
+            if len(A.shape) == 3:
+                At = A.transpose(2, 1).contiguous()
+            else:
+                At = A.transpose(1, 0).contiguous()
+            # cA, state = F.quantize(At.float(), code=ctx.fw_code)
+            # fp8At = F.dequantize(cA, state).to(A.dtype)
+            grad_B = torch.matmul(At.to(grad_output.dtype), grad_output).to(B.dtype)
+
+        return grad_A, grad_B, None, None, None, None, None
+
+
+class MatMulFP8Global(torch.autograd.Function):
+    # forward is the same, but we added the fallback for pre-turing GPUs
+    # backward is mostly the same, but adds one extra clause (see "elif state.CxB is not None")
+
+    @staticmethod
+    def forward(ctx, A, B, out=None, fw_code=None, bw_code=None, bsz=1024, bsz2=1024):
+        # default of pytorch behavior if inputs are empty
+        ctx.is_empty = False
+        if prod(A.shape) == 0:
+            ctx.is_empty = True
+            ctx.A = A
+            ctx.B = B
+
+            B_shape = B.shape
+            if A.shape[-1] == B_shape[0]:
+                return torch.empty(A.shape[:-1] + B_shape[1:], dtype=A.dtype, device=A.device)
+            else:
+                return torch.empty(A.shape[:-1] + B_shape[:1], dtype=A.dtype, device=A.device)
+
+        # 1. Dequantize
+        # 2. MatmulnN
+        cA, state = F.quantize(A.float(), code=fw_code)
+        fp8A = F.dequantize(cA, state).to(A.dtype)
+
+        cB, state = F.quantize(B.float(), code=fw_code)
+        fp8B = F.dequantize(cB, state).to(B.dtype)
+
+        output = torch.matmul(fp8A, fp8B)
+
+        # output is half
+
+        # 3. Save state
+        ctx.fw_code = fw_code
+        ctx.bw_code = bw_code
+        ctx.bsz = bsz
+        ctx.bsz2 = bsz2
+        ctx.dtype_A, ctx.dtype_B = A.dtype, B.dtype
+
+        if any(ctx.needs_input_grad[:2]):
+            # NOTE: we send back A, and re-quant.
+            ctx.tensors = (A, fp8B)
+        else:
+            ctx.tensors = (None, None)
+
+        return output
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        if ctx.is_empty:
+            return torch.zeros_like(ctx.A), torch.zeros_like(ctx.B), None, None, None, None, None
+
+        req_gradA, req_gradB, _, _, _, _, _ = ctx.needs_input_grad
+        A, B = ctx.tensors
+
+        grad_A, grad_B = None, None
+
+        # TODO: Fix blocksize to be output_dim
+        cgrad_out, state = F.quantize(grad_output.float(), code=ctx.bw_code)
+        fp8out = F.dequantize(cgrad_out, state).to(grad_output.dtype)
+
+        # cgrad_output_2, state_2 = F.quantize(grad_output.float(), code=ctx.bw_code)
+        # fp8out_2 = F.dequantize(cgrad_output_2, state_2).to(grad_output.dtype)
+
+        # grad_output_reshape = grad_output.reshape(-1, grad_output.shape[-1]).contiguous()
+        # fp8grad_transpose, stategrad_transpose = F.vectorwise_quant(grad_output_reshape, dim=0, quant_type='vector')
+        # fp8out_transpose = (fp8grad_transpose / 7) * stategrad_transpose
+        # fp8out_transpose = fp8out_transpose.view(grad_output.shape[0], grad_output.shape[1], grad_output.shape[2])
+
+        # not supported by PyTorch. TODO: create work-around
+        if req_gradA:
+            grad_A = torch.matmul(fp8out, B.t().to(fp8out.dtype)).to(A.dtype)
+
+        if req_gradB:
+            if len(A.shape) == 3:
+                At = A.transpose(2, 1).contiguous()
+            else:
+                At = A.transpose(1, 0).contiguous()
+            cA, state = F.quantize(At.float(), code=ctx.fw_code)
+            fp8At = F.dequantize(cA, state).to(A.dtype)
+            grad_B = torch.matmul(fp8At.to(fp8out.dtype), fp8out).to(B.dtype)
+
+        return grad_A, grad_B, None, None, None, None, None
+
+
+class SwitchBackBnb(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, A, B, out=None, bias=None, state: Optional[MatmulLtState] = None):
+        state = state or MatmulLtState()
+
+        # default to pytorch behavior if inputs are empty
+        ctx.is_empty = False
+        if prod(A.shape) == 0:
+            ctx.is_empty = True
+            ctx.A = A
+            ctx.B = B
+            ctx.bias = bias
+            if A.shape[-1] == B.shape[0]:
+                return torch.empty(A.shape[:-1] + B.shape[1:], dtype=A.dtype, device=A.device)
+            else:
+                return torch.empty(A.shape[:-1] + B.shape[:1], dtype=A.dtype, device=A.device)
+
+        # 1. Quantize A
+        # 2. Quantize B
+        # 3. Matmul
+        # 4. Mixed-precision decomposition matmul
+        # 5. Save state
+        input_shape = A.shape
+        if state.outlier_pool is None:
+            state.outlier_pool = GlobalOutlierPooler.get_instance()
+
+        # Cast A to fp16
+        if A.dtype != torch.float16:
+            warnings.warn(f"MatMul8bitLt: inputs will be cast from {A.dtype} to float16 during quantization")
