Text Generation
Transformers
Safetensors
qwen3
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Instructions to use DCAgent2/stack-nl2bashseq with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use DCAgent2/stack-nl2bashseq with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="DCAgent2/stack-nl2bashseq") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("DCAgent2/stack-nl2bashseq") model = AutoModelForCausalLM.from_pretrained("DCAgent2/stack-nl2bashseq") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use DCAgent2/stack-nl2bashseq with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "DCAgent2/stack-nl2bashseq" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "DCAgent2/stack-nl2bashseq", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/DCAgent2/stack-nl2bashseq
- SGLang
How to use DCAgent2/stack-nl2bashseq with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "DCAgent2/stack-nl2bashseq" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "DCAgent2/stack-nl2bashseq", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "DCAgent2/stack-nl2bashseq" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "DCAgent2/stack-nl2bashseq", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use DCAgent2/stack-nl2bashseq with Docker Model Runner:
docker model run hf.co/DCAgent2/stack-nl2bashseq
| { | |
| "best_global_step": null, | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 7.0, | |
| "eval_steps": 500, | |
| "global_step": 6664, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.005252100840336135, | |
| "grad_norm": 11.793528240093083, | |
| "learning_rate": 2.39880059970015e-07, | |
| "loss": 0.7356, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.750144362449646, | |
| "step": 5, | |
| "valid_targets_mean": 3398.8, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 0.01050420168067227, | |
| "grad_norm": 9.544117410382237, | |
| "learning_rate": 5.397301349325338e-07, | |
| "loss": 0.6877, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5979838371276855, | |
| "step": 10, | |
| "valid_targets_mean": 4347.4, | |
| "valid_targets_min": 1651 | |
| }, | |
| { | |
| "epoch": 0.015756302521008403, | |
| "grad_norm": 8.563267976325044, | |
| "learning_rate": 8.395802098950526e-07, | |
| "loss": 0.6892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.607923150062561, | |
| "step": 15, | |
| "valid_targets_mean": 4633.2, | |
| "valid_targets_min": 1560 | |
| }, | |
| { | |
| "epoch": 0.02100840336134454, | |
| "grad_norm": 8.188017827775615, | |
| "learning_rate": 1.1394302848575713e-06, | |
| "loss": 0.6695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6504963040351868, | |
| "step": 20, | |
| "valid_targets_mean": 3938.1, | |
| "valid_targets_min": 1841 | |
| }, | |
| { | |
| "epoch": 0.026260504201680673, | |
| "grad_norm": 6.523486790143655, | |
| "learning_rate": 1.43928035982009e-06, | |
| "loss": 0.6875, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7029379606246948, | |
| "step": 25, | |
| "valid_targets_mean": 3724.8, | |
| "valid_targets_min": 1374 | |
| }, | |
| { | |
| "epoch": 0.031512605042016806, | |
| "grad_norm": 5.51646121673247, | |
| "learning_rate": 1.7391304347826088e-06, | |
| "loss": 0.6512, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6437787413597107, | |
| "step": 30, | |
| "valid_targets_mean": 3264.8, | |
| "valid_targets_min": 1089 | |
| }, | |
| { | |
| "epoch": 0.03676470588235294, | |
| "grad_norm": 4.319516286938232, | |
| "learning_rate": 2.0389805097451275e-06, | |
| "loss": 0.5769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6190094947814941, | |
| "step": 35, | |
| "valid_targets_mean": 3374.9, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 0.04201680672268908, | |
| "grad_norm": 3.7446317656639048, | |
| "learning_rate": 2.3388305847076464e-06, | |
| "loss": 0.5953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6246620416641235, | |
| "step": 40, | |
| "valid_targets_mean": 3514.9, | |
| "valid_targets_min": 1574 | |
| }, | |
| { | |
| "epoch": 0.04726890756302521, | |
| "grad_norm": 2.652464883959148, | |
| "learning_rate": 2.6386806596701653e-06, | |
| "loss": 0.5982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6014484167098999, | |
| "step": 45, | |
| "valid_targets_mean": 3956.7, | |
| "valid_targets_min": 1899 | |
| }, | |
| { | |
| "epoch": 0.052521008403361345, | |
| "grad_norm": 1.9872458383108076, | |
| "learning_rate": 2.9385307346326843e-06, | |
| "loss": 0.5479, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5268915891647339, | |
| "step": 50, | |
| "valid_targets_mean": 4397.6, | |
| "valid_targets_min": 2261 | |
| }, | |
| { | |
| "epoch": 0.05777310924369748, | |
| "grad_norm": 1.6856790315793377, | |
| "learning_rate": 3.2383808095952024e-06, | |
| "loss": 0.5517, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.535837709903717, | |
| "step": 55, | |
| "valid_targets_mean": 3263.8, | |
| "valid_targets_min": 1873 | |
| }, | |
| { | |
| "epoch": 0.06302521008403361, | |
| "grad_norm": 1.2418730665732476, | |
| "learning_rate": 3.5382308845577213e-06, | |
| "loss": 0.5094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5285685658454895, | |
| "step": 60, | |
| "valid_targets_mean": 3430.2, | |
| "valid_targets_min": 1013 | |
| }, | |
| { | |
| "epoch": 0.06827731092436974, | |
| "grad_norm": 0.8868118364710539, | |
| "learning_rate": 3.83808095952024e-06, | |
| "loss": 0.4861, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4722723364830017, | |
| "step": 65, | |
| "valid_targets_mean": 4041.6, | |
| "valid_targets_min": 1366 | |
| }, | |
| { | |
| "epoch": 0.07352941176470588, | |
| "grad_norm": 0.9428896232462224, | |
| "learning_rate": 4.137931034482759e-06, | |
| "loss": 0.5037, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.53287672996521, | |
| "step": 70, | |
| "valid_targets_mean": 3515.2, | |
| "valid_targets_min": 2050 | |
| }, | |
| { | |
| "epoch": 0.07878151260504201, | |
| "grad_norm": 0.9383925766489141, | |
| "learning_rate": 4.437781109445278e-06, | |
| "loss": 0.5135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.524086058139801, | |
| "step": 75, | |
| "valid_targets_mean": 3048.9, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 0.08403361344537816, | |
| "grad_norm": 0.9340541727723428, | |
| "learning_rate": 4.737631184407796e-06, | |
| "loss": 0.489, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5150220990180969, | |
| "step": 80, | |
| "valid_targets_mean": 2982.6, | |
| "valid_targets_min": 855 | |
| }, | |
| { | |
| "epoch": 0.08928571428571429, | |
| "grad_norm": 0.7737514488505065, | |
| "learning_rate": 5.037481259370315e-06, | |
| "loss": 0.4923, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.498151570558548, | |
| "step": 85, | |
| "valid_targets_mean": 3753.4, | |
| "valid_targets_min": 2125 | |
| }, | |
| { | |
| "epoch": 0.09453781512605042, | |
| "grad_norm": 0.8574731436330563, | |
| "learning_rate": 5.337331334332834e-06, | |
| "loss": 0.5013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4901542663574219, | |
| "step": 90, | |
| "valid_targets_mean": 2946.8, | |
| "valid_targets_min": 1173 | |
| }, | |
| { | |
| "epoch": 0.09978991596638656, | |
| "grad_norm": 0.8207916191722505, | |
| "learning_rate": 5.6371814092953526e-06, | |
| "loss": 0.4644, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.45455610752105713, | |
| "step": 95, | |
| "valid_targets_mean": 3284.2, | |
| "valid_targets_min": 1438 | |
| }, | |
| { | |
| "epoch": 0.10504201680672269, | |
| "grad_norm": 0.6890899996201467, | |
| "learning_rate": 5.937031484257871e-06, | |
| "loss": 0.4503, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3856987953186035, | |
| "step": 100, | |
| "valid_targets_mean": 3818.4, | |
| "valid_targets_min": 1133 | |
| }, | |
| { | |
| "epoch": 0.11029411764705882, | |
| "grad_norm": 0.7375624963619357, | |
| "learning_rate": 6.2368815592203904e-06, | |
| "loss": 0.4597, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.42143797874450684, | |
| "step": 105, | |
| "valid_targets_mean": 3339.8, | |
| "valid_targets_min": 1051 | |
| }, | |
| { | |
| "epoch": 0.11554621848739496, | |
| "grad_norm": 0.7422023625324775, | |
| "learning_rate": 6.536731634182909e-06, | |
| "loss": 0.449, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.47655802965164185, | |
| "step": 110, | |
| "valid_targets_mean": 3285.1, | |
| "valid_targets_min": 1642 | |
| }, | |
| { | |
| "epoch": 0.1207983193277311, | |
| "grad_norm": 0.6767508461063624, | |
| "learning_rate": 6.8365817091454274e-06, | |
| "loss": 0.4463, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.42775917053222656, | |
| "step": 115, | |
| "valid_targets_mean": 3340.8, | |
| "valid_targets_min": 1747 | |
| }, | |
| { | |
| "epoch": 0.12605042016806722, | |
| "grad_norm": 0.7262880447276591, | |
| "learning_rate": 7.136431784107947e-06, | |
| "loss": 0.4491, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4421280324459076, | |
| "step": 120, | |
| "valid_targets_mean": 3218.2, | |
| "valid_targets_min": 1765 | |
| }, | |
| { | |
| "epoch": 0.13130252100840337, | |
| "grad_norm": 0.7791375198049785, | |
| "learning_rate": 7.436281859070465e-06, | |
| "loss": 0.4446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4363992214202881, | |
| "step": 125, | |
| "valid_targets_mean": 3182.9, | |
| "valid_targets_min": 1233 | |
| }, | |
| { | |
| "epoch": 0.13655462184873948, | |
| "grad_norm": 0.7376851052403786, | |
| "learning_rate": 7.736131934032984e-06, | |
| "loss": 0.4195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4291246831417084, | |
| "step": 130, | |
| "valid_targets_mean": 3020.8, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 0.14180672268907563, | |
| "grad_norm": 0.7251699149877672, | |
| "learning_rate": 8.035982008995503e-06, | |
| "loss": 0.4383, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4199424982070923, | |
| "step": 135, | |
| "valid_targets_mean": 3269.2, | |
| "valid_targets_min": 926 | |
| }, | |
| { | |
| "epoch": 0.14705882352941177, | |
| "grad_norm": 0.7174356234657762, | |
| "learning_rate": 8.335832083958023e-06, | |
| "loss": 0.4209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3915764093399048, | |
| "step": 140, | |
| "valid_targets_mean": 3256.4, | |
| "valid_targets_min": 1016 | |
| }, | |
| { | |
| "epoch": 0.15231092436974789, | |
| "grad_norm": 0.7429933172485804, | |
| "learning_rate": 8.63568215892054e-06, | |
| "loss": 0.4246, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4099087417125702, | |
| "step": 145, | |
| "valid_targets_mean": 3058.4, | |
| "valid_targets_min": 1367 | |
| }, | |
| { | |
| "epoch": 0.15756302521008403, | |
| "grad_norm": 0.6609884978656597, | |
| "learning_rate": 8.93553223388306e-06, | |
| "loss": 0.4275, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38431763648986816, | |
| "step": 150, | |
| "valid_targets_mean": 3513.8, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 0.16281512605042017, | |
| "grad_norm": 0.7414117819519422, | |
| "learning_rate": 9.235382308845579e-06, | |
| "loss": 0.4149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.41205519437789917, | |
| "step": 155, | |
| "valid_targets_mean": 2834.2, | |
| "valid_targets_min": 1480 | |
| }, | |
| { | |
| "epoch": 0.16806722689075632, | |
| "grad_norm": 0.766190877937914, | |
| "learning_rate": 9.535232383808097e-06, | |
| "loss": 0.4221, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.40908199548721313, | |
| "step": 160, | |
| "valid_targets_mean": 2854.1, | |
| "valid_targets_min": 1229 | |
| }, | |
| { | |
| "epoch": 0.17331932773109243, | |
| "grad_norm": 0.7322988121738653, | |
| "learning_rate": 9.835082458770614e-06, | |
| "loss": 0.4219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.409344881772995, | |
| "step": 165, | |
| "valid_targets_mean": 3215.8, | |
| "valid_targets_min": 1727 | |
| }, | |
| { | |
| "epoch": 0.17857142857142858, | |
| "grad_norm": 0.7036664531547118, | |
| "learning_rate": 1.0134932533733135e-05, | |
| "loss": 0.4117, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3934198021888733, | |
| "step": 170, | |
| "valid_targets_mean": 3578.9, | |
| "valid_targets_min": 1860 | |
| }, | |
| { | |
| "epoch": 0.18382352941176472, | |
| "grad_norm": 0.7504645575443613, | |
| "learning_rate": 1.0434782608695653e-05, | |
| "loss": 0.4176, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.41233259439468384, | |
| "step": 175, | |
| "valid_targets_mean": 3121.1, | |
| "valid_targets_min": 1145 | |
| }, | |
| { | |
| "epoch": 0.18907563025210083, | |
| "grad_norm": 0.8540338092213114, | |
| "learning_rate": 1.0734632683658172e-05, | |
| "loss": 0.4206, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4257834851741791, | |
| "step": 180, | |
| "valid_targets_mean": 2910.1, | |
| "valid_targets_min": 1195 | |
| }, | |
| { | |
| "epoch": 0.19432773109243698, | |
| "grad_norm": 0.672825231347177, | |
| "learning_rate": 1.103448275862069e-05, | |
| "loss": 0.4046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.37902510166168213, | |
| "step": 185, | |
| "valid_targets_mean": 3215.2, | |
| "valid_targets_min": 1087 | |
| }, | |
| { | |
| "epoch": 0.19957983193277312, | |
| "grad_norm": 0.7299680458741075, | |
| "learning_rate": 1.1334332833583211e-05, | |
| "loss": 0.4173, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3878703713417053, | |
| "step": 190, | |
| "valid_targets_mean": 3842.3, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 0.20483193277310924, | |
| "grad_norm": 0.7452069066845327, | |
| "learning_rate": 1.1634182908545729e-05, | |
| "loss": 0.4012, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.40164220333099365, | |
| "step": 195, | |
| "valid_targets_mean": 3112.6, | |
| "valid_targets_min": 1097 | |
| }, | |
| { | |
| "epoch": 0.21008403361344538, | |
| "grad_norm": 0.6701617672469308, | |
| "learning_rate": 1.1934032983508246e-05, | |
| "loss": 0.3889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4067269563674927, | |
| "step": 200, | |
| "valid_targets_mean": 3587.6, | |
| "valid_targets_min": 1833 | |
| }, | |
| { | |
| "epoch": 0.21533613445378152, | |
| "grad_norm": 0.6999892185242177, | |
| "learning_rate": 1.2233883058470766e-05, | |
| "loss": 0.4046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.42524561285972595, | |
| "step": 205, | |
| "valid_targets_mean": 3660.6, | |
| "valid_targets_min": 1182 | |
| }, | |
| { | |
| "epoch": 0.22058823529411764, | |
| "grad_norm": 0.6936596594842792, | |
| "learning_rate": 1.2533733133433283e-05, | |
| "loss": 0.3979, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4341742992401123, | |
| "step": 210, | |
| "valid_targets_mean": 3839.4, | |
| "valid_targets_min": 1586 | |
| }, | |
| { | |
| "epoch": 0.22584033613445378, | |
| "grad_norm": 0.7203275927233951, | |
| "learning_rate": 1.2833583208395803e-05, | |
| "loss": 0.4122, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4256739020347595, | |
| "step": 215, | |
| "valid_targets_mean": 3062.7, | |
| "valid_targets_min": 1714 | |
| }, | |
| { | |
| "epoch": 0.23109243697478993, | |
| "grad_norm": 0.8564716523005904, | |
| "learning_rate": 1.313343328335832e-05, | |
| "loss": 0.4007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4049715995788574, | |
| "step": 220, | |
| "valid_targets_mean": 3611.1, | |
| "valid_targets_min": 1504 | |
| }, | |
| { | |
| "epoch": 0.23634453781512604, | |
| "grad_norm": 0.6563920907148093, | |
| "learning_rate": 1.3433283358320841e-05, | |
| "loss": 0.3885, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3269047141075134, | |
| "step": 225, | |
| "valid_targets_mean": 3457.1, | |
| "valid_targets_min": 1170 | |
| }, | |
| { | |
| "epoch": 0.2415966386554622, | |
| "grad_norm": 0.7196899057898143, | |
| "learning_rate": 1.3733133433283359e-05, | |
| "loss": 0.3893, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38853204250335693, | |
| "step": 230, | |
| "valid_targets_mean": 3089.3, | |
| "valid_targets_min": 1436 | |
| }, | |
| { | |
| "epoch": 0.24684873949579833, | |
| "grad_norm": 0.7287732338618192, | |
| "learning_rate": 1.4032983508245878e-05, | |
| "loss": 0.3919, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38699156045913696, | |
| "step": 235, | |
| "valid_targets_mean": 3257.2, | |
| "valid_targets_min": 1302 | |
| }, | |
| { | |
| "epoch": 0.25210084033613445, | |
| "grad_norm": 0.7214311214899903, | |
| "learning_rate": 1.4332833583208396e-05, | |
| "loss": 0.3841, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.41121745109558105, | |
| "step": 240, | |
| "valid_targets_mean": 3425.1, | |
| "valid_targets_min": 2023 | |
| }, | |
| { | |
| "epoch": 0.25735294117647056, | |
| "grad_norm": 0.648625366284687, | |
| "learning_rate": 1.4632683658170917e-05, | |
| "loss": 0.361, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3404223918914795, | |
| "step": 245, | |
| "valid_targets_mean": 3196.4, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 0.26260504201680673, | |
| "grad_norm": 0.7442737796844059, | |
| "learning_rate": 1.4932533733133435e-05, | |
| "loss": 0.3958, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.37922531366348267, | |
| "step": 250, | |
| "valid_targets_mean": 3197.8, | |
| "valid_targets_min": 1059 | |
| }, | |
| { | |
| "epoch": 0.26785714285714285, | |
| "grad_norm": 0.7290285381798371, | |
| "learning_rate": 1.5232383808095954e-05, | |
| "loss": 0.3759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33683687448501587, | |
| "step": 255, | |
