onecomp 1.3.2__tar.gz → 1.3.3__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {onecomp-1.3.2/onecomp.egg-info → onecomp-1.3.3}/PKG-INFO +1 -1
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/__version__.py +1 -1
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/_lpcd_config.py +3 -2
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/qep/_qep_config.py +2 -2
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/_quantizer.py +26 -6
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/gptq/gptq_layer.py +0 -6
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/runner.py +4 -0
- {onecomp-1.3.2 → onecomp-1.3.3/onecomp.egg-info}/PKG-INFO +1 -1
- {onecomp-1.3.2 → onecomp-1.3.3}/LICENSE +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/README.md +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/llama3-8b-gptq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/llama3-8b-jointq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/llama3-8b-lpcd-gptq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/llama3-8b-qep-gptq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/llama3-8b-various/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/qwen3-14b-gptq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/qwen3-14b-jointq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/qwen3-8b-gptq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/benchmark/qwen3-8b-jointq/quant_benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/api/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/api/jobs.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/constants.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/core/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/core/config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/core/database.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/main.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/models/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/models/job.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/schemas/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/schemas/job.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/services/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/services/huggingface.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/services/inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/services/job_store.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/worker/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/worker/celery_app.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/app/worker/tasks.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/cpu_patch.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/start_backend.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/dashboard/backend/start_worker.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/envs/vllm/v0_24_0_rocm/src/onecomp_vllm_v0_24_0_rocm/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/envs/vllm/v0_24_0_rocm/src/onecomp_vllm_v0_24_0_rocm/patch.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/cpu_inference/example_gptq_gguf_cpu.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/cpu_inference/example_mixed_gptq_gguf_cpu.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/cpu_inference/example_serve_cpu.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_auto_run.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_autobit.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_custom_calibration.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_gptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_jointq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_lpcd_gptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_mdbf.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_qep_gptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/example_save_load.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_blockwise_global_ptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_blockwise_global_ptq_staged.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_blockwise_ptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_global_ptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_global_ptq_dbf.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_global_ptq_distributed.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_lora_gptq_vllm_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_lora_sft.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_lora_sft_knowledge.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_lora_sft_knowledge_jointq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/post_process/example_reload_post_process_resave.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/pre_process/example_llama_preprocess_rtn.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/pre_process/example_preprocess_save_load.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_autobit_vllm_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_dbf_vllm_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_gptq_vllm_gptoss_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_gptq_vllm_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_gptq_vllm_qwen36_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_jointq_vllm_inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/llamacpp_plugins/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/llamacpp_plugins/gptq/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/llamacpp_plugins/gptq/constants.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/llamacpp_plugins/gptq/llamacpp_plugin.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/autobit/validate_autobit.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/autobit_qep/validate_autobit.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/gptq/validate_gptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/gptq/validate_load.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/gptq/validate_vllm.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/jointq/validate_jointq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/model_validation/qep_gptq/validate_gptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/__main__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/analyzer/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/analyzer/cumulative_error.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/analyzer/quantization_error.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/analyzer/weight_outlier.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/_cache.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/c4.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/calibration_config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/calibration_data_loader.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/chunking.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/custom.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/calibration/wikitext.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cli.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/cli.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/eval/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/eval/benchmark.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/eval/inspect_gguf.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/eval/parity.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/eval/perplexity.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/auto.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/blocks.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/checkpoint.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/dequantize.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/direct.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/fallback.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/rotation.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/export/skeleton.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/inference.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/llama_tooling.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/cpu/serve.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/__main__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/conf/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/conf/eval_config.yaml +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/base.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/adapter.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/data.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/gen_answer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/judge.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/radar_chart.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/run.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/mt_bench/show_result.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/throughput/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/throughput/adapter.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/throughput/bench.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/evals/throughput/run.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/orchestrator/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/orchestrator/aggregator.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/orchestrator/runner.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/orchestrator/server.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/orchestrator/subprocess_runner.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/run_evaluate.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/schema.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/utils/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/utils/model_utils.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/utils/ports.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/utils/resources.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/eval/utils/secrets.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/export/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/export/gguf_export.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/export/gguf_reader.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/export/gguf_writer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/export/hub.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/export/model_card.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/log.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/_gradient_solver.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/_lpcd_runner.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/_metric.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/_refiner.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/arch/_llama.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/arch/_llama_cf.