compressed-tensors 0.4.0__py3-none-any.whl → 0.6.0__py3-none-any.whl

{compressed_tensors-0.4.0.dist-info → compressed_tensors-0.6.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: compressed-tensors
-Version: 0.4.0
+Version: 0.6.0
 Summary: Library for utilization of compressed safetensors of neural network models
 Home-page: https://github.com/neuralmagic/compressed-tensors
 Author: Neuralmagic, Inc.
@@ -8,43 +8,56 @@ Author-email: support@neuralmagic.com
 License: Apache 2.0
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: torch >=1.7.0
+Requires-Dist: torch>=1.7.0
 Requires-Dist: transformers
-Requires-Dist: pydantic >=2.0
+Requires-Dist: pydantic>=2.0
+Provides-Extra: accelerate
+Requires-Dist: accelerate; extra == "accelerate"
 Provides-Extra: dev
-Requires-Dist: black ==22.12.0 ; extra == 'dev'
-Requires-Dist: isort ==5.8.0 ; extra == 'dev'
-Requires-Dist: wheel >=0.36.2 ; extra == 'dev'
-Requires-Dist: flake8 >=3.8.3 ; extra == 'dev'
-Requires-Dist: pytest >=6.0.0 ; extra == 'dev'
-Requires-Dist: nbconvert >=7.16.3 ; extra == 'dev'
+Requires-Dist: black==22.12.0; extra == "dev"
+Requires-Dist: isort==5.8.0; extra == "dev"
+Requires-Dist: wheel>=0.36.2; extra == "dev"
+Requires-Dist: flake8>=3.8.3; extra == "dev"
+Requires-Dist: pytest>=6.0.0; extra == "dev"
+Requires-Dist: nbconvert>=7.16.3; extra == "dev"
 
-# compressed_tensors
+# compressed-tensors
 
-This repository extends a [safetensors](https://github.com/huggingface/safetensors) format to efficiently store sparse and/or quantized tensors on disk. `compressed-tensors` format supports multiple compression types to minimize the disk space and facilitate the tensor manipulation.
+The `compressed-tensors` library extends the [safetensors](https://github.com/huggingface/safetensors) format, providing a versatile and efficient way to store and manage compressed tensor data. This library supports various quantization and sparsity schemes, making it a unified format for handling different model optimizations like GPTQ, AWQ, SmoothQuant, INT8, FP8, SparseGPT, and more.
 
-## Motivation
+## Why `compressed-tensors`?
 
-### Reduce disk space by saving sparse tensors in a compressed format
+As model compression becomes increasingly important for efficient deployment of LLMs, the landscape of quantization and compression techniques has become increasingly fragmented.
+Each method often comes with its own storage format and loading procedures, making it challenging to work with multiple techniques or switch between them.
+`compressed-tensors` addresses this by providing a single, extensible format that can represent a wide variety of compression schemes. 
 
-The compressed format stores the data much more efficiently by taking advantage of two properties of tensors:
+* **Unified Checkpoint Format**: Supports various compression schemes in a single, consistent format.
+* **Wide Compatibility**: Works with popular quantization methods like GPTQ, SmoothQuant, and FP8. See [llm-compressor](https://github.com/vllm-project/llm-compressor)
+* **Flexible Quantization Support**: 
+  * Weight-only quantization (e.g., W4A16, W8A16, WnA16)
+  * Activation quantization (e.g., W8A8)
+  * KV cache quantization
+  * Non-uniform schemes (different layers can be quantized in different ways!)
+* **Sparsity Support**: Handles both unstructured and semi-structured (e.g., 2:4) sparsity patterns.
+* **Open-Source Integration**: Designed to work seamlessly with Hugging Face models and PyTorch.
 
-- Sparse tensors -> due to a large number of entries that are equal to zero.
-- Quantized -> due to their low precision representation.