| "valid_targets_mean": 3301.7, | |
| "valid_targets_min": 964 | |
| }, | |
| { | |
| "epoch": 0.27310924369747897, | |
| "grad_norm": 0.7658760733120059, | |
| "learning_rate": 1.5532233883058472e-05, | |
| "loss": 0.3864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3716367483139038, | |
| "step": 260, | |
| "valid_targets_mean": 3694.2, | |
| "valid_targets_min": 1105 | |
| }, | |
| { | |
| "epoch": 0.27836134453781514, | |
| "grad_norm": 0.8430133851017164, | |
| "learning_rate": 1.5832083958020993e-05, | |
| "loss": 0.3997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4212614893913269, | |
| "step": 265, | |
| "valid_targets_mean": 2612.6, | |
| "valid_targets_min": 1491 | |
| }, | |
| { | |
| "epoch": 0.28361344537815125, | |
| "grad_norm": 0.7343320220234513, | |
| "learning_rate": 1.613193403298351e-05, | |
| "loss": 0.4027, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3935287594795227, | |
| "step": 270, | |
| "valid_targets_mean": 3110.5, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 0.28886554621848737, | |
| "grad_norm": 0.7657356714112876, | |
| "learning_rate": 1.6431784107946028e-05, | |
| "loss": 0.3703, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36865124106407166, | |
| "step": 275, | |
| "valid_targets_mean": 3259.2, | |
| "valid_targets_min": 1822 | |
| }, | |
| { | |
| "epoch": 0.29411764705882354, | |
| "grad_norm": 0.6955693394366955, | |
| "learning_rate": 1.6731634182908546e-05, | |
| "loss": 0.3727, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36471718549728394, | |
| "step": 280, | |
| "valid_targets_mean": 3547.6, | |
| "valid_targets_min": 1625 | |
| }, | |
| { | |
| "epoch": 0.29936974789915966, | |
| "grad_norm": 0.7042195932323547, | |
| "learning_rate": 1.7031484257871064e-05, | |
| "loss": 0.3672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3420675992965698, | |
| "step": 285, | |
| "valid_targets_mean": 3628.2, | |
| "valid_targets_min": 1415 | |
| }, | |
| { | |
| "epoch": 0.30462184873949577, | |
| "grad_norm": 0.8809660140825649, | |
| "learning_rate": 1.7331334332833585e-05, | |
| "loss": 0.3791, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35459285974502563, | |
| "step": 290, | |
| "valid_targets_mean": 3398.9, | |
| "valid_targets_min": 1460 | |
| }, | |
| { | |
| "epoch": 0.30987394957983194, | |
| "grad_norm": 0.7752770337054249, | |
| "learning_rate": 1.7631184407796102e-05, | |
| "loss": 0.3838, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3816995620727539, | |
| "step": 295, | |
| "valid_targets_mean": 2947.4, | |
| "valid_targets_min": 1079 | |
| }, | |
| { | |
| "epoch": 0.31512605042016806, | |
| "grad_norm": 0.7665849286642425, | |
| "learning_rate": 1.7931034482758623e-05, | |
| "loss": 0.3576, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36882513761520386, | |
| "step": 300, | |
| "valid_targets_mean": 3003.0, | |
| "valid_targets_min": 1274 | |
| }, | |
| { | |
| "epoch": 0.32037815126050423, | |
| "grad_norm": 0.7763875661157622, | |
| "learning_rate": 1.823088455772114e-05, | |
| "loss": 0.3735, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3900800943374634, | |
| "step": 305, | |
| "valid_targets_mean": 3306.3, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 0.32563025210084034, | |
| "grad_norm": 0.7724334584719763, | |
| "learning_rate": 1.853073463268366e-05, | |
| "loss": 0.3894, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36286020278930664, | |
| "step": 310, | |
| "valid_targets_mean": 3332.2, | |
| "valid_targets_min": 1312 | |
| }, | |
| { | |
| "epoch": 0.33088235294117646, | |
| "grad_norm": 0.8144297942259716, | |
| "learning_rate": 1.8830584707646176e-05, | |
| "loss": 0.377, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.395212858915329, | |
| "step": 315, | |
| "valid_targets_mean": 2601.3, | |
| "valid_targets_min": 987 | |
| }, | |
| { | |
| "epoch": 0.33613445378151263, | |
| "grad_norm": 0.6210630540146885, | |
| "learning_rate": 1.9130434782608697e-05, | |
| "loss": 0.3653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34900736808776855, | |
| "step": 320, | |
| "valid_targets_mean": 4341.5, | |
| "valid_targets_min": 1856 | |
| }, | |
| { | |
| "epoch": 0.34138655462184875, | |
| "grad_norm": 0.7382149230509586, | |
| "learning_rate": 1.9430284857571215e-05, | |
| "loss": 0.3724, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.37840932607650757, | |
| "step": 325, | |
| "valid_targets_mean": 3290.9, | |
| "valid_targets_min": 1601 | |
| }, | |
| { | |
| "epoch": 0.34663865546218486, | |
| "grad_norm": 0.7441687764233078, | |
| "learning_rate": 1.9730134932533736e-05, | |
| "loss": 0.3807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3847064971923828, | |
| "step": 330, | |
| "valid_targets_mean": 3119.4, | |
| "valid_targets_min": 1986 | |
| }, | |
| { | |
| "epoch": 0.35189075630252103, | |
| "grad_norm": 0.6427174614521525, | |
| "learning_rate": 2.0029985007496254e-05, | |
| "loss": 0.3556, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38938313722610474, | |
| "step": 335, | |
| "valid_targets_mean": 3777.1, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 0.35714285714285715, | |
| "grad_norm": 0.6767174257258602, | |
| "learning_rate": 2.0329835082458775e-05, | |
| "loss": 0.3632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33448052406311035, | |
| "step": 340, | |
| "valid_targets_mean": 2878.0, | |
| "valid_targets_min": 512 | |
| }, | |
| { | |
| "epoch": 0.36239495798319327, | |
| "grad_norm": 0.7046783787151958, | |
| "learning_rate": 2.0629685157421292e-05, | |
| "loss": 0.3697, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35749202966690063, | |
| "step": 345, | |
| "valid_targets_mean": 3215.4, | |
| "valid_targets_min": 1202 | |
| }, | |
| { | |
| "epoch": 0.36764705882352944, | |
| "grad_norm": 0.7219690552414987, | |
| "learning_rate": 2.092953523238381e-05, | |
| "loss": 0.3881, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.37818586826324463, | |
| "step": 350, | |
| "valid_targets_mean": 3693.1, | |
| "valid_targets_min": 1455 | |
| }, | |
| { | |
| "epoch": 0.37289915966386555, | |
| "grad_norm": 0.6860266938886983, | |
| "learning_rate": 2.1229385307346328e-05, | |
| "loss": 0.3675, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34528088569641113, | |
| "step": 355, | |
| "valid_targets_mean": 3477.5, | |
| "valid_targets_min": 1571 | |
| }, | |
| { | |
| "epoch": 0.37815126050420167, | |
| "grad_norm": 0.7609425900554888, | |
| "learning_rate": 2.152923538230885e-05, | |
| "loss": 0.3949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.429862916469574, | |
| "step": 360, | |
| "valid_targets_mean": 2912.1, | |
| "valid_targets_min": 1450 | |
| }, | |
| { | |
| "epoch": 0.38340336134453784, | |
| "grad_norm": 0.6848089283903739, | |
| "learning_rate": 2.1829085457271363e-05, | |
| "loss": 0.3669, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3468002676963806, | |
| "step": 365, | |
| "valid_targets_mean": 3323.0, | |
| "valid_targets_min": 1431 | |
| }, | |
| { | |
| "epoch": 0.38865546218487396, | |
| "grad_norm": 0.7419166645990429, | |
| "learning_rate": 2.2128935532233884e-05, | |
| "loss": 0.3744, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3837374746799469, | |
| "step": 370, | |
| "valid_targets_mean": 3461.4, | |
| "valid_targets_min": 1147 | |
| }, | |
| { | |
| "epoch": 0.3939075630252101, | |
| "grad_norm": 0.8576299436135565, | |
| "learning_rate": 2.2428785607196405e-05, | |
| "loss": 0.393, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.39932119846343994, | |
| "step": 375, | |
| "valid_targets_mean": 2641.6, | |
| "valid_targets_min": 998 | |
| }, | |
| { | |
| "epoch": 0.39915966386554624, | |
| "grad_norm": 0.7247001275212052, | |
| "learning_rate": 2.2728635682158923e-05, | |
| "loss": 0.3684, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3718080520629883, | |
| "step": 380, | |
| "valid_targets_mean": 3260.9, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 0.40441176470588236, | |
| "grad_norm": 0.7885722026955804, | |
| "learning_rate": 2.302848575712144e-05, | |
| "loss": 0.3797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.39757227897644043, | |
| "step": 385, | |
| "valid_targets_mean": 2796.2, | |
| "valid_targets_min": 1005 | |
| }, | |
| { | |
| "epoch": 0.4096638655462185, | |
| "grad_norm": 0.7828630662690874, | |
| "learning_rate": 2.3328335832083958e-05, | |
| "loss": 0.3793, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3814929127693176, | |
| "step": 390, | |
| "valid_targets_mean": 3183.6, | |
| "valid_targets_min": 2298 | |
| }, | |
| { | |
| "epoch": 0.41491596638655465, | |
| "grad_norm": 0.8409621672836618, | |
| "learning_rate": 2.362818590704648e-05, | |
| "loss": 0.3763, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3965279459953308, | |
| "step": 395, | |
| "valid_targets_mean": 2821.2, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 0.42016806722689076, | |
| "grad_norm": 0.6764711962883958, | |
| "learning_rate": 2.3928035982009e-05, | |
| "loss": 0.355, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3417738080024719, | |
| "step": 400, | |
| "valid_targets_mean": 3265.2, | |
| "valid_targets_min": 1665 | |
| }, | |
| { | |
| "epoch": 0.4254201680672269, | |
| "grad_norm": 0.7365452834319242, | |
| "learning_rate": 2.4227886056971515e-05, | |
| "loss": 0.3735, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3877049684524536, | |
| "step": 405, | |
| "valid_targets_mean": 3019.3, | |
| "valid_targets_min": 1076 | |
| }, | |
| { | |
| "epoch": 0.43067226890756305, | |
| "grad_norm": 0.6737132753666218, | |
| "learning_rate": 2.4527736131934036e-05, | |
| "loss": 0.3566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36130669713020325, | |
| "step": 410, | |
| "valid_targets_mean": 3610.8, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 0.43592436974789917, | |
| "grad_norm": 0.9130730152687944, | |
| "learning_rate": 2.4827586206896553e-05, | |
| "loss": 0.3783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3609384000301361, | |
| "step": 415, | |
| "valid_targets_mean": 3255.8, | |
| "valid_targets_min": 1728 | |
| }, | |
| { | |
| "epoch": 0.4411764705882353, | |
| "grad_norm": 0.8464202746590961, | |
| "learning_rate": 2.5127436281859074e-05, | |
| "loss": 0.3788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33865395188331604, | |
| "step": 420, | |
| "valid_targets_mean": 3377.2, | |
| "valid_targets_min": 981 | |
| }, | |
| { | |
| "epoch": 0.44642857142857145, | |
| "grad_norm": 0.7384506566000549, | |
| "learning_rate": 2.542728635682159e-05, | |
| "loss": 0.3462, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3339518904685974, | |
| "step": 425, | |
| "valid_targets_mean": 2839.1, | |
| "valid_targets_min": 1351 | |
| }, | |
| { | |
| "epoch": 0.45168067226890757, | |
| "grad_norm": 0.6110714780946447, | |
| "learning_rate": 2.572713643178411e-05, | |
| "loss": 0.3677, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3625059425830841, | |
| "step": 430, | |
| "valid_targets_mean": 4396.7, | |
| "valid_targets_min": 2416 | |
| }, | |
| { | |
| "epoch": 0.4569327731092437, | |
| "grad_norm": 0.5790619763281034, | |
| "learning_rate": 2.602698650674663e-05, | |
| "loss": 0.3524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32317501306533813, | |
| "step": 435, | |
| "valid_targets_mean": 4668.6, | |
| "valid_targets_min": 1233 | |
| }, | |
| { | |
| "epoch": 0.46218487394957986, | |
| "grad_norm": 0.642678687454446, | |
| "learning_rate": 2.632683658170915e-05, | |
| "loss": 0.3433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3185131549835205, | |
| "step": 440, | |
| "valid_targets_mean": 3936.4, | |
| "valid_targets_min": 1524 | |
| }, | |
| { | |
| "epoch": 0.46743697478991597, | |
| "grad_norm": 0.5867900140587048, | |
| "learning_rate": 2.6626686656671666e-05, | |
| "loss": 0.3446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3317568600177765, | |
| "step": 445, | |
| "valid_targets_mean": 4389.3, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 0.4726890756302521, | |
| "grad_norm": 0.6996864572340856, | |
| "learning_rate": 2.6926536731634184e-05, | |
| "loss": 0.3573, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35135549306869507, | |
| "step": 450, | |
| "valid_targets_mean": 3241.3, | |
| "valid_targets_min": 906 | |
| }, | |
| { | |
| "epoch": 0.47794117647058826, | |
| "grad_norm": 0.6713569052572641, | |
| "learning_rate": 2.7226386806596705e-05, | |
| "loss": 0.3452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3594322204589844, | |
| "step": 455, | |
| "valid_targets_mean": 3680.4, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 0.4831932773109244, | |
| "grad_norm": 0.6528374934626601, | |
| "learning_rate": 2.752623688155922e-05, | |
| "loss": 0.3433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33547288179397583, | |
| "step": 460, | |
| "valid_targets_mean": 3168.9, | |
| "valid_targets_min": 1484 | |
| }, | |
| { | |
| "epoch": 0.4884453781512605, | |
| "grad_norm": 0.7585928062211041, | |
| "learning_rate": 2.782608695652174e-05, | |
| "loss": 0.3517, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34602272510528564, | |
| "step": 465, | |
| "valid_targets_mean": 3667.5, | |
| "valid_targets_min": 1440 | |
| }, | |
| { | |
| "epoch": 0.49369747899159666, | |
| "grad_norm": 0.7549275178598076, | |
| "learning_rate": 2.812593703148426e-05, | |
| "loss": 0.3689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.40832167863845825, | |
| "step": 470, | |
| "valid_targets_mean": 3025.6, | |
| "valid_targets_min": 1349 | |
| }, | |
| { | |
| "epoch": 0.4989495798319328, | |
| "grad_norm": 0.8305062206700754, | |
| "learning_rate": 2.8425787106446782e-05, | |
| "loss": 0.3684, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36201930046081543, | |
| "step": 475, | |
| "valid_targets_mean": 2935.8, | |
| "valid_targets_min": 1156 | |
| }, | |
| { | |
| "epoch": 0.5042016806722689, | |
| "grad_norm": 0.7148304561333289, | |
| "learning_rate": 2.8725637181409296e-05, | |
| "loss": 0.358, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3326527178287506, | |
| "step": 480, | |
| "valid_targets_mean": 3190.4, | |
| "valid_targets_min": 993 | |
| }, | |
| { | |
| "epoch": 0.509453781512605, | |
| "grad_norm": 0.8146565793649729, | |
| "learning_rate": 2.9025487256371818e-05, | |
| "loss": 0.3431, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3617240786552429, | |
| "step": 485, | |
| "valid_targets_mean": 2785.6, | |
| "valid_targets_min": 1275 | |
| }, | |
| { | |
| "epoch": 0.5147058823529411, | |
| "grad_norm": 0.7209287114298926, | |
| "learning_rate": 2.9325337331334335e-05, | |
| "loss": 0.352, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36016473174095154, | |
| "step": 490, | |
| "valid_targets_mean": 3266.8, | |
| "valid_targets_min": 1669 | |
| }, | |
| { | |
| "epoch": 0.5199579831932774, | |
| "grad_norm": 0.7172263463358715, | |
| "learning_rate": 2.9625187406296856e-05, | |
| "loss": 0.3609, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36014240980148315, | |
| "step": 495, | |
| "valid_targets_mean": 3357.5, | |
| "valid_targets_min": 1710 | |
| }, | |
| { | |
| "epoch": 0.5252100840336135, | |
| "grad_norm": 0.6972857671515527, | |
| "learning_rate": 2.992503748125937e-05, | |
| "loss": 0.3617, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34464967250823975, | |
| "step": 500, | |
| "valid_targets_mean": 3468.1, | |
| "valid_targets_min": 1685 | |
| }, | |
| { | |
| "epoch": 0.5304621848739496, | |
| "grad_norm": 0.7173833022077326, | |
| "learning_rate": 3.022488755622189e-05, | |
| "loss": 0.3608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35650789737701416, | |
| "step": 505, | |
| "valid_targets_mean": 3659.6, | |
| "valid_targets_min": 1931 | |
| }, | |
| { | |
| "epoch": 0.5357142857142857, | |
| "grad_norm": 0.7592796407843059, | |
| "learning_rate": 3.052473763118441e-05, | |
| "loss": 0.3512, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3427087068557739, | |
| "step": 510, | |
| "valid_targets_mean": 3862.3, | |
| "valid_targets_min": 2009 | |
| }, | |
| { | |
| "epoch": 0.5409663865546218, | |
| "grad_norm": 1.1258165798910273, | |
| "learning_rate": 3.0824587706146934e-05, | |
| "loss": 0.361, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.367844820022583, | |
| "step": 515, | |
| "valid_targets_mean": 1792.6, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 0.5462184873949579, | |
| "grad_norm": 1.1349526835427646, | |
| "learning_rate": 3.112443778110945e-05, | |
| "loss": 0.3252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3122398853302002, | |
| "step": 520, | |
| "valid_targets_mean": 1354.7, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 0.5514705882352942, | |
| "grad_norm": 1.8110765728106655, | |
| "learning_rate": 3.142428785607197e-05, | |
| "loss": 0.3052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3012329041957855, | |
| "step": 525, | |
| "valid_targets_mean": 1556.9, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 0.5567226890756303, | |
| "grad_norm": 1.1404918052705433, | |
| "learning_rate": 3.172413793103448e-05, | |
| "loss": 0.2933, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2664037346839905, | |
| "step": 530, | |
| "valid_targets_mean": 1397.1, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 0.5619747899159664, | |
| "grad_norm": 0.9458053300681202, | |
| "learning_rate": 3.2023988005997004e-05, | |