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/lpcd/arch/_qwen3.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/model_config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_base.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/dbf_block_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/dbf_cbq_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/generic_block_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/gptq_block_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/gptq_cbq_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/helpers.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/onebit_block_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_blockwise/onebit_cbq_optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/core.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/dbf_adapter.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/gptq_adapter.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/helpers.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/losses.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_global_ptq/trainer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/_runtime.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/blockwise_ptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/global_ptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/global_ptq_distributed.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/post_process/post_process_lora_sft.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/hadamard_utils.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/modeling_llama.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/modeling_qwen3.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/optimizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/prepare_rotated_model.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/preprocess_args.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/quant_models.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/rotation_utils.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/pre_process/train_rotation.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/qep/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/qep/_quantize_with_qep.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/qep/_quantize_with_qep_arch.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantized_model_loader.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/arb/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/arb/_arb.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/arb/arb_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/_autobit.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/activation_stats.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/dbf_fallback.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/ilp.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/manual.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/autobit/visualize.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/cq/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/cq/_cq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/cq/cq_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/_dbf.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/admm_extended.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/balance.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/dbf_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/dbf_layer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/dbf_original.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/fine_tune.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/dbf/middle.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/gemlite.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/gptq/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/gptq/_gptq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/gptq/config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/_jointq.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/__version__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/clip.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/error_propagation/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/error_propagation/local_search_advanced.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/error_propagation/quantize_advanced.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/error_propagation/quantizer_advanced.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/local_search.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/quantize.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/quantize_multi_gpu.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/quantizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/jointq/core/solution.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/_mdbf.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/admm.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/gradient_refine.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/initialize.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/mdbf_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/mdbf_layer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/mdbf/utils.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/onebit/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/onebit/_onebit.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/onebit/onebit_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/onebit/onebit_layer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/qbb/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/qbb/_qbb.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/qbb/qbb_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/_quip.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/quant_quip.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/quip_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/utils.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/utils_had.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/quip/vector_balance.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/rtn/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/rtn/_rtn.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/rtn/quantizer.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/quantizer/rtn/rtn_impl.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/rotated_model_config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/runner_methods/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/runner_methods/chunked_quantization.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/runner_methods/jointq_error_propagation.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/runner_methods/multi_gpu_quantization.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/accuracy.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/activation_capture.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/activation_check.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/blockwise.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/device.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/dtype.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/lora.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/model_inputs.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/mxfp4_compat.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/perplexity.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/quant_config.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/quantization_progress.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/unfuse_moe.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp/utils/vram_estimator.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp.egg-info/SOURCES.txt +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp.egg-info/dependency_links.txt +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp.egg-info/entry_points.txt +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp.egg-info/requires.txt +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/onecomp.egg-info/top_level.txt +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/pyproject.toml +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/scripts/check_copyright_header.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/scripts/check_no_japanese.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/setup.cfg +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/dbf/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/dbf/modules/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/dbf/modules/gemlite_linear.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/dbf/modules/naive.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/dbf/vllm_plugin.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/gptq/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/gptq/constants.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/gptq/gptoss_wna16_moe.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/gptq/vllm_plugin.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/patches/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/patches/_paths.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/patches/apply_all.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/patches/gpt_oss_gptq_moe.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/patches/gpt_oss_wna16_bias.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/utils/__init__.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/utils/module.py +0 -0
- {onecomp-1.3.2 → onecomp-1.3.3}/vllm_plugins/utils/rotation.py +0 -0
|
@@ -25,7 +25,8 @@ class LPCDConfig:
|
|
|
25
25
|
gd_steps: Number of gradient-descent epochs per sub-problem.
|
|
26
26
|
gd_batch_size: Effective batch size for gradient accumulation.
|
|
27
27
|
gd_base_lr: Base learning rate for gradient-descent solver.
|
|
28
|
-
device: Device to perform LPCD optimisation on.
|
|
28
|
+
device: Device to perform LPCD optimisation on. Default is None.
|
|
29
|
+
When None, Runner synchronises it with ModelConfig.device.
|
|
29
30
|
|
|
30
31
|
Examples:
|
|
31
32
|
Minimal (residual correction only, fast)::
|
|
@@ -53,4 +54,4 @@ class LPCDConfig:
|
|
|
53
54
|
gd_steps: int = 20
|
|
54
55
|
gd_batch_size: int = 16
|
|
55
56
|
gd_base_lr: float = 1e-4
|
|
56
|
-
device: str =
|
|
57
|
+
device: str = None
|
|
@@ -25,7 +25,7 @@ class QEPConfig:
|
|
|
25
25
|
Default is 0.5.
|
|
26
26
|
device (str): Device to use for QEP computations
|
|
27
27
|
(e.g., "cuda", "mps", "cpu").
|
|
28
|
-
Default is
|
|
28
|
+
Default is None. When None, Runner synchronises it with ModelConfig.device.
|
|
29
29
|
exclude_layer_keywords (list[str]): List of keywords to identify
|
|
30
30
|
layers excluded from error propagation. Layers whose names
|
|
31
31
|
contain any of these keywords will be excluded.
