axolotl
Go ahead and axolotl questions
파일 탐색기
최종 버전 다운로드 (.zip)- bug-report.yaml
- config.yml
- docs.yml
- feature-request.yaml
- base.yml
- docker-e2e.yml
- docs.yml
- lint.yml
- main.yml
- multi-gpu-e2e.yml
- nightlies.yml
- precommit-autoupdate.yml
- preview-docs.yml
- pypi.yml
- tests-nightly.yml
- tests.yml
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- dependabot.yml
- FUNDING.yml
- PULL_REQUEST_TEMPLATE.md
- release-drafter.yml
- SECURITY.md
- SUPPORT.md
- config.yaml
- handler.py
- test_input.json
- train.py
- utils.py
- .gitignore
- Dockerfile
- hub.json
- README.md
- requirements.txt
- test-input.json
- tests.json
- launch.json
- README.md
- tasks.json
- __init__.py
- cicd.sh
- cicd_cuda_kernels.sh
- cleanup.py
- cleanup.sh
- Dockerfile-uv.jinja
- e2e_cuda_kernels.py
- e2e_tests.py
- multigpu.py
- multigpu.sh
- single_gpu.py
- zero1.json
- zero1_torch_compile.json
- zero2.json
- zero2_torch_compile.json
- zero3.json
- zero3_bf16.json
- zero3_bf16_cpuoffload_all.json
- zero3_bf16_cpuoffload_params.json
- dev_chat_template.yml
- README.md
- Dockerfile-cloud-no-tmux-uv
- Dockerfile-cloud-uv
- Dockerfile-uv
- Dockerfile-uv-base
- grpo.md
- model_architectures.md
- new_model_support.md
- preference_tuning.md
- pretraining.md
- reward_modelling.md
- sft.md
- conversation.qmd
- index.qmd
- inst_tune.qmd
- pretraining.qmd
- stepwise_supervised.qmd
- template_free.qmd
- tokenized.qmd
- 4d-mask.png
- ray-cluster-dashboard.png
- examples-allowlist.yml
- generate_config_docs.py
- generate_examples_docs.py
- .gitignore
- 1_58bit_finetuning.qmd
- amd_hpc.qmd
- attention.qmd
- batch_vs_grad.qmd
- checkpoint_saving.qmd
- choosing_method.qmd
- cli.qmd
- custom_integrations.qmd
- dataset_loading.qmd
- dataset_preprocessing.qmd
- debugging.qmd
- docker.qmd
- ebft.qmd
- expert_quantization.qmd
- faq.qmd
- fsdp_qlora.qmd
- getting-started.qmd
- gradient_checkpointing.qmd
- grpo.qmd
- inference.qmd
- input_output.qmd
- installation.qmd
- lora.qmd
- lora_optims.qmd
- lr_groups.qmd
- mac.qmd
- mixed_precision.qmd
- multi-gpu.qmd
- multi-node.qmd
- multimodal.qmd
- multimodal_assistant_mask.md
- multipack.qmd
- nccl.qmd
- nd_parallelism.qmd
- nvfp4_lora.qmd
- optimizations.qmd
- optimizers.qmd
- qat.qmd
- quantize.qmd
- ray-integration.qmd
- reward_modelling.qmd
- rlhf.qmd
- sequence_parallelism.qmd
- streaming.qmd
- support-matrix.qmd
- telemetry.qmd
- torchao.qmd
- training_stability.qmd
- vllm_serving.qmd
- llama3-8b-deepspeed-alst.yaml
- llama3-8b-fsdp2-alst.yaml
- README.md
- apertus-8b-qlora.yaml
- README.md
- afm-4.5b-qlora.yaml
- README.md
- btlm-ft.yml
- qlora.yml
- lora.yml
- qlora.yml
- lora.yml
- qlora.yml
- lora.yml
- qlora.yml
- README.md
- 16bit-lora.yaml
- 8bit-lora.yaml
- fft-ds-zero3.yaml
- README.md
- deepcoder-14B-preview-lora.yml
- config-7b-lora.yml
- config-7b-qlora.yml
- config-7b.yml
- qlora.yml
- qlora.yml
- config.yml
- config.yml
- README.md
- config.yml
- lora.yml
- qlora.yml
- README.md
- lora.yml
- config.yml
- README.md
- lora.yml
- qlora.yml
- qwen2-moe-lora.yaml
- qwen2-moe-qlora.yaml
- README.md
- config-3b.yml
- README.md
- config-lora.yml
- fft.yml
- lora.yml
- README.md
- qlora.yml
- lora-mps.yml
- lora.yml
- pretrain.yml
- qlora.yml
- README.md
- xgen-7b-8k-qlora.yml
- qlora.yml
- README.md
- README.md
- baseten.yaml
- modal.yaml
- command-r-7b-qlora.yml
- fft.yaml
- qlora-vision.yaml
- qlora.yaml
- README.md
- colab-axolotl-example.ipynb
- cogito-v1-preview-llama-3B-lora.yml
- cogito-v1-preview-qwen-14B-lora.yml
- fft-fsdp-16b.yaml
- qlora-fsdp-2_5.yaml
- README.md
- v4-flash-nvfp4-lora.yaml
- devstral-small-qlora.yml
- README.md
- llama-3_1-8b-hsdp-tp.yaml
- qwen3-8b-fsdp-tp-cp.yaml
- README.md
- eaft-example.yml
- llama-1b-ebft-opencode-novllm.yaml
- llama-1b-ebft-opencode.yaml