+
+        # 1. Quantize A
+        if len(A.shape) == 3:
+            A = A.view(-1, A.shape[-1]).contiguous()
+        CA, CAt, SCA, SCAt, outlier_cols = F.int8_double_quant(A.to(torch.float16), threshold=state.threshold)
+
+        if state.threshold > 0.0 and outlier_cols is not None:
+            if state.has_fp16_weights:
+                idx = outlier_cols
+                CA[:, idx] = 0
+                subA = A[:, idx]
+                state.subB = B[:, idx].t().contiguous()
+                state.idx = idx
+            else:
+                if state.SB is None:
+                    state.SB = (state.CB.shape, "row")
+        else:
+            if not state.has_fp16_weights and state.SB is None:
+                state.SB = (state.CB.shape, "row")
+            subA = None
+
+        # 2. Quantize B
+        if state.has_fp16_weights:
+            # print('B shape', B.shape)
+            has_grad = True if (getattr(B, "grad", None) is not None) else False
+            is_transposed = not B.is_contiguous() and B.shape[0] == B.stride(1)
+            if is_transposed:
+                B = B.contiguous()
+
+            if (state.is_training and not has_grad) or state.SB is None:
+                state.reset_grads()
+                (
+                    state.CB,
+                    state.CBt,
+                    state.SCB,
+                    state.SCBt,
+                    _,
+                ) = F.int8_double_quant(B.to(torch.float16))
+                state.SB = (state.CB.shape, "row")
+        else:
+            has_grad = False
+
+        if outlier_cols is not None and not state.has_fp16_weights:
+            # extract outliers
+            state.idx = outlier_cols
+            outliers = state.CB[:, state.idx.long()].clone()
+            state.subB = (outliers * state.SCB.view(-1, 1) / 127.0).t().contiguous().to(A.dtype)
+            CA[:, state.idx.long()] = 0
+
+            subA = A[:, state.idx.long()]
+
+        shapeB = state.SB[0]
+
+        if len(input_shape) == 3:
+            output_shape = (input_shape[0], input_shape[1], shapeB[0])
+        else:
+            output_shape = (input_shape[0], shapeB[0])
+
+        # 3. Matmul
+        out32 = F.int8_linear_matmul(CA, state.CB)
+        # we apply the fused bias here
+
+        if bias is None or bias.dtype == torch.float16:
+            output = F.int8_mm_dequant(out32, SCA, state.SCB, bias=bias).to(A.dtype)
+        else:  # apply bias separately
+            output = F.int8_mm_dequant(out32, SCA, state.SCB, bias=None).to(A.dtype)
+            output.add_(bias)
+
+        # 4. Mixed-precision decomposition matmul
+        if outlier_cols is not None and subA is not None:
+            output += torch.matmul(subA, state.subB)
+
+        # 5. Save state
+        ctx.state = state
+
+        ctx.grad_shape = input_shape
+        ctx.dtype_A, ctx.dtype_B, ctx.dtype_bias = A.dtype, B.dtype, None if bias is None else bias.dtype
+
+        if any(ctx.needs_input_grad[:2]):
+            ctx.tensors = (CAt, subA, A)
+            ctx.tensor_states = (SCAt, state.idx)
+        else:
+            ctx.tensors = [None, None, None]
+            ctx.tensor_states = (None, None)
+            ctx.save_for_backward(None, None)
+
+        clone_func = torch.clone if len(output_shape) == 3 else lambda x: x
+        return clone_func(output.view(output_shape))
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        if ctx.is_empty:
+            bias_grad = None if ctx.bias is None else torch.zeros_like(ctx.bias)
+            return torch.zeros_like(ctx.A), torch.zeros_like(ctx.B), None, bias_grad, None
+
+        req_gradA, req_gradB, _, req_gradBias, _ = ctx.needs_input_grad
+        CAt, subA, A = ctx.tensors
+        SCAt, idx = ctx.tensor_states
+        state = ctx.state
+        grad_A = grad_B = grad_bias = None
+
+        if req_gradBias:
+            # compute grad_bias first before changing grad_output dtype
+            grad_bias = grad_output.sum(0, dtype=ctx.dtype_bias)
+
+        # Cast grad_output to fp16
+        if len(grad_output.shape) == 3:
+            grad_output = grad_output.reshape(-1, grad_output.shape[-1]).contiguous()
+
+        Cgrad, Cgradt, SCgrad, SCgradt, outlier_cols = F.int8_double_quant(grad_output.to(torch.float16))
+
+        if req_gradB:
+            # print('back A shape', A.shape)
+            # print('grad output t shape', grad_output.t().shape)
+            grad_B = torch.matmul(grad_output.t(), A)
+
+        if req_gradA:
+            if state.CB is not None:
+                CB = state.CB.to(ctx.dtype_A, copy=True).mul_(state.SCB.unsqueeze(1).mul(1.0 / 127.0))
+                grad_A = torch.matmul(grad_output, CB).view(ctx.grad_shape).to(ctx.dtype_A)
+            else:
+                raise Exception("State must contain either CBt or CB matrix for backward")