|
|
@@ -52,6 +52,6 @@ class QEPConfig:
|
|
|
52
52
|
general: bool = False
|
|
53
53
|
percdamp: float = 0.01
|
|
54
54
|
perccorr: float = 0.5
|
|
55
|
-
device: str =
|
|
55
|
+
device: str = None
|
|
56
56
|
exclude_layer_keywords: list[str] = field(default_factory=lambda: ["mlp.down_proj"])
|
|
57
57
|
# TODO: exclude_layer_keywords depends on the architecture and needs to be fixed
|
|
@@ -400,12 +400,32 @@ class Quantizer(metaclass=ABCMeta):
|
|
|
400
400
|
weight[:, dead] = 0
|
|
401
401
|
|
|
402
402
|
# QEP correction
|
|
403
|
-
|
|
404
|
-
|
|
405
|
-
|
|
406
|
-
|
|
407
|
-
|
|
408
|
-
|
|
403
|
+
damp_scale = 1.0
|
|
404
|
+
max_retries = 5
|
|
405
|
+
for attempt in range(max_retries):
|
|
406
|
+
try:
|
|
407
|
+
damp = percdamp * torch.mean(torch.diag(hessian))
|
|
408
|
+
diag = torch.arange(hessian.shape[0], device=hessian.device)
|
|
409
|
+
hessian[diag, diag] += damp
|
|
410
|
+
rhs = weight @ delta_hatX
|
|
411
|
+
delta_weight = _safe_cholesky_and_solve(hessian, rhs).t()
|
|
412
|
+
weight = weight + (perccorr * delta_weight)
|
|
413
|
+
break
|
|
414
|
+
except torch._C._LinAlgError:
|
|
415
|
+
damp_scale *= 10.0
|
|
416
|
+
extra = damp_scale * damp
|
|
417
|
+
hessian[diag, diag] += extra
|
|
418
|
+
self.logger.warning(
|
|
419
|
+
"Cholesky failed (attempt %d/%d); adding extra damping %.2e",
|
|
420
|
+
attempt + 1,
|
|
421
|
+
max_retries,
|
|
422
|
+
extra,
|
|
423
|
+
)
|
|
424
|
+
else:
|
|
425
|
+
raise RuntimeError(
|
|
426
|
+
"Cholesky decomposition failed after %d damping attempts. "
|
|
427
|
+
"The Hessian may be severely ill-conditioned." % max_retries
|
|
428
|
+
)
|
|
409
429
|
|
|
410
430
|
if isinstance(module, Conv1d):
|
|
411
431
|
weight = weight.t()
|
|
@@ -414,12 +414,6 @@ class GPTQLinear(nn.Module):
|
|
|
414
414
|
if self._gemlite_layer is not None:
|
|
415
415
|
self.using_gemlite = True
|
|
416
416
|
|
|
417
|
-
# Permutation order
|
|
418
|
-
if perm is not None and actorder:
|
|
419
|
-
self.register_buffer("perm", perm.to(device))
|
|
420
|
-
else:
|
|
421
|
-
self.perm = None
|
|
422
|
-
|
|
423
417
|
# Bias
|
|
424
418
|
if bias is not None:
|
|
425
419
|
self.register_buffer("bias", bias.to(torch.float16).to(device))
|
|
@@ -261,9 +261,13 @@ class Runner:
|
|
|
261
261
|
self.qep_config = None
|
|
262
262
|
if qep:
|
|
263
263
|
self.qep_config = qep_config if qep_config is not None else QEPConfig()
|
|
264
|
+
if self.qep_config.device is None and self.model_config is not None:
|
|
265
|
+
self.qep_config.device = str(self.model_config.get_device())
|
|
264
266
|
self.lpcd_config = None
|
|
265
267
|
if lpcd:
|
|
266
268
|
self.lpcd_config = lpcd_config if lpcd_config is not None else LPCDConfig()
|
|
269
|
+
if self.lpcd_config.device is None and self.model_config is not None:
|
|
270
|
+
self.lpcd_config.device = str(self.model_config.get_device())
|
|
267
271
|
self.report_progress = report_progress
|
|
268
272
|
|
|
269
273
|
def check(self):
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{onecomp-1.3.2 → onecomp-1.3.3}/envs/vllm/v0_24_0_rocm/src/onecomp_vllm_v0_24_0_rocm/__init__.py
RENAMED
|
File without changes
|
{onecomp-1.3.2 → onecomp-1.3.3}/envs/vllm/v0_24_0_rocm/src/onecomp_vllm_v0_24_0_rocm/patch.py
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_gptq_vllm_gptoss_inference.py
RENAMED
|
File without changes
|
|
File without changes
|
{onecomp-1.3.2 → onecomp-1.3.3}/example/vllm_inference/example_gptq_vllm_qwen36_inference.py
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|