- llama-1b-ebft-strided-structured.yaml
- llama-1b-ebft-strided.yaml
- llama-3b-ebft-strided-fft.yaml
- llama-8b-ebft-strided-fft.yaml
- qwen35-4b-ebft-structured-async.yaml
- qwen35-4b-ebft-structured.yaml
- qwen35-9b-ebft-structured.yaml
- README.md
- qwen3_30ba3b_ep_fft_4gpu.yaml
- qwen3_30ba3b_ep_fsdp_fft_4gpu.yaml
- qwen3_30ba3b_ep_lora_4gpu.yaml
- falcon-e-3b-dpo.yaml
- falcon-e-3b-ft.yaml
- falcon-h1-1b-deep-qlora.yaml
- falcon-h1-1b-qlora-cp.yaml
- falcon-h1-1b-qlora.yaml
- falcon-h1-34b-qlora.yaml
- falcon-h1-3b-qlora.yaml
- falcon-h1-500m-qlora.yaml
- falcon-h1-7b-qlora.yaml
- qlora.yml
- reward-model.yaml
- gemma-3-1b-qlora.yml
- gemma-3-270m-qlora.yml
- gemma-3-4b-qlora.yml
- gemma-3-4b-vision-qlora.yml
- gemma-3n-e2b-qlora.yml
- gemma-3n-e2b-vision-audio-qlora.yml
- gemma-3n-e2b-vision-qlora.yml
- README.md
- 26b-a4b-moe-bnb-lora.yaml
- 26b-a4b-moe-nvfp4-lora.yaml
- 26b-a4b-moe-qlora.yaml
- 31b-qlora.yaml
- e2b-vision-lora.yaml
- README.md
- 12b-text-lora.yaml
- 12b-vision-lora.yaml
- README.md
- qlora-32b.yaml
- glm-45-air-qlora.yaml
- README.md
- glm-4-6v-flash-ddp.yaml
- glm-4-6v-flash-qlora.yaml
- README.md
- lora.yaml
- lora_fsdp.yaml
- qlora.yaml
- qlora_fsdp.yaml
- README.md
- glm-5.2-nvfp4-lora.yaml
- gpt-oss-120b-fft-fsdp2-offload.yaml
- gpt-oss-20b-fft-deepspeed-zero3.yaml
- gpt-oss-20b-fft-fsdp2-offload.yaml
- gpt-oss-20b-fft-fsdp2.yaml
- gpt-oss-20b-sft-lora-singlegpu.yaml
- gpt-oss-safeguard-20b-sft-lora-singlegpu.yaml
- README.md
- granite-4.0-tiny-fft.yaml
- README.md
- hunyuan-v1-dense-qlora.yaml
- README.md
- internvl3_5-8b-qlora.yml
- README.md
- qlora.yaml
- qlora_deepspeed.yaml
- qlora_fsdp_large.yaml
- README.md
- kimi-48b-lora.yaml
- README.md
- lfm2-350m-fft.yaml
- lfm2-8b-a1b-lora.yaml
- lfm2-vl-lora.yaml
- README.md
- fft_optimized.yml
- gptq-lora.yml
- lisa.yml
- loftq.yml
- lora.yml
- qlora-fsdp.yml
- qlora.yml
- README.md
- relora.yml
- pretrain-1b.yaml
- sft-1b.yaml
- 3b-fp8-fsdp2.yaml
- 3b-qat-fsdp2.yaml
- 3b-qat-mxfp4.yaml
- 3b-qat-nvfp4.yaml
- fft-8b-liger-fsdp.yaml
- fft-8b.yaml
- instruct-dpo-lora-8b.yml
- instruct-lora-8b.yml
- lora-1b-deduplicate-dpo.yml
- lora-1b-deduplicate-sft.yml
- lora-1b-kernels.yml
- lora-1b-ray.yml
- lora-1b-sample-packing-sequentially.yml
- lora-1b.yml
- lora-8b.yml
- opentelemetry-qlora.yml
- qlora-1b-gdpo.yaml
- qlora-1b-kto.yaml
- qlora-1b.yml
- qlora-fsdp-405b.yaml
- qlora-fsdp-70b.yaml
- qlora.yml
- README.md
- sparse-finetuning.yaml
- lora-11b.yaml
- maverick-qlora-fsdp1.yaml
- scout-qlora-fsdp1.yaml
- scout-qlora-single-h100.yaml
- scout-vision-qlora-fsdp.yaml
- README.md
- scout-qlora-flexattn-fsdp2.yaml
- scout-qlora-single-h100-flex.yaml
- scout-vision-qlora-fsdp2-flex.yaml
- lora-7b.yaml
- magistral-small-think-qlora.yaml
- README.md
- magistral-small-vision-24B-qlora.yml
- README.md
- magistral-small-fsdp-qlora.yaml
- magistral-small-qlora.yaml
- README.md
- config.yml
- mimo-7b-qlora.yaml
- README.md
- m2-qlora.yaml
- ministral-small-qlora.yaml
- README.md
- ministral3-3b-think-qlora.yaml
- README.md
- ministral3-3b-vision-qlora.yml
- README.md
- ministral3-3b-qlora.yaml
- README.md
- bigstral-ds-zero3.yaml
- mistral-dpo-qlora.yml
- mixtral-8x22b-qlora-fsdp.yml
- mixtral-qlora-fsdp.yml
- mixtral.yml
- mixtral_22.yml
- lora-mps.yml
- mistral-qlora-orpo.yml
- config.yml
- lora.yml
- mistral-qlora-fsdp.yml
- qlora.yml
- README.md
- qlora-text.yml
- qlora-vision.yml
- README.md
- mistral-small-3.1-24B-lora.yml
- README.md
- fft-text.yml
- fft-vision.yml
- qlora-text.yml
- qlora-vision.yml
- README.md
- qlora-vision.yaml
- qlora.yaml
- README.md
- nemotron-mini-4b-qlora.yaml
- 120b-a12b-qlora.yaml
- nano-30b-a3b-qlora-cp.yaml
- nano-30b-a3b-qlora.yaml
- README.md
- olmo3-7b-qlora.yaml
- README.md
- finetune.yml
- README.md
- paddleocr-vl-1_6-full-finetune.yaml