-
-### Introduce an elegant interface to save/load compressed tensors
-
-The library provides the user with the ability to compress/decompress tensors. The properties of tensors are defined by human-readable configs, allowing the users to understand the compression format at a quick glance.
+This allows developers and researchers to easily experiment with composing different quantization methods, simplify model deployment pipelines, and reduce the overhead of supporting multiple compression formats in inference engines.
 
 ## Installation
 
-### Pip
+### From [PyPI](https://pypi.org/project/compressed-tensors)
 
+Stable release:
 ```bash
 pip install compressed-tensors
 ```
 
-### From source
+Nightly release:
+```bash
+pip install compressed-tensors-nightly
+```
+
+### From Source
 
 ```bash
 git clone https://github.com/neuralmagic/compressed-tensors

compressed_tensors-0.6.0.dist-info/RECORD

@@ -0,0 +1,52 @@
+compressed_tensors/__init__.py,sha256=SV1csvHUVCd8kHXz6UDZim1HZ_fAVG3vfk-j_4Bb6hY,789
+compressed_tensors/base.py,sha256=Mq4mfVQcJhNpha-BXzpOfpmFIdl01o09BJE7D2oQ_00,796
+compressed_tensors/version.py,sha256=qWZbaQsbr0_5Y1qtWGstNPljw-ggMLBGZBvSKfzB9pw,1585
+compressed_tensors/compressors/__init__.py,sha256=wmX4VnkUTS63xBwK5-6w8FP78bNZpcdcqvf2KOEC5E4,1133
+compressed_tensors/compressors/base.py,sha256=NfVkhq6PRiq2cvAXaUXLoqC_nVYWdSrkE12c9AXYSMo,9956
+compressed_tensors/compressors/dense.py,sha256=xcWECjcRY4INN6jC7vHx5wvUX3NmnKlxA9SVE1A6m2Q,1267
+compressed_tensors/compressors/helpers.py,sha256=k9avlkmeYj6vkOAvl-MgcixtP7ib24SCfhzZ-RusXfw,5403
+compressed_tensors/compressors/marlin_24.py,sha256=e7fGUyZbjUpA5VUMCPxqcYPGNiwoDKupHJaXWCoVKRw,9410
+compressed_tensors/compressors/model_compressor.py,sha256=gI6KKtH3eeWi2540Ayx-4bg9o8qjrvxlF4Gd_sqltGA,16678
+compressed_tensors/compressors/naive_quantized.py,sha256=z3h3ca5xKCN69mahutxcbzdv-OysiaxaM8P-Qum6zUQ,4823
+compressed_tensors/compressors/pack_quantized.py,sha256=27RVmJ2wg2dvCoawj407HSmKT3VPGJ6ujAMHlT26WlI,7571
+compressed_tensors/compressors/sparse_bitmask.py,sha256=kiDwBlFV0sJGLcIdDYxIiuF64ccgwDfqq1hWRQThYDc,8647
+compressed_tensors/config/__init__.py,sha256=ZBqWn3r6ku1qfmlHHYp0mQueY0i7Pwhr9rbQk9dDlMc,704
+compressed_tensors/config/base.py,sha256=BNTFKy12isY7qblwxdi_R1f00EzgrNOXLrfxqLCPT8w,1903
+compressed_tensors/config/dense.py,sha256=NgSxnFCnckU9-iunxEaqiFwqgdO7YYxlWKR74jNbjks,1317
+compressed_tensors/config/sparse_bitmask.py,sha256=pZUboRNZTu6NajGOQEFExoPknak5ynVAUeiiYpS1Gt8,1308
+compressed_tensors/linear/__init__.py,sha256=fH6rjBYAxuwrTzBTlTjTgCYNyh6TCvCqajCz4Im4YrA,617
+compressed_tensors/linear/compressed_linear.py,sha256=G0gEFfxLAUsgRcnfSV-PKz1ZBNTVokOauOoup7SE1mw,3210
+compressed_tensors/quantization/__init__.py,sha256=83J5bPB7PavN2TfCoW7_vEDhfYpm4TDrqYO9vdSQ5bk,760
