cpu-inference.yml
Commit
1 year ago
- Files157
- .github/workflows
- cpu-torch-latest.yml
- hpu-gaudi2.yml
- nv-a6000.yml
- nv-accelerate-v100.yml
- nv-ds-chat.yml
- nv-inference.yml
- nv-mii.yml
- nv-nightly.yml
- nv-pre-compile-ops.yml
- nv-torch-latest-v100.yml
- nv-torch-nightly-v100.yml
- nv-transformers-v100.yml
README.md- accelerator
cpu_accelerator.py- real_accelerator.py
- blogs
- deepspeed-fp6/03-05-2024
README-Chinese.md- README.md
- assets
fp6-design.png- hero-figure.png
- servingllm
- 100-1000.png
- 100-250.png
- 100-500.png
- deepspeed-ulysses
- README.md
- csrc
- cpu/comm
ccl.cpp- shm.cpp
shm.h
- fp_quantizer
- includes
- context.h
- quantize.h
- quantize.cpp
quantize.cu
- includes
- memory_access_utils.h
- reduction_utils.h
- deepspeed
- __init__.py
- autotuning
- utils.py
- checkpoint
- constants.py
- ds_to_universal.py
- reshape_utils.py
- universal_checkpoint.py
- zero_checkpoint.py
- elasticity
- elastic_agent.py
- env_report.py
- git_version_info.py
- inference/v2
- config_v2.py
- engine_factory.py
- kernels/core_ops
- __init__.py
- blas_kernels
- blas_utils.h
- core_ops.cpp
- cuda_linear
- __init__.py
- cuda_linear.py
- cuda_linear_kernels.cpp
- cuda_linear_kernels.h
- fp6_linear.cu
- fp6_linear.cuh
- include
- configs.h
- kernel_matmul.cuh
- kernel_reduction.cuh
- ptx_cp.async.cuh
- ptx_mma.cuh
- utils_core.cuh
- utils_gmem.cuh
- utils_paralleldequant.cuh
- weight_prepacking.h
- gated_activations
- gated_activation_kernels_cuda.cu
- model_implementations
- __init__.py
- flat_model_helpers.py
- qwen_v2
- __init__.py
- container.py
- model.py
- policy.py
- modules
- heuristics.py
- implementations/linear
- __init__.py
- quantized_linear.py
- ragged/csrc
- ragged_ops.cpp
- launcher
- multinode_runner.py
- module_inject
- auto_tp.py
- containers
- bloom.py
- fusedqkv_utils.py
- moe
- layer.py
- sharded_moe.py
- utils.py
- ops
- __init__.py
- fp_quantizer
- __init__.py
- quantize.py
- transformer/inference
- config.py
- ds_attention.py
- triton
- matmul_ext.py
- runtime
- __init__.py
- base_optimizer.py
- bf16_optimizer.py
- comm
- coalesced_collectives.py
- config.py
- engine.py
- fp16
- fused_optimizer.py
- unfused_optimizer.py
- utils.py
- zero
- partition_parameters.py
- stage3.py
- stage_1_and_2.py
- utils
- __init__.py
- tensor_fragment.py
- docs
- _tutorials
- accelerator-abstraction-interface.md
- index.md
- op_builder
- all_ops.py
- builder.py
- cpu
- comm.py
- fp_quantizer.py
- inference_core_ops.py
- xpu
- builder.py
- requirements
requirements-dev.txt- requirements-inf.txt
- setup.py
- tests/unit
- checkpoint
- common.py
- test_latest_checkpoint.py
- test_lr_scheduler.py
- test_moe_checkpoint.py
- test_other_optimizer.py
- test_pipeline.py
- test_universal_checkpoint.py
- test_zero_optimizer.py
- common.py
- compression
- test_dequantization.py
- elasticity
- test_elastic.py
- inference
- test_inference.py
- v2
- modules
- test_quantized_linear_module.py
- ragged
- test_manager_configs.py
- launcher
- test_user_args.py
- moe
- test_moe.py
- multi_output_model.py
- ops
- accelerators
- test_accelerator_backward.py
- fp_quantizer
- test_fp_quant.py
- runtime
- activation_checkpointing
- test_activation_checkpointing.py
- comm
- test_coalesced_collectives.py
- compile
- test_compile_wrapper.py
- test_compile_zero.py
- test_load_config.py
- half_precision
- onebit
- test_onebit.py
- test_bf16.py
- test_dynamic_loss_scale.py
- test_fp16.py
- sparse_tensor
- test_averaging_sparse_gradients.py
- test_sparse_grads.py
- test_data_efficiency.py
- test_ds_config_dict.py
- test_ds_initialize.py
- test_multi_output_model.py
- test_mup_optimizers.py
- test_pld.py
- zero
- test_ignore_unused_parameters.py
- test_zero.py
- test_zero_context.py
- test_zero_context_return.py
- test_zero_leaf_module.py
- test_zero_tensor_fragment.py
- simple_model.py
- util.py
- version.txt