- paddleocr-vl-1_6-qlora.yaml
- README.md
- lora-3.5.yaml
- phi-ft.yml
- phi-qlora.yml
- phi2-ft.yml
- phi3-ft-fsdp.yml
- phi3-ft.yml
- README.md
- lora-12b.yml
- plano-4b-qlora.yaml
- README.md
- Gemma3-12B_baseline.yml
- Gemma3-12B_qat.yml
- Math-Gemma3-12B_baseline.yml
- Math-Gemma3-12B_qat.yml
- Math-Gemma3-27B_baseline.yml
- Math-Gemma3-27B_qat.yml
- Math-Qwen2.5-72B_baseline.yml
- Math-Qwen2.5-72B_qat.yml
- Qwen2.5-72B_baseline.yml
- Qwen2.5-72B_qat.yml
- adamw-pretrain-fsdp2.yaml
- dpo.yaml
- muon-pretrain-fsdp2.yaml
- prm.yaml
- qlora-fsdp.yaml
- reward-model.yaml
- lora-7b.yaml
- lora-7b.yaml
- 30b-a3b-nvfp4-lora.yaml
- 32b-qlora.yaml
- 8b-lora-fused-attn-compile.yaml
- 8b-lora-fused-attn.yaml
- 8b-qat-fsdp2.yml
- qlora-fsdp.yaml
- README.md
- reward-model.yaml
- qwen3-next-80b-a3b-nvfp4-lora.yaml
- qwen3-next-80b-a3b-qlora.yaml
- README.md
- 122b-a10b-moe-qlora-fsdp.yaml
- 122b-a10b-moe-qlora.yaml
- 27b-fft.yaml
- 27b-qlora-fsdp.yaml
- 27b-qlora.yaml
- 35b-a3b-moe-qlora-fsdp.yaml
- 35b-a3b-moe-qlora.yaml
- 35b-a3b-moe-vision-lora.yaml
- 9b-fft-vision.yaml
- 9b-lora-vision.yaml
- README.md
- README.md
- seed-oss-36b-qlora.yaml
- README.md
- shieldstral-3b-lora.yaml
- shieldstral-3b-vision-lora.yaml
- axolotl.slurm
- README.md
- README.md
- smolvlm2-2B-lora.yaml
- pretrain.yaml
- README.md
- sft.yaml
- custom_trainer_profiling.py
- dpo-swanlab-completions.yml
- dpo-swanlab-full-featured.yml
- lora-swanlab-profiling.yml
- README.md
- README.md
- trinity-nano-preview-qlora.yaml
- README.md
- voxtral-mini-audio-qlora.yml
- voxtral-mini-qlora.yml
- axolotl-badge-web-legacy.png
- axolotl-badge-web.png
- axolotl.png
- axolotl_logo_digital_black.svg
- axolotl_logo_digital_white.svg
- axolotl_symbol_digital_black.svg
- axolotl_symbol_digital_white.svg
- axolotl_wordmark_digital_black.svg
- axolotl_wordmark_digital_white.svg
- sticker_fixed.png
- analyze_profile.py
- chat_datasets.py
- cloud-entrypoint-term.sh
- cloud-entrypoint.sh
- cuda13_env.sh
- cutcrossentropy_install.py
- generate_cli_config_options.py
- motd
- uv-entrypoint.sh
- __init__.py
- run.sh
- train_sft.py
- __init__.py
- __init__.py
- base.py
- modal_.py
- __init__.py
- args.py
- diffusion.py
- fetch.py
- load.py
- lora_merge.py
- sweeps.py
- train.py
- __init__.py
- args.py
- art.py
- chat.py
- checks.py
- config.py
- config_options.py
- delinearize_llama4.py
- evaluate.py
- generate_config_options.py
- inference.py
- main.py
- merge_lora.py
- merge_sharded_fsdp_weights.py
- plugins.py
- preprocess.py
- quantize.py
- train.py
- vllm_serve.py
- __init__.py
- architectures.py
- const.py
- datasets.py
- __init__.py
- __init__.py
- base.py
- causal.py
- rl.py
- __init__.py
- chatml.py
- llama3x.py
- shared.py
- __init__.py
- messages.py
- __init__.py
- chat_builder.py
- __init__.py
- chat.py
- __init__.py
- args.py
- trainer.py
- __init__.py
- args.py
- kernels.py
- rewards.py
- strided.py
- trainer.py
- __init__.py
- args.py
- async_trainer.py
- fast_async_trainer.py
- replay_buffer.py
- sampler.py
- trainer.py
- __init__.py
- activation_checkpointing.py
- checkpoints.py
- distributed_parallel.py
- layer_offloading.py
- optimizer.py
- packing.py
- rng_state_loader.py
- scheduler.py
- __init__.py
- base.py
- constants.py
- mamba.py
- trl.py
- utils.py
- __init__.py
- training_args.py
- training_args_base.py
- __init__.py
- ACKNOWLEDGEMENTS.md
- args.py
- LICENSE
- README.md
- __init__.py
- args.py
- plugin.py
- README.md
- __init__.py
- args.py
- callbacks.py
- generation.py
- plugin.py
- README.md
- trainer.py
- utils.py
- __init__.py
- args.py
- buffer.py
- experts_fn.py
- plugin.py
- README.md
- shard.py
- __init__.py
- args.py
- LICENSE
- optimizer.py
- README.md
- prep_math_rl.py
- tinker_rl.yaml
- tinker_sft.yaml
- __init__.py
- math_reward.py
- __init__.py
- args.py
- data.py