+compressed_tensors/quantization/quant_args.py,sha256=CmyVtjJeHlqCW-7R5Z7tIw6lXUrzCX6Y9bwgmMxEudY,8069
+compressed_tensors/quantization/quant_config.py,sha256=NpVu8YJ4Xw2pIQW_PGaNaml8kx1bUnxkvb0jBYWbKdE,9971
+compressed_tensors/quantization/quant_scheme.py,sha256=HmR1DcFZcjuqX7KHUYI0NFXsCIzJ8sxFGH6zhYuHmEs,5870
+compressed_tensors/quantization/lifecycle/__init__.py,sha256=MXE2E7GfIfRRfhrdGy2Og3AZOz5N59B0ZGFcsD89y6c,821
+compressed_tensors/quantization/lifecycle/apply.py,sha256=uftWFunr_CpCZM_qWfo2O1USXKB2qSYD1pBJsO8BuCU,15285
+compressed_tensors/quantization/lifecycle/calibration.py,sha256=PlS_EqCOPqJD3QKuLPXO9AOtDzXtQWvEBTynFv-FFVw,2698
+compressed_tensors/quantization/lifecycle/compressed.py,sha256=Fj9n66IN0EWsOAkBHg3O0GlOQpxstqjCcs0ttzMXrJ0,2296
+compressed_tensors/quantization/lifecycle/forward.py,sha256=PljD9pzATILEOiC3ZdHUTsfSbZdAa6iSIxWmvAHLG9I,13688
+compressed_tensors/quantization/lifecycle/frozen.py,sha256=h1XYt89MouBTf3jTYLG_6OdFxIu5q2N8tPjsy6J4E6Y,1726
+compressed_tensors/quantization/lifecycle/helpers.py,sha256=TmLY_G5VP_Fg2Ywio_dxoHRTxOKZdT7_aG5S9WtD4zI,2424
+compressed_tensors/quantization/lifecycle/initialize.py,sha256=S5Kwy16Da8WUIIpa1xVKc72MijJ5C_rqM6JjanZ7MGk,7133
+compressed_tensors/quantization/observers/__init__.py,sha256=4Sa7rqi5RB_S5bPO8KmncETiqDsoMBhwP37arlQym8s,764
+compressed_tensors/quantization/observers/base.py,sha256=5ovQicWPYHjIxr6-EkQ4lgOX0PpI9g23iSzKpxjM1Zg,8420
+compressed_tensors/quantization/observers/helpers.py,sha256=s_A23Qa_BLfOdHJCN5bm-qPWkhjjj_RIVrhSp1Y9Dtk,4211
+compressed_tensors/quantization/observers/memoryless.py,sha256=jH_c6K3gxf4W3VNXQ7tbnP-J_86QTrEfjBn6Kh1C-H8,2165
+compressed_tensors/quantization/observers/min_max.py,sha256=sQXqU3z-voxIDfR_9mQzwQUflZj2sASm_G8CYaXntFw,3865
+compressed_tensors/quantization/observers/mse.py,sha256=Aeh-253Vbab1F8cYuBiGNn4OXWJ67wXQ_JVfl3mu2a8,6034
+compressed_tensors/quantization/utils/__init__.py,sha256=VdtEmP0bvuND_IGQnyqUPc5lnFp-1_yD7StKSX4x80w,656
+compressed_tensors/quantization/utils/helpers.py,sha256=pwvU613XRvMDtI5b39II5jukBl5OUCqoX0ofVRpOFRY,8633
+compressed_tensors/registry/__init__.py,sha256=FwLSNYqfIrb5JD_6OK_MT4_svvKTN_nEhpgQlQvGbjI,658
+compressed_tensors/registry/registry.py,sha256=fxjOjh2wklCvJhQxwofdy-zV8q7MkQ85SLG77nml2iA,11890
+compressed_tensors/utils/__init__.py,sha256=gS4gSU2pwcAbsKj-6YMaqhm25udFy6ISYaWBf-myRSM,808
+compressed_tensors/utils/helpers.py,sha256=bh4G8mj_YCRf8Bo2FQ9FkIIZXY8xqqPjckNnVYB0gBA,3557
+compressed_tensors/utils/offload.py,sha256=d9q8LNe8HyF8tOjgjA7QGLD3HRysmNp0d8eBbdqBgIM,4089
+compressed_tensors/utils/permutations_24.py,sha256=kx6fsfDHebx94zsSzhXGyCyuC9sVyah6BUUir_StT28,2530
+compressed_tensors/utils/permute.py,sha256=V6tJLKo3Syccj-viv4F7ZKZgJeCB-hl-dK8RKI_kBwI,2355