| "loss": 0.2891, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.289004921913147, | |
| "step": 535, | |
| "valid_targets_mean": 1698.1, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 0.5672268907563025, | |
| "grad_norm": 1.0086441843661589, | |
| "learning_rate": 3.2323838080959525e-05, | |
| "loss": 0.2798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26696014404296875, | |
| "step": 540, | |
| "valid_targets_mean": 1393.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 0.5724789915966386, | |
| "grad_norm": 1.0260360672408402, | |
| "learning_rate": 3.262368815592204e-05, | |
| "loss": 0.29, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2623750567436218, | |
| "step": 545, | |
| "valid_targets_mean": 1443.1, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 0.5777310924369747, | |
| "grad_norm": 0.9985074269719646, | |
| "learning_rate": 3.292353823088456e-05, | |
| "loss": 0.2606, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26140543818473816, | |
| "step": 550, | |
| "valid_targets_mean": 1392.2, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 0.582983193277311, | |
| "grad_norm": 0.9985105336106163, | |
| "learning_rate": 3.3223388305847075e-05, | |
| "loss": 0.2812, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25588417053222656, | |
| "step": 555, | |
| "valid_targets_mean": 1386.6, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 0.5882352941176471, | |
| "grad_norm": 1.1959754188425515, | |
| "learning_rate": 3.3523238380809596e-05, | |
| "loss": 0.2603, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27921628952026367, | |
| "step": 560, | |
| "valid_targets_mean": 1442.9, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 0.5934873949579832, | |
| "grad_norm": 1.001393308800324, | |
| "learning_rate": 3.382308845577212e-05, | |
| "loss": 0.2606, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2664732336997986, | |
| "step": 565, | |
| "valid_targets_mean": 1536.1, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 0.5987394957983193, | |
| "grad_norm": 0.9505908357818789, | |
| "learning_rate": 3.412293853073464e-05, | |
| "loss": 0.2688, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2818218767642975, | |
| "step": 570, | |
| "valid_targets_mean": 1463.1, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 0.6039915966386554, | |
| "grad_norm": 1.0707192051376093, | |
| "learning_rate": 3.442278860569715e-05, | |
| "loss": 0.2615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2692759931087494, | |
| "step": 575, | |
| "valid_targets_mean": 1292.8, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 0.6092436974789915, | |
| "grad_norm": 1.1070031492080266, | |
| "learning_rate": 3.4722638680659673e-05, | |
| "loss": 0.2482, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2560975253582001, | |
| "step": 580, | |
| "valid_targets_mean": 1542.4, | |
| "valid_targets_min": 878 | |
| }, | |
| { | |
| "epoch": 0.6144957983193278, | |
| "grad_norm": 0.898740421578359, | |
| "learning_rate": 3.5022488755622194e-05, | |
| "loss": 0.2447, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23506242036819458, | |
| "step": 585, | |
| "valid_targets_mean": 1506.8, | |
| "valid_targets_min": 942 | |
| }, | |
| { | |
| "epoch": 0.6197478991596639, | |
| "grad_norm": 0.9882747404397235, | |
| "learning_rate": 3.5322338830584716e-05, | |
| "loss": 0.2549, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2667955756187439, | |
| "step": 590, | |
| "valid_targets_mean": 1401.9, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 0.625, | |
| "grad_norm": 0.9973429121391257, | |
| "learning_rate": 3.562218890554723e-05, | |
| "loss": 0.2558, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2597658038139343, | |
| "step": 595, | |
| "valid_targets_mean": 1483.8, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 0.6302521008403361, | |
| "grad_norm": 1.1184131263781747, | |
| "learning_rate": 3.592203898050975e-05, | |
| "loss": 0.2616, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2559979557991028, | |
| "step": 600, | |
| "valid_targets_mean": 1356.3, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 0.6355042016806722, | |
| "grad_norm": 1.1551828432063649, | |
| "learning_rate": 3.6221889055472265e-05, | |
| "loss": 0.259, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2352353036403656, | |
| "step": 605, | |
| "valid_targets_mean": 1340.6, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 0.6407563025210085, | |
| "grad_norm": 0.9979079936759186, | |
| "learning_rate": 3.6521739130434786e-05, | |
| "loss": 0.2565, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.257199227809906, | |
| "step": 610, | |
| "valid_targets_mean": 1528.6, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 0.6460084033613446, | |
| "grad_norm": 0.9872520856471761, | |
| "learning_rate": 3.68215892053973e-05, | |
| "loss": 0.2478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24192388355731964, | |
| "step": 615, | |
| "valid_targets_mean": 1268.9, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 0.6512605042016807, | |
| "grad_norm": 1.0290619368672818, | |
| "learning_rate": 3.712143928035982e-05, | |
| "loss": 0.2568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.250153124332428, | |
| "step": 620, | |
| "valid_targets_mean": 1360.1, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 0.6565126050420168, | |
| "grad_norm": 0.9161121685195928, | |
| "learning_rate": 3.742128935532234e-05, | |
| "loss": 0.2579, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25210872292518616, | |
| "step": 625, | |
| "valid_targets_mean": 1452.3, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 0.6617647058823529, | |
| "grad_norm": 1.231558662443427, | |
| "learning_rate": 3.772113943028486e-05, | |
| "loss": 0.2355, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2508789896965027, | |
| "step": 630, | |
| "valid_targets_mean": 1391.3, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 0.667016806722689, | |
| "grad_norm": 0.8973325739286778, | |
| "learning_rate": 3.802098950524738e-05, | |
| "loss": 0.2461, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22503423690795898, | |
| "step": 635, | |
| "valid_targets_mean": 1344.4, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 0.6722689075630253, | |
| "grad_norm": 0.9602158780480545, | |
| "learning_rate": 3.83208395802099e-05, | |
| "loss": 0.2427, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2667675018310547, | |
| "step": 640, | |
| "valid_targets_mean": 1545.2, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 0.6775210084033614, | |
| "grad_norm": 0.9207285003363753, | |
| "learning_rate": 3.862068965517242e-05, | |
| "loss": 0.239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23141629993915558, | |
| "step": 645, | |
| "valid_targets_mean": 1354.6, | |
| "valid_targets_min": 779 | |
| }, | |
| { | |
| "epoch": 0.6827731092436975, | |
| "grad_norm": 0.8852191091332312, | |
| "learning_rate": 3.8920539730134934e-05, | |
| "loss": 0.2563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2596163749694824, | |
| "step": 650, | |
| "valid_targets_mean": 1434.1, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 0.6880252100840336, | |
| "grad_norm": 0.912601154449096, | |
| "learning_rate": 3.9220389805097455e-05, | |
| "loss": 0.2314, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22037877142429352, | |
| "step": 655, | |
| "valid_targets_mean": 1388.2, | |
| "valid_targets_min": 868 | |
| }, | |
| { | |
| "epoch": 0.6932773109243697, | |
| "grad_norm": 1.015856823111782, | |
| "learning_rate": 3.9520239880059976e-05, | |
| "loss": 0.2452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23023347556591034, | |
| "step": 660, | |
| "valid_targets_mean": 1282.3, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 0.6985294117647058, | |
| "grad_norm": 0.9855004619555965, | |
| "learning_rate": 3.982008995502249e-05, | |
| "loss": 0.2467, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2527748942375183, | |
| "step": 665, | |
| "valid_targets_mean": 1457.2, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 0.7037815126050421, | |
| "grad_norm": 0.9177092400221334, | |
| "learning_rate": 3.9999989022799437e-05, | |
| "loss": 0.2481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26127058267593384, | |
| "step": 670, | |
| "valid_targets_mean": 1448.5, | |
| "valid_targets_min": 806 | |
| }, | |
| { | |
| "epoch": 0.7090336134453782, | |
| "grad_norm": 0.8689313261706375, | |
| "learning_rate": 3.9999865529431466e-05, | |
| "loss": 0.2298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2190496325492859, | |
| "step": 675, | |
| "valid_targets_mean": 1380.7, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 0.7142857142857143, | |
| "grad_norm": 0.8662595043279143, | |
| "learning_rate": 3.9999604822044886e-05, | |
| "loss": 0.2387, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22773924469947815, | |
| "step": 680, | |
| "valid_targets_mean": 1515.7, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 0.7195378151260504, | |
| "grad_norm": 0.9275572542299572, | |
| "learning_rate": 3.999920690242835e-05, | |
| "loss": 0.2194, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21551215648651123, | |
| "step": 685, | |
| "valid_targets_mean": 1402.0, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 0.7247899159663865, | |
| "grad_norm": 0.8209341148096483, | |
| "learning_rate": 3.999867177331189e-05, | |
| "loss": 0.2359, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22521734237670898, | |
| "step": 690, | |
| "valid_targets_mean": 1574.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 0.7300420168067226, | |
| "grad_norm": 0.9448587281311374, | |
| "learning_rate": 3.9997999438366895e-05, | |
| "loss": 0.2429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23695966601371765, | |
| "step": 695, | |
| "valid_targets_mean": 1430.1, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 0.7352941176470589, | |
| "grad_norm": 0.9348888815574206, | |
| "learning_rate": 3.9997189902206065e-05, | |
| "loss": 0.2388, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24235057830810547, | |
| "step": 700, | |
| "valid_targets_mean": 1372.3, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 0.740546218487395, | |
| "grad_norm": 0.898775745123375, | |
| "learning_rate": 3.999624317038344e-05, | |
| "loss": 0.234, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20886942744255066, | |
| "step": 705, | |
| "valid_targets_mean": 1192.8, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 0.7457983193277311, | |
| "grad_norm": 0.8896703345645318, | |
| "learning_rate": 3.9995159249394303e-05, | |
| "loss": 0.2447, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2359616905450821, | |
| "step": 710, | |
| "valid_targets_mean": 1508.0, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 0.7510504201680672, | |
| "grad_norm": 0.9044856336058926, | |
| "learning_rate": 3.999393814667517e-05, | |
| "loss": 0.2258, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22560223937034607, | |
| "step": 715, | |
| "valid_targets_mean": 1430.2, | |
| "valid_targets_min": 1046 | |
| }, | |
| { | |
| "epoch": 0.7563025210084033, | |
| "grad_norm": 0.7809432057780077, | |
| "learning_rate": 3.9992579870603695e-05, | |
| "loss": 0.2447, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2628813087940216, | |
| "step": 720, | |
| "valid_targets_mean": 1651.8, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 0.7615546218487395, | |
| "grad_norm": 0.8587913679452784, | |
| "learning_rate": 3.999108443049869e-05, | |
| "loss": 0.2396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24335074424743652, | |
| "step": 725, | |
| "valid_targets_mean": 1540.4, | |
| "valid_targets_min": 846 | |
| }, | |
| { | |
| "epoch": 0.7668067226890757, | |
| "grad_norm": 0.9548503644681028, | |
| "learning_rate": 3.998945183661997e-05, | |
| "loss": 0.2532, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2418360412120819, | |
| "step": 730, | |
| "valid_targets_mean": 1455.1, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 0.7720588235294118, | |
| "grad_norm": 0.8057865606417343, | |
| "learning_rate": 3.9987682100168376e-05, | |
| "loss": 0.2286, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21932202577590942, | |
| "step": 735, | |
| "valid_targets_mean": 1497.4, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 0.7773109243697479, | |
| "grad_norm": 0.8569864825362092, | |
| "learning_rate": 3.99857752332856e-05, | |
| "loss": 0.2326, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24598419666290283, | |
| "step": 740, | |
| "valid_targets_mean": 1543.8, | |
| "valid_targets_min": 982 | |
| }, | |
| { | |
| "epoch": 0.782563025210084, | |
| "grad_norm": 0.8966461968047943, | |
| "learning_rate": 3.998373124905418e-05, | |
| "loss": 0.2414, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24538162350654602, | |
| "step": 745, | |
| "valid_targets_mean": 1346.3, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 0.7878151260504201, | |
| "grad_norm": 0.9787609428455722, | |
| "learning_rate": 3.9981550161497375e-05, | |
| "loss": 0.2329, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22620789706707, | |
| "step": 750, | |
| "valid_targets_mean": 1395.7, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 0.7930672268907563, | |
| "grad_norm": 0.8647918297737699, | |
| "learning_rate": 3.9979231985579074e-05, | |
| "loss": 0.2291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22276586294174194, | |
| "step": 755, | |
| "valid_targets_mean": 1290.6, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 0.7983193277310925, | |
| "grad_norm": 0.8550351182526175, | |
| "learning_rate": 3.99767767372037e-05, | |
| "loss": 0.2309, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24379882216453552, | |
| "step": 760, | |
| "valid_targets_mean": 1568.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 0.8035714285714286, | |
| "grad_norm": 0.8680279094247919, | |
| "learning_rate": 3.997418443321609e-05, | |
| "loss": 0.2389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23442883789539337, | |
| "step": 765, | |
| "valid_targets_mean": 1703.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 0.8088235294117647, | |
| "grad_norm": 0.9333899796386794, | |
| "learning_rate": 3.997145509140138e-05, | |
| "loss": 0.2392, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25179609656333923, | |
| "step": 770, | |
| "valid_targets_mean": 1652.8, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 0.8140756302521008, | |
| "grad_norm": 0.7856961011675193, | |
| "learning_rate": 3.9968588730484896e-05, | |
| "loss": 0.221, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.224946990609169, | |
| "step": 775, | |
| "valid_targets_mean": 1481.8, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 0.819327731092437, | |
| "grad_norm": 0.7761881827856689, | |
| "learning_rate": 3.9965585370132013e-05, | |
| "loss": 0.2298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21191829442977905, | |
| "step": 780, | |
| "valid_targets_mean": 1686.5, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 0.8245798319327731, | |
| "grad_norm": 0.8321605983306755, | |
| "learning_rate": 3.996244503094804e-05, | |
| "loss": 0.2275, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20524507761001587, | |
| "step": 785, | |
| "valid_targets_mean": 1345.1, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 0.8298319327731093, | |
| "grad_norm": 0.8224185017405469, | |
| "learning_rate": 3.995916773447804e-05, | |
| "loss": 0.2291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22673243284225464, | |
| "step": 790, | |
| "valid_targets_mean": 1619.1, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 0.8350840336134454, | |
| "grad_norm": 1.5699722200850348, | |
| "learning_rate": 3.995575350320671e-05, | |
| "loss": 0.2223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2145438939332962, | |
| "step": 795, | |
| "valid_targets_mean": 1423.2, | |
| "valid_targets_min": 674 | |
| }, | |
| { | |
| "epoch": 0.8403361344537815, | |
| "grad_norm": 0.8647740397931104, | |
| "learning_rate": 3.995220236055824e-05, | |
| "loss": 0.2149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19764792919158936, | |
| "step": 800, | |
| "valid_targets_mean": 1207.4, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 0.8455882352941176, | |
| "grad_norm": 0.8208866711497894, | |
| "learning_rate": 3.99485143308961e-05, | |
| "loss": 0.2219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22692370414733887, | |
| "step": 805, | |
| "valid_targets_mean": 1542.4, | |
| "valid_targets_min": 912 | |
| }, | |
| { | |
| "epoch": 0.8508403361344538, | |
| "grad_norm": 0.7660210546255793, | |
| "learning_rate": 3.994468943952296e-05, | |
| "loss": 0.2316, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23304234445095062, | |
| "step": 810, | |
| "valid_targets_mean": 1591.0, | |
| "valid_targets_min": 897 | |
| }, | |
| { | |
| "epoch": 0.8560924369747899, | |
| "grad_norm": 0.8316792538373139, | |
| "learning_rate": 3.994072771268041e-05, | |
| "loss": 0.2177, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2084471583366394, | |
| "step": 815, | |
| "valid_targets_mean": 1290.1, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 0.8613445378151261, | |
| "grad_norm": 0.8668112604053329, | |
| "learning_rate": 3.993662917754885e-05, | |