- plugin.py
- rl_trainer.py
- trainer.py
- __init__.py
- liger.py
- __init__.py
- forward_kl.py
- __init__.py
- args.py
- callbacks.py
- chat_template.py
- collator.py
- collator_online_teacher.py
- README.md
- trainer.py
- utils.py
- __init__.py
- dsv4.py
- gemma4.py
- glm_moe_dsa.py
- nvfp4_moe.py
- qwen3_moe.py
- __init__.py
- attention.py
- attention_csa.py
- attention_gather.py
- gated_pool.py
- indexer.py
- lora_fp8.py
- lora_mlp.py
- mhc.py
- patch.py
- rope.py
- __init__.py
- _autotune.py
- attention_mla_absorb.py
- attention_topk.py
- config.py
- context_parallel.py
- dispatch.py
- indexer.py
- patch.py
- __init__.py
- blockscaled_gemm_dispatch.py
- fused_dx.py
- grouped.py
- grouped_kernel.py
- quant_mxfp8.py
- quant_nvfp4.py
- swiglu.py
- __init__.py
- grouped_gram.py
- lora_ops.py
- ops.py
- single.py
- scalar_type.hpp
- kernel.h
- kernel_selector.h
- marlin_template.h
- ops_standalone.cu
- repack_standalone.cu
- sm80_kernel_bfloat16_fe2m1f_bfloat16.cu
- dequant.h
- marlin.cuh
- marlin_dtypes.cuh
- marlin_mma.h
- launcher_body.inc
- PROVENANCE.md
- __init__.py
- backend.py
- fused_dequant.py
- nonexpert_linear.py
- prep.py
- __init__.py
- chunked_bnb.py
- dequant_grouped.py
- experts.py
- experts_lora_fastpath.py
- gemma4_fp8_nonexpert.py
- gemma4_nf4_nonexpert.py
- grouped_lora.py
- grouped_moe.py
- grouped_train.py
- layers.py
- lora_ops.py
- metadata.json
- multi_lora.py
- mx_weights.py
- nvfp4_fp8_quantizer.py
- nvfp4_fsdp.py
- nvfp4_moe_loading.py
- nvfp4_nonexpert.py
- nvfp4_weight_converter.py
- parallel_experts.py
- parallel_linear_lora.py
- runtime.py
- selective_dequant.py
- selective_dequant_kernel.py
- torchao_fp4_add.py
- __init__.py
- experts.py
- experts_lora_fastpath.py
- fp4_cute.py
- fp4_cute_ops.py
- fp8_bwd.py
- lora.py
- multi_lora.py
- nvfp4.py
- nvfp4_lora.py
- nvfp4_quant.py
- sf_layout.py
- triton_nvfp4.py
- __init__.py
- __init__.py
- args.py
- autotune_callback.py
- autotune_collector.py
- constants.py
- merge_aware_callback.py
- merge_aware_linear.py
- plugin.py
- quant_training_guard.py
- README.md
- __init__.py
- base.py
- deepseekv2.py
- jamba.py
- qwen3_5.py
- qwen3_5_moe.py
- __init__.py
- args.py
- LICENSE
- plugin.py
- README.md
- utils.py
- __init__.py
- args.py
- plugin.py
- README.md
- utils.py
- __init__.py
- args.py
- cli.py
- README.md
- __init__.py
- args.py
- plugin.py
- nemo_gym_multi_env.yaml
- nemo_gym_multi_turn.yaml
- nemo_gym_sudoku.yaml
- __init__.py
- args.py
- data_producer.py
- dataset.py
- multi_turn.py
- plugin.py
- README.md
- rewards.py
- server.py
- snr_results_google-gemma-2-2b.json
- snr_results_meta-llama-Llama-3.2-1B-Instruct.json
- snr_results_meta-llama-Llama-3.2-1B.json
- snr_results_meta-llama-Llama-3.2-3B-Instruct.json
- snr_results_meta-llama-Llama-3.2-3B.json
- snr_results_Qwen-Qwen2.5-1.5B-Instruct.json
- snr_results_Qwen-Qwen2.5-1.5B.json
- snr_results_Qwen-Qwen2.5-3B-Instruct.json
- snr_results_Qwen-Qwen2.5-3B.json
- snr_results_Qwen-Qwen2.5-7B-Instruct.json
- snr_results_Qwen-Qwen2.5-7B.json
- __init__.py
- args.py
- LICENSE
- README.md
- __init__.py
- args.py
- callbacks.py
- completion_logger.py
- plugins.py
- profiling.py
- README.md
- __init__.py
- base.py
- config.py
- LICENSE.md
- __init__.py
- autotune_telemetry.py
- dora.py
- geglu.py
- gemma4_fused_rope.py
- lora.py
- op_registry.py
- quantize.py
- rms_norm_gated.py
- swiglu.py
- utils.py
- __init__.py
- __init__.py
- adapter.py
- constants.py
- model.py
- patch_manager.py
- processor.py
- tokenizer.py
- utils.py
- __init__.py
- processing.py
- __init__.py
- configuration_kimi.py
- modeling_kimi.py
- patch_kimi_linear.py
- tokenization_kimi.py
- __init__.py
- processing.py
- __init__.py
- processing.py
- __init__.py
- base.py
- profile.py
- registry.py
- templates.py
- __init__.py