+compressed_tensors/utils/safetensors_load.py,sha256=m08ANVuTBxQdoa6LufDgcNJ7wCLDJolyZljB8VEybAU,8578
+compressed_tensors/utils/semi_structured_conversions.py,sha256=XKNffPum54kPASgqKzgKvyeqWPAkair2XEQXjkp7ho8,13489
+compressed_tensors-0.6.0.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+compressed_tensors-0.6.0.dist-info/METADATA,sha256=opV5oXtNn7EmogPQaMiv9ddOLHv-XObt_NuUSwR5Bz4,6782
+compressed_tensors-0.6.0.dist-info/WHEEL,sha256=eOLhNAGa2EW3wWl_TU484h7q1UNgy0JXjjoqKoxAAQc,92
+compressed_tensors-0.6.0.dist-info/top_level.txt,sha256=w2i-GyPs2s1UwVxvutSvN_lM22SXC2hQFBmoMcPnV7Y,19
+compressed_tensors-0.6.0.dist-info/RECORD,,

{compressed_tensors-0.4.0.dist-info → compressed_tensors-0.6.0.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: bdist_wheel (0.43.0)
+Generator: bdist_wheel (0.44.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

compressed_tensors/compressors/int_quantized.py

@@ -1,126 +0,0 @@
-# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#    http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import logging
-from typing import Dict, Generator, Tuple
-
-import torch
-from compressed_tensors.compressors import Compressor
-from compressed_tensors.config import CompressionFormat
-from compressed_tensors.quantization import QuantizationArgs
-from compressed_tensors.quantization.lifecycle.forward import dequantize, quantize
-from compressed_tensors.quantization.utils import can_quantize
-from compressed_tensors.utils import get_nested_weight_mappings, merge_names
-from safetensors import safe_open
-from torch import Tensor
-from tqdm import tqdm
-
-
-__all__ = ["IntQuantizationCompressor"]
-
-_LOGGER: logging.Logger = logging.getLogger(__name__)
-
-
-@Compressor.register(name=CompressionFormat.int_quantized.value)
-class IntQuantizationCompressor(Compressor):
-    """
-    Integer compression for quantized models. Weight of each quantized layer is
-    converted from its original float type to the format specified by the layer's
-    quantization scheme.
-    """
-
-    COMPRESSION_PARAM_NAMES = ["weight", "weight_scale", "weight_zero_point"]
-
-    def compress(
-        self,
-        model_state: Dict[str, Tensor],
-        model_quant_args: Dict[str, QuantizationArgs],
-        **kwargs,
-    ) -> Dict[str, Tensor]:
-        """
-        Compresses a dense state dict
-
-        :param model_state: state dict of uncompressed model
-        :param model_quant_args: quantization args for each quantized weight, needed for
-        quantize function to calculate bit depth
-        :return: compressed state dict
-        """
-        compressed_dict = {}
-        weight_suffix = ".weight"
-        _LOGGER.debug(
-            f"Compressing model with {len(model_state)} parameterized layers..."
