{ "run_info": { "created_at": "2026-04-14T17:06:46+00:00", "total_time": 1255.1344404510019, "experiment_name": "prefixtuning/llama-3.2-3B-lr_0.005_zero_init", "peft_branch": "main", "train_config": { "model_id": "meta-llama/Llama-3.2-3B", "dtype": "bfloat16", "max_seq_length": 768, "batch_size": 4, "batch_size_eval": 50, "max_steps": 5000, "eval_steps": 250, "compile": false, "query_template": "Question: {query} Think step by step.\nAnswer:", "seed": 0, "grad_norm_clip": 1.0, "optimizer_type": "AdamW", "optimizer_kwargs": { "lr": 0.005 }, "lr_scheduler": "cosine", "use_amp": false, "autocast_adapter_dtype": true, "generation_kwargs": { "max_length": 800, "max_new_tokens": 300 }, "attn_implementation": null, "init_kv_cache_prefix": null }, "peft_config": { "task_type": "CAUSAL_LM", "peft_type": "PREFIX_TUNING", "auto_mapping": null, "peft_version": "0.19.0", "base_model_name_or_path": "meta-llama/Llama-3.2-3B", "revision": null, "inference_mode": false, "num_virtual_tokens": 200, "token_dim": 1024, "num_transformer_submodules": 1, "num_attention_heads": 8, "num_layers": 28, "modules_to_save": null, "init_weights": "zero", "encoder_hidden_size": 3072, "prefix_projection": false }, "error_msg": "" }, "train_info": { "accelerator_memory_reserved_avg": 13987667876, "accelerator_memory_max": 21969764352, "accelerator_memory_reserved_99th": 19157483520, "train_time": 970.3953097560006, "file_size": 45875328, "num_trainable_params": 11468800, "num_total_params": 3224218624, "status": "success", "metrics": [ { "step": 250, "valid accuracy": 0.2, "train loss": 1.2678344011306764, "train samples": 1000, "train time": 27.194034429907333, "eval time": 15.960887206005282, "tokens / sec": 7785.494298233168, "mem allocated avg": 7055727894.528, "mem reserved avg": 14077342515.2, "elapsed time": 65.7957478990138 }, { "step": 500, "valid accuracy": 0.36, "train loss": 0.7902288379669189, "train samples": 2000, "train time": 27.01687965914607, "eval time": 11.780722531999345, "tokens / sec": 7698.705499085536, "mem allocated avg": 7047839463.424, "mem reserved avg": 13768356528.128, "elapsed time": 107.10377580400382 }, { "step": 750, "valid accuracy": 0.24, "train loss": 0.7467671196460723, "train samples": 3000, "train time": 27.93710034090327, "eval time": 13.317983273998834, "tokens / sec": 7674.418511003849, "mem allocated avg": 7058360565.76, "mem reserved avg": 13883792162.816, "elapsed time": 150.8143266110128 }, { "step": 1000, "valid accuracy": 0.28, "train loss": 0.7247283264398575, "train samples": 4000, "train time": 28.07561205892125, "eval time": 15.844377710003755, "tokens / sec": 7420.532794183539, "mem allocated avg": 7050874298.368, "mem reserved avg": 13931053580.288, "elapsed time": 197.35486783200759 }, { "step": 1250, "valid accuracy": 0.28, "train loss": 0.7230025169849396, "train samples": 5000, "train time": 27.4366716020304, "eval time": 15.843770207007765, "tokens / sec": 7600.703285910508, "mem allocated avg": 7049694097.408, "mem reserved avg": 13987190145.024, "elapsed time": 243.09947748400737 }, { "step": 1500, "valid accuracy": 0.3, "train loss": 0.7165617660284043, "train samples": 6000, "train time": 27.41693946205487, "eval time": 15.844231765004224, "tokens / sec": 7635.097283185629, "mem allocated avg": 7052062963.712, "mem reserved avg": 13884471640.064, "elapsed time": 288.92823548801243 }, { "step": 1750, "valid accuracy": 0.38, "train loss": 0.7077521693706512, "train samples": 7000, "train time": 27.86099931402714, "eval time": 15.84966672999144, "tokens / sec": 7514.267440313827, "mem allocated avg": 7052561504.256, "mem reserved avg": 14377218473.984, "elapsed time": 335.2041009560053 }, { "step": 2000, "valid accuracy": 0.42, "train loss": 0.7111810308694839, "train samples": 8000, "train time": 27.7813215659844, "eval time": 13.623913794013788, "tokens / sec": 7476.102225975604, "mem allocated avg": 7049145475.072, "mem reserved avg": 14036481605.632, "elapsed time": 379.2373954960058 }, { "step": 2250, "valid accuracy": 0.38, "train loss": 0.7035314469337464, "train samples": 9000, "train time": 28.324825873816735, "eval time": 15.338107262999984, "tokens / sec": 7588.678601505416, "mem allocated avg": 7059892746.24, "mem reserved avg": 14239712411.648, "elapsed time": 425.4388579290098 }, { "step": 2500, "valid accuracy": 0.3, "train loss": 0.70130124771595, "train samples": 10000, "train time": 27.76416820794111, "eval time": 15.847382225998444, "tokens / sec": 7418.4466272283025, "mem allocated avg": 7045819248.64, "mem reserved avg": 13771493867.52, "elapsed time": 471.59908634400927 }, { "step": 2750, "valid accuracy": 0.38, "train loss": 0.6942626680135727, "train samples": 11000, "train time": 27.77380103613541, "eval time": 11.25709281899617, "tokens / sec": 7628.808160767404, "mem allocated avg": 7056581769.216, "mem reserved avg": 14109697376.256, "elapsed time": 513.1207982520136 }, { "step": 3000, "valid accuracy": 0.34, "train loss": 0.6865150318145752, "train samples": 12000, "train time": 27.40294734176132, "eval time": 15.83896046000882, "tokens / sec": 7617.1003577377915, "mem allocated avg": 7051527202.816, "mem reserved avg": 14113497415.68, "elapsed time": 559.0147692530008 }, { "step": 3250, "valid accuracy": 0.32, "train loss": 0.6951180123090744, "train samples": 13000, "train time": 27.608605705041555, "eval time": 15.836226497005555, "tokens / sec": 7638.958745442468, "mem allocated avg": 7052913205.248, "mem reserved avg": 13957486084.096, "elapsed time": 604.9435471830075 }, { "step": 3500, "valid accuracy": 0.46, "train loss": 0.6807289460897445, "train samples": 14000, "train time": 28.273788013044395, "eval time": 15.85275076299149, "tokens / sec": 7418.5319598219285, "mem allocated avg": 7052796133.376, "mem reserved avg": 13872937304.064, "elapsed time": 651.6589886930014 }, { "step": 3750, "valid accuracy": 0.38, "train loss": 0.6770447121858597, "train samples": 15000, "train time": 28.624993037970853, "eval time": 10.463159157996415, "tokens / sec": 7570.412321587118, "mem allocated avg": 7062565990.4, "mem reserved avg": 14240719044.608, "elapsed time": 693.3172026100074 }, { "step": 4000, "valid accuracy": 0.44, "train loss": 0.6930302439928054, "train samples": 16000, "train time": 27.629406421037856, "eval time": 9.406142670995905, "tokens / sec": 7396.937772951368, "mem allocated avg": 7044787052.544, "mem reserved avg": 13913278119.936, "elapsed time": 732.9555247080134 }, { "step": 4250, "valid accuracy": 0.46, "train loss": 0.672576385140419, "train samples": 17000, "train time": 28.082130014998256, "eval time": 10.856074246999924, "tokens / sec": 7527.527288247017, "mem allocated avg": 7054679074.816, "mem reserved avg": 14022187417.6, "elapsed time": 774.335282400003 }, { "step": 4500, "valid accuracy": 0.36, "train loss": 0.6801181032657623, "train samples": 18000, "train time": 27.873706766869873, "eval time": 15.860146595994593, "tokens / sec": 7455.700159944578, "mem allocated avg": 7049662265.344, "mem reserved avg": 13916012806.144, "elapsed time": 