- configuration_mamba.py
- modeling_mamba.py
- __init__.py
- __init__.py
- float8_fsdp.py
- float8_moe_filter.py
- fsdp2.py
- fsdp2_quantized.py
- parallelism_config.py
- __init__.py
- flash_attn_4.py
- flash_attn_d512.py
- flex_attn.py
- fp8_attn.py
- large_head.py
- sage_attn.py
- sdpa_varlen.py
- xformers.py
- __init__.py
- batch_dataset_fetcher.py
- __init__.py
- offload_cpu.py
- offload_disk.py
- __init__.py
- chunked.py
- eaft.py
- __init__.py
- __init__.py
- activation.py
- __init__.py
- modeling.py
- fused_attn.py
- __init__.py
- fused_attn.py
- __init__.py
- modeling.py
- __init__.py
- modeling.py
- __init__.py
- mistral_common_tokenizer.py
- __init__.py
- modeling.py
- __init__.py
- modeling_flash_attention_utils.py
- __init__.py
- fused_attn.py
- __init__.py
- fused_attn.py
- modeling.py
- __init__.py
- fused_attn.py
- __init__.py
- fused_attn.py
- __init__.py
- modeling.py
- __init__.py
- fused_attn.py
- __init__.py
- modeling.py
- __init__.py
- mamba_utils.py
- __init__.py
- utils.py
- __init__.py
- batch.py
- __init__.py
- patch.py
- __init__.py
- base.py
- patch.py
- __init__.py
- lr.py
- trl.py
- trl_vllm.py
- utils.py
- __init__.py
- trainer_loss_calc.py
- __init__.py
- __init__.py
- activation_offload_checkpoint.py
- btlm_attn_hijack_flash.py
- checkpoint_activation_offload.py
- deepspeed_utils.py
- fsdp2_qlora.py
- gemma4_hybrid_mask.py
- gemma4_loss_kwargs.py
- kernelize_fixes.py
- llama_attn_hijack_flash.py
- llama_attn_hijack_xformers.py
- lora_kernels.py
- mistral_attn_hijack_flash.py
- moe_quant.py
- multipack.py
- relora.py
- scaled_softmax_attn.py
- selective_checkpointing.py
- selective_checkpointing_offload.py
- stablelm_attn_hijack_flash.py
- torchao_optim.py
- trainer_accelerator_args.py
- transformers_fa_utils.py
- utils.py
- __init__.py
- chat_template.py
- llama3.py
- README.md
- __init__.py
- chat_template.py
- chatml.py
- llama3.py
- passthrough.py
- user_defined.py
- zephyr.py
- __init__.py
- ebft_chat_multiturn.py
- ebft_opencode.py
- ebft_pretrain.py
- ebft_reasoning.py
- ebft_strided_chat.py
- ebft_strided_structured.py
- __init__.py
- chatml.py
- llama3.py
- user_defined.py
- __init__.py
- chat.py
- __init__.py
- chat_template.py
- __init__.py
- _synthetic.py
- alpaca_chat.py
- alpaca_instruct.py
- alpaca_w_system.py
- base.py
- chat_template.py
- completion.py
- context_qa.py
- creative_acr.py
- input_output.py
- jinja_template_analyzer.py
- llama2_chat.py
- metharme.py
- orcamini.py
- pretrain.py
- pygmalion.py
- stepwise_supervised.py
- user_defined.py
- __init__.py
- process_cleanup.py
- vllm_serve_lora.py
- vllm_worker_ext.py
- __init__.py
- callbacks.py
- errors.py
- manager.py
- runtime_metrics.py
- whitelist.yaml
- __init__.py
- comet_.py
- dynamic_checkpoint.py
- generation.py
- lisa.py
- mlflow_.py
- models.py
- opentelemetry.py
- perplexity.py
- profiler.py
- qat.py
- swanlab.py
- tokens_per_second.py
- trackio_.py
- alpaca.jinja
- aya.jinja
- chatml.jinja
- cohere.jinja
- command_a.jinja
- command_a_rag.jinja
- command_a_tool_use.jinja
- deepseek_v2.jinja
- deepseek_v3.jinja
- exaone.jinja
- exaone4.jinja
- falcon_h1.jinja
- gemma.jinja
- gemma3.jinja
- gemma3n.jinja
- gemma4.jinja
- gemma4_unified.jinja
- jamba.jinja
- llama3.jinja
- llama3_2_vision.jinja
- llama4.jinja
- llava.jinja
- metharme.jinja
- mistral_v1.jinja
- mistral_v2v3.jinja
- mistral_v3_tekken.jinja
- mistral_v7_tekken.jinja
- nemotron_h.jinja
- phi_3.jinja
- phi_35.jinja
- phi_4.jinja
- pixtral.jinja
- qwen2_vl.jinja
- qwen3.jinja
- qwen3_5.jinja
- qwen_25.jinja
- __init__.py
- base.py
- __init__.py
- batching.py
- core.py
- dpo.py
- mamba.py
- mm_chat.py
- __init__.py
- __init__.py
- __init__.py
- sequence_parallel.py
- __init__.py
- lock.py
- rl.py
- sft.py
- shared.py
- streaming.py
- utils.py
- wrappers.py
- __init__.py