-        )
-
-        for name, value in tqdm(model_state.items(), desc="Compressing model"):
-            if name.endswith(weight_suffix):
-                prefix = name[: -(len(weight_suffix))]
-                scale = model_state.get(merge_names(prefix, "weight_scale"), None)
-                zp = model_state.get(merge_names(prefix, "weight_zero_point"), None)
-                if scale is not None and zp is not None:
-                    # weight is quantized, compress it
-                    quant_args = model_quant_args[prefix]
-                    if can_quantize(value, quant_args):
-                        # only quantize if not already quantized
-                        value = quantize(
-                            x=value,
-                            scale=scale,
-                            zero_point=zp,
-                            args=quant_args,
-                            dtype=torch.int8,
-                        )
-            elif name.endswith("zero_point"):
-                if torch.all(value == 0):
-                    # all zero_points are 0, no need to include in
-                    # compressed state_dict
-                    continue
-            compressed_dict[name] = value.to("cpu")
-
-        return compressed_dict
-
-    def decompress(
-        self, path_to_model_or_tensors: str, device: str = "cpu"
-    ) -> Generator[Tuple[str, Tensor], None, None]:
-        """
-        Reads a compressed state dict located at path_to_model_or_tensors
-        and returns a generator for sequentially decompressing back to a
-        dense state dict
-
-        :param model_path: path to compressed safetensors model (directory with
-            one or more safetensors files) or compressed tensors file
-        :param device: optional device to load intermediate weights into
-        :return: compressed state dict
-        """
-        weight_mappings = get_nested_weight_mappings(
-            path_to_model_or_tensors, self.COMPRESSION_PARAM_NAMES
-        )
-        for weight_name in weight_mappings.keys():
-            weight_data = {}
-            for param_name, safe_path in weight_mappings[weight_name].items():
-                full_name = merge_names(weight_name, param_name)
-                with safe_open(safe_path, framework="pt", device=device) as f:
-                    weight_data[param_name] = f.get_tensor(full_name)
-
-            if "weight_scale" in weight_data:
-                zero_point = weight_data.get("weight_zero_point", None)
-                scale = weight_data["weight_scale"]
-                if zero_point is None:
-                    # zero_point assumed to be 0 if not included in state_dict
-                    zero_point = torch.zeros_like(scale)
-
-                decompressed = dequantize(
-                    x_q=weight_data["weight"],
-                    scale=scale,
-                    zero_point=zero_point,
-                )
-                yield merge_names(weight_name, "weight"), decompressed

compressed_tensors/compressors/utils/helpers.py

@@ -1,43 +0,0 @@
-# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#    http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import torch
-
-
-__all__ = ["tensor_follows_mask_structure"]
-
-
-def tensor_follows_mask_structure(tensor, mask: str = "2:4") -> bool:
-    """
-    :param tensor: tensor to check
-    :param mask: mask structure to check for, in the format "n:m"
-    :return: True if the tensor follows the mask structure, False otherwise.