820.668733826009 }, { "step": 4750, "valid accuracy": 0.42, "train loss": 0.6726169712543487, "train samples": 19000, "train time": 27.956917096104007, "eval time": 15.841582774999551, "tokens / sec": 7509.375918607867, "mem allocated avg": 7052527755.264, "mem reserved avg": 14004571340.8, "elapsed time": 866.9247034900036 }, { "step": 5000, "valid accuracy": 0.4, "train loss": 0.6787571088075638, "train samples": 20000, "train time": 27.523053067998262, "eval time": 10.57767208799487, "tokens / sec": 7567.474418096891, "mem allocated avg": 7049240582.144, "mem reserved avg": 13645857685.504, "elapsed time": 907.492314247007 }, { "step": 5000, "test accuracy": 0.36694465504169826, "train loss": 0.6787571088075638, "train samples": 20000, "train total tokens": 4198051, "forgetting": 2.828566551208496 } ] }, "meta_info": { "model_info": { "sha": "13afe5124825b4f3751f836b40dafda64c1ed062", "created_at": "2024-09-18T15:23:48+00:00" }, "dataset_info": { "metamath": { "sha": "aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18", "created_at": "2023-09-21T17:22:46+00:00" }, "gsm8k": { "sha": "740312add88f781978c0658806c59bc2815b9866", "created_at": "2022-04-12T10:22:10+00:00" } }, "package_info": { "transformers-version": "5.5.4", "transformers-commit-hash": null, "peft-version": "0.19.0", "peft-commit-hash": "6d5a6f4f2f902dbf13d21d2661d57c3c05df1dae", "datasets-version": "4.2.0", "datasets-commit-hash": null, "bitsandbytes-version": "0.46.0", "bitsandbytes-commit-hash": null, "torch-version": "2.9.0+cu128", "torch-commit-hash": null }, "system_info": { "system": "Linux", "release": "6.17.0-1009-aws", "version": "#9~24.04.2-Ubuntu SMP Fri Mar 6 23:50:29 UTC 2026", "machine": "x86_64", "processor": "x86_64", "accelerator": "NVIDIA L40S" }, "pytorch_info": "PyTorch built with:\n - GCC 13.3\n - C++ Version: 201703\n - Intel(R) oneAPI Math Kernel Library Version 2024.2-Product Build 20240605 for Intel(R) 64 architecture applications\n - Intel(R) MKL-DNN v3.7.1 (Git Hash 8d263e693366ef8db40acc569cc7d8edf644556d)\n - OpenMP 201511 (a.k.a. OpenMP 4.5)\n - LAPACK is enabled (usually provided by MKL)\n - NNPACK is enabled\n - CPU capability usage: AVX2\n - CUDA Runtime 12.8\n - NVCC architecture flags: -gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_90,code=sm_90;-gencode;arch=compute_100,code=sm_100;-gencode;arch=compute_120,code=sm_120\n - CuDNN 90.7.1\n - Built with CuDNN 90.8\n - Magma 2.6.1\n - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, COMMIT_SHA=0fabc3ba44823f257e70ce397d989c8de5e362c1, CUDA_VERSION=12.8, CUDNN_VERSION=9.8.0, CXX_COMPILER=/opt/rh/gcc-toolset-13/root/usr/bin/c++, CXX_FLAGS= -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DLIBKINETO_NOXPUPTI=ON -DUSE_FBGEMM -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -DC10_NODEPRECATED -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=range-loop-construct -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-unknown-pragmas -Wno-unused-parameter -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wsuggest-override -Wno-psabi -Wno-error=old-style-cast -faligned-new -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-dangling-reference -Wno-error=dangling-reference -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, TORCH_VERSION=2.9.0, USE_CUDA=ON, USE_CUDNN=ON, USE_CUSPARSELT=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=ON, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=1, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF, USE_XCCL=OFF, USE_XPU=OFF, \n" } }