MANIFEST.in
Makefile
README.md
pyproject.toml
setup.cfg
setup.py
csrc/__init__.py
csrc/activation_kernels.metal
csrc/cache_kernels.metal
csrc/custom_all_reduce.metal
csrc/custom_all_reduce_test.metal
csrc/custom_quickreduce.metal
csrc/layernorm_kernels.metal
csrc/layernorm_quant_kernels.metal
csrc/metal_utils_kernels.metal
csrc/metal_views.metal
csrc/permute_cols.metal
csrc/pos_encoding_kernels.metal
csrc/sampler.metal
csrc/type_convert.metal
csrc/all2all/__init__.py
csrc/all2all/internode_combine.metal
csrc/all2all/internode_dispatch.metal
csrc/all2all/intranode.metal
csrc/all2all/intranode_combine.metal
csrc/all2all/intranode_dispatch.metal
csrc/attention/__init__.py
csrc/attention/__init__.pyi
csrc/attention/attention_generic.metal
csrc/attention/attention_kernels.metal
csrc/attention/attention_utils.metal
csrc/attention/dtype_bfloat16.metal
csrc/attention/dtype_float16.metal
csrc/attention/dtype_float32.metal
csrc/attention/dtype_fp8.metal
csrc/attention/merge_attn_states.metal
csrc/attention/paged_attention.metal
csrc/attention/vertical_slash_index.metal
csrc/attention/mla/sm100_metalcutlass_mla_kernel.metal
csrc/core/__init__.py
csrc/core/__init__.pyi
csrc/cpu/__init__.py
csrc/mamba/__init__.py
csrc/mamba/mamba_ssm/__init__.py
csrc/mamba/mamba_ssm/selective_scan_fwd.metal
csrc/metalcutlass_extensions/__init__.py
csrc/metalcutlass_extensions/metalcutlass_utils.metal
csrc/metalcutlass_extensions/mlxllm_collective_builder.metal
csrc/metalcutlass_extensions/mlxllm_custom_types.metal
csrc/metalcutlass_extensions/mlxllm_metalcutlass_library_extension.py
csrc/metalcutlass_extensions/test_integration.py
csrc/moe/__init__.py
csrc/moe/__init__.pyi
csrc/moe/grouped_topk_kernels.metal
csrc/moe/moe_align_sum_kernels.metal
csrc/moe/moe_lora_align_sum_kernels.metal
csrc/moe/moe_permute_unpermute_op.metal
csrc/moe/moe_wna16.metal
csrc/moe/topk_softmax_kernels.metal
csrc/moe/marlin_moe_wna16/ops.metal
csrc/moe/permute_unpermute_kernels/moe_permute_unpermute_kernel.metal
csrc/quickreduce/quick_reduce_impl.metal
csrc/rocm/attention.metal
csrc/rocm/skinny_gemms.metal
csrc/sparse/metalcutlass/sparse_compressor_c3x.metal
csrc/sparse/metalcutlass/sparse_scaled_mm_c3x.metal
csrc/sparse/metalcutlass/sparse_scaled_mm_entry.metal
mlxllm/__init__.py
mlxllm/_bc_linter.py
mlxllm/_custom_ops.py
mlxllm/_ipex_ops.py
mlxllm/beam_search.py
mlxllm/collect_env.py
mlxllm/connections.py
mlxllm/csrc_ops.py
mlxllm/env_override.py
mlxllm/envs.py
mlxllm/forward_context.py
mlxllm/logits_process.py
mlxllm/logprobs.py
mlxllm/metal_ops.py
mlxllm/mlx_inference_engine.py
mlxllm/mlx_metrics.py
mlxllm/mlx_ops.py
mlxllm/outputs.py
mlxllm/pooling_params.py
mlxllm/sampling_params.py
mlxllm/scalar_types.py
mlxllm/scripts.py
mlxllm/sequence.py
mlxllm/tasks.py
mlxllm/test_mlxllm_basic.py
mlxllm/torch_ops.py
mlxllm/tracing.py
mlxllm/version.py
mlxllm/attention/backends/__init__.py
mlxllm/attention/backends/abstract.py
mlxllm/attention/backends/metal.py
mlxllm/attention/backends/registry.py
mlxllm/attention/backends/utils.py
mlxllm/attention/layers/__init__.py
mlxllm/attention/layers/chunked_local_attention.py
mlxllm/attention/layers/cross_attention.py
mlxllm/attention/layers/encoder_only_attention.py
mlxllm/attention/layers/logits_processor.py