| "loss": 0.2325, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24037593603134155, | |
| "step": 820, | |
| "valid_targets_mean": 1407.2, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 0.8665966386554622, | |
| "grad_norm": 0.9891212862372226, | |
| "learning_rate": 3.993239386224732e-05, | |
| "loss": 0.2266, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22597835958003998, | |
| "step": 825, | |
| "valid_targets_mean": 1600.2, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 0.8718487394957983, | |
| "grad_norm": 0.8365460249899657, | |
| "learning_rate": 3.992802179583322e-05, | |
| "loss": 0.2294, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22174695134162903, | |
| "step": 830, | |
| "valid_targets_mean": 1442.4, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 0.8771008403361344, | |
| "grad_norm": 0.955688819196928, | |
| "learning_rate": 3.9923513008302204e-05, | |
| "loss": 0.2233, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23063375055789948, | |
| "step": 835, | |
| "valid_targets_mean": 1426.6, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 0.8823529411764706, | |
| "grad_norm": 0.7775298281309749, | |
| "learning_rate": 3.991886753058792e-05, | |
| "loss": 0.2239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21736116707324982, | |
| "step": 840, | |
| "valid_targets_mean": 1430.9, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 0.8876050420168067, | |
| "grad_norm": 0.9219183595495097, | |
| "learning_rate": 3.991408539456182e-05, | |
| "loss": 0.2397, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2496243417263031, | |
| "step": 845, | |
| "valid_targets_mean": 1412.6, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 0.8928571428571429, | |
| "grad_norm": 0.9447703661465241, | |
| "learning_rate": 3.990916663303293e-05, | |
| "loss": 0.2292, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2532111406326294, | |
| "step": 850, | |
| "valid_targets_mean": 1372.1, | |
| "valid_targets_min": 667 | |
| }, | |
| { | |
| "epoch": 0.898109243697479, | |
| "grad_norm": 0.7856700847441219, | |
| "learning_rate": 3.990411127974762e-05, | |
| "loss": 0.235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2314455211162567, | |
| "step": 855, | |
| "valid_targets_mean": 1628.1, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 0.9033613445378151, | |
| "grad_norm": 0.9349349696144393, | |
| "learning_rate": 3.989891936938939e-05, | |
| "loss": 0.2331, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2500951886177063, | |
| "step": 860, | |
| "valid_targets_mean": 1464.7, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 0.9086134453781513, | |
| "grad_norm": 0.9119192470614107, | |
| "learning_rate": 3.9893590937578634e-05, | |
| "loss": 0.2337, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2547591030597687, | |
| "step": 865, | |
| "valid_targets_mean": 1745.9, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 0.9138655462184874, | |
| "grad_norm": 0.9202449016918255, | |
| "learning_rate": 3.9888126020872375e-05, | |
| "loss": 0.2197, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20298632979393005, | |
| "step": 870, | |
| "valid_targets_mean": 1129.1, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 0.9191176470588235, | |
| "grad_norm": 0.8174032516425299, | |
| "learning_rate": 3.988252465676401e-05, | |
| "loss": 0.2296, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25348424911499023, | |
| "step": 875, | |
| "valid_targets_mean": 1624.6, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 0.9243697478991597, | |
| "grad_norm": 0.7819162668790818, | |
| "learning_rate": 3.98767868836831e-05, | |
| "loss": 0.23, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27575284242630005, | |
| "step": 880, | |
| "valid_targets_mean": 1755.8, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 0.9296218487394958, | |
| "grad_norm": 0.8141268835699211, | |
| "learning_rate": 3.987091274099504e-05, | |
| "loss": 0.2229, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2335035800933838, | |
| "step": 885, | |
| "valid_targets_mean": 1516.4, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 0.9348739495798319, | |
| "grad_norm": 0.7937591348329995, | |
| "learning_rate": 3.986490226900084e-05, | |
| "loss": 0.2232, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22677749395370483, | |
| "step": 890, | |
| "valid_targets_mean": 1391.4, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 0.9401260504201681, | |
| "grad_norm": 0.8638827876332873, | |
| "learning_rate": 3.985875550893684e-05, | |
| "loss": 0.2125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2247205227613449, | |
| "step": 895, | |
| "valid_targets_mean": 1412.4, | |
| "valid_targets_min": 778 | |
| }, | |
| { | |
| "epoch": 0.9453781512605042, | |
| "grad_norm": 0.7937334625060325, | |
| "learning_rate": 3.9852472502974386e-05, | |
| "loss": 0.2369, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25422534346580505, | |
| "step": 900, | |
| "valid_targets_mean": 1599.6, | |
| "valid_targets_min": 851 | |
| }, | |
| { | |
| "epoch": 0.9506302521008403, | |
| "grad_norm": 0.7815070706059014, | |
| "learning_rate": 3.984605329421961e-05, | |
| "loss": 0.2211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21097871661186218, | |
| "step": 905, | |
| "valid_targets_mean": 1398.4, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 0.9558823529411765, | |
| "grad_norm": 0.7673078400464335, | |
| "learning_rate": 3.983949792671307e-05, | |
| "loss": 0.2148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20964649319648743, | |
| "step": 910, | |
| "valid_targets_mean": 1477.9, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 0.9611344537815126, | |
| "grad_norm": 0.851415365699838, | |
| "learning_rate": 3.9832806445429486e-05, | |
| "loss": 0.2131, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20876117050647736, | |
| "step": 915, | |
| "valid_targets_mean": 1299.4, | |
| "valid_targets_min": 639 | |
| }, | |
| { | |
| "epoch": 0.9663865546218487, | |
| "grad_norm": 0.866799347417237, | |
| "learning_rate": 3.982597889627742e-05, | |
| "loss": 0.2056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21439066529273987, | |
| "step": 920, | |
| "valid_targets_mean": 1277.9, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 0.9716386554621849, | |
| "grad_norm": 0.7220607995686592, | |
| "learning_rate": 3.981901532609896e-05, | |
| "loss": 0.2239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2239415943622589, | |
| "step": 925, | |
| "valid_targets_mean": 1529.9, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 0.976890756302521, | |
| "grad_norm": 0.7774847164190153, | |
| "learning_rate": 3.9811915782669406e-05, | |
| "loss": 0.2182, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20755037665367126, | |
| "step": 930, | |
| "valid_targets_mean": 1331.6, | |
| "valid_targets_min": 695 | |
| }, | |
| { | |
| "epoch": 0.9821428571428571, | |
| "grad_norm": 0.7514140462848654, | |
| "learning_rate": 3.980468031469691e-05, | |
| "loss": 0.2219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2222648710012436, | |
| "step": 935, | |
| "valid_targets_mean": 1581.9, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 0.9873949579831933, | |
| "grad_norm": 0.7556883989929587, | |
| "learning_rate": 3.9797308971822206e-05, | |
| "loss": 0.2155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2113157957792282, | |
| "step": 940, | |
| "valid_targets_mean": 1327.6, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 0.9926470588235294, | |
| "grad_norm": 0.7916476167495613, | |
| "learning_rate": 3.978980180461821e-05, | |
| "loss": 0.2245, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23679675161838531, | |
| "step": 945, | |
| "valid_targets_mean": 1346.4, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 0.9978991596638656, | |
| "grad_norm": 0.7026077039339454, | |
| "learning_rate": 3.9782158864589696e-05, | |
| "loss": 0.2291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20915091037750244, | |
| "step": 950, | |
| "valid_targets_mean": 1483.2, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 1.0031512605042017, | |
| "grad_norm": 0.5638798620218901, | |
| "learning_rate": 3.977438020417293e-05, | |
| "loss": 0.3126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33122366666793823, | |
| "step": 955, | |
| "valid_targets_mean": 4404.5, | |
| "valid_targets_min": 1323 | |
| }, | |
| { | |
| "epoch": 1.0084033613445378, | |
| "grad_norm": 0.5755600203933873, | |
| "learning_rate": 3.9766465876735354e-05, | |
| "loss": 0.3596, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3643394112586975, | |
| "step": 960, | |
| "valid_targets_mean": 4483.4, | |
| "valid_targets_min": 1005 | |
| }, | |
| { | |
| "epoch": 1.013655462184874, | |
| "grad_norm": 0.5433500453139155, | |
| "learning_rate": 3.9758415936575155e-05, | |
| "loss": 0.3387, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3266267776489258, | |
| "step": 965, | |
| "valid_targets_mean": 4502.1, | |
| "valid_targets_min": 1509 | |
| }, | |
| { | |
| "epoch": 1.01890756302521, | |
| "grad_norm": 0.6267141368025664, | |
| "learning_rate": 3.975023043892094e-05, | |
| "loss": 0.3404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3264230787754059, | |
| "step": 970, | |
| "valid_targets_mean": 3430.2, | |
| "valid_targets_min": 1561 | |
| }, | |
| { | |
| "epoch": 1.0241596638655461, | |
| "grad_norm": 0.607104749067981, | |
| "learning_rate": 3.974190943993133e-05, | |
| "loss": 0.329, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3339942693710327, | |
| "step": 975, | |
| "valid_targets_mean": 3385.4, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 1.0294117647058822, | |
| "grad_norm": 0.536136082897943, | |
| "learning_rate": 3.973345299669461e-05, | |
| "loss": 0.3558, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3149448335170746, | |
| "step": 980, | |
| "valid_targets_mean": 3982.6, | |
| "valid_targets_min": 1413 | |
| }, | |
| { | |
| "epoch": 1.0346638655462186, | |
| "grad_norm": 0.5075659675082035, | |
| "learning_rate": 3.972486116722828e-05, | |
| "loss": 0.3169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32082343101501465, | |
| "step": 985, | |
| "valid_targets_mean": 4911.7, | |
| "valid_targets_min": 2160 | |
| }, | |
| { | |
| "epoch": 1.0399159663865547, | |
| "grad_norm": 0.5581426071520623, | |
| "learning_rate": 3.971613401047872e-05, | |
| "loss": 0.3346, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35363245010375977, | |
| "step": 990, | |
| "valid_targets_mean": 3782.6, | |
| "valid_targets_min": 1511 | |
| }, | |
| { | |
| "epoch": 1.0451680672268908, | |
| "grad_norm": 0.698413730384984, | |
| "learning_rate": 3.970727158632075e-05, | |
| "loss": 0.3462, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.374415785074234, | |
| "step": 995, | |
| "valid_targets_mean": 3030.0, | |
| "valid_targets_min": 1462 | |
| }, | |
| { | |
| "epoch": 1.050420168067227, | |
| "grad_norm": 0.6206079287451801, | |
| "learning_rate": 3.969827395555721e-05, | |
| "loss": 0.3457, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3208198845386505, | |
| "step": 1000, | |
| "valid_targets_mean": 3772.4, | |
| "valid_targets_min": 874 | |
| }, | |
| { | |
| "epoch": 1.055672268907563, | |
| "grad_norm": 0.6660693366788901, | |
| "learning_rate": 3.968914117991857e-05, | |
| "loss": 0.3445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3395417630672455, | |
| "step": 1005, | |
| "valid_targets_mean": 3337.1, | |
| "valid_targets_min": 953 | |
| }, | |
| { | |
| "epoch": 1.0609243697478992, | |
| "grad_norm": 0.6001541700551754, | |
| "learning_rate": 3.967987332206249e-05, | |
| "loss": 0.3465, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3561500310897827, | |
| "step": 1010, | |
| "valid_targets_mean": 4284.9, | |
| "valid_targets_min": 1265 | |
| }, | |
| { | |
| "epoch": 1.0661764705882353, | |
| "grad_norm": 0.5433768685257184, | |
| "learning_rate": 3.967047044557341e-05, | |
| "loss": 0.3387, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3220909833908081, | |
| "step": 1015, | |
| "valid_targets_mean": 5010.6, | |
| "valid_targets_min": 1596 | |
| }, | |
| { | |
| "epoch": 1.0714285714285714, | |
| "grad_norm": 0.6300742894051691, | |
| "learning_rate": 3.966093261496208e-05, | |
| "loss": 0.3393, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3596876561641693, | |
| "step": 1020, | |
| "valid_targets_mean": 3490.4, | |
| "valid_targets_min": 1759 | |
| }, | |
| { | |
| "epoch": 1.0766806722689075, | |
| "grad_norm": 0.6791244172980975, | |
| "learning_rate": 3.9651259895665146e-05, | |
| "loss": 0.3504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36252090334892273, | |
| "step": 1025, | |
| "valid_targets_mean": 3264.3, | |
| "valid_targets_min": 1353 | |
| }, | |
| { | |
| "epoch": 1.0819327731092436, | |
| "grad_norm": 0.5749237683976328, | |
| "learning_rate": 3.9641452354044687e-05, | |
| "loss": 0.3478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31524378061294556, | |
| "step": 1030, | |
| "valid_targets_mean": 3814.6, | |
| "valid_targets_min": 1643 | |
| }, | |
| { | |
| "epoch": 1.0871848739495797, | |
| "grad_norm": 0.6067338424067318, | |
| "learning_rate": 3.9631510057387765e-05, | |
| "loss": 0.345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32914769649505615, | |
| "step": 1035, | |
| "valid_targets_mean": 3166.1, | |
| "valid_targets_min": 1269 | |
| }, | |
| { | |
| "epoch": 1.092436974789916, | |
| "grad_norm": 0.6139160246013105, | |
| "learning_rate": 3.962143307390596e-05, | |
| "loss": 0.3648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3532394766807556, | |
| "step": 1040, | |
| "valid_targets_mean": 3617.4, | |
| "valid_targets_min": 1681 | |
| }, | |
| { | |
| "epoch": 1.0976890756302522, | |
| "grad_norm": 0.5852424169834737, | |
| "learning_rate": 3.961122147273491e-05, | |
| "loss": 0.3475, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.306821346282959, | |
| "step": 1045, | |
| "valid_targets_mean": 3919.8, | |
| "valid_targets_min": 1477 | |
| }, | |
| { | |
| "epoch": 1.1029411764705883, | |
| "grad_norm": 0.6196200689994448, | |
| "learning_rate": 3.9600875323933826e-05, | |
| "loss": 0.3494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34865230321884155, | |
| "step": 1050, | |
| "valid_targets_mean": 3418.5, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 1.1081932773109244, | |
| "grad_norm": 0.5913392876658434, | |
| "learning_rate": 3.959039469848502e-05, | |
| "loss": 0.3432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3537408113479614, | |
| "step": 1055, | |
| "valid_targets_mean": 3706.7, | |
| "valid_targets_min": 2077 | |
| }, | |
| { | |
| "epoch": 1.1134453781512605, | |
| "grad_norm": 0.6771276206524633, | |
| "learning_rate": 3.9579779668293416e-05, | |
| "loss": 0.3249, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3203701972961426, | |
| "step": 1060, | |
| "valid_targets_mean": 2878.1, | |
| "valid_targets_min": 1505 | |
| }, | |
| { | |
| "epoch": 1.1186974789915967, | |
| "grad_norm": 0.6786962469374573, | |
| "learning_rate": 3.956903030618605e-05, | |
| "loss": 0.3523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35575729608535767, | |
| "step": 1065, | |
| "valid_targets_mean": 3418.0, | |
| "valid_targets_min": 1957 | |
| }, | |
| { | |
| "epoch": 1.1239495798319328, | |
| "grad_norm": 0.5934416223711014, | |
| "learning_rate": 3.955814668591156e-05, | |
| "loss": 0.3487, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38778382539749146, | |
| "step": 1070, | |
| "valid_targets_mean": 4106.4, | |
| "valid_targets_min": 1704 | |
| }, | |
| { | |
| "epoch": 1.129201680672269, | |
| "grad_norm": 1.3852530686455735, | |
| "learning_rate": 3.9547128882139735e-05, | |
| "loss": 0.3349, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3465884327888489, | |
| "step": 1075, | |
| "valid_targets_mean": 3046.9, | |
| "valid_targets_min": 1422 | |
| }, | |
| { | |
| "epoch": 1.134453781512605, | |
| "grad_norm": 0.5980418747067316, | |
| "learning_rate": 3.9535976970460925e-05, | |
| "loss": 0.3252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3092625141143799, | |
| "step": 1080, | |
| "valid_targets_mean": 3447.4, | |
| "valid_targets_min": 1603 | |
| }, | |
| { | |
| "epoch": 1.1397058823529411, | |
| "grad_norm": 0.7839568612134259, | |
| "learning_rate": 3.9524691027385585e-05, | |
| "loss": 0.3504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3626514971256256, | |
| "step": 1085, | |
| "valid_targets_mean": 2597.8, | |
| "valid_targets_min": 1685 | |
| }, | |
| { | |
| "epoch": 1.1449579831932772, | |
| "grad_norm": 0.6257049596683059, | |
| "learning_rate": 3.95132711303437e-05, | |
| "loss": 0.3294, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3430899977684021, | |
| "step": 1090, | |
| "valid_targets_mean": 3290.6, | |
| "valid_targets_min": 1208 | |
| }, | |
| { | |
| "epoch": 1.1502100840336134, | |
| "grad_norm": 0.9029010729944724, | |
| "learning_rate": 3.9501717357684315e-05, | |
| "loss": 0.3368, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.326965868473053, | |
| "step": 1095, | |
| "valid_targets_mean": 2661.8, | |
| "valid_targets_min": 1501 | |
| }, | |
| { | |
| "epoch": 1.1554621848739495, | |
| "grad_norm": 0.6012247577775289, | |
| "learning_rate": 3.9490029788674934e-05, | |
| "loss": 0.3438, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3628232479095459, | |
| "step": 1100, | |
