EdgeRazor
EdgeRazor: A Lightweight Framework for Large Language Models via Mixed-Precision Quantization-Aware Distillation
파일 탐색기
최종 버전 다운로드 (.zip)- Architeacture.png
- Demo.gif
- EdgeRazor-Framework.png
- Logo-full.png
- Logo-Full.svg
- Logo.png
- Logo.svg
- PerformanceComparison.png
- Playground.png
- Demo.md
- Deployment.md
- Evaluation.md
- local_server_q4_0.sh
- local_server_tq1_0.sh
- local_server_tq2_0.sh
- kd_kldc_fd.yaml
- qat_w4_a8_kd_fd.json
- qat_w4_a8_kd_fd.py
- qat_w4_a8_kd_fd.yaml
- qat_w1.58mp4_a8.yaml
- distill_qwen2.5-omni.sh
- requirements.txt
- cache_old_datasets.sh
- data_prepare.sh
- distill_qwen3.py
- distill_qwen3.sh
- README.md
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- run.sh
- train.yaml
- eval_fp.sh
- convert_qweight.py
- __init__.py
- config.py
- dataset.py
- ds_z2_config_qwen3.json
- ds_z3_config_qwen3.json
- main.py
- modeling.py
- trainer.py
- config.json
- configuration_llama.py
- modeling_llama.py
- tokenization_llama.py
- tokenization_llama_fast.py
- generate_mobilellm-350m_template.sh
- config.json
- configuration_qwen3.py
- modeling_qwen3.py
- modular_qwen3.py
- README.md
- generate_qwen3-0.6b_template.sh
- config.json
- configuration_qwen3.py
- modeling_qwen3.py
- modular_qwen3.py
- generate_qwen3-1.7b_template.sh
- README.md
- prepare.sh
- README.md
- requirements.txt
- eval.py
- eval.sh
- convert-pth-to-gguf.py
- convert-pth-to-gguf.sh
- quantize.sh
- __init__.py
- arg.py
- q_vit_w4_a16.yaml
- q_vit_w5_a16.yaml
- q_vit_w8_a16.yaml
- train.py
- train.sh
- .gitignore
- Quantization.md
- README.md
- requirements.txt
- vit.cpp
- partial_combine.py
- run.sh
- train_w4.yaml
- convert_qweight.py
- test_qweight_equivalence.py
- __init__.py
- config.py
- dataset.py
- ds_z2_config_qwen2_5omni.json
- ds_z3_config_qwen2_5omni.json
- main.py
- modeling.py
- train.yaml
- trainer.py
- __init__.py
- config.json
- configuration_qwen2_5_omni.py
- modeling_qwen2_5_omni.py
- modular_qwen2_5_omni.py
- processing_qwen2_5_omni.py
- README.md
- generate_qwen2.5-omni-7b_template.sh
- README.md
- requirements.txt
- __init__.py
- __main__.py
- arg.py
- ResNet-18.json
- train.py
- count_params.py
- inspect_layers.py
- q_resnet_w1.58_a16.yaml
- q_resnet_w1.58_a16_with_overrides.json
- q_resnet_w1.58_a16_with_overrides.yaml
- q_resnet_w1.58_a16_with_wildcard.yaml
- q_resnet_w1.58_a4.yaml
- q_resnet_w1.58_a8.yaml
- q_resnet_w4_a16.yaml
- q_resnet_w4_a4.yaml
- q_resnet_w4_a8.yaml
- quant_prop.py
- README.md
- requirements.txt
- run.sh
- perf-f1-1.png
- perf-f1-2.png
- __init__.py
- __main__.py
- arg.py
- test.py
- train.py
- ViT-S-16.json
- count_params.py
- generate_table.py
- inspect_layers.py
- q_vit_w1.58_a16.yaml
- q_vit_w1.58_a4.yaml
- q_vit_w1.58_a8.yaml
- q_vit_w1.58mp4_a16.yaml
- q_vit_w1.58mp4_a16_kldr_fd.yaml
- q_vit_w1.58mp4_a4.yaml
- q_vit_w1.58mp4_a4_kldr_fd.yaml
- q_vit_w1.58mp4_a8.yaml
- q_vit_w1.58mp4_a8_kldr_fd.yaml
- q_vit_w4_a16.yaml
- q_vit_w4_a4.yaml
- q_vit_w4_a8.yaml
- quant_prop.py
- README.md
- requirements.txt
- run.sh
- run1.sh
- run2.sh
- test.sh
- vit_hidden_states_layer.py
- vit_outputs.py
- app.py
- config.py
- README.md
- requirements.txt
- style.css
- __init__.py
- distill_config.py
- distill_function.py
- distill_function_config.py
- layer_select.py
- moe_loss.py
- __init__.py
- kd.py
- __init__.py
- qattn.py
- qkv_cache.py
- __init__.py
- qconv1d.py
- qconv2d.py
- qconv3d.py
- qembedding.py
- qlinear.py
- __init__.py
- quant_config.py
- quant_function.py
- quant_function_config.py
- quant_function_decorator.py
- quant_selector.py
- __init__.py
- map.py
- qat.py
- quantize.py
- modeling_edgerazor.py
- __init__.py
- edgerazor_quant.py
- embedding.py
- linear_marlin.py
- linear_py.py
- quant_mode_parse.py
- quant_ops.py
- __init__.py
- convert.py
- edgerazor.py
- edgerazor_config.py
- log.py
- trainer.py
- mobilellm_base.yaml
- qwen3_instruct.yaml
- qwen3_instruct_gen.yaml
- ifeval.yaml
- ifeval_qwen3_instruct.yaml
- instructions.py
- instructions_registry.py
- instructions_util.py
- README.md
- utils.py
- qwen2_5_omni.sh
- README.md
- __init__.py
- test_full_pipeline.py
- test_trainer_pipeline.py
- __init__.py
- test_edgerazor_kd_integration.py
- test_edgerazor_qat_integration.py
- test_kd_integration.py
- test_kv_cache_yaml_integration.py
- test_qad_integration.py
- test_qat_integration.py
- test_qkv_cache_integration.py
- test_trainer_integration.py
- test_v134_export_and_load.py
- w1.58-a8-kv8.yaml
- w4-a8-kv8.yaml
- __init__.py
- test_convert.py
- test_distill_config.py
- test_edgerazor.py
- test_edgerazor_config.py
- test_embedding.py
- test_kd.py
- test_log.py
- test_map.py
- test_mixed_precision.py
- test_qat.py
- test_qconv2d.py
- test_qembedding.py
- test_qkv_cache.py
- test_qlinear.py
- test_quant_config.py
- test_quant_emb_lm_head.py
- test_quant_function.py
- test_quant_function_decorator.py
- test_quant_mode_parse.py
- test_quant_selector.py
- test_replace_weights.py
- test_trainer.py
- test_vllm.py
- test_vllm_w158.py
- __init__.py
- conftest.py
- .gitignore
- LICENSE
- noxfile.py
- pyproject.toml
- README.md
- README_ZH.md
// repository documentation
Was this content helpful?
(0 ratings)