mlxllm/attention/ops/__init__.py
mlxllm/attention/ops/chunked_prefill_paged_decode.py
mlxllm/attention/ops/common.py
mlxllm/attention/ops/flashmla.py
mlxllm/attention/ops/merge_attn_states.py
mlxllm/attention/ops/metal_decode_attention.py
mlxllm/attention/ops/metal_flash_attention.py
mlxllm/attention/ops/metal_merge_attn_states.py
mlxllm/attention/ops/metal_reshape_and_cache_flash.py
mlxllm/attention/ops/metal_unified_attention.py
mlxllm/attention/ops/paged_attn.py
mlxllm/attention/ops/pallas_kv_cache_update.py
mlxllm/attention/ops/prefix_prefill.py
mlxllm/attention/ops/rocm_aiter_mla.py
mlxllm/attention/ops/rocm_aiter_paged_attn.py
mlxllm/attention/utils/__init__.py
mlxllm/attention/utils/fa_utils.py
mlxllm/attention/utils/kv_sharing_utils.py
mlxllm/autotuner/__init__.py
mlxllm/autotuner/callbacks.py
mlxllm/autotuner/datasets.py
mlxllm/autotuner/dora.py
mlxllm/autotuner/lora.py
mlxllm/autotuner/losses.py
mlxllm/autotuner/trainer.py
mlxllm/autotuner/utils.py
mlxllm/config/__init__.py
mlxllm/config/cache.py
mlxllm/config/device.py
mlxllm/config/mlx.py
mlxllm/config/model.py
mlxllm/config/scheduler.py
mlxllm/core/__init__.py
mlxllm/core/block_pool.py
mlxllm/core/encoder_cache_manager.py
mlxllm/core/kv_cache_coordinator.py
mlxllm/core/kv_cache_manager.py
mlxllm/core/kv_cache_utils.py
mlxllm/core/single_type_kv_cache_manager.py
mlxllm/core/sched/__init__.py
mlxllm/core/sched/interface.py
mlxllm/core/sched/output.py
mlxllm/core/sched/request_queue.py
mlxllm/core/sched/scheduler.py
mlxllm/engine/__init__.py
mlxllm/engine/arg_utils.py
mlxllm/engine/async_llm_engine.py
mlxllm/engine/coordinator.py
mlxllm/engine/core.py
mlxllm/engine/core_client.py
mlxllm/engine/detokenizer.py
mlxllm/engine/exceptions.py
mlxllm/engine/llm_engine.py
mlxllm/engine/logprobs.py
mlxllm/engine/output_processor.py
mlxllm/engine/parallel_sampling.py
mlxllm/engine/processor.py
mlxllm/engine/protocol.py
mlxllm/engine/utils.py
mlxllm/entry_points/__init__.py
mlxllm/entry_points/api_server.py
mlxllm/entry_points/chat_utils.py
mlxllm/entry_points/constants.py
mlxllm/entry_points/context.py
mlxllm/entry_points/harmony_utils.py
mlxllm/entry_points/launcher.py
mlxllm/entry_points/llm.py
mlxllm/entry_points/logger.py
mlxllm/entry_points/renderer.py
mlxllm/entry_points/score_utils.py
mlxllm/entry_points/ssl.py
mlxllm/entry_points/tool.py
mlxllm/entry_points/tool_server.py
mlxllm/entry_points/utils.py
mlxllm/entry_points/anthropic/__init__.py
mlxllm/entry_points/anthropic/api_server.py
mlxllm/entry_points/anthropic/protocol.py
mlxllm/entry_points/anthropic/serving_messages.py
mlxllm/entry_points/cli/__init__.py
mlxllm/entry_points/cli/benchmark.py
mlxllm/entry_points/cli/chat.py
mlxllm/entry_points/cli/collect_env.py
mlxllm/entry_points/cli/convert.py
mlxllm/entry_points/cli/embed.py
mlxllm/entry_points/cli/generate.py
mlxllm/entry_points/cli/main.py
mlxllm/entry_points/cli/models.py
mlxllm/entry_points/cli/openai.py
mlxllm/entry_points/cli/run_batch.py
mlxllm/entry_points/cli/serve.py
mlxllm/entry_points/cli/tokenize.py
mlxllm/entry_points/cli/types.py
mlxllm/entry_points/cli/benchmark/__init__.py
mlxllm/entry_points/cli/benchmark/base.py
mlxllm/entry_points/cli/benchmark/latency.py
mlxllm/entry_points/cli/benchmark/main.py
mlxllm/entry_points/cli/benchmark/serve.py