| "valid_targets_mean": 3849.0, | |
| "valid_targets_min": 1866 | |
| }, | |
| { | |
| "epoch": 1.1607142857142858, | |
| "grad_norm": 0.614388950135581, | |
| "learning_rate": 3.9478208503501e-05, | |
| "loss": 0.3242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31284618377685547, | |
| "step": 1105, | |
| "valid_targets_mean": 3397.9, | |
| "valid_targets_min": 1239 | |
| }, | |
| { | |
| "epoch": 1.165966386554622, | |
| "grad_norm": 0.653467443640438, | |
| "learning_rate": 3.946625358326538e-05, | |
| "loss": 0.3439, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.346976101398468, | |
| "step": 1110, | |
| "valid_targets_mean": 3155.5, | |
| "valid_targets_min": 1177 | |
| }, | |
| { | |
| "epoch": 1.171218487394958, | |
| "grad_norm": 0.579694254639963, | |
| "learning_rate": 3.945416510998775e-05, | |
| "loss": 0.3355, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32414424419403076, | |
| "step": 1115, | |
| "valid_targets_mean": 3594.5, | |
| "valid_targets_min": 1270 | |
| }, | |
| { | |
| "epoch": 1.1764705882352942, | |
| "grad_norm": 0.6219350779760418, | |
| "learning_rate": 3.944194316660406e-05, | |
| "loss": 0.3396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3272247314453125, | |
| "step": 1120, | |
| "valid_targets_mean": 3571.2, | |
| "valid_targets_min": 1216 | |
| }, | |
| { | |
| "epoch": 1.1817226890756303, | |
| "grad_norm": 0.6292410803179813, | |
| "learning_rate": 3.942958783696598e-05, | |
| "loss": 0.3334, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32565245032310486, | |
| "step": 1125, | |
| "valid_targets_mean": 3290.8, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 1.1869747899159664, | |
| "grad_norm": 0.678979812159635, | |
| "learning_rate": 3.94170992058403e-05, | |
| "loss": 0.3349, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3311670422554016, | |
| "step": 1130, | |
| "valid_targets_mean": 2986.1, | |
| "valid_targets_min": 1292 | |
| }, | |
| { | |
| "epoch": 1.1922268907563025, | |
| "grad_norm": 0.6456475043572049, | |
| "learning_rate": 3.9404477358908354e-05, | |
| "loss": 0.3445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3214397132396698, | |
| "step": 1135, | |
| "valid_targets_mean": 3485.4, | |
| "valid_targets_min": 1639 | |
| }, | |
| { | |
| "epoch": 1.1974789915966386, | |
| "grad_norm": 0.6695778532642306, | |
| "learning_rate": 3.9391722382765445e-05, | |
| "loss": 0.3281, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35410648584365845, | |
| "step": 1140, | |
| "valid_targets_mean": 2959.4, | |
| "valid_targets_min": 1394 | |
| }, | |
| { | |
| "epoch": 1.2027310924369747, | |
| "grad_norm": 0.622191556528526, | |
| "learning_rate": 3.937883436492025e-05, | |
| "loss": 0.3351, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3593181073665619, | |
| "step": 1145, | |
| "valid_targets_mean": 3428.9, | |
| "valid_targets_min": 2280 | |
| }, | |
| { | |
| "epoch": 1.2079831932773109, | |
| "grad_norm": 0.6333880919334702, | |
| "learning_rate": 3.9365813393794186e-05, | |
| "loss": 0.3265, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35325169563293457, | |
| "step": 1150, | |
| "valid_targets_mean": 3233.9, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 1.213235294117647, | |
| "grad_norm": 0.6223354482954172, | |
| "learning_rate": 3.9352659558720836e-05, | |
| "loss": 0.3196, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3268436789512634, | |
| "step": 1155, | |
| "valid_targets_mean": 3211.6, | |
| "valid_targets_min": 1654 | |
| }, | |
| { | |
| "epoch": 1.2184873949579833, | |
| "grad_norm": 0.659296380097006, | |
| "learning_rate": 3.933937294994535e-05, | |
| "loss": 0.3341, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.320861279964447, | |
| "step": 1160, | |
| "valid_targets_mean": 3057.8, | |
| "valid_targets_min": 584 | |
| }, | |
| { | |
| "epoch": 1.2237394957983194, | |
| "grad_norm": 0.636761914351194, | |
| "learning_rate": 3.932595365862379e-05, | |
| "loss": 0.3322, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34962353110313416, | |
| "step": 1165, | |
| "valid_targets_mean": 3572.6, | |
| "valid_targets_min": 2498 | |
| }, | |
| { | |
| "epoch": 1.2289915966386555, | |
| "grad_norm": 0.6592752073339605, | |
| "learning_rate": 3.9312401776822504e-05, | |
| "loss": 0.3367, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32008883357048035, | |
| "step": 1170, | |
| "valid_targets_mean": 3161.1, | |
| "valid_targets_min": 1350 | |
| }, | |
| { | |
| "epoch": 1.2342436974789917, | |
| "grad_norm": 0.6063230189221366, | |
| "learning_rate": 3.9298717397517546e-05, | |
| "loss": 0.3365, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3388460874557495, | |
| "step": 1175, | |
| "valid_targets_mean": 3486.1, | |
| "valid_targets_min": 1624 | |
| }, | |
| { | |
| "epoch": 1.2394957983193278, | |
| "grad_norm": 0.687253883898937, | |
| "learning_rate": 3.928490061459396e-05, | |
| "loss": 0.3067, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3046797513961792, | |
| "step": 1180, | |
| "valid_targets_mean": 2769.6, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 1.2447478991596639, | |
| "grad_norm": 0.5602035515910363, | |
| "learning_rate": 3.927095152284521e-05, | |
| "loss": 0.331, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3130587935447693, | |
| "step": 1185, | |
| "valid_targets_mean": 3734.1, | |
| "valid_targets_min": 2209 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "grad_norm": 0.6913362496913676, | |
| "learning_rate": 3.925687021797249e-05, | |
| "loss": 0.3228, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32542842626571655, | |
| "step": 1190, | |
| "valid_targets_mean": 3107.0, | |
| "valid_targets_min": 1259 | |
| }, | |
| { | |
| "epoch": 1.2552521008403361, | |
| "grad_norm": 0.5736429239269779, | |
| "learning_rate": 3.924265679658407e-05, | |
| "loss": 0.3142, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30502641201019287, | |
| "step": 1195, | |
| "valid_targets_mean": 3685.6, | |
| "valid_targets_min": 1183 | |
| }, | |
| { | |
| "epoch": 1.2605042016806722, | |
| "grad_norm": 0.577437241781624, | |
| "learning_rate": 3.922831135619462e-05, | |
| "loss": 0.3109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3453516364097595, | |
| "step": 1200, | |
| "valid_targets_mean": 3891.3, | |
| "valid_targets_min": 1101 | |
| }, | |
| { | |
| "epoch": 1.2657563025210083, | |
| "grad_norm": 0.5958378130959316, | |
| "learning_rate": 3.9213833995224605e-05, | |
| "loss": 0.3321, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3231067359447479, | |
| "step": 1205, | |
| "valid_targets_mean": 3877.6, | |
| "valid_targets_min": 1658 | |
| }, | |
| { | |
| "epoch": 1.2710084033613445, | |
| "grad_norm": 0.6735908115281981, | |
| "learning_rate": 3.919922481299952e-05, | |
| "loss": 0.3188, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3365190923213959, | |
| "step": 1210, | |
| "valid_targets_mean": 3050.1, | |
| "valid_targets_min": 1423 | |
| }, | |
| { | |
| "epoch": 1.2762605042016806, | |
| "grad_norm": 0.6601685873762229, | |
| "learning_rate": 3.918448390974928e-05, | |
| "loss": 0.3281, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3299792408943176, | |
| "step": 1215, | |
| "valid_targets_mean": 3202.4, | |
| "valid_targets_min": 1185 | |
| }, | |
| { | |
| "epoch": 1.2815126050420167, | |
| "grad_norm": 0.6705034899134344, | |
| "learning_rate": 3.9169611386607476e-05, | |
| "loss": 0.3418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34724968671798706, | |
| "step": 1220, | |
| "valid_targets_mean": 3779.1, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 1.2867647058823528, | |
| "grad_norm": 0.5833278508200939, | |
| "learning_rate": 3.9154607345610746e-05, | |
| "loss": 0.3194, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3058725893497467, | |
| "step": 1225, | |
| "valid_targets_mean": 3628.9, | |
| "valid_targets_min": 1352 | |
| }, | |
| { | |
| "epoch": 1.2920168067226891, | |
| "grad_norm": 0.5696798688900944, | |
| "learning_rate": 3.913947188969801e-05, | |
| "loss": 0.3224, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32415255904197693, | |
| "step": 1230, | |
| "valid_targets_mean": 4061.2, | |
| "valid_targets_min": 1535 | |
| }, | |
| { | |
| "epoch": 1.2972689075630253, | |
| "grad_norm": 0.6650203053880857, | |
| "learning_rate": 3.912420512270981e-05, | |
| "loss": 0.3106, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3156277537345886, | |
| "step": 1235, | |
| "valid_targets_mean": 2753.1, | |
| "valid_targets_min": 854 | |
| }, | |
| { | |
| "epoch": 1.3025210084033614, | |
| "grad_norm": 0.6404333915329682, | |
| "learning_rate": 3.910880714938757e-05, | |
| "loss": 0.3168, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30156928300857544, | |
| "step": 1240, | |
| "valid_targets_mean": 2942.5, | |
| "valid_targets_min": 1179 | |
| }, | |
| { | |
| "epoch": 1.3077731092436975, | |
| "grad_norm": 0.604288417212929, | |
| "learning_rate": 3.909327807537288e-05, | |
| "loss": 0.3221, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30336058139801025, | |
| "step": 1245, | |
| "valid_targets_mean": 3598.9, | |
| "valid_targets_min": 1290 | |
| }, | |
| { | |
| "epoch": 1.3130252100840336, | |
| "grad_norm": 0.875497325929702, | |
| "learning_rate": 3.90776180072068e-05, | |
| "loss": 0.3127, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32053911685943604, | |
| "step": 1250, | |
| "valid_targets_mean": 2332.6, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 1.3182773109243697, | |
| "grad_norm": 0.6151615426766832, | |
| "learning_rate": 3.906182705232909e-05, | |
| "loss": 0.3084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3063713312149048, | |
| "step": 1255, | |
| "valid_targets_mean": 3670.2, | |
| "valid_targets_min": 1764 | |
| }, | |
| { | |
| "epoch": 1.3235294117647058, | |
| "grad_norm": 0.6686081770965902, | |
| "learning_rate": 3.904590531907751e-05, | |
| "loss": 0.3375, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3719046711921692, | |
| "step": 1260, | |
| "valid_targets_mean": 3528.6, | |
| "valid_targets_min": 1410 | |
| }, | |
| { | |
| "epoch": 1.328781512605042, | |
| "grad_norm": 0.7395507920156082, | |
| "learning_rate": 3.9029852916687034e-05, | |
| "loss": 0.3145, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3327389657497406, | |
| "step": 1265, | |
| "valid_targets_mean": 3147.7, | |
| "valid_targets_min": 1683 | |
| }, | |
| { | |
| "epoch": 1.334033613445378, | |
| "grad_norm": 0.709030289881482, | |
| "learning_rate": 3.9013669955289136e-05, | |
| "loss": 0.3181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3167096972465515, | |
| "step": 1270, | |
| "valid_targets_mean": 3632.5, | |
| "valid_targets_min": 996 | |
| }, | |
| { | |
| "epoch": 1.3392857142857144, | |
| "grad_norm": 0.6586192793197465, | |
| "learning_rate": 3.899735654591101e-05, | |
| "loss": 0.3171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3190867602825165, | |
| "step": 1275, | |
| "valid_targets_mean": 3135.3, | |
| "valid_targets_min": 1349 | |
| }, | |
| { | |
| "epoch": 1.3445378151260505, | |
| "grad_norm": 0.682140733121633, | |
| "learning_rate": 3.898091280047486e-05, | |
| "loss": 0.3167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33190667629241943, | |
| "step": 1280, | |
| "valid_targets_mean": 3391.6, | |
| "valid_targets_min": 1248 | |
| }, | |
| { | |
| "epoch": 1.3497899159663866, | |
| "grad_norm": 0.5957692884205409, | |
| "learning_rate": 3.896433883179703e-05, | |
| "loss": 0.3131, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3161918520927429, | |
| "step": 1285, | |
| "valid_targets_mean": 3323.4, | |
| "valid_targets_min": 1445 | |
| }, | |
| { | |
| "epoch": 1.3550420168067228, | |
| "grad_norm": 0.6350286869649353, | |
| "learning_rate": 3.894763475358736e-05, | |
| "loss": 0.3115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3040814995765686, | |
| "step": 1290, | |
| "valid_targets_mean": 3134.3, | |
| "valid_targets_min": 2000 | |
| }, | |
| { | |
| "epoch": 1.3602941176470589, | |
| "grad_norm": 0.6934797525878172, | |
| "learning_rate": 3.8930800680448275e-05, | |
| "loss": 0.3064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32412275671958923, | |
| "step": 1295, | |
| "valid_targets_mean": 2828.3, | |
| "valid_targets_min": 1197 | |
| }, | |
| { | |
| "epoch": 1.365546218487395, | |
| "grad_norm": 0.6077199732980291, | |
| "learning_rate": 3.891383672787411e-05, | |
| "loss": 0.331, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32446929812431335, | |
| "step": 1300, | |
| "valid_targets_mean": 3219.0, | |
| "valid_targets_min": 1223 | |
| }, | |
| { | |
| "epoch": 1.370798319327731, | |
| "grad_norm": 0.6275476086489861, | |
| "learning_rate": 3.889674301225025e-05, | |
| "loss": 0.3223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29918792843818665, | |
| "step": 1305, | |
| "valid_targets_mean": 3199.2, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 1.3760504201680672, | |
| "grad_norm": 0.6746694949290901, | |
| "learning_rate": 3.8879519650852356e-05, | |
| "loss": 0.3256, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33456575870513916, | |
| "step": 1310, | |
| "valid_targets_mean": 2817.8, | |
| "valid_targets_min": 1603 | |
| }, | |
| { | |
| "epoch": 1.3813025210084033, | |
| "grad_norm": 0.6421028373174448, | |
| "learning_rate": 3.886216676184555e-05, | |
| "loss": 0.3302, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3090163767337799, | |
| "step": 1315, | |
| "valid_targets_mean": 2942.2, | |
| "valid_targets_min": 1135 | |
| }, | |
| { | |
| "epoch": 1.3865546218487395, | |
| "grad_norm": 0.5573926446318049, | |
| "learning_rate": 3.8844684464283614e-05, | |
| "loss": 0.3167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33767062425613403, | |
| "step": 1320, | |
| "valid_targets_mean": 4077.0, | |
| "valid_targets_min": 1994 | |
| }, | |
| { | |
| "epoch": 1.3918067226890756, | |
| "grad_norm": 0.637126836573775, | |
| "learning_rate": 3.882707287810817e-05, | |
| "loss": 0.3286, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30095210671424866, | |
| "step": 1325, | |
| "valid_targets_mean": 3060.1, | |
| "valid_targets_min": 1066 | |
| }, | |
| { | |
| "epoch": 1.3970588235294117, | |
| "grad_norm": 0.6321058379302512, | |
| "learning_rate": 3.880933212414786e-05, | |
| "loss": 0.3285, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33812808990478516, | |
| "step": 1330, | |
| "valid_targets_mean": 3687.4, | |
| "valid_targets_min": 1211 | |
| }, | |
| { | |
| "epoch": 1.4023109243697478, | |
| "grad_norm": 0.6375112145565428, | |
| "learning_rate": 3.87914623241175e-05, | |
| "loss": 0.3213, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3038761019706726, | |
| "step": 1335, | |
| "valid_targets_mean": 2972.2, | |
| "valid_targets_min": 1195 | |
| }, | |
| { | |
| "epoch": 1.407563025210084, | |
| "grad_norm": 0.7592514428689119, | |
| "learning_rate": 3.877346360061728e-05, | |
| "loss": 0.3291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3361939489841461, | |
| "step": 1340, | |
| "valid_targets_mean": 2429.5, | |
| "valid_targets_min": 484 | |
| }, | |
| { | |
| "epoch": 1.41281512605042, | |
| "grad_norm": 0.6066430879087057, | |
| "learning_rate": 3.8755336077131894e-05, | |
| "loss": 0.3227, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3248080015182495, | |
| "step": 1345, | |
| "valid_targets_mean": 3851.9, | |
| "valid_targets_min": 1257 | |
| }, | |
| { | |
| "epoch": 1.4180672268907564, | |
| "grad_norm": 0.5858554643624372, | |
| "learning_rate": 3.873707987802967e-05, | |
| "loss": 0.3208, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.304246723651886, | |
| "step": 1350, | |
| "valid_targets_mean": 3483.2, | |
| "valid_targets_min": 1372 | |
| }, | |
| { | |
| "epoch": 1.4233193277310925, | |
| "grad_norm": 0.6612075948640241, | |
| "learning_rate": 3.871869512856179e-05, | |
| "loss": 0.3129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3278389275074005, | |
| "step": 1355, | |
| "valid_targets_mean": 3186.8, | |
| "valid_targets_min": 1769 | |
| }, | |
| { | |
| "epoch": 1.4285714285714286, | |
| "grad_norm": 0.6552998713136254, | |
| "learning_rate": 3.870018195486138e-05, | |
| "loss": 0.3158, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31067150831222534, | |
| "step": 1360, | |
| "valid_targets_mean": 3407.6, | |
| "valid_targets_min": 1074 | |
| }, | |
| { | |
| "epoch": 1.4338235294117647, | |
| "grad_norm": 0.5966743275691717, | |
| "learning_rate": 3.868154048394262e-05, | |
| "loss": 0.3239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3010474741458893, | |
| "step": 1365, | |
| "valid_targets_mean": 3796.4, | |
| "valid_targets_min": 980 | |
| }, | |
| { | |
| "epoch": 1.4390756302521008, | |
| "grad_norm": 0.7359011060228771, | |
| "learning_rate": 3.8662770843699944e-05, | |
| "loss": 0.3306, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3496723771095276, | |
| "step": 1370, | |
| "valid_targets_mean": 2808.1, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 1.444327731092437, | |
| "grad_norm": 0.65437369178061, | |
| "learning_rate": 3.8643873162907086e-05, | |
| "loss": 0.3076, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3403715491294861, | |
| "step": 1375, | |
| "valid_targets_mean": 3677.5, | |
| "valid_targets_min": 1471 | |
| }, | |
| { | |
| "epoch": 1.449579831932773, | |
| "grad_norm": 0.6531132095155441, | |