- sft.py
- __init__.py
- mistral3_processor.py
- mistral_tokenizer.py
- __init__.py
- adopt.py
- qgalore.py
- sinkgd.py
- sinkgd_triton.py
- __init__.py
- multipack.py
- utils.py
- __init__.py
- __init__.py
- config.py
- datasets.py
- deprecated.py
- dynamic_checkpoint.py
- enums.py
- fsdp.py
- integrations.py
- model.py
- multimodal.py
- peft.py
- quantization.py
- training.py
- trl.py
- utils.py
- validation.py
- vllm.py
- __init__.py
- bench.py
- comet_.py
- cuda13.py
- datasets.py
- dict.py
- distributed.py
- environment.py
- fp32_norms.py
- freeze.py
- import_helper.py
- logging.py
- lora.py
- mlflow_.py
- model_shard_quant.py
- quantization.py
- schedulers.py
- tee.py
- tokenization.py
- trackio_.py
- train.py
- trainer.py
- wandb_.py
- weight_serde.py
- __init__.py
- convert.py
- datasets.py
- evaluate.py
- logging_config.py
- processing_strategies.py
- prompt_tokenizers.py
- prompters.py
- train.py
- __init__.py
- conftest.py
- test_chat_repl.py
- test_cli_base.py
- test_cli_evaluate.py
- test_cli_fetch.py
- test_cli_inference.py
- test_cli_interface.py
- test_cli_merge_lora.py
- test_cli_merge_sharded_fsdp_weights.py
- test_cli_plugins.py
- test_cli_preprocess.py
- test_cli_sweeps.py
- test_cli_train.py
- test_cli_version.py
- test_cli_vllm_serve.py
- test_load_cfg_capabilities.py
- test_nested_options.py
- test_utils.py
- __init__.py
- __init__.py
- test_messages.py
- conftest.py
- test_async_grpo.py
- test_builders.py
- test_builders_rl.py
- test_gc_steps_wiring.py
- test_quantized_lora_checkpoint_resume.py
- test_cut_cross_entropy.py
- test_fp8.py
- test_hooks.py
- test_kd.py
- test_liger.py
- test_llm_compressor.py
- test_scattermoe_lora_kernels.py
- test_scattermoe_lora_olmoe.py
- test_sonicmoe.py
- test_sonicmoe_lora.py
- test_geglu.py
- test_lora.py
- test_lora_features.py
- test_lora_mlp_geglu_moe_correctness.py
- test_quantize.py
- test_scattermoe_int64_offsets.py
- test_swiglu.py
- __init__.py
- test_sp.py
- __init__.py
- test_flex.py
- test_gdpo.py
- test_grpo.py
- __init__.py
- _fp32_norms_dtype_capture.py
- test_dist_muon_fsdp2.py
- test_eval.py
- test_fp8_fsdp2.py
- test_fsdp1.py
- test_fsdp2.py
- test_fsdp2_fp32_norms.py
- test_fsdp2_lora_kernels.py
- test_gemma3.py
- test_llama.py
- test_locking.py
- test_ray.py
- test_sinkgd_fsdp2.py
- test_tiled_mlp_fsdp2.py
- test_tp.py
- __init__.py
- test_lora_kernel_patching.py
- __init__.py
- test_4d_multipack_llama.py
- test_activation_checkpointing.py
- test_cli_integrations.py
- test_fa_xentropy.py
- test_falcon_samplepack.py
- test_flattening.py
- test_fsdp2_qlora.py
- test_fused_llama.py
- test_lora_llama_multipack.py
- test_mistral_samplepack.py
- test_mixtral_samplepack.py
- test_model_patches.py
- test_multipack_packing_equivalence.py
- test_peft_embeddings.py
- test_phi_multipack.py
- test_resume.py
- __init__.py
- test_batch_flattening.py
- test_flex.py
- test_relora_llama.py
- test_reward_model_smollm2.py
- test_trainer_loss_calc.py
- .gitignore
- __init__.py
- test_activation_offloading.py
- test_deepseekv3.py
- test_diffusion.py
- test_dpo.py
- test_ebft.py
- test_embeddings_lr.py
- test_evaluate.py
- test_falcon.py
- test_gemma2.py
- test_gemma3_text.py
- test_imports.py
- test_llama.py
- test_llama_pretrain.py
- test_llama_vision.py
- test_load_model.py
- test_lora_llama.py
- test_mamba.py
- test_mistral.py
- test_mixtral.py
- test_optimizers.py
- test_packing_loss.py
- test_phi.py
- test_preprocess.py
- test_process_reward_model_smollm2.py
- test_profiler.py
- test_qat.py
- test_quantization.py
- test_qwen.py
- test_save_first_step.py
- test_schedulers.py
- test_streaming.py
- test_tokenizer.py
- utils.py
- alpaca.json
- conversation.json
- conversation.missingturns.json
- conversation.tokenized.json
- conversation.tokenized_llama2chat.json