mlxllm/entry_points/cli/benchmark/throughput.py
mlxllm/entry_points/openai/__init__.py
mlxllm/entry_points/openai/api_server.py
mlxllm/entry_points/openai/cli_args.py
mlxllm/entry_points/openai/logits_processors.py
mlxllm/entry_points/openai/protocol.py
mlxllm/entry_points/openai/run_batch.py
mlxllm/entry_points/openai/serving_chat.py
mlxllm/entry_points/openai/serving_classification.py
mlxllm/entry_points/openai/serving_completion.py
mlxllm/entry_points/openai/serving_embedding.py
mlxllm/entry_points/openai/serving_engine.py
mlxllm/entry_points/openai/serving_models.py
mlxllm/entry_points/openai/serving_pooling.py
mlxllm/entry_points/openai/serving_responses.py
mlxllm/entry_points/openai/serving_score.py
mlxllm/entry_points/openai/serving_tokenization.py
mlxllm/entry_points/openai/serving_transcription.py
mlxllm/entry_points/openai/speech_to_text.py
mlxllm/entry_points/openai/tool_parsers/__init__.py
mlxllm/entry_points/tool_parsers/__init__.py
mlxllm/entry_points/tool_parsers/abstract_tool_parser.py
mlxllm/entry_points/tool_parsers/llama_tool_parser.py
mlxllm/entry_points/tool_parsers/mistral_tool_parser.py
mlxllm/entry_points/tool_parsers/openai_tool_parser.py
mlxllm/inputs/__init__.py
mlxllm/inputs/data.py
mlxllm/inputs/parse.py
mlxllm/inputs/preprocess.py
mlxllm/kv_offload/__init__.py
mlxllm/kv_offload/abstract.py
mlxllm/kv_offload/backend.py
mlxllm/kv_offload/factory.py
mlxllm/kv_offload/lru_manager.py
mlxllm/kv_offload/m4.py
mlxllm/kv_offload/mediums.py
mlxllm/kv_offload/spec.py
mlxllm/kv_offload/backends/__init__.py
mlxllm/kv_offload/backends/m4metal.py
mlxllm/kv_offload/worker/__init__.py
mlxllm/kv_offload/worker/m4cpu_m4gpu_metal.py
mlxllm/kv_offload/worker/worker.py
mlxllm/model_executor/__init__.py
mlxllm/model_executor/custom_ops.py
mlxllm/model_executor/executor.py
mlxllm/model_executor/parameters.py
mlxllm/model_executor/utils.py
mlxllm/model_executor/worker.py
mlxllm/model_executor/layers/__init__.py
mlxllm/model_executor/layers/activation.py
mlxllm/model_executor/layers/attention_layer_base.py
mlxllm/model_executor/layers/batch_invariant.py
mlxllm/model_executor/layers/layernorm.py
mlxllm/model_executor/layers/lightning_attn.py
mlxllm/model_executor/layers/linear.py
mlxllm/model_executor/layers/logits_processor.py
mlxllm/model_executor/layers/mla.py
mlxllm/model_executor/layers/pooler.py
mlxllm/model_executor/layers/resampler.py
mlxllm/model_executor/layers/rotary_embedding.py
mlxllm/model_executor/layers/utils.py
mlxllm/model_executor/layers/vocab_parallel_embedding.py
mlxllm/model_executor/layers/fla/__init__.py
mlxllm/model_executor/layers/fla/based.py
mlxllm/model_executor/layers/fla/gla.py
mlxllm/model_executor/layers/fla/retnet.py
mlxllm/model_executor/layers/fla/rwkv.py
mlxllm/model_executor/layers/fused_moe/__init__.py
mlxllm/model_executor/layers/fused_moe/experts.py
mlxllm/model_executor/layers/fused_moe/fused_moe.py
mlxllm/model_executor/layers/fused_moe/routing.py
mlxllm/model_executor/layers/mamba/__init__.py
mlxllm/model_executor/layers/mamba/mamba.py
mlxllm/model_executor/layers/mamba/ssm.py
mlxllm/model_executor/layers/multimodal_embedding/__init__.py
mlxllm/model_executor/layers/multimodal_embedding/audio.py
mlxllm/model_executor/layers/multimodal_embedding/fusion.py