| "learning_rate": 3.862484757121627e-05, | |
| "loss": 0.3079, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28045761585235596, | |
| "step": 1380, | |
| "valid_targets_mean": 3032.6, | |
| "valid_targets_min": 1353 | |
| }, | |
| { | |
| "epoch": 1.4548319327731092, | |
| "grad_norm": 0.7024121158190606, | |
| "learning_rate": 3.860569419915727e-05, | |
| "loss": 0.3108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31772103905677795, | |
| "step": 1385, | |
| "valid_targets_mean": 2601.9, | |
| "valid_targets_min": 1084 | |
| }, | |
| { | |
| "epoch": 1.4600840336134453, | |
| "grad_norm": 0.6174258612511311, | |
| "learning_rate": 3.858641317813653e-05, | |
| "loss": 0.303, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3092023730278015, | |
| "step": 1390, | |
| "valid_targets_mean": 3280.8, | |
| "valid_targets_min": 1428 | |
| }, | |
| { | |
| "epoch": 1.4653361344537816, | |
| "grad_norm": 0.6773260761366405, | |
| "learning_rate": 3.8567004640436234e-05, | |
| "loss": 0.2957, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30377262830734253, | |
| "step": 1395, | |
| "valid_targets_mean": 3229.4, | |
| "valid_targets_min": 1193 | |
| }, | |
| { | |
| "epoch": 1.4705882352941178, | |
| "grad_norm": 0.6442894382965205, | |
| "learning_rate": 3.8547468719213476e-05, | |
| "loss": 0.303, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2951875925064087, | |
| "step": 1400, | |
| "valid_targets_mean": 3519.2, | |
| "valid_targets_min": 1614 | |
| }, | |
| { | |
| "epoch": 1.4758403361344539, | |
| "grad_norm": 0.6233027503211065, | |
| "learning_rate": 3.852780554849925e-05, | |
| "loss": 0.296, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2837674617767334, | |
| "step": 1405, | |
| "valid_targets_mean": 3699.1, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 1.48109243697479, | |
| "grad_norm": 0.6558871789844729, | |
| "learning_rate": 3.850801526319761e-05, | |
| "loss": 0.3004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.277635395526886, | |
| "step": 1410, | |
| "valid_targets_mean": 2730.4, | |
| "valid_targets_min": 1575 | |
| }, | |
| { | |
| "epoch": 1.486344537815126, | |
| "grad_norm": 0.5993355650257447, | |
| "learning_rate": 3.84880979990847e-05, | |
| "loss": 0.3028, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.289740651845932, | |
| "step": 1415, | |
| "valid_targets_mean": 3345.0, | |
| "valid_targets_min": 1697 | |
| }, | |
| { | |
| "epoch": 1.4915966386554622, | |
| "grad_norm": 0.5660470420295868, | |
| "learning_rate": 3.846805389280782e-05, | |
| "loss": 0.3022, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2910107374191284, | |
| "step": 1420, | |
| "valid_targets_mean": 3602.2, | |
| "valid_targets_min": 1970 | |
| }, | |
| { | |
| "epoch": 1.4968487394957983, | |
| "grad_norm": 0.6727298285320141, | |
| "learning_rate": 3.844788308188452e-05, | |
| "loss": 0.3296, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3241168260574341, | |
| "step": 1425, | |
| "valid_targets_mean": 3128.4, | |
| "valid_targets_min": 887 | |
| }, | |
| { | |
| "epoch": 1.5021008403361344, | |
| "grad_norm": 0.7266333636248493, | |
| "learning_rate": 3.8427585704701634e-05, | |
| "loss": 0.3085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3230634331703186, | |
| "step": 1430, | |
| "valid_targets_mean": 2712.2, | |
| "valid_targets_min": 1279 | |
| }, | |
| { | |
| "epoch": 1.5073529411764706, | |
| "grad_norm": 0.625826335328431, | |
| "learning_rate": 3.840716190051433e-05, | |
| "loss": 0.2988, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28237617015838623, | |
| "step": 1435, | |
| "valid_targets_mean": 3052.6, | |
| "valid_targets_min": 1021 | |
| }, | |
| { | |
| "epoch": 1.5126050420168067, | |
| "grad_norm": 0.643231670297035, | |
| "learning_rate": 3.8386611809445155e-05, | |
| "loss": 0.3098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30906054377555847, | |
| "step": 1440, | |
| "valid_targets_mean": 3352.1, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 1.5178571428571428, | |
| "grad_norm": 0.7439322282160008, | |
| "learning_rate": 3.8365935572483095e-05, | |
| "loss": 0.3074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3927016258239746, | |
| "step": 1445, | |
| "valid_targets_mean": 2860.6, | |
| "valid_targets_min": 1039 | |
| }, | |
| { | |
| "epoch": 1.523109243697479, | |
| "grad_norm": 0.5906966913938343, | |
| "learning_rate": 3.834513333148258e-05, | |
| "loss": 0.3107, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29975053668022156, | |
| "step": 1450, | |
| "valid_targets_mean": 3742.9, | |
| "valid_targets_min": 2459 | |
| }, | |
| { | |
| "epoch": 1.528361344537815, | |
| "grad_norm": 0.5784262635567486, | |
| "learning_rate": 3.832420522916252e-05, | |
| "loss": 0.3133, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27920496463775635, | |
| "step": 1455, | |
| "valid_targets_mean": 3435.8, | |
| "valid_targets_min": 1276 | |
| }, | |
| { | |
| "epoch": 1.5336134453781511, | |
| "grad_norm": 0.5304379397196218, | |
| "learning_rate": 3.830315140910534e-05, | |
| "loss": 0.3032, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2916795015335083, | |
| "step": 1460, | |
| "valid_targets_mean": 4335.8, | |
| "valid_targets_min": 1572 | |
| }, | |
| { | |
| "epoch": 1.5388655462184873, | |
| "grad_norm": 0.690684220200614, | |
| "learning_rate": 3.8281972015755965e-05, | |
| "loss": 0.309, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32439422607421875, | |
| "step": 1465, | |
| "valid_targets_mean": 2975.5, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 1.5441176470588234, | |
| "grad_norm": 0.9031172365787135, | |
| "learning_rate": 3.826066719442086e-05, | |
| "loss": 0.2345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22157755494117737, | |
| "step": 1470, | |
| "valid_targets_mean": 1444.9, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 1.5493697478991597, | |
| "grad_norm": 0.8412510599148572, | |
| "learning_rate": 3.823923709126701e-05, | |
| "loss": 0.2161, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22126498818397522, | |
| "step": 1475, | |
| "valid_targets_mean": 1456.5, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 1.5546218487394958, | |
| "grad_norm": 1.1541206078120574, | |
| "learning_rate": 3.821768185332095e-05, | |
| "loss": 0.2097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2107081413269043, | |
| "step": 1480, | |
| "valid_targets_mean": 1432.6, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 1.559873949579832, | |
| "grad_norm": 0.8048779001890545, | |
| "learning_rate": 3.81960016284677e-05, | |
| "loss": 0.2089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.192092627286911, | |
| "step": 1485, | |
| "valid_targets_mean": 1432.6, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 1.565126050420168, | |
| "grad_norm": 0.8091682143521767, | |
| "learning_rate": 3.817419656544979e-05, | |
| "loss": 0.2141, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21445538103580475, | |
| "step": 1490, | |
| "valid_targets_mean": 1387.0, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 1.5703781512605042, | |
| "grad_norm": 0.8028634331257293, | |
| "learning_rate": 3.815226681386626e-05, | |
| "loss": 0.2181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21761010587215424, | |
| "step": 1495, | |
| "valid_targets_mean": 1527.4, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 1.5756302521008403, | |
| "grad_norm": 0.7439177233189905, | |
| "learning_rate": 3.8130212524171576e-05, | |
| "loss": 0.2007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19787126779556274, | |
| "step": 1500, | |
| "valid_targets_mean": 1353.6, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 1.5808823529411766, | |
| "grad_norm": 0.8142970979200826, | |
| "learning_rate": 3.810803384767465e-05, | |
| "loss": 0.2207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1932760328054428, | |
| "step": 1505, | |
| "valid_targets_mean": 1309.9, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 1.5861344537815127, | |
| "grad_norm": 0.7340679359153581, | |
| "learning_rate": 3.808573093653777e-05, | |
| "loss": 0.2028, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19831883907318115, | |
| "step": 1510, | |
| "valid_targets_mean": 1465.4, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 1.5913865546218489, | |
| "grad_norm": 0.7860661366128959, | |
| "learning_rate": 3.806330394377556e-05, | |
| "loss": 0.2069, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19419294595718384, | |
| "step": 1515, | |
| "valid_targets_mean": 1379.4, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 1.596638655462185, | |
| "grad_norm": 0.8568638447176881, | |
| "learning_rate": 3.8040753023253956e-05, | |
| "loss": 0.2049, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2107909917831421, | |
| "step": 1520, | |
| "valid_targets_mean": 1158.2, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 1.601890756302521, | |
| "grad_norm": 0.6995773260097756, | |
| "learning_rate": 3.801807832968912e-05, | |
| "loss": 0.2085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1988203525543213, | |
| "step": 1525, | |
| "valid_targets_mean": 1543.1, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 1.6071428571428572, | |
| "grad_norm": 0.7456785948304131, | |
| "learning_rate": 3.799528001864637e-05, | |
| "loss": 0.1984, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17537815868854523, | |
| "step": 1530, | |
| "valid_targets_mean": 1300.1, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 1.6123949579831933, | |
| "grad_norm": 0.8087741971757628, | |
| "learning_rate": 3.797235824653918e-05, | |
| "loss": 0.201, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20719020068645477, | |
| "step": 1535, | |
| "valid_targets_mean": 1401.3, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 1.6176470588235294, | |
| "grad_norm": 0.7888860351461294, | |
| "learning_rate": 3.7949313170628006e-05, | |
| "loss": 0.1935, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2095479965209961, | |
| "step": 1540, | |
| "valid_targets_mean": 1436.8, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 1.6228991596638656, | |
| "grad_norm": 0.767345470182285, | |
| "learning_rate": 3.79261449490193e-05, | |
| "loss": 0.2106, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22027720510959625, | |
| "step": 1545, | |
| "valid_targets_mean": 1531.6, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 1.6281512605042017, | |
| "grad_norm": 0.8137805205710857, | |
| "learning_rate": 3.7902853740664356e-05, | |
| "loss": 0.2073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21532106399536133, | |
| "step": 1550, | |
| "valid_targets_mean": 1469.6, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 1.6334033613445378, | |
| "grad_norm": 0.9016471374390324, | |
| "learning_rate": 3.7879439705358286e-05, | |
| "loss": 0.2139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23087842762470245, | |
| "step": 1555, | |
| "valid_targets_mean": 1437.8, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 1.638655462184874, | |
| "grad_norm": 0.8353128948745288, | |
| "learning_rate": 3.785590300373884e-05, | |
| "loss": 0.2004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2092447727918625, | |
| "step": 1560, | |
| "valid_targets_mean": 1327.3, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 1.64390756302521, | |
| "grad_norm": 0.868105384855138, | |
| "learning_rate": 3.7832243797285385e-05, | |
| "loss": 0.203, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21071121096611023, | |
| "step": 1565, | |
| "valid_targets_mean": 1685.2, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 1.6491596638655461, | |
| "grad_norm": 0.7919143845079746, | |
| "learning_rate": 3.780846224831775e-05, | |
| "loss": 0.2109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2265183925628662, | |
| "step": 1570, | |
| "valid_targets_mean": 1595.3, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 1.6544117647058822, | |
| "grad_norm": 0.7549544772773371, | |
| "learning_rate": 3.7784558519995135e-05, | |
| "loss": 0.203, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18298515677452087, | |
| "step": 1575, | |
| "valid_targets_mean": 1381.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 1.6596638655462184, | |
| "grad_norm": 0.804144396397427, | |
| "learning_rate": 3.776053277631496e-05, | |
| "loss": 0.197, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1860344111919403, | |
| "step": 1580, | |
| "valid_targets_mean": 1345.7, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 1.6649159663865545, | |
| "grad_norm": 0.9064506425487462, | |
| "learning_rate": 3.7736385182111785e-05, | |
| "loss": 0.1997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20600754022598267, | |
| "step": 1585, | |
| "valid_targets_mean": 1342.1, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 1.6701680672268906, | |
| "grad_norm": 0.8588049760182106, | |
| "learning_rate": 3.771211590305614e-05, | |
| "loss": 0.1956, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1966136246919632, | |
| "step": 1590, | |
| "valid_targets_mean": 1450.6, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 1.675420168067227, | |
| "grad_norm": 0.7948243288514434, | |
| "learning_rate": 3.768772510565342e-05, | |
| "loss": 0.1987, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2040775567293167, | |
| "step": 1595, | |
| "valid_targets_mean": 1461.3, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 1.680672268907563, | |
| "grad_norm": 0.8255403927136926, | |
| "learning_rate": 3.76632129572427e-05, | |
| "loss": 0.1992, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20673060417175293, | |
| "step": 1600, | |
| "valid_targets_mean": 1723.1, | |
| "valid_targets_min": 1075 | |
| }, | |
| { | |
| "epoch": 1.6859243697478992, | |
| "grad_norm": 0.7484827437017396, | |
| "learning_rate": 3.763857962599565e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20060554146766663, | |
| "step": 1605, | |
| "valid_targets_mean": 1669.1, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 1.6911764705882353, | |
| "grad_norm": 0.7780511290517912, | |
| "learning_rate": 3.761382528091531e-05, | |
| "loss": 0.1956, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21109002828598022, | |
| "step": 1610, | |
| "valid_targets_mean": 1521.4, | |
| "valid_targets_min": 957 | |
| }, | |
| { | |
| "epoch": 1.6964285714285714, | |
| "grad_norm": 0.8261801428366944, | |
| "learning_rate": 3.7588950091834986e-05, | |
| "loss": 0.1987, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18024760484695435, | |
| "step": 1615, | |
| "valid_targets_mean": 1282.1, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 1.7016806722689075, | |
| "grad_norm": 0.7650069206711103, | |
| "learning_rate": 3.756395422941706e-05, | |
| "loss": 0.2052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20178046822547913, | |
| "step": 1620, | |
| "valid_targets_mean": 1499.7, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 1.7069327731092439, | |
| "grad_norm": 0.7087222035820604, | |
| "learning_rate": 3.753883786515181e-05, | |
| "loss": 0.1919, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19103536009788513, | |
| "step": 1625, | |
| "valid_targets_mean": 1621.1, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 1.71218487394958, | |
| "grad_norm": 0.8887563433958335, | |
| "learning_rate": 3.751360117135628e-05, | |
| "loss": 0.1956, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20276349782943726, | |
| "step": 1630, | |
| "valid_targets_mean": 1390.9, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 1.717436974789916, | |
| "grad_norm": 0.8362395348050953, | |
| "learning_rate": 3.7488244321173025e-05, | |
| "loss": 0.1854, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18552342057228088, | |
| "step": 1635, | |
| "valid_targets_mean": 1254.4, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 1.7226890756302522, | |
| "grad_norm": 0.8655931910128335, | |
| "learning_rate": 3.746276748856898e-05, | |
| "loss": 0.1951, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18865495920181274, | |
| "step": 1640, | |
| "valid_targets_mean": 1479.2, | |
| "valid_targets_min": 817 | |
| }, | |
| { | |
| "epoch": 1.7279411764705883, | |
| "grad_norm": 0.792892975194453, | |
| "learning_rate": 3.743717084833425e-05, | |
| "loss": 0.1969, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19585290551185608, | |
| "step": 1645, | |
| "valid_targets_mean": 1464.4, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 1.7331932773109244, | |
| "grad_norm": 0.779810593913834, | |
| "learning_rate": 3.741145457608093e-05, | |
| "loss": 0.1966, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20523177087306976, | |
| "step": 1650, | |
| "valid_targets_mean": 1381.2, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 1.7384453781512605, | |
| "grad_norm": 0.7826687066418191, | |
| "learning_rate": 3.738561884824183e-05, | |
| "loss": 0.1997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20436634123325348, | |
| "step": 1655, | |
| "valid_targets_mean": 1486.0, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 1.7436974789915967, | |
| "grad_norm": 0.7652636064297311, | |
| "learning_rate": 3.735966384206936e-05, | |
| "loss": 0.1986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20646733045578003, | |
| "step": 1660, | |
| "valid_targets_mean": 1459.7, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 1.7489495798319328, | |