+
+        return grad_A, grad_B, None, grad_bias, None
+
+
+def get_block_sizes(input_matrix, weight_matrix):
+    input_features = input_matrix.shape[-1]
+    output_features = weight_matrix.shape[0] if weight_matrix.shape[1] == input_features else weight_matrix.shape[1]
+    array = [4096, 2048, 1024, 512, 256, 128, 64, 0]
+    bsz, bsz2 = 1024, 1024
+    for i, k in enumerate(array):
+        if input_features > array[i + 1]:
+            bsz = k
+            break
+    for i, k in enumerate(array):
+        if output_features > array[i + 1]:
+            bsz2 = k
+            break
+
+    return bsz, bsz2
+
+
+def matmul_fp8_global(
+    A: torch.Tensor,
+    B: torch.Tensor,
+    fw_code: torch.Tensor,
+    bw_code: torch.Tensor,
+    out: Optional[torch.Tensor] = None,
+    bsz: int = -1,
+    bsz2: int = -1,
+):
+    if bsz == -1 or bsz2 == -1:
+        bsz, bsz2 = get_block_sizes(A, B)
+    return MatMulFP8Global.apply(A, B, out, fw_code, bw_code, bsz, bsz2)
+
+
+def matmul_fp8_mixed(
+    A: torch.Tensor,
+    B: torch.Tensor,
+    fw_code: torch.Tensor,
+    bw_code: torch.Tensor,
+    out: Optional[torch.Tensor] = None,
+    bsz: int = -1,
+    bsz2: int = -1,
+):
+    if bsz == -1 or bsz2 == -1:
+        bsz, bsz2 = get_block_sizes(A, B)
+    return MatMulFP8Mixed.apply(A, B, out, fw_code, bw_code, bsz, bsz2)
+
+
+def switchback_bnb(
+    A: torch.Tensor,
+    B: torch.Tensor,
+    out: Optional[torch.Tensor] = None,
+    state: Optional[MatmulLtState] = None,
+    threshold=0.0,
+    bias=None,
+):
+    state = state or MatmulLtState()
+    if threshold > 0.0:
+        state.threshold = threshold
+    return SwitchBackBnb.apply(A, B, out, bias, state)

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/research/nn/__init__.py

@@ -0,0 +1 @@
+from .modules import LinearFP8Global, LinearFP8Mixed

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/research/nn/modules.py

@@ -0,0 +1,76 @@
+from typing import TypeVar
+
+import torch
+from torch import nn
+
+import bitsandbytes as bnb
+
+T = TypeVar("T", bound="torch.nn.Module")
+
+
+class LinearFP8Mixed(nn.Linear):
+    def __init__(self, input_features, output_features, bias=True):
+        super().__init__(input_features, output_features, bias)
+        self.bw_code = None
+        self.fw_code = None
+        array = [4096, 2048, 1024, 512, 256, 128, 64, 0]
+        for i, k in enumerate(array):
+            if input_features > array[i + 1]:
+                self.bsz = k
+                break
+        for i, k in enumerate(array):
+            if output_features > array[i + 1]:
+                self.bsz2 = k
+                break
+
+    def forward(self, x: torch.Tensor):
+        if self.fw_code is None:
+            self.bw_code = bnb.functional.create_fp8_map(True, 5, 2, 8).to(x.device)
+            self.fw_code = bnb.functional.create_fp8_map(True, 4, 3, 8).to(x.device)
+
+        out = bnb.research.matmul_fp8_mixed(
+            x,
+            self.weight.t(),
+            fw_code=self.fw_code,
+            bw_code=self.bw_code,
+            bsz=self.bsz,
+            bsz2=self.bsz2,
+        )
+        if self.bias is not None:
+            out += self.bias
+
+        return out
+
+
+class LinearFP8Global(nn.Linear):
+    def __init__(self, input_features, output_features, bias=True):
+        super().__init__(input_features, output_features, bias)
+        self.bw_code = None
+        self.fw_code = None
+        array = [4096, 2048, 1024, 512, 256, 128, 64, 0]
+        for i, k in enumerate(array):
+            if input_features > array[i + 1]:
+                self.bsz = k
+                break
+        for i, k in enumerate(array):
+            if output_features > array[i + 1]:
+                self.bsz2 = k
+                break
+
+    def forward(self, x: torch.Tensor):
+        if self.fw_code is None:
+            self.bw_code = bnb.functional.create_fp8_map(True, 5, 2, 8).to(x.device)
+            self.fw_code = bnb.functional.create_fp8_map(True, 4, 3, 8).to(x.device)
+
+        out = bnb.matmul_fp8_global(
+            x,
+            self.weight.t(),
+            fw_code=self.fw_code,
+            bw_code=self.bw_code,
+            bsz=self.bsz,
+            bsz2=self.bsz2,
+        )
+        if self.bias is not None:
+            out += self.bias
+
+        return out

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/triton/dequantize_rowwise.py

@@ -0,0 +1,64 @@
+import math
+
+import torch
+
+from bitsandbytes.triton.triton_utils import is_triton_available
+
+if not is_triton_available():
+
+    def dequantize_rowwise(x: torch.Tensor, state_x: torch.Tensor):
+        return None
+else:
+    import triton
+    import triton.language as tl
+
+    # rowwise quantize
+
+    # TODO: autotune this better.