mlxllm/model_executor/layers/multimodal_embedding/projection.py
mlxllm/model_executor/layers/multimodal_embedding/vision.py
mlxllm/model_executor/layers/quantization/__init__.py
mlxllm/model_executor/layers/quantization/kv_cache.py
mlxllm/model_executor/layers/quantization/config/__init__.py
mlxllm/model_executor/layers/quantization/config/quant_config.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/MPLinearKernel.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/__init__.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/allspark.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/bitblas.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/conch.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/cutlass.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/dynamic_4bit.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/exllama.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/machete.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/marlin.py
mlxllm/model_executor/layers/quantization/kernals/mixed_precision/metalcutlass.py
mlxllm/model_executor/layers/quantization/kernals/scaled_mm/ScaledMMLinearKernel.py
mlxllm/model_executor/layers/quantization/kernals/scaled_mm/__init__.py
mlxllm/model_executor/layers/quantization/kernals/scaled_mm/aiter.py
mlxllm/model_executor/layers/quantization/kernals/scaled_mm/m4.py
mlxllm/model_executor/layers/quantization/mamba/__init__.py
mlxllm/model_executor/layers/quantization/mamba/mamba_quant.py
mlxllm/model_executor/layers/quantization/quark/__init__.py
mlxllm/model_executor/layers/quantization/quark/base_config.py
mlxllm/model_executor/layers/quantization/quark/quark.py
mlxllm/model_executor/layers/quantization/quark/quark_moe.py
mlxllm/model_executor/layers/quantization/quark/utils.py
mlxllm/model_executor/layers/quantization/quark/schemas/__init__.py
mlxllm/model_executor/layers/quantization/quark/schemas/quark_ocp_mx.py
mlxllm/model_executor/layers/quantization/quark/schemas/quark_scheme.py
mlxllm/model_executor/layers/quantization/quark/schemas/quark_w8a8_fp8.py
mlxllm/model_executor/layers/quantization/quark/schemas/quark_w8a8_int8.py
mlxllm/model_executor/model_loader/__init__.py
mlxllm/model_executor/model_loader/base_loader.py
mlxllm/model_executor/model_loader/bitsandbytes_loader.py
mlxllm/model_executor/model_loader/default_loader.py
mlxllm/model_executor/model_loader/gguf_loader.py
mlxllm/model_executor/model_loader/online_quantization.py
mlxllm/model_executor/model_loader/sharded_state_loader.py
mlxllm/model_executor/model_loader/tensorizer_loader.py
mlxllm/model_executor/model_loader/utils.py
mlxllm/model_executor/model_loader/weight_utils.py
mlxllm/model_executor/models/__init__.py
mlxllm/model_executor/models/deepseek.py
mlxllm/model_executor/models/llama.py
mlxllm/model_executor/models/qwen2.py
mlxllm/models/__init__.py
mlxllm/models/config_helper.py
mlxllm/models/convert.py
mlxllm/models/download.py
mlxllm/models/info.py
mlxllm/models/manager.py
mlxllm/multimodal/__init__.py
mlxllm/multimodal/audio.py
mlxllm/multimodal/base.py
mlxllm/multimodal/cache.py
mlxllm/multimodal/evs.py
mlxllm/multimodal/hasher.py
mlxllm/multimodal/image.py
mlxllm/multimodal/inputs.py
mlxllm/multimodal/parse.py
mlxllm/multimodal/parsing.py
mlxllm/multimodal/processing.py