- __init__.py
- test_dsv4_attention_ops.py
- __init__.py
- test_dsa_attention_eager.py
- test_glm_dsa_cpu.py
- test_mla_absorb_gather.py
- test_sparse_topk_attn.py
- __init__.py
- bench_int64_kernel.py
- bench_int64_kernel_results.md
- bench_mxfp4.py
- bench_mxfp4_results.md
- conftest.py
- test_bnb_experts_forward.py
- test_ep_sentinel_skip.py
- test_gates_backward_autocast_dtype.py
- test_gptoss_layout.py
- test_grouped_fp4_dequant_path.py
- test_grouped_fp4_guards.py
- test_grouped_fp4_perf.py
- test_grouped_fp4_skew.py
- test_grouped_fp4_train.py
- test_grouped_fp4_weight_scale_2.py
- test_mx_lora_grouped_gram.py
- test_mxfp4_cache_fsdp.py
- test_mxfp4_expert_weights.py
- test_mxfp4_experts_forward.py
- test_mxfp4_integration.py
- test_nvfp4_experts_forward.py
- test_nvfp4_fsdp_sharding.py
- test_nvfp4_weight_converter.py
- test_parallel_experts_large_batch_repro.py
- test_scattermoe_lora_int64_indices.py
- test_scattermoe_lora_m_bucket.py
- test_shared_dequant_helper.py
- __init__.py
- test_grouped_backend.py
- test_kernels_adapters.py
- test_kernels_config.py
- test_kernels_config_intent.py
- test_merge_aware_callback.py
- test_merge_aware_linear.py
- test_quant_training_guard.py
- __init__.py
- test_tiled_mlp_moe.py
- test_mora.py
- __init__.py
- test_adapter_plugin_registry.py
- test_diffusion.py
- test_diffusion_callback.py
- test_expert_parallel.py
- test_gemma4_moe.py
- test_kd_chat_template.py
- test_kd_liger.py
- test_kd_topk_forward_kl.py
- test_kd_trainer_direct_loss.py
- test_liger.py
- test_liger_qwen_vl_rope_default.py
- test_nemo_gym.py
- test_scattermoe_autotune_telemetry.py
- test_scattermoe_lora.py
- test_scattermoe_lora_kernels.py
- test_scattermoe_multi_lora.py
- test_sonicmoe.py
- test_sonicmoe_fp4_cute_prep.py
- test_sonicmoe_lora.py
- test_sonicmoe_multi_lora.py
- test_sonicmoe_nvfp4_base.py
- test_sonicmoe_nvfp4_forward.py
- test_sonicmoe_nvfp4_gradients.py
- test_swanlab.py
- test_fused_rope_autotune_telemetry.py
- test_gemma4_fused_rope.py
- test_gemma4_fused_rope_compile.py
- test_gemma4_fused_rope_unit_offset.py
- test_op_registry.py
- test_rms_norm_gated.py
- test_patch_manager_gemma4_unified.py
- test_checkpoint_activation_offload.py
- test_flash_attn_4.py
- test_flash_attn_d512.py
- test_float8_fsdp_sharding.py
- test_float8_moe_filter.py
- test_fsdp2_quantized.py
- test_gemma4_fused_attn.py
- test_gemma4_fused_attn_patch.py
- test_gemma4_hybrid_mask.py
- test_gemma4_unified_fused_attn.py
- test_kernelize_fixes.py
- test_large_head_attention.py
- test_large_head_compile_hoist.py
- test_llama_attn_hijack_flash.py
- test_lora_kernels_gemma4_unified.py
- test_lora_mlp_routed_expert_guard.py
- test_mamba_utils.py
- test_pixtral_flash_attention_patch.py
- test_qwen3_5_fused_attn.py
- test_qwen3_5_packing_patch.py
- test_qwen3_fused_attn.py
- test_qwen3_fused_attn_defensive.py
- test_qwen3_fused_attn_robustness.py
- test_qwen3_next_modeling_patch.py
- test_qwen3_vl_fused_attn.py
- test_relora.py
- test_sdpa_varlen.py
- test_selective_checkpointing.py
- test_selective_checkpointing_offload.py
- test_trainer_accelerator_args.py
- test_trl_vllm.py
- test_voxtral_modeling_patch.py
- test_validation.py
- __init__.py
- test_chat.py
- __init__.py
- conftest.py
- test_alpaca.py
- test_chat_template_ds_schema_unification.py
- test_chat_template_utils.py
- test_chat_templates.py
- test_chat_templates_advanced.py
- test_chat_templates_mistral.py
- test_chat_templates_muse_glimmer.py
- test_chat_templates_snapshot.py
- test_chat_templates_thinking.py
- test_chat_templates_tool_call_string_arguments.py
- test_dpo_chat_templates.py
- test_dpo_chatml.py
- test_dpo_user_defined.py
- test_get_messages_str_format.py
- test_jinja_template_analyzer.py
- test_kto_chatml.py
- test_kto_llama3.py
- test_kto_user_defined.py