+    @triton.autotune(
+        configs=[
+            triton.Config({}, num_stages=1, num_warps=8),
+            triton.Config({}, num_stages=2, num_warps=8),
+            triton.Config({}, num_stages=4, num_warps=8),
+            triton.Config({}, num_stages=8, num_warps=8),
+            triton.Config({}, num_stages=1),
+            triton.Config({}, num_stages=2),
+            triton.Config({}, num_stages=4),
+            triton.Config({}, num_stages=8),
+            triton.Config({}, num_warps=1),
+            triton.Config({}, num_warps=2),
+            triton.Config({}, num_warps=4),
+            triton.Config({}, num_warps=8),
+        ],
+        key=["n_elements"],
+    )
+    @triton.jit
+    def _dequantize_rowwise(
+        x_ptr,
+        state_x,
+        output_ptr,
+        inv_127,
+        n_elements,
+        BLOCK_SIZE: tl.constexpr,
+        P2: tl.constexpr,
+    ):
+        pid = tl.program_id(axis=0)
+        block_start = pid * BLOCK_SIZE
+        arange = tl.arange(0, P2)
+        offsets = block_start + arange
+        row_mask = arange < BLOCK_SIZE
+        x = tl.load(x_ptr + offsets, mask=row_mask)
+        max_val = tl.load(state_x + pid)
+        output = max_val * x * inv_127
+        tl.store(output_ptr + offsets, output, mask=row_mask)
+
+    def dequantize_rowwise(x: torch.Tensor, state_x: torch.Tensor):
+        output = torch.empty(*x.shape, device=x.device, dtype=torch.float16)
+
+        P2 = int(2 ** (math.ceil(math.log2(x.shape[1]))))
+
+        assert x.is_cuda and output.is_cuda
+        n_elements = output.numel()
+        grid = lambda meta: (x.shape[0],)
+        _dequantize_rowwise[grid](x, state_x, output, 1.0 / 127, n_elements, BLOCK_SIZE=x.shape[1], P2=P2)
+        return output

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/triton/int8_matmul_mixed_dequantize.py

@@ -0,0 +1,205 @@
+import torch
+
+from bitsandbytes.triton.triton_utils import is_triton_available
+
+if not is_triton_available():
+
+    def int8_matmul_mixed_dequantize(a, b, state_x, state_w, bias):
+        return None
+else:
+    import triton
+    import triton.language as tl
+    from triton.ops.matmul_perf_model import early_config_prune, estimate_matmul_time
+
+    # This is a matmul kernel based on triton.ops.matmul
+    # It is modified to support rowwise quantized input and global quantized weight
+    # It's purpose is fused matmul then dequantize
+    # It does support bias.
+
+    def init_to_zero(name):
+        return lambda nargs: nargs[name].zero_()
+
+    def get_configs_io_bound():
+        configs = []
+        for num_stages in [2, 3, 4, 5, 6]:
+            for block_m in [16, 32]:
+                for block_k in [32, 64]:
+                    for block_n in [32, 64, 128, 256]:
+                        num_warps = 2 if block_n <= 64 else 4
+                        configs.append(
+                            triton.Config(
+                                {"BLOCK_M": block_m, "BLOCK_N": block_n, "BLOCK_K": block_k, "SPLIT_K": 1},
+                                num_stages=num_stages,
+                                num_warps=num_warps,
+                            ),
+                        )
+                        # split_k
+                        for split_k in [2, 4, 8, 16]:
+                            configs.append(
+                                triton.Config(
+                                    {"BLOCK_M": block_m, "BLOCK_N": block_n, "BLOCK_K": block_k, "SPLIT_K": split_k},
+                                    num_stages=num_stages,
+                                    num_warps=num_warps,
+                                    pre_hook=init_to_zero("C"),
+                                ),
+                            )
+        return configs
+
+    @triton.autotune(
+        configs=[
+            # basic configs for compute-bound matmuls
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 256, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 128, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 64, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 256, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 128, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 64, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 128, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 32, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 32, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=5, num_warps=2),
+            # good for int8
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 256, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 128, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 64, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 256, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 128, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 64, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 128, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 32, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 32, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=5, num_warps=2),
+            *get_configs_io_bound(),
+        ],
+        key=["M", "N", "K"],
+        prune_configs_by={"early_config_prune": early_config_prune, "perf_model": estimate_matmul_time, "top_k": 10},
+    )
+    @triton.heuristics(
+        {