mlxllm/multimodal/profiling.py
mlxllm/multimodal/registry.py
mlxllm/multimodal/utils.py
mlxllm/multimodal/video.py
mlxllm/tokenization/__init__.py
mlxllm/tokenization/tokenizer.py
requirements/requirements.dev.txt
requirements/requirements.docs.txt
requirements/requirements.prod.txt
requirements/requirements.txt
tests/conftest.py
tests/unit/attention/test_layer.py
tests/unit/attention/test_selector.py
tests/unit/attention/backends/test_abstract.py
tests/unit/attention/backends/test_logits_processor.py
tests/unit/attention/backends/test_registry.py
tests/unit/attention/backends/test_utils.py
tests/unit/attention/layers/test_chunked_local_attention.py
tests/unit/attention/layers/test_cross_attention.py
tests/unit/attention/layers/test_encoder_only_attention.py
tests/unit/attention/ops/test_chunked_prefill_paged_decode.py
tests/unit/attention/ops/test_common.py
tests/unit/attention/ops/test_flashmla.py
tests/unit/attention/ops/test_merge_attn_states.py
tests/unit/attention/ops/test_metal_decode_attention.py
tests/unit/attention/ops/test_metal_flash_attention.py
tests/unit/attention/ops/test_metal_merge_attn_states.py
tests/unit/attention/ops/test_metal_reshape_and_cache_flash.py
tests/unit/attention/ops/test_metal_unified_attention.py
tests/unit/attention/ops/test_paged_attn.py
tests/unit/attention/ops/test_pallas_kv_cache_update.py
tests/unit/attention/ops/test_prefix_prefill.py
tests/unit/attention/utils/test_fa_utils.py
tests/unit/attention/utils/test_kv_sharing_utils.py
tests/unit/config/test_cache.py
tests/unit/config/test_config_scheduler.py
tests/unit/config/test_device.py
tests/unit/config/test_mlx.py
tests/unit/config/test_model.py
tests/unit/core/test_block_pool.py
tests/unit/core/test_encoder_cache_manager.py
tests/unit/core/test_kv_cache_coordinator.py
tests/unit/core/test_kv_cache_manager.py
tests/unit/core/test_kv_cache_utils.py
tests/unit/core/test_single_type_kv_cache_manager.py
tests/unit/core/sched/test_core_scheduler.py
tests/unit/core/sched/test_interface.py
tests/unit/core/sched/test_output.py
tests/unit/core/sched/test_request_queue.py
tests/unit/engine/test_arg_utils.py
tests/unit/engine/test_async_llm_engine.py
tests/unit/engine/test_coordinator.py
tests/unit/engine/test_core_client.py
tests/unit/engine/test_detokenizer.py
tests/unit/engine/test_engine_utils.py
tests/unit/engine/test_exceptions.py
tests/unit/engine/test_llm_engine.py
tests/unit/engine/test_logprobs.py
tests/unit/engine/test_output_processor.py
tests/unit/engine/test_parallel_sampling.py
tests/unit/engine/test_processor.py
tests/unit/engine/test_protocol.py
tests/unit/entry_points/__init__.py
tests/unit/entry_points/conftest.py
tests/unit/entry_points/test_constants.py
tests/unit/entry_points/test_context.py
tests/unit/entry_points/test_logger.py
tests/unit/entry_points/anthropic/__init__.py
tests/unit/entry_points/anthropic/test_api_server.py
tests/unit/entry_points/anthropic/test_protocol.py
tests/unit/entry_points/anthropic/test_serving_messages.py
tests/unit/entry_points/cli/__init__.py
tests/unit/entry_points/cli/test_collect_env.py
tests/unit/entry_points/cli/test_main.py
tests/unit/entry_points/cli/test_models.py
tests/unit/entry_points/cli/test_openai.py
tests/unit/entry_points/cli/test_run_batch.py