| "grad_norm": 0.7655125678378137, | |
| "learning_rate": 3.733358973563425e-05, | |
| "loss": 0.1921, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18100708723068237, | |
| "step": 1665, | |
| "valid_targets_mean": 1397.0, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 1.754201680672269, | |
| "grad_norm": 0.7963580538812527, | |
| "learning_rate": 3.730739670782435e-05, | |
| "loss": 0.1914, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19578437507152557, | |
| "step": 1670, | |
| "valid_targets_mean": 1288.8, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 1.759453781512605, | |
| "grad_norm": 0.8662683116944544, | |
| "learning_rate": 3.72810849383434e-05, | |
| "loss": 0.2051, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18982970714569092, | |
| "step": 1675, | |
| "valid_targets_mean": 1446.1, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 1.7647058823529411, | |
| "grad_norm": 0.783041949999934, | |
| "learning_rate": 3.725465460770978e-05, | |
| "loss": 0.209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25431910157203674, | |
| "step": 1680, | |
| "valid_targets_mean": 1673.2, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 1.7699579831932772, | |
| "grad_norm": 0.7217643791824649, | |
| "learning_rate": 3.7228105897255324e-05, | |
| "loss": 0.195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2004881650209427, | |
| "step": 1685, | |
| "valid_targets_mean": 1598.7, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 1.7752100840336134, | |
| "grad_norm": 0.7436378042988409, | |
| "learning_rate": 3.7201438989124e-05, | |
| "loss": 0.1917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1999681144952774, | |
| "step": 1690, | |
| "valid_targets_mean": 1445.2, | |
| "valid_targets_min": 551 | |
| }, | |
| { | |
| "epoch": 1.7804621848739495, | |
| "grad_norm": 0.7501915605313334, | |
| "learning_rate": 3.717465406627074e-05, | |
| "loss": 0.2001, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20576688647270203, | |
| "step": 1695, | |
| "valid_targets_mean": 1552.2, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 1.7857142857142856, | |
| "grad_norm": 0.7628246639482904, | |
| "learning_rate": 3.714775131246011e-05, | |
| "loss": 0.2009, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20440012216567993, | |
| "step": 1700, | |
| "valid_targets_mean": 1511.6, | |
| "valid_targets_min": 825 | |
| }, | |
| { | |
| "epoch": 1.7909663865546217, | |
| "grad_norm": 0.7246985646816825, | |
| "learning_rate": 3.71207309122651e-05, | |
| "loss": 0.1889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20958301424980164, | |
| "step": 1705, | |
| "valid_targets_mean": 1489.3, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 1.7962184873949578, | |
| "grad_norm": 0.8076167406390439, | |
| "learning_rate": 3.709359305106585e-05, | |
| "loss": 0.1866, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18848812580108643, | |
| "step": 1710, | |
| "valid_targets_mean": 1356.9, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 1.8014705882352942, | |
| "grad_norm": 0.7618349813747334, | |
| "learning_rate": 3.7066337915048354e-05, | |
| "loss": 0.201, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19808264076709747, | |
| "step": 1715, | |
| "valid_targets_mean": 1359.9, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 1.8067226890756303, | |
| "grad_norm": 0.8211165193393286, | |
| "learning_rate": 3.7038965691203205e-05, | |
| "loss": 0.2016, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20665648579597473, | |
| "step": 1720, | |
| "valid_targets_mean": 1443.4, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 1.8119747899159664, | |
| "grad_norm": 1.5195639624253412, | |
| "learning_rate": 3.701147656732431e-05, | |
| "loss": 0.1928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18661421537399292, | |
| "step": 1725, | |
| "valid_targets_mean": 1376.3, | |
| "valid_targets_min": 944 | |
| }, | |
| { | |
| "epoch": 1.8172268907563025, | |
| "grad_norm": 1.3672377372582318, | |
| "learning_rate": 3.6983870732007596e-05, | |
| "loss": 0.1883, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.171901673078537, | |
| "step": 1730, | |
| "valid_targets_mean": 1597.7, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 1.8224789915966386, | |
| "grad_norm": 0.7029694603857538, | |
| "learning_rate": 3.695614837464972e-05, | |
| "loss": 0.2015, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18585661053657532, | |
| "step": 1735, | |
| "valid_targets_mean": 1419.2, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 1.8277310924369747, | |
| "grad_norm": 0.7128079074101666, | |
| "learning_rate": 3.692830968544675e-05, | |
| "loss": 0.192, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1792781949043274, | |
| "step": 1740, | |
| "valid_targets_mean": 1570.5, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 1.832983193277311, | |
| "grad_norm": 0.7873525110103127, | |
| "learning_rate": 3.690035485539291e-05, | |
| "loss": 0.1863, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18637944757938385, | |
| "step": 1745, | |
| "valid_targets_mean": 1177.1, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 1.8382352941176472, | |
| "grad_norm": 0.812792057458698, | |
| "learning_rate": 3.6872284076279205e-05, | |
| "loss": 0.1911, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18359485268592834, | |
| "step": 1750, | |
| "valid_targets_mean": 1280.1, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 1.8434873949579833, | |
| "grad_norm": 0.7516261504287793, | |
| "learning_rate": 3.684409754069215e-05, | |
| "loss": 0.1801, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18630534410476685, | |
| "step": 1755, | |
| "valid_targets_mean": 1253.6, | |
| "valid_targets_min": 604 | |
| }, | |
| { | |
| "epoch": 1.8487394957983194, | |
| "grad_norm": 0.7873312358062879, | |
| "learning_rate": 3.681579544201244e-05, | |
| "loss": 0.1939, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1809154450893402, | |
| "step": 1760, | |
| "valid_targets_mean": 1232.8, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 1.8539915966386555, | |
| "grad_norm": 0.7042638950908449, | |
| "learning_rate": 3.6787377974413614e-05, | |
| "loss": 0.1927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1879248321056366, | |
| "step": 1765, | |
| "valid_targets_mean": 1484.8, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 1.8592436974789917, | |
| "grad_norm": 0.8708411897151677, | |
| "learning_rate": 3.6758845332860734e-05, | |
| "loss": 0.1897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2018292248249054, | |
| "step": 1770, | |
| "valid_targets_mean": 1383.6, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 1.8644957983193278, | |
| "grad_norm": 0.727245204127273, | |
| "learning_rate": 3.673019771310903e-05, | |
| "loss": 0.1971, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19675494730472565, | |
| "step": 1775, | |
| "valid_targets_mean": 1386.2, | |
| "valid_targets_min": 840 | |
| }, | |
| { | |
| "epoch": 1.8697478991596639, | |
| "grad_norm": 0.769303157042911, | |
| "learning_rate": 3.670143531170258e-05, | |
| "loss": 0.1945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.200996533036232, | |
| "step": 1780, | |
| "valid_targets_mean": 1375.6, | |
| "valid_targets_min": 772 | |
| }, | |
| { | |
| "epoch": 1.875, | |
| "grad_norm": 0.6945199303988268, | |
| "learning_rate": 3.667255832597294e-05, | |
| "loss": 0.1922, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19052547216415405, | |
| "step": 1785, | |
| "valid_targets_mean": 1612.4, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 1.8802521008403361, | |
| "grad_norm": 0.7679602226980033, | |
| "learning_rate": 3.664356695403781e-05, | |
| "loss": 0.1919, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18414106965065002, | |
| "step": 1790, | |
| "valid_targets_mean": 1427.0, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 1.8855042016806722, | |
| "grad_norm": 0.7707612142751945, | |
| "learning_rate": 3.661446139479965e-05, | |
| "loss": 0.1971, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20061108469963074, | |
| "step": 1795, | |
| "valid_targets_mean": 1479.3, | |
| "valid_targets_min": 766 | |
| }, | |
| { | |
| "epoch": 1.8907563025210083, | |
| "grad_norm": 0.8561024756443094, | |
| "learning_rate": 3.658524184794436e-05, | |
| "loss": 0.1953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16889707744121552, | |
| "step": 1800, | |
| "valid_targets_mean": 1241.5, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 1.8960084033613445, | |
| "grad_norm": 0.7618847795685194, | |
| "learning_rate": 3.655590851393984e-05, | |
| "loss": 0.1998, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20724733173847198, | |
| "step": 1805, | |
| "valid_targets_mean": 1410.9, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 1.9012605042016806, | |
| "grad_norm": 0.7752684468128507, | |
| "learning_rate": 3.652646159403468e-05, | |
| "loss": 0.206, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20427414774894714, | |
| "step": 1810, | |
| "valid_targets_mean": 1569.1, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 1.9065126050420167, | |
| "grad_norm": 0.7353501689368992, | |
| "learning_rate": 3.649690129025675e-05, | |
| "loss": 0.1986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2081902176141739, | |
| "step": 1815, | |
| "valid_targets_mean": 1378.0, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 1.9117647058823528, | |
| "grad_norm": 0.7289002335110303, | |
| "learning_rate": 3.6467227805411824e-05, | |
| "loss": 0.1979, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21570971608161926, | |
| "step": 1820, | |
| "valid_targets_mean": 1620.0, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 1.917016806722689, | |
| "grad_norm": 0.7240815176489389, | |
| "learning_rate": 3.6437441343082174e-05, | |
| "loss": 0.1885, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18155962228775024, | |
| "step": 1825, | |
| "valid_targets_mean": 1355.1, | |
| "valid_targets_min": 564 | |
| }, | |
| { | |
| "epoch": 1.9222689075630253, | |
| "grad_norm": 0.7462727456723074, | |
| "learning_rate": 3.6407542107625167e-05, | |
| "loss": 0.189, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18159963190555573, | |
| "step": 1830, | |
| "valid_targets_mean": 1419.1, | |
| "valid_targets_min": 651 | |
| }, | |
| { | |
| "epoch": 1.9275210084033614, | |
| "grad_norm": 0.9386783714270683, | |
| "learning_rate": 3.637753030417192e-05, | |
| "loss": 0.2, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18306277692317963, | |
| "step": 1835, | |
| "valid_targets_mean": 1291.7, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 1.9327731092436975, | |
| "grad_norm": 0.725039331848806, | |
| "learning_rate": 3.6347406138625805e-05, | |
| "loss": 0.1896, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17465978860855103, | |
| "step": 1840, | |
| "valid_targets_mean": 1236.8, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 1.9380252100840336, | |
| "grad_norm": 0.7526079231589902, | |
| "learning_rate": 3.631716981766111e-05, | |
| "loss": 0.186, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16068580746650696, | |
| "step": 1845, | |
| "valid_targets_mean": 1258.9, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 1.9432773109243697, | |
| "grad_norm": 0.7599278884583646, | |
| "learning_rate": 3.6286821548721594e-05, | |
| "loss": 0.1901, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1918371468782425, | |
| "step": 1850, | |
| "valid_targets_mean": 1495.5, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 1.9485294117647058, | |
| "grad_norm": 0.7979519291571235, | |
| "learning_rate": 3.625636154001904e-05, | |
| "loss": 0.2073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21574504673480988, | |
| "step": 1855, | |
| "valid_targets_mean": 1421.0, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 1.9537815126050422, | |
| "grad_norm": 0.7338015635554208, | |
| "learning_rate": 3.622579000053186e-05, | |
| "loss": 0.1853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20440888404846191, | |
| "step": 1860, | |
| "valid_targets_mean": 1487.2, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 1.9590336134453783, | |
| "grad_norm": 0.7129631803753865, | |
| "learning_rate": 3.619510714000368e-05, | |
| "loss": 0.1916, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2010660171508789, | |
| "step": 1865, | |
| "valid_targets_mean": 1512.4, | |
| "valid_targets_min": 823 | |
| }, | |
| { | |
| "epoch": 1.9642857142857144, | |
| "grad_norm": 0.6958574230399347, | |
| "learning_rate": 3.616431316894181e-05, | |
| "loss": 0.1671, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15293735265731812, | |
| "step": 1870, | |
| "valid_targets_mean": 1375.4, | |
| "valid_targets_min": 595 | |
| }, | |
| { | |
| "epoch": 1.9695378151260505, | |
| "grad_norm": 0.8984376849635196, | |
| "learning_rate": 3.613340829861592e-05, | |
| "loss": 0.1902, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2001931369304657, | |
| "step": 1875, | |
| "valid_targets_mean": 1401.2, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 1.9747899159663866, | |
| "grad_norm": 0.796278651195792, | |
| "learning_rate": 3.61023927410565e-05, | |
| "loss": 0.1886, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18896615505218506, | |
| "step": 1880, | |
| "valid_targets_mean": 1185.6, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 1.9800420168067228, | |
| "grad_norm": 0.6919750871597347, | |
| "learning_rate": 3.607126670905346e-05, | |
| "loss": 0.1871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16711284220218658, | |
| "step": 1885, | |
| "valid_targets_mean": 1364.2, | |
| "valid_targets_min": 685 | |
| }, | |
| { | |
| "epoch": 1.9852941176470589, | |
| "grad_norm": 0.6805728110467865, | |
| "learning_rate": 3.60400304161546e-05, | |
| "loss": 0.1838, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17011727392673492, | |
| "step": 1890, | |
| "valid_targets_mean": 1345.2, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 1.990546218487395, | |
| "grad_norm": 0.7601178723938317, | |
| "learning_rate": 3.600868407666426e-05, | |
| "loss": 0.1895, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18218779563903809, | |
| "step": 1895, | |
| "valid_targets_mean": 1431.4, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 1.995798319327731, | |
| "grad_norm": 0.6963519557748621, | |
| "learning_rate": 3.597722790564172e-05, | |
| "loss": 0.1984, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19570836424827576, | |
| "step": 1900, | |
| "valid_targets_mean": 1505.8, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 2.0010504201680672, | |
| "grad_norm": 0.8543332754648679, | |
| "learning_rate": 3.5945662118899824e-05, | |
| "loss": 0.2233, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3345944881439209, | |
| "step": 1905, | |
| "valid_targets_mean": 2493.1, | |
| "valid_targets_min": 954 | |
| }, | |
| { | |
| "epoch": 2.0063025210084033, | |
| "grad_norm": 0.6540945057835557, | |
| "learning_rate": 3.5913986933003455e-05, | |
| "loss": 0.3021, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2893906235694885, | |
| "step": 1910, | |
| "valid_targets_mean": 3336.9, | |
| "valid_targets_min": 533 | |
| }, | |
| { | |
| "epoch": 2.0115546218487395, | |
| "grad_norm": 0.5859587843451456, | |
| "learning_rate": 3.5882202565268046e-05, | |
| "loss": 0.2897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30491194128990173, | |
| "step": 1915, | |
| "valid_targets_mean": 3515.8, | |
| "valid_targets_min": 1635 | |
| }, | |
| { | |
| "epoch": 2.0168067226890756, | |
| "grad_norm": 0.5848592413314749, | |
| "learning_rate": 3.58503092337581e-05, | |
| "loss": 0.2927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3176591992378235, | |
| "step": 1920, | |
| "valid_targets_mean": 3744.6, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 2.0220588235294117, | |
| "grad_norm": 0.5277378066045798, | |
| "learning_rate": 3.5818307157285705e-05, | |
| "loss": 0.2788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2889269292354584, | |
| "step": 1925, | |
| "valid_targets_mean": 4188.5, | |
| "valid_targets_min": 2187 | |
| }, | |
| { | |
| "epoch": 2.027310924369748, | |
| "grad_norm": 0.6598583479222365, | |
| "learning_rate": 3.5786196555409006e-05, | |
| "loss": 0.2979, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30424559116363525, | |
| "step": 1930, | |
| "valid_targets_mean": 3188.1, | |
| "valid_targets_min": 1432 | |
| }, | |
| { | |
| "epoch": 2.032563025210084, | |
| "grad_norm": 0.6035540307689007, | |
| "learning_rate": 3.575397764843073e-05, | |
| "loss": 0.2808, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2632938027381897, | |
| "step": 1935, | |
| "valid_targets_mean": 3703.5, | |
| "valid_targets_min": 767 | |
| }, | |
| { | |
| "epoch": 2.03781512605042, | |
| "grad_norm": 0.6545986679555346, | |
| "learning_rate": 3.572165065739664e-05, | |
| "loss": 0.2807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2869631052017212, | |
| "step": 1940, | |
| "valid_targets_mean": 3793.3, | |
| "valid_targets_min": 1488 | |
| }, | |
| { | |
| "epoch": 2.043067226890756, | |
| "grad_norm": 0.5266355858459322, | |
| "learning_rate": 3.5689215804094053e-05, | |
| "loss": 0.2847, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2697930932044983, | |