+            "EVEN_K": lambda args: args["K"] % (args["BLOCK_K"] * args["SPLIT_K"]) == 0,
+        },
+    )
+    @triton.jit
+    def _int8_matmul_mixed_dequantize(
+        A,
+        B,
+        C,
+        bias,
+        state_x_ptr,
+        state_w_ptr,
+        M,
+        N,
+        K,
+        divfactor: tl.constexpr,
+        has_bias: tl.constexpr,
+        stride_am,
+        stride_ak,
+        stride_bk,
+        stride_bn,
+        stride_cm,
+        stride_cn,
+        BLOCK_M: tl.constexpr,
+        BLOCK_N: tl.constexpr,
+        BLOCK_K: tl.constexpr,
+        GROUP_M: tl.constexpr,
+        SPLIT_K: tl.constexpr,
+        EVEN_K: tl.constexpr,
+        ACC_TYPE: tl.constexpr,
+    ):
+        # matrix multiplication
+        pid = tl.program_id(0)
+        pid_z = tl.program_id(1)
+        grid_m = tl.cdiv(M, BLOCK_M)
+        grid_n = tl.cdiv(N, BLOCK_N)
+        # re-order program ID for better L2 performance
+        width = GROUP_M * grid_n
+        group_id = pid // width
+        group_size = min(grid_m - group_id * GROUP_M, GROUP_M)
+        pid_m = group_id * GROUP_M + (pid % group_size)
+        pid_n = (pid % width) // (group_size)
+        # do matrix multiplication
+        rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+        rn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+        ram = tl.max_contiguous(tl.multiple_of(rm % M, BLOCK_M), BLOCK_M)
+        rbn = tl.max_contiguous(tl.multiple_of(rn % N, BLOCK_N), BLOCK_N)
+        rk = pid_z * BLOCK_K + tl.arange(0, BLOCK_K)
+        # pointers
+        A = A + (ram[:, None] * stride_am + rk[None, :] * stride_ak)
+        B = B + (rk[:, None] * stride_bk + rbn[None, :] * stride_bn)
+
+        # rematerialize rm and rn to save registers
+        rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+        rn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+
+        w_factor = tl.load(state_w_ptr)
+        x_factor = tl.load(state_x_ptr + ram)[:, None]
+
+        # acc = tl.zeros((BLOCK_M, BLOCK_N), dtype=ACC_TYPE)
+        acc = tl.zeros((BLOCK_M, BLOCK_N), dtype=tl.int32)
+        for k in range(0, tl.cdiv(K, BLOCK_K * SPLIT_K)):
+            if EVEN_K:
+                a = tl.load(A)
+                b = tl.load(B)
+            else:
+                k_remaining = K - k * (BLOCK_K * SPLIT_K)
+                a = tl.load(A, mask=rk[None, :] < k_remaining, other=0.0)
+                b = tl.load(B, mask=rk[:, None] < k_remaining, other=0.0)
+            acc += tl.dot(a, b)
+            A += BLOCK_K * SPLIT_K * stride_ak
+            B += BLOCK_K * SPLIT_K * stride_bk
+
+        acc = w_factor * (x_factor * (acc * divfactor))
+        acc = acc.to(C.dtype.element_ty)
+
+        # conditionally add bias
+        if has_bias:
+            bias = tl.load(bias + rn).to(C.dtype.element_ty)
+            acc = acc + bias[None, :]
+
+        C = C + (rm[:, None] * stride_cm + rn[None, :] * stride_cn)
+        mask = (rm < M)[:, None] & (rn < N)[None, :]
+        # handles write-back with reduction-splitting
+        if SPLIT_K == 1:
+            tl.store(C, acc, mask=mask)
+        else:
+            tl.atomic_add(C, acc, mask=mask)
+
+    def int8_matmul_mixed_dequantize(a, b, state_x, state_w, bias):
+        device = a.device
+        divfactor = 1.0 / (127.0 * 127.0)
+        has_bias = 0 if bias is None else 1
+        # handle non-contiguous inputs if necessary
+        if a.stride(0) > 1 and a.stride(1) > 1:
+            a = a.contiguous()
+        if b.stride(0) > 1 and b.stride(1) > 1:
+            b = b.contiguous()
+        # checks constraints
+        assert a.shape[1] == b.shape[0], "incompatible dimensions"
+        M, K = a.shape
+        _, N = b.shape
+        # allocates output
+        c = torch.empty((M, N), device=device, dtype=torch.float16)
+        # accumulator types
+        ACC_TYPE = tl.float32  # if a.dtype in [torch.float16, torch.bfloat16, torch.float32] else tl.int32
+        # launch int8_matmul_mixed_dequantize kernel
+        grid = lambda META: (triton.cdiv(M, META["BLOCK_M"]) * triton.cdiv(N, META["BLOCK_N"]), META["SPLIT_K"])
+        _int8_matmul_mixed_dequantize[grid](
+            a,
+            b,
+            c,
+            bias,
+            state_x,
+            state_w,
+            M,
+            N,
+            K,
+            divfactor,
+            has_bias,
+            a.stride(0),
+            a.stride(1),
+            b.stride(0),
+            b.stride(1),
+            c.stride(0),
+            c.stride(1),
+            GROUP_M=8,
+            ACC_TYPE=ACC_TYPE,
+        )
+        return c

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/triton/int8_matmul_rowwise_dequantize.py

@@ -0,0 +1,206 @@
+import torch
+
+from bitsandbytes.triton.triton_utils import is_triton_available
+
+if not is_triton_available():
+
+    def int8_matmul_rowwise_dequantize(a, b, state_x, state_w, bias):
+        return None
+else:
+    import triton
+    import triton.language as tl
+    from triton.ops.matmul_perf_model import early_config_prune, estimate_matmul_time
+
+    # This is a matmul kernel based on triton.ops.matmul
+    # It is modified to support rowwise quantized input and columnwise quantized weight
+    # It's purpose is fused matmul then dequantize
+    # It does support bias.