tests/unit/entry_points/cli/test_types.py
tests/unit/entry_points/openai/__init__.py
tests/unit/entry_points/openai/test_protocol.py
tests/unit/entry_points/openai/test_serving_engine.py
tests/unit/kv_offload/test_backend.py
tests/unit/kv_offload/test_factory.py
tests/unit/kv_offload/test_kv_offload_abstract.py
tests/unit/kv_offload/test_lru_manager.py
tests/unit/kv_offload/test_m4.py
tests/unit/kv_offload/test_mediums.py
tests/unit/kv_offload/test_spec.py
tests/unit/kv_offload/backends/test_m4metal.py
tests/unit/kv_offload/worker/test_m4cpu_m4gpu_metal.py
tests/unit/kv_offload/worker/test_worker.py
tests/unit/model_executor/__init__.py
tests/unit/model_executor/test_custom_ops.py
tests/unit/model_executor/test_executor.py
tests/unit/model_executor/test_parameters.py
tests/unit/model_executor/test_utils.py
tests/unit/model_executor/test_worker.py
tests/unit/model_executor/layers/__init__.py
tests/unit/model_executor/layers/fla/__init__.py
tests/unit/model_executor/layers/fla/test_based.py
tests/unit/model_executor/layers/fla/test_gla.py
tests/unit/model_executor/layers/fla/test_retnet.py
tests/unit/model_executor/layers/fla/test_rwkv.py
tests/unit/model_executor/layers/fused_moe/__init__.py
tests/unit/model_executor/layers/fused_moe/test_experts.py
tests/unit/model_executor/layers/fused_moe/test_fused_moe.py
tests/unit/model_executor/layers/fused_moe/test_routing.py
tests/unit/model_executor/layers/mamba/__init__.py
tests/unit/model_executor/layers/mamba/test_mamba.py
tests/unit/model_executor/layers/mamba/test_ssm.py
tests/unit/model_executor/layers/multimodal_embedding/__init__.py
tests/unit/model_executor/layers/multimodal_embedding/test_audio.py
tests/unit/model_executor/layers/multimodal_embedding/test_fusion.py
tests/unit/model_executor/layers/multimodal_embedding/test_projection.py
tests/unit/model_executor/layers/multimodal_embedding/test_vision.py
tests/unit/model_executor/model_loader/__init__.py
tests/unit/model_executor/model_loader/test_base_loader.py
tests/unit/model_executor/model_loader/test_default_loader.py
tests/unit/model_executor/model_loader/test_weight_utils.py
tests/unit/model_executor/models/__init__.py
tests/unit/model_executor/models/test_deepseek.py
tests/unit/model_executor/models/test_llama.py
tests/unit/model_executor/models/test_qwen2.py
tests/unit/model_executor/warmup/__init__.py
tests/unit/model_executor/warmup/test_deep_gemm_warmup.py
tests/unit/model_executor/warmup/test_kernel_warmup.py
tests/unit/test_csrc/__init__.py
tests/unit/test_csrc/conftest.py
tests/unit/test_csrc/mlx_torch_bridge.py
tests/unit/test_csrc/test_activation_ops.py
tests/unit/test_csrc/test_bindings_availability.py
tests/unit/test_csrc/test_moe_ops.py
tests/unit/test_csrc/test_utility_ops.py
tools/__init__.py
tools/base_tool.py
tools/tool_factory.py
tools/tool_manager.py
tools/tool_registry.py
tools/tools.py
tools/mlx_conversion/__init__.py
tools/mlx_conversion/mlx_converter.py
tools/mlx_conversion/mlx_model_benchmark.py
tools/mlx_conversion/mlx_model_metrics.py
tools/mlx_conversion/mlx_optimizer.py
tools/mlx_conversion/mlx_probe_model.py
tools/mlx_packaging/__init__.py
tools/mlx_packaging/mlx_bundler.py
tools/mlxllm_versioning/__init__.py
tools/validation/__init__.py
tools/validation/validate_config.py