Aluode/PerceptionLabPortable

1# Copyright 2023 The HuggingFace Team. All rights reserved.2#3# Licensed under the Apache License, Version 2.0 (the "License");4# you may not use this file except in compliance with the License.5# You may obtain a copy of the License at6#7#     http://www.apache.org/licenses/LICENSE-2.08#9# Unless required by applicable law or agreed to in writing, software10# distributed under the License is distributed on an "AS IS" BASIS,11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.12# See the License for the specific language governing permissions and13# limitations under the License.14from typing import TYPE_CHECKING15 16from ..utils import OptionalDependencyNotAvailable, _LazyModule, is_torch_available, is_torch_greater_or_equal17 18 19_import_structure = {20    "aqlm": ["replace_with_aqlm_linear"],21    "awq": [22        "fuse_awq_modules",23        "post_init_awq_exllama_modules",24        "post_init_awq_ipex_modules",25        "replace_quantization_scales",26        "replace_with_awq_linear",27    ],28    "bitnet": [29        "BitLinear",30        "pack_weights",31        "replace_with_bitnet_linear",32        "unpack_weights",33    ],34    "bitsandbytes": [35        "dequantize_and_replace",36        "get_keys_to_not_convert",37        "replace_8bit_linear",38        "replace_with_bnb_linear",39        "set_module_8bit_tensor_to_device",40        "set_module_quantized_tensor_to_device",41        "validate_bnb_backend_availability",42    ],43    "deepspeed": [44        "HfDeepSpeedConfig",45        "HfTrainerDeepSpeedConfig",46        "deepspeed_config",47        "deepspeed_init",48        "deepspeed_load_checkpoint",49        "deepspeed_optim_sched",50        "is_deepspeed_available",51        "is_deepspeed_zero3_enabled",52        "set_hf_deepspeed_config",53        "unset_hf_deepspeed_config",54    ],55    "eetq": ["replace_with_eetq_linear"],56    "fbgemm_fp8": ["FbgemmFp8Linear", "FbgemmFp8Llama4TextExperts", "replace_with_fbgemm_fp8_linear"],57    "finegrained_fp8": ["FP8Linear", "replace_with_fp8_linear"],58    "fsdp": ["is_fsdp_enabled", "is_fsdp_managed_module"],59    "ggml": [60        "GGUF_CONFIG_MAPPING",61        "GGUF_TOKENIZER_MAPPING",62        "_gguf_parse_value",63        "load_dequant_gguf_tensor",64        "load_gguf",65    ],66    "higgs": [67        "HiggsLinear",68        "dequantize_higgs",69        "quantize_with_higgs",70        "replace_with_higgs_linear",71    ],72    "hqq": ["prepare_for_hqq_linear"],73    "hub_kernels": [74        "LayerRepository",75        "register_kernel_mapping",76        "replace_kernel_forward_from_hub",77        "use_kernel_forward_from_hub",78    ],79    "integration_utils": [80        "INTEGRATION_TO_CALLBACK",81        "AzureMLCallback",82        "ClearMLCallback",83        "CodeCarbonCallback",84        "CometCallback",85        "DagsHubCallback",86        "DVCLiveCallback",87        "FlyteCallback",88        "MLflowCallback",89        "NeptuneCallback",90        "NeptuneMissingConfiguration",91        "SwanLabCallback",92        "TensorBoardCallback",93        "TrackioCallback",94        "WandbCallback",95        "get_available_reporting_integrations",96        "get_reporting_integration_callbacks",97        "hp_params",98        "is_azureml_available",99        "is_clearml_available",100        "is_codecarbon_available",101        "is_comet_available",102        "is_dagshub_available",103        "is_dvclive_available",104        "is_flyte_deck_standard_available",105        "is_flytekit_available",106        "is_mlflow_available",107        "is_neptune_available",108        "is_optuna_available",109        "is_ray_available",110        "is_ray_tune_available",111        "is_sigopt_available",112        "is_swanlab_available",113        "is_tensorboard_available",114        "is_trackio_available",115        "is_wandb_available",116        "rewrite_logs",117        "run_hp_search_optuna",118        "run_hp_search_ray",119        "run_hp_search_sigopt",120        "run_hp_search_wandb",121    ],122    "mxfp4": [123        "Mxfp4GptOssExperts",124        "convert_moe_packed_tensors",125        "dequantize",126        "load_and_swizzle_mxfp4",127        "quantize_to_mxfp4",128        "replace_with_mxfp4_linear",129        "swizzle_mxfp4",130    ],131    "peft": ["PeftAdapterMixin"],132    "quanto": ["replace_with_quanto_layers"],133    "spqr": ["replace_with_spqr_linear"],134    "vptq": ["replace_with_vptq_linear"],135}136 137try:138    if not is_torch_available():139        raise OptionalDependencyNotAvailable()140except OptionalDependencyNotAvailable:141    pass142else:143    _import_structure["executorch"] = [144        "TorchExportableModuleWithStaticCache",145        "convert_and_export_with_cache",146    ]147 148try:149    if not is_torch_greater_or_equal("2.3"):150        raise OptionalDependencyNotAvailable()151except OptionalDependencyNotAvailable:152    pass153else:154    _import_structure["tensor_parallel"] = [155        "shard_and_distribute_module",156        "ALL_PARALLEL_STYLES",157        "translate_to_torch_parallel_style",158    ]159try:160    if not is_torch_greater_or_equal("2.5"):161        raise OptionalDependencyNotAvailable()162except OptionalDependencyNotAvailable:163    pass164else:165    _import_structure["flex_attention"] = [166        "make_flex_block_causal_mask",167    ]168 169if TYPE_CHECKING:170    from .aqlm import replace_with_aqlm_linear171    from .awq import (172        fuse_awq_modules,173        post_init_awq_exllama_modules,174        post_init_awq_ipex_modules,175        replace_quantization_scales,176        replace_with_awq_linear,177    )178    from .bitnet import (179        BitLinear,180        pack_weights,181        replace_with_bitnet_linear,182        unpack_weights,183    )184    from .bitsandbytes import (185        dequantize_and_replace,186        get_keys_to_not_convert,187        replace_8bit_linear,188        replace_with_bnb_linear,189        set_module_8bit_tensor_to_device,190        set_module_quantized_tensor_to_device,191        validate_bnb_backend_availability,192    )193    from .deepspeed import (194        HfDeepSpeedConfig,195        HfTrainerDeepSpeedConfig,196        deepspeed_config,197        deepspeed_init,198        deepspeed_load_checkpoint,199        deepspeed_optim_sched,200        is_deepspeed_available,201        is_deepspeed_zero3_enabled,202        set_hf_deepspeed_config,203        unset_hf_deepspeed_config,204    )205    from .eetq import replace_with_eetq_linear206    from .fbgemm_fp8 import FbgemmFp8Linear, FbgemmFp8Llama4TextExperts, replace_with_fbgemm_fp8_linear207    from .finegrained_fp8 import FP8Linear, replace_with_fp8_linear208    from .fsdp import is_fsdp_enabled, is_fsdp_managed_module209    from .ggml import (210        GGUF_CONFIG_MAPPING,211        GGUF_TOKENIZER_MAPPING,212        _gguf_parse_value,213        load_dequant_gguf_tensor,214        load_gguf,215    )216    from .higgs import HiggsLinear, dequantize_higgs, quantize_with_higgs, replace_with_higgs_linear217    from .hqq import prepare_for_hqq_linear218    from .hub_kernels import (219        LayerRepository,220        register_kernel_mapping,221        replace_kernel_forward_from_hub,222        use_kernel_forward_from_hub,223    )224    from .integration_utils import (225        INTEGRATION_TO_CALLBACK,226        AzureMLCallback,227        ClearMLCallback,228        CodeCarbonCallback,229        CometCallback,230        DagsHubCallback,231        DVCLiveCallback,232        FlyteCallback,233        MLflowCallback,234        NeptuneCallback,235        NeptuneMissingConfiguration,236        SwanLabCallback,237        TensorBoardCallback,238        TrackioCallback,239        WandbCallback,240        get_available_reporting_integrations,241        get_reporting_integration_callbacks,242        hp_params,243        is_azureml_available,244        is_clearml_available,245        is_codecarbon_available,246        is_comet_available,247        is_dagshub_available,248        is_dvclive_available,249        is_flyte_deck_standard_available,250        is_flytekit_available,251        is_mlflow_available,252        is_neptune_available,253        is_optuna_available,254        is_ray_available,255        is_ray_tune_available,256        is_sigopt_available,257        is_swanlab_available,258        is_tensorboard_available,259        is_trackio_available,260        is_wandb_available,261        rewrite_logs,262        run_hp_search_optuna,263        run_hp_search_ray,264        run_hp_search_sigopt,265        run_hp_search_wandb,266    )267    from .mxfp4 import (268        Mxfp4GptOssExperts,269        dequantize,270        load_and_swizzle_mxfp4,271        quantize_to_mxfp4,272        replace_with_mxfp4_linear,273        swizzle_mxfp4,274    )275    from .peft import PeftAdapterMixin276    from .quanto import replace_with_quanto_layers277    from .spqr import replace_with_spqr_linear278    from .vptq import replace_with_vptq_linear279 280    try:281        if not is_torch_available():282            raise OptionalDependencyNotAvailable()283    except OptionalDependencyNotAvailable:284        pass285    else:286        from .executorch import TorchExportableModuleWithStaticCache, convert_and_export_with_cache287 288    try:289        if not is_torch_greater_or_equal("2.3"):290            raise OptionalDependencyNotAvailable()291    except OptionalDependencyNotAvailable:292        pass293    else:294        from .tensor_parallel import (295            ALL_PARALLEL_STYLES,296            shard_and_distribute_module,297            translate_to_torch_parallel_style,298        )299 300    try:301        if not is_torch_greater_or_equal("2.5"):302            raise OptionalDependencyNotAvailable()303    except OptionalDependencyNotAvailable:304        pass305    else:306        from .flex_attention import make_flex_block_causal_mask307else:308    import sys309 310    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)311