+
+    def init_to_zero(name):
+        return lambda nargs: nargs[name].zero_()
+
+    def get_configs_io_bound():
+        configs = []
+        for num_stages in [2, 3, 4, 5, 6]:
+            for block_m in [16, 32]:
+                for block_k in [32, 64]:
+                    for block_n in [32, 64, 128, 256]:
+                        num_warps = 2 if block_n <= 64 else 4
+                        configs.append(
+                            triton.Config(
+                                {"BLOCK_M": block_m, "BLOCK_N": block_n, "BLOCK_K": block_k, "SPLIT_K": 1},
+                                num_stages=num_stages,
+                                num_warps=num_warps,
+                            ),
+                        )
+                        # split_k
+                        for split_k in [2, 4, 8, 16]:
+                            configs.append(
+                                triton.Config(
+                                    {"BLOCK_M": block_m, "BLOCK_N": block_n, "BLOCK_K": block_k, "SPLIT_K": split_k},
+                                    num_stages=num_stages,
+                                    num_warps=num_warps,
+                                    pre_hook=init_to_zero("C"),
+                                ),
+                            )
+        return configs
+
+    @triton.autotune(
+        configs=[
+            # basic configs for compute-bound matmuls
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 256, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 128, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 64, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 256, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 128, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 64, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 128, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 32, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 32, "BLOCK_K": 32, "SPLIT_K": 1}, num_stages=5, num_warps=2),
+            # good for int8
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 256, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 128, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=3, num_warps=8),
+            triton.Config({"BLOCK_M": 256, "BLOCK_N": 64, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 256, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 128, "BLOCK_K": 128, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 64, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 128, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 32, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=4, num_warps=4),
+            triton.Config({"BLOCK_M": 64, "BLOCK_N": 32, "BLOCK_K": 64, "SPLIT_K": 1}, num_stages=5, num_warps=2),
+            *get_configs_io_bound(),
+        ],
+        key=["M", "N", "K"],
+        prune_configs_by={"early_config_prune": early_config_prune, "perf_model": estimate_matmul_time, "top_k": 10},
+    )
+    @triton.heuristics(
+        {
+            "EVEN_K": lambda args: args["K"] % (args["BLOCK_K"] * args["SPLIT_K"]) == 0,
+        },
+    )
+    @triton.jit
+    def _int8_matmul_rowwise_dequantize(
+        A,
+        B,
+        C,
+        bias,
+        state_x_ptr,
+        state_w_ptr,
+        M,
+        N,
+        K,
+        divfactor,
+        has_bias: tl.constexpr,
+        stride_am,
+        stride_ak,
+        stride_bk,
+        stride_bn,
+        stride_cm,
+        stride_cn,
+        BLOCK_M: tl.constexpr,
+        BLOCK_N: tl.constexpr,
+        BLOCK_K: tl.constexpr,
+        GROUP_M: tl.constexpr,
+        SPLIT_K: tl.constexpr,
+        EVEN_K: tl.constexpr,
+        ACC_TYPE: tl.constexpr,
+    ):
+        # matrix multiplication
+        pid = tl.program_id(0)
+        pid_z = tl.program_id(1)
+        grid_m = tl.cdiv(M, BLOCK_M)
+        grid_n = tl.cdiv(N, BLOCK_N)
+        # re-order program ID for better L2 performance
+        width = GROUP_M * grid_n
+        group_id = pid // width
+        group_size = min(grid_m - group_id * GROUP_M, GROUP_M)
+        pid_m = group_id * GROUP_M + (pid % group_size)
+        pid_n = (pid % width) // (group_size)
+        # do matrix multiplication
+        rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+        rn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+        ram = tl.max_contiguous(tl.multiple_of(rm % M, BLOCK_M), BLOCK_M)
+        rbn = tl.max_contiguous(tl.multiple_of(rn % N, BLOCK_N), BLOCK_N)
+        rk = pid_z * BLOCK_K + tl.arange(0, BLOCK_K)
+        # pointers
+        A = A + (ram[:, None] * stride_am + rk[None, :] * stride_ak)
+        B = B + (rk[:, None] * stride_bk + rbn[None, :] * stride_bn)
+
+        # rematerialize rm and rn to save registers
+        rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+        rn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+