-        Note, some weights can incidentally be zero, so we check for
-        atleast n zeros in each chunk of size m
-    """
-
-    n, m = tuple(map(int, mask.split(":")))
-    # Reshape the tensor into chunks of size m
-    tensor = tensor.view(-1, m)
-
-    # Count the number of zeros in each chunk
-    zero_counts = (tensor == 0).sum(dim=1)
-
-    # Check if the number of zeros in each chunk atleast n
-    # Greater than sign is needed as some weights can incidentally
-    # be zero
-    if not torch.all(zero_counts >= n).item():
-        raise ValueError()
-
-    return True

compressed_tensors-0.4.0.dist-info/RECORD

@@ -1,48 +0,0 @@
-compressed_tensors/__init__.py,sha256=SV1csvHUVCd8kHXz6UDZim1HZ_fAVG3vfk-j_4Bb6hY,789
-compressed_tensors/base.py,sha256=OA2TOLP1gP3LSH7gp508eqr2ZtDQ-pqRHElCp-aB0vs,755
-compressed_tensors/version.py,sha256=_nj1yS4msz1OXd0H1v1m-z1JkMOuy19M9lFDTWP5xf0,1585
-compressed_tensors/compressors/__init__.py,sha256=rhqPp3YXFxCJRLZs1KRNSHTIxK2rNU--sYwDI8MW47w,1061
-compressed_tensors/compressors/base.py,sha256=LWEgbpgTxzmoqQ7Xhq2OQszUgWoDtFuGCiV1Y8nlBGw,2134
-compressed_tensors/compressors/dense.py,sha256=G_XHbvuENyupIKlXSITOQgvPkNkcMEOLcLWQr70V9EE,1257
-compressed_tensors/compressors/helpers.py,sha256=k9avlkmeYj6vkOAvl-MgcixtP7ib24SCfhzZ-RusXfw,5403
-compressed_tensors/compressors/int_quantized.py,sha256=Ct2vCK0yoPm6vkIFlzDMGQ7m14xT1GyURsSwH9DP770,5242
-compressed_tensors/compressors/marlin_24.py,sha256=X_BjtFB3Mn0hqiLz56UM3jGX2eNmGLnvEIPfbg7di6U,9444
-compressed_tensors/compressors/model_compressor.py,sha256=h3ixQtfzt6HxSNtdnB9OVdpCucTmIo4paDoaM7XYZXE,12559
-compressed_tensors/compressors/pack_quantized.py,sha256=VPiLlgJlDgARrn7YmiQoLqUfxErKBfj54epMYWRsF8k,8451
-compressed_tensors/compressors/sparse_bitmask.py,sha256=H9oZSTYI1oRCzAMbd4zThUnZd1h2rfs8DmA3tPcvuNE,8637
-compressed_tensors/compressors/utils/__init__.py,sha256=-mbGDZh1hd9T6u62Ht_iBIK255UmMg0f5bLkSs1f9Cc,731
-compressed_tensors/compressors/utils/helpers.py,sha256=4fq7KclSIK__jemCG9pwYlgWLrQjsaAMxhIrhjdw0BQ,1506
-compressed_tensors/compressors/utils/permutations_24.py,sha256=kx6fsfDHebx94zsSzhXGyCyuC9sVyah6BUUir_StT28,2530
-compressed_tensors/compressors/utils/semi_structured_conversions.py,sha256=g1EZHzdv-ko7ufPX430dp7wE33o6FWJXuSP4zZydCu0,13488
-compressed_tensors/config/__init__.py,sha256=ZBqWn3r6ku1qfmlHHYp0mQueY0i7Pwhr9rbQk9dDlMc,704
-compressed_tensors/config/base.py,sha256=ZnpuOevCE0pXdA8OJfIJnxj-ccproH7o1EOwRY8_hUU,1482
-compressed_tensors/config/dense.py,sha256=NgSxnFCnckU9-iunxEaqiFwqgdO7YYxlWKR74jNbjks,1317
-compressed_tensors/config/sparse_bitmask.py,sha256=pZUboRNZTu6NajGOQEFExoPknak5ynVAUeiiYpS1Gt8,1308
-compressed_tensors/quantization/__init__.py,sha256=83J5bPB7PavN2TfCoW7_vEDhfYpm4TDrqYO9vdSQ5bk,760