- test_raw_io.py
- test_stepwise.py
- test_synthetic.py
- __init__.py
- conftest.py
- test_callbacks.py
- test_errors.py
- test_manager.py
- test_runtime_metrics.py
- test_dynamic_checkpoint.py
- test_gc_callback.py
- test_skip_eval_on_resume.py
- test_tokens_per_second.py
- test_hash.py
- test_rl.py
- test_utils.py
- test_config_validation_lora.py
- test_freeze_lora.py
- test_merge_lora.py
- test_modules_to_save_kwargs.py
- test_mora_validation.py
- test_activation_offloading.py
- test_config_validators.py
- test_default_values.py
- test_ebft.py
- test_fsdp.py
- test_moe_quant.py
- test_qgalore.py
- test_selective_checkpointing_validation.py
- test_cuda13.py
- test_grpo_rw_fnc.py
- test_import_helper.py
- test_mistral3_processor.py
- test_mistral_tokenizer.py
- test_sinkgd.py
- test_train.py
- __init__.py
- conftest.py
- constants.py
- hf_offline_utils.py
- test_attn_implementation.py
- test_calculate_total_num_steps.py
- test_chunked_xentropy.py
- test_context_parallel_batch_size.py
- test_convert.py
- test_data.py
- test_datasets.py
- test_dict.py
- test_ebft_kernels.py
- test_ebft_strided_structured.py
- test_empty_tokenization.py
- test_exact_deduplication.py
- test_fp32_norms.py
- test_freeze.py
- test_glm_dsa_cp_validation.py
- test_http_weight_sync.py
- test_loaders.py
- test_logging_config_file_capture.py
- test_lora.py
- test_mm_chat_collator.py
- test_mm_collator_warnings.py
- test_model_support.py
- test_model_support_callsites.py
- test_model_support_profiles.py
- test_model_support_registrations.py
- test_model_support_registry.py
- test_no_legacy_attn_reads.py
- test_normalize_config.py
- test_opentelemetry_callback.py
- test_packed_batch_sampler.py
- test_packed_dataset.py
- test_packed_pretraining.py
- test_perplexity.py
- test_process_labels_fusion.py
- test_processing_strategies.py
- test_prompt_tokenizers.py
- test_prompters.py
- test_revision_parameter.py
- test_save_deduplicated.py
- test_schedulers.py
- test_streaming.py
- test_tensor_parallel_batch_size.py
- test_tokenizers.py
- test_train.py
- test_triton_kernels.py
- test_utils_tee.py
- test_validation_dataset.py
- test_vectorized_scanner.py
- test_weighted_loss.py
- .axolotl-complete.bash
- .bandit
- .coderabbit.yaml
- .coveragerc
- .editorconfig
- .gitattributes
- .gitignore
- .mypy.ini
- .pre-commit-config.yaml
- _quarto.yml
- AGENTS.md
- CITATION.cff
- CLAUDE.md
- CNAME
- codecov.yml
- docker-compose.yaml
- FAQS.md
- favicon.jpg
- index.qmd
- LICENSE
- MANIFEST.in
- pyproject.toml
- README.md
- styles.css
- VERSION
# 설치 가이드
1. 코드 내려받기
git clone https://github.com/axolotl-ai-cloud/axolotl
깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.
cd axolotl
방금 내려받은 프로젝트 폴더 안으로 이동합니다.
2. 공식 설치 스크립트
쉬움 추천curl -LsSf https://astral.sh/uv/install.sh | sh
공식 설치 스크립트를 다운로드해서 바로 실행합니다. 이 한 줄로 필요한 게 자동으로 설치됩니다.
설치 후 새 터미널을 열고, 프로그램의 버전 확인 명령(예: --version)으로 정상 설치됐는지 확인하세요.
이 레포의 README에 적힌 실제 명령어를 그대로 가져왔습니다.
3. Docker
쉬움사전 준비물
- Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
- Docker Desktop 컨테이너를 빌드하고 실행하려면 필요합니다. 설치 후 실행해서 백그라운드에 켜두세요.
docker run --gpus '"all"' --ipc=host --rm -it axolotlai/axolotl:main-latest
빌드된 이미지를 실제 컨테이너로 실행합니다.
<img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/docker-e2e.yml/badge.svg" alt="docker-e2e-tests">
이 명령어를 터미널에 그대로 입력해 실행하세요.
터미널에 docker compose ps 를 입력해 컨테이너들이 Up 상태인지 확인하세요. README에 포트 번호가 적혀있다면 브라우저에서 http://localhost:포트번호 로 접속해보세요.
이 레포의 README에 적힌 실제 명령어를 그대로 가져왔습니다.
4. Python
쉬움사전 준비물
uv pip install torch==2.12.0 torchvision
requirements.txt 등에 명시된 파이썬 라이브러리를 설치합니다.
uv pip install --no-build-isolation axolotl[deepspeed]
requirements.txt 등에 명시된 파이썬 라이브러리를 설치합니다.
에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.
이 레포의 README에 적힌 실제 명령어를 그대로 가져왔습니다.
// repository documentation
Was this content helpful?
(0 ratings)