+        w_factor = tl.load(state_w_ptr + rbn)[None, :]
+        x_factor = tl.load(state_x_ptr + ram)[:, None]
+
+        # acc = tl.zeros((BLOCK_M, BLOCK_N), dtype=ACC_TYPE)
+        acc = tl.zeros((BLOCK_M, BLOCK_N), dtype=tl.int32)
+        for k in range(0, tl.cdiv(K, BLOCK_K * SPLIT_K)):
+            if EVEN_K:
+                a = tl.load(A)
+                b = tl.load(B)
+            else:
+                k_remaining = K - k * (BLOCK_K * SPLIT_K)
+                a = tl.load(A, mask=rk[None, :] < k_remaining, other=0.0)
+                b = tl.load(B, mask=rk[:, None] < k_remaining, other=0.0)
+            acc += tl.dot(a, b)
+            A += BLOCK_K * SPLIT_K * stride_ak
+            B += BLOCK_K * SPLIT_K * stride_bk
+
+        acc = w_factor * (x_factor * (acc * divfactor))
+        acc = acc.to(C.dtype.element_ty)
+
+        if has_bias:
+            bias = tl.load(bias + rn).to(C.dtype.element_ty)
+            acc = acc + bias[None, :]
+
+        C = C + (rm[:, None] * stride_cm + rn[None, :] * stride_cn)
+        mask = (rm < M)[:, None] & (rn < N)[None, :]
+        # handles write-back with reduction-splitting
+        if SPLIT_K == 1:
+            tl.store(C, acc, mask=mask)
+        else:
+            tl.atomic_add(C, acc, mask=mask)
+
+    def int8_matmul_rowwise_dequantize(a, b, state_x, state_w, bias):
+        divfactor = 1.0 / (127.0 * 127.0)
+
+        has_bias = 0 if bias is None else 1
+
+        device = a.device
+        # handle non-contiguous inputs if necessary
+        if a.stride(0) > 1 and a.stride(1) > 1:
+            a = a.contiguous()
+        if b.stride(0) > 1 and b.stride(1) > 1:
+            b = b.contiguous()
+        # checks constraints
+        assert a.shape[1] == b.shape[0], "incompatible dimensions"
+        M, K = a.shape
+        _, N = b.shape
+        # allocates output
+        c = torch.empty((M, N), device=device, dtype=torch.float16)
+        # accumulator types
+        ACC_TYPE = tl.float32  # if a.dtype in [torch.float16, torch.bfloat16, torch.float32] else tl.int32
+        # launch int8_matmul_rowwise_dequantize kernel
+        grid = lambda META: (triton.cdiv(M, META["BLOCK_M"]) * triton.cdiv(N, META["BLOCK_N"]), META["SPLIT_K"])
+        _int8_matmul_rowwise_dequantize[grid](
+            a,
+            b,
+            c,
+            bias,
+            state_x,
+            state_w,
+            M,
+            N,
+            K,
+            divfactor,
+            has_bias,
+            a.stride(0),
+            a.stride(1),
+            b.stride(0),
+            b.stride(1),
+            c.stride(0),
+            c.stride(1),
+            GROUP_M=8,
+            ACC_TYPE=ACC_TYPE,
+        )
+        return c

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/triton/quantize_columnwise_and_transpose.py

@@ -0,0 +1,75 @@
+import math
+
+import torch
+
+from bitsandbytes.triton.triton_utils import is_triton_available
+
+if not is_triton_available():
+
+    def quantize_columnwise_and_transpose(x: torch.Tensor):
+        return None
+else:
+    import triton
+    import triton.language as tl
+
+    # This kernel does fused columnwise quantization and transpose.
+
+    # TODO: autotune this better.
+    @triton.autotune(
+        configs=[
+            triton.Config({}, num_stages=1),
+            triton.Config({}, num_stages=2),
+            triton.Config({}, num_stages=4),
+            triton.Config({}, num_stages=8),
+            triton.Config({}, num_stages=16),
+            triton.Config({}, num_stages=1, num_warps=8),
+            triton.Config({}, num_stages=2, num_warps=8),
+            triton.Config({}, num_stages=4, num_warps=8),
+            triton.Config({}, num_stages=8, num_warps=8),
+            triton.Config({}, num_stages=16, num_warps=8),
+            triton.Config({}, num_warps=1),
+            triton.Config({}, num_warps=2),
+            triton.Config({}, num_warps=4),
+            triton.Config({}, num_warps=8),
+        ],
+        key=["n_elements"],
+    )
+    @triton.jit
+    def _quantize_columnwise_and_transpose(
+        x_ptr,
+        output_ptr,
+        output_maxs,
+        n_elements,
+        M: tl.constexpr,
+        N: tl.constexpr,
+        BLOCK_SIZE: tl.constexpr,
+        P2: tl.constexpr,
+    ):
+        pid = tl.program_id(axis=0)
+        block_start = pid
+        p2_arange = tl.arange(0, P2)
+        p2_arange_mask = p2_arange < M
+        arange = p2_arange * N
+        offsets = block_start + arange
+        x = tl.load(x_ptr + offsets, mask=p2_arange_mask)
+        abs_x = tl.abs(x)
+        max_val = tl.max(tl.where(p2_arange_mask, abs_x, 0), axis=0)
+        output = tl.libdevice.llrint(127.0 * (x / max_val))
+
+        new_start = pid * M
+        new_offsets = new_start + p2_arange
+        tl.store(output_ptr + new_offsets, output, mask=p2_arange_mask)