-compressed_tensors/quantization/quant_args.py,sha256=Z9Zu20ooAwEWlliAdUw1f1zwSrheuD6vqm3YXgJ1Lws,4388
-compressed_tensors/quantization/quant_config.py,sha256=hL42sXp1wAZxyrkHarw7tAMRcwSVEr0MT3wmrmL3NhE,8285
-compressed_tensors/quantization/quant_scheme.py,sha256=aX4h8t8RDqrWeUqoqrYMOxc0xkWcu8Ue_CHLoG-fRjQ,3569
-compressed_tensors/quantization/lifecycle/__init__.py,sha256=ggRGWRqhCxCaTTDWRcgTVX3axnS2xV6rc5YvdzK7fSg,798
-compressed_tensors/quantization/lifecycle/apply.py,sha256=aZrglJ5mR3Xaxwj51-1BVVB1JGVkKQEeHxGfBaVmsHI,8881
-compressed_tensors/quantization/lifecycle/calibration.py,sha256=mLns4jlaWmBwOW8Jtlm5bMX-JET1AiZYUBO7qa-XuxI,1776
-compressed_tensors/quantization/lifecycle/compressed.py,sha256=VreB10xPwgSLQQlTu20UCrFpRS--cA7-lx5s7nrPPrg,2247
-compressed_tensors/quantization/lifecycle/forward.py,sha256=0T817yzYqFR1wUjk2XCtOISwr4u7cdkKqAv13jjfu24,11113
-compressed_tensors/quantization/lifecycle/frozen.py,sha256=h1XYt89MouBTf3jTYLG_6OdFxIu5q2N8tPjsy6J4E6Y,1726
-compressed_tensors/quantization/lifecycle/initialize.py,sha256=9xgPzHejQUO_AkZcc_SH5kqFeieG-9uo0fMRYV51i7Y,4577
-compressed_tensors/quantization/observers/__init__.py,sha256=DNH31NQYrIBBcmHsMyFA6whh4pbRsLwuNa6L8AeXaGc,745
-compressed_tensors/quantization/observers/base.py,sha256=z_JC-CRz-PY7WlpSoyOoSQQWz5ekTEd5LbXt0iHQRes,5239
-compressed_tensors/quantization/observers/helpers.py,sha256=FUyYUNd-3LbXt0-8Lwr7EPI2m-LXXBTXW1l5iOajNhA,2272
-compressed_tensors/quantization/observers/memoryless.py,sha256=jH_c6K3gxf4W3VNXQ7tbnP-J_86QTrEfjBn6Kh1C-H8,2165
-compressed_tensors/quantization/observers/min_max.py,sha256=UK7zCMzxv9GGn6BflBxdajV20RiWaCY2RHcvZodCP1w,3669
-compressed_tensors/quantization/utils/__init__.py,sha256=VdtEmP0bvuND_IGQnyqUPc5lnFp-1_yD7StKSX4x80w,656
-compressed_tensors/quantization/utils/helpers.py,sha256=NzAH18Cn_-mTAR87y6IlcQU5gC393XSjgNKC9CRkr78,6017
-compressed_tensors/registry/__init__.py,sha256=FwLSNYqfIrb5JD_6OK_MT4_svvKTN_nEhpgQlQvGbjI,658
-compressed_tensors/registry/registry.py,sha256=fxjOjh2wklCvJhQxwofdy-zV8q7MkQ85SLG77nml2iA,11890
-compressed_tensors/utils/__init__.py,sha256=5DrYjoZbaEvSkJcC-GRSbM_RBHVF4tG9gMd3zsJnjLw,665
-compressed_tensors/utils/helpers.py,sha256=5ull5yFT31M2zVxKeFvpvvlvX5f1Sk1LGuj_wrfZWCY,2267
-compressed_tensors/utils/safetensors_load.py,sha256=0MheXwx1jeY12PeISppiSIZHs6rmN2YddwPpFb9V67I,8527
-compressed_tensors-0.4.0.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-compressed_tensors-0.4.0.dist-info/METADATA,sha256=NtnK_A9ck3KPmh4syGcGtMBGX-_2FyFa7ntCAdf-KGo,5651
-compressed_tensors-0.4.0.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
-compressed_tensors-0.4.0.dist-info/top_level.txt,sha256=w2i-GyPs2s1UwVxvutSvN_lM22SXC2hQFBmoMcPnV7Y,19
-compressed_tensors-0.4.0.dist-info/RECORD,,