+        tl.store(output_maxs + pid, max_val)
+
+    def quantize_columnwise_and_transpose(x: torch.Tensor):
+        M, N = x.shape
+        output = torch.empty(N, M, device=x.device, dtype=torch.int8)
+        output_maxs = torch.empty(x.shape[1], device=x.device, dtype=torch.float16)
+
+        P2 = int(2 ** (math.ceil(math.log2(M))))
+
+        assert x.is_cuda and output.is_cuda
+        n_elements = output.numel()
+        grid = lambda meta: (triton.cdiv(n_elements, meta["BLOCK_SIZE"]),)
+        _quantize_columnwise_and_transpose[grid](x, output, output_maxs, n_elements, M, N, BLOCK_SIZE=M, P2=P2)
+        return output, output_maxs

evalkit_eagle/lib/python3.10/site-packages/bitsandbytes/triton/quantize_global.py

@@ -0,0 +1,124 @@
+import torch
+
+from bitsandbytes.triton.triton_utils import is_triton_available
+
+if not is_triton_available():
+
+    def quantize_global_transpose(input):
+        return None
+
+    def quantize_global(x: torch.Tensor):
+        return None
+else:
+    import triton
+    import triton.language as tl
+
+    # global quantize
+    @triton.autotune(
+        configs=[
+            triton.Config({"BLOCK_SIZE": 1024}, num_warps=4),
+            triton.Config({"BLOCK_SIZE": 2048}, num_stages=1),
+        ],
+        key=["n_elements"],
+    )
+    @triton.jit
+    def _quantize_global(
+        x_ptr,
+        absmax_inv_ptr,
+        output_ptr,
+        n_elements,
+        BLOCK_SIZE: tl.constexpr,
+    ):
+        pid = tl.program_id(axis=0)
+        block_start = pid * BLOCK_SIZE
+        offsets = block_start + tl.arange(0, BLOCK_SIZE)
+        mask = offsets < n_elements
+        x = tl.load(x_ptr + offsets, mask=mask)
+        absmax_inv = tl.load(absmax_inv_ptr)
+        output = tl.libdevice.llrint(127.0 * (x * absmax_inv))
+        tl.store(output_ptr + offsets, output, mask=mask)
+
+    def quantize_global(x: torch.Tensor):
+        absmax = x.abs().max().unsqueeze(0)
+        absmax_inv = 1.0 / absmax
+        output = torch.empty(*x.shape, device="cuda", dtype=torch.int8)
+        assert x.is_cuda and output.is_cuda
+        n_elements = output.numel()
+        grid = lambda meta: (triton.cdiv(n_elements, meta["BLOCK_SIZE"]),)
+        _quantize_global[grid](x, absmax_inv, output, n_elements)
+        return output, absmax
+
+    # global quantize and transpose
+    @triton.autotune(
+        configs=[
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 128, "GROUP_M": 8}, num_warps=4),
+            triton.Config({"BLOCK_M": 128, "BLOCK_N": 128, "GROUP_M": 8}, num_warps=4),
+            # ...
+        ],
+        key=["M", "N"],
+    )
+    @triton.jit
+    def _quantize_global_transpose(
+        A,
+        absmax_inv_ptr,
+        B,
+        stride_am,
+        stride_an,
+        stride_bn,
+        stride_bm,
+        M,
+        N,
+        BLOCK_M: tl.constexpr,
+        BLOCK_N: tl.constexpr,
+        GROUP_M: tl.constexpr,
+    ):
+        pid = tl.program_id(0)
+        grid_m = (M + BLOCK_M - 1) // BLOCK_M
+        grid_n = (N + BLOCK_N - 1) // BLOCK_N
+
+        width = GROUP_M * grid_n
+        group_id = pid // width
+        group_size = min(grid_m - group_id * GROUP_M, GROUP_M)
+        pid_m = group_id * GROUP_M + (pid % group_size)
+        pid_n = (pid % width) // group_size
+
+        rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+        rn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+        A = A + (rm[:, None] * stride_am + rn[None, :] * stride_an)
+        mask = (rm < M)[:, None] & (rn < N)[None, :]
+        a = tl.load(A, mask=mask)
+        absmax_inv = tl.load(absmax_inv_ptr)
+
+        # rematerialize to save registers
+        rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+        rn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+        B = B + (rm[:, None] * stride_bm + rn[None, :] * stride_bn)
+        mask = (rm < M)[:, None] & (rn < N)[None, :]
+
+        output = tl.libdevice.llrint(127.0 * (a * absmax_inv))
+
+        tl.store(B, output, mask=mask)
+
+    def quantize_global_transpose(input):
+        absmax = input.abs().max().unsqueeze(0)
+        absmax_inv = 1.0 / absmax
+        M, N = input.shape
+        out = torch.empty(N, M, device="cuda", dtype=torch.int8)
+
+        assert out.size(0) == N and out.size(1) == M
+        assert input.stride(0) == 1 or input.stride(1) == 1
+        assert out.stride(0) == 1 or out.stride(1) == 1
+
+        grid = lambda META: (triton.cdiv(M, META["BLOCK_M"]) * triton.cdiv(N, META["BLOCK_N"]),)
+        _quantize_global_transpose[grid](
+            input,
+            absmax_inv,
+            out,
+            input.stride(0),
+            input.stride(1),
+            out.stride(0),
+            out.stride(1),
+            M,
+            N,
+        )
+        return out, absmax