| "step": 1945, | |
| "valid_targets_mean": 3814.2, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 2.0483193277310923, | |
| "grad_norm": 0.5529414922643042, | |
| "learning_rate": 3.5656673311050285e-05, | |
| "loss": 0.3058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.298675537109375, | |
| "step": 1950, | |
| "valid_targets_mean": 3997.1, | |
| "valid_targets_min": 1241 | |
| }, | |
| { | |
| "epoch": 2.0535714285714284, | |
| "grad_norm": 0.5889438843034916, | |
| "learning_rate": 3.562402340153115e-05, | |
| "loss": 0.2915, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3097384572029114, | |
| "step": 1955, | |
| "valid_targets_mean": 3797.0, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 2.0588235294117645, | |
| "grad_norm": 0.4888383801769635, | |
| "learning_rate": 3.559126629953942e-05, | |
| "loss": 0.2944, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2800885736942291, | |
| "step": 1960, | |
| "valid_targets_mean": 5091.3, | |
| "valid_targets_min": 2061 | |
| }, | |
| { | |
| "epoch": 2.064075630252101, | |
| "grad_norm": 0.5549566154812772, | |
| "learning_rate": 3.5558402229813287e-05, | |
| "loss": 0.2964, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28934812545776367, | |
| "step": 1965, | |
| "valid_targets_mean": 4230.9, | |
| "valid_targets_min": 1329 | |
| }, | |
| { | |
| "epoch": 2.069327731092437, | |
| "grad_norm": 0.5836584192752005, | |
| "learning_rate": 3.552543141782483e-05, | |
| "loss": 0.2831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2896938621997833, | |
| "step": 1970, | |
| "valid_targets_mean": 3577.5, | |
| "valid_targets_min": 1412 | |
| }, | |
| { | |
| "epoch": 2.0745798319327733, | |
| "grad_norm": 0.6187434500702056, | |
| "learning_rate": 3.549235408977843e-05, | |
| "loss": 0.2955, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2899538576602936, | |
| "step": 1975, | |
| "valid_targets_mean": 3247.0, | |
| "valid_targets_min": 1334 | |
| }, | |
| { | |
| "epoch": 2.0798319327731094, | |
| "grad_norm": 0.5443173852803067, | |
| "learning_rate": 3.545917047260928e-05, | |
| "loss": 0.3036, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27411580085754395, | |
| "step": 1980, | |
| "valid_targets_mean": 4207.9, | |
| "valid_targets_min": 1847 | |
| }, | |
| { | |
| "epoch": 2.0850840336134455, | |
| "grad_norm": 0.6049544183292745, | |
| "learning_rate": 3.542588079398179e-05, | |
| "loss": 0.2962, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2940400540828705, | |
| "step": 1985, | |
| "valid_targets_mean": 3505.8, | |
| "valid_targets_min": 1040 | |
| }, | |
| { | |
| "epoch": 2.0903361344537816, | |
| "grad_norm": 0.6603766972468182, | |
| "learning_rate": 3.539248528228801e-05, | |
| "loss": 0.3076, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3235289454460144, | |
| "step": 1990, | |
| "valid_targets_mean": 3247.1, | |
| "valid_targets_min": 1269 | |
| }, | |
| { | |
| "epoch": 2.0955882352941178, | |
| "grad_norm": 0.6507849332876462, | |
| "learning_rate": 3.535898416664611e-05, | |
| "loss": 0.304, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29736390709877014, | |
| "step": 1995, | |
| "valid_targets_mean": 2779.6, | |
| "valid_targets_min": 1041 | |
| }, | |
| { | |
| "epoch": 2.100840336134454, | |
| "grad_norm": 0.6263414769134444, | |
| "learning_rate": 3.5325377676898766e-05, | |
| "loss": 0.2963, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30523204803466797, | |
| "step": 2000, | |
| "valid_targets_mean": 2963.0, | |
| "valid_targets_min": 1291 | |
| }, | |
| { | |
| "epoch": 2.10609243697479, | |
| "grad_norm": 0.7168587293071478, | |
| "learning_rate": 3.529166604361161e-05, | |
| "loss": 0.2933, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3294321298599243, | |
| "step": 2005, | |
| "valid_targets_mean": 2588.5, | |
| "valid_targets_min": 1285 | |
| }, | |
| { | |
| "epoch": 2.111344537815126, | |
| "grad_norm": 0.7408150111837036, | |
| "learning_rate": 3.525784949807162e-05, | |
| "loss": 0.2856, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28167465329170227, | |
| "step": 2010, | |
| "valid_targets_mean": 3554.5, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 2.116596638655462, | |
| "grad_norm": 0.6431903275157029, | |
| "learning_rate": 3.522392827228557e-05, | |
| "loss": 0.2898, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2769734561443329, | |
| "step": 2015, | |
| "valid_targets_mean": 2872.8, | |
| "valid_targets_min": 958 | |
| }, | |
| { | |
| "epoch": 2.1218487394957983, | |
| "grad_norm": 0.6159459227356286, | |
| "learning_rate": 3.518990259897841e-05, | |
| "loss": 0.2991, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30407899618148804, | |
| "step": 2020, | |
| "valid_targets_mean": 3317.6, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 2.1271008403361344, | |
| "grad_norm": 0.5523195230692622, | |
| "learning_rate": 3.515577271159167e-05, | |
| "loss": 0.2962, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2792849838733673, | |
| "step": 2025, | |
| "valid_targets_mean": 3849.6, | |
| "valid_targets_min": 1631 | |
| }, | |
| { | |
| "epoch": 2.1323529411764706, | |
| "grad_norm": 0.5889011467855694, | |
| "learning_rate": 3.512153884428188e-05, | |
| "loss": 0.2929, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2551637589931488, | |
| "step": 2030, | |
| "valid_targets_mean": 3065.6, | |
| "valid_targets_min": 908 | |
| }, | |
| { | |
| "epoch": 2.1376050420168067, | |
| "grad_norm": 0.6584415672877822, | |
| "learning_rate": 3.5087201231918944e-05, | |
| "loss": 0.2884, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2937019467353821, | |
| "step": 2035, | |
| "valid_targets_mean": 2962.0, | |
| "valid_targets_min": 1264 | |
| }, | |
| { | |
| "epoch": 2.142857142857143, | |
| "grad_norm": 0.5843888599963277, | |
| "learning_rate": 3.505276011008454e-05, | |
| "loss": 0.2907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29179146885871887, | |
| "step": 2040, | |
| "valid_targets_mean": 3853.1, | |
| "valid_targets_min": 1889 | |
| }, | |
| { | |
| "epoch": 2.148109243697479, | |
| "grad_norm": 0.6018556070826566, | |
| "learning_rate": 3.501821571507048e-05, | |
| "loss": 0.2867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27817726135253906, | |
| "step": 2045, | |
| "valid_targets_mean": 3645.2, | |
| "valid_targets_min": 1075 | |
| }, | |
| { | |
| "epoch": 2.153361344537815, | |
| "grad_norm": 0.7369039817083102, | |
| "learning_rate": 3.498356828387714e-05, | |
| "loss": 0.2908, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28218644857406616, | |
| "step": 2050, | |
| "valid_targets_mean": 2874.9, | |
| "valid_targets_min": 1036 | |
| }, | |
| { | |
| "epoch": 2.158613445378151, | |
| "grad_norm": 0.5740508320378505, | |
| "learning_rate": 3.494881805421176e-05, | |
| "loss": 0.2923, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2825021743774414, | |
| "step": 2055, | |
| "valid_targets_mean": 3808.9, | |
| "valid_targets_min": 1162 | |
| }, | |
| { | |
| "epoch": 2.1638655462184873, | |
| "grad_norm": 0.6677586403458295, | |
| "learning_rate": 3.4913965264486894e-05, | |
| "loss": 0.2876, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2949675917625427, | |
| "step": 2060, | |
| "valid_targets_mean": 3164.9, | |
| "valid_targets_min": 1281 | |
| }, | |
| { | |
| "epoch": 2.1691176470588234, | |
| "grad_norm": 0.583170841394605, | |
| "learning_rate": 3.48790101538187e-05, | |
| "loss": 0.284, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2632118761539459, | |
| "step": 2065, | |
| "valid_targets_mean": 3503.5, | |
| "valid_targets_min": 1841 | |
| }, | |
| { | |
| "epoch": 2.1743697478991595, | |
| "grad_norm": 0.6659128870236808, | |
| "learning_rate": 3.484395296202537e-05, | |
| "loss": 0.3006, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30187827348709106, | |
| "step": 2070, | |
| "valid_targets_mean": 3121.7, | |
| "valid_targets_min": 1629 | |
| }, | |
| { | |
| "epoch": 2.1796218487394956, | |
| "grad_norm": 0.7188591131831868, | |
| "learning_rate": 3.480879392962541e-05, | |
| "loss": 0.2885, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2901453971862793, | |
| "step": 2075, | |
| "valid_targets_mean": 2668.6, | |
| "valid_targets_min": 1254 | |
| }, | |
| { | |
| "epoch": 2.184873949579832, | |
| "grad_norm": 0.7182569920584896, | |
| "learning_rate": 3.477353329783606e-05, | |
| "loss": 0.2927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30515027046203613, | |
| "step": 2080, | |
| "valid_targets_mean": 3032.6, | |
| "valid_targets_min": 1774 | |
| }, | |
| { | |
| "epoch": 2.190126050420168, | |
| "grad_norm": 0.678604840038963, | |
| "learning_rate": 3.4738171308571595e-05, | |
| "loss": 0.2928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2986491322517395, | |
| "step": 2085, | |
| "valid_targets_mean": 3079.8, | |
| "valid_targets_min": 1643 | |
| }, | |
| { | |
| "epoch": 2.1953781512605044, | |
| "grad_norm": 0.6392734138959271, | |
| "learning_rate": 3.470270820444168e-05, | |
| "loss": 0.2875, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3071242570877075, | |
| "step": 2090, | |
| "valid_targets_mean": 3021.6, | |
| "valid_targets_min": 1244 | |
| }, | |
| { | |
| "epoch": 2.2006302521008405, | |
| "grad_norm": 0.585164572193224, | |
| "learning_rate": 3.4667144228749697e-05, | |
| "loss": 0.2878, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2744428515434265, | |
| "step": 2095, | |
| "valid_targets_mean": 3588.4, | |
| "valid_targets_min": 959 | |
| }, | |
| { | |
| "epoch": 2.2058823529411766, | |
| "grad_norm": 0.5843477775486158, | |
| "learning_rate": 3.463147962549109e-05, | |
| "loss": 0.2826, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2655976414680481, | |
| "step": 2100, | |
| "valid_targets_mean": 3404.8, | |
| "valid_targets_min": 2037 | |
| }, | |
| { | |
| "epoch": 2.2111344537815127, | |
| "grad_norm": 0.5703037332493616, | |
| "learning_rate": 3.459571463935168e-05, | |
| "loss": 0.2817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2898704409599304, | |
| "step": 2105, | |
| "valid_targets_mean": 3745.2, | |
| "valid_targets_min": 1168 | |
| }, | |
| { | |
| "epoch": 2.216386554621849, | |
| "grad_norm": 0.6261977462323101, | |
| "learning_rate": 3.455984951570598e-05, | |
| "loss": 0.2819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2655256390571594, | |
| "step": 2110, | |
| "valid_targets_mean": 3479.8, | |
| "valid_targets_min": 1700 | |
| }, | |
| { | |
| "epoch": 2.221638655462185, | |
| "grad_norm": 0.6273371865248328, | |
| "learning_rate": 3.452388450061556e-05, | |
| "loss": 0.2832, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25678104162216187, | |
| "step": 2115, | |
| "valid_targets_mean": 2905.8, | |
| "valid_targets_min": 1261 | |
| }, | |
| { | |
| "epoch": 2.226890756302521, | |
| "grad_norm": 0.6154152263889715, | |
| "learning_rate": 3.448781984082727e-05, | |
| "loss": 0.2983, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2871716022491455, | |
| "step": 2120, | |
| "valid_targets_mean": 3363.2, | |
| "valid_targets_min": 1124 | |
| }, | |
| { | |
| "epoch": 2.232142857142857, | |
| "grad_norm": 0.6237253157214722, | |
| "learning_rate": 3.4451655783771625e-05, | |
| "loss": 0.2874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2972951829433441, | |
| "step": 2125, | |
| "valid_targets_mean": 3210.6, | |
| "valid_targets_min": 1361 | |
| }, | |
| { | |
| "epoch": 2.2373949579831933, | |
| "grad_norm": 0.5944182232171416, | |
| "learning_rate": 3.4415392577561064e-05, | |
| "loss": 0.2727, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26949986815452576, | |
| "step": 2130, | |
| "valid_targets_mean": 2948.4, | |
| "valid_targets_min": 1555 | |
| }, | |
| { | |
| "epoch": 2.2426470588235294, | |
| "grad_norm": 0.6239663761256676, | |
| "learning_rate": 3.4379030470988285e-05, | |
| "loss": 0.2888, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31491410732269287, | |
| "step": 2135, | |
| "valid_targets_mean": 3311.2, | |
| "valid_targets_min": 1739 | |
| }, | |
| { | |
| "epoch": 2.2478991596638656, | |
| "grad_norm": 0.6358976127732551, | |
| "learning_rate": 3.4342569713524496e-05, | |
| "loss": 0.278, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2920129895210266, | |
| "step": 2140, | |
| "valid_targets_mean": 3192.8, | |
| "valid_targets_min": 1278 | |
| }, | |
| { | |
| "epoch": 2.2531512605042017, | |
| "grad_norm": 0.534574837437247, | |
| "learning_rate": 3.4306010555317735e-05, | |
| "loss": 0.2728, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2571936249732971, | |
| "step": 2145, | |
| "valid_targets_mean": 3951.6, | |
| "valid_targets_min": 1522 | |
| }, | |
| { | |
| "epoch": 2.258403361344538, | |
| "grad_norm": 0.6364635502715449, | |
| "learning_rate": 3.4269353247191135e-05, | |
| "loss": 0.2662, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28747859597206116, | |
| "step": 2150, | |
| "valid_targets_mean": 3641.8, | |
| "valid_targets_min": 464 | |
| }, | |
| { | |
| "epoch": 2.263655462184874, | |
| "grad_norm": 0.5854458561363409, | |
| "learning_rate": 3.423259804064122e-05, | |
| "loss": 0.2933, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2922632098197937, | |
| "step": 2155, | |
| "valid_targets_mean": 3598.4, | |
| "valid_targets_min": 1003 | |
| }, | |
| { | |
| "epoch": 2.26890756302521, | |
| "grad_norm": 0.6600515244262581, | |
| "learning_rate": 3.419574518783616e-05, | |
| "loss": 0.2741, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2910274267196655, | |
| "step": 2160, | |
| "valid_targets_mean": 3158.1, | |
| "valid_targets_min": 1632 | |
| }, | |
| { | |
| "epoch": 2.274159663865546, | |
| "grad_norm": 0.6099216472379927, | |
| "learning_rate": 3.4158794941614055e-05, | |
| "loss": 0.2822, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2801975905895233, | |
| "step": 2165, | |
| "valid_targets_mean": 3131.9, | |
| "valid_targets_min": 1256 | |
| }, | |
| { | |
| "epoch": 2.2794117647058822, | |
| "grad_norm": 0.6173217093072781, | |
| "learning_rate": 3.4121747555481216e-05, | |
| "loss": 0.2932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2909969091415405, | |
| "step": 2170, | |
| "valid_targets_mean": 3894.9, | |
| "valid_targets_min": 1336 | |
| }, | |
| { | |
| "epoch": 2.2846638655462184, | |
| "grad_norm": 0.5395127043883584, | |
| "learning_rate": 3.4084603283610365e-05, | |
| "loss": 0.2944, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2728331685066223, | |
| "step": 2175, | |
| "valid_targets_mean": 4360.4, | |
| "valid_targets_min": 1651 | |
| }, | |
| { | |
| "epoch": 2.2899159663865545, | |
| "grad_norm": 0.6332732127944597, | |
| "learning_rate": 3.404736238083897e-05, | |
| "loss": 0.2742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2798515558242798, | |
| "step": 2180, | |
| "valid_targets_mean": 2921.1, | |
| "valid_targets_min": 644 | |
| }, | |
| { | |
| "epoch": 2.2951680672268906, | |
| "grad_norm": 0.7961507640330674, | |
| "learning_rate": 3.401002510266744e-05, | |
| "loss": 0.2707, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26711398363113403, | |
| "step": 2185, | |
| "valid_targets_mean": 2848.1, | |
| "valid_targets_min": 1397 | |
| }, | |
| { | |
| "epoch": 2.3004201680672267, | |
| "grad_norm": 0.6593038634993043, | |
| "learning_rate": 3.397259170525739e-05, | |
| "loss": 0.2726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28789085149765015, | |
| "step": 2190, | |
| "valid_targets_mean": 2673.2, | |
| "valid_targets_min": 1557 | |
| }, | |
| { | |
| "epoch": 2.3056722689075633, | |
| "grad_norm": 0.6958476283104791, | |
| "learning_rate": 3.393506244542991e-05, | |
| "loss": 0.2786, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.292041540145874, | |
| "step": 2195, | |
| "valid_targets_mean": 2930.8, | |
| "valid_targets_min": 1405 | |
| }, | |
| { | |
| "epoch": 2.310924369747899, | |
| "grad_norm": 0.6218694808083172, | |
| "learning_rate": 3.389743758066374e-05, | |
| "loss": 0.2754, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2553059458732605, | |
| "step": 2200, | |
| "valid_targets_mean": 3421.8, | |
| "valid_targets_min": 1042 | |
| }, | |
| { | |
| "epoch": 2.3161764705882355, | |
| "grad_norm": 0.6961575232099882, | |
| "learning_rate": 3.385971736909357e-05, | |
| "loss": 0.2664, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27729037404060364, | |
| "step": 2205, | |
| "valid_targets_mean": 3172.2, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 2.3214285714285716, | |
| "grad_norm": 0.6655059432652326, | |
| "learning_rate": 3.382190206950822e-05, | |
| "loss": 0.2769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27774959802627563, | |
| "step": 2210, | |
| "valid_targets_mean": 3166.3, | |
| "valid_targets_min": 1288 | |
| }, | |
| { | |
| "epoch": 2.3266806722689077, | |
| "grad_norm": 0.5920405087936736, | |
| "learning_rate": 3.3783991941348916e-05, | |
| "loss": 0.2892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2696320414543152, | |
| "step": 2215, | |
| "valid_targets_mean": 3645.1, | |
| "valid_targets_min": 1915 | |
| }, | |
| { | |
| "epoch": 2.331932773109244, | |
| "grad_norm": 0.6019837190284246, | |
| "learning_rate": 3.374598724470745e-05, | |
| "loss": 0.282, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2848033010959625, | |
| "step": 2220, | |
| "valid_targets_mean": 3868.2, | |
| "valid_targets_min": 1501 | |
| }, | |
| { | |
| "epoch": 2.33718487394958, | |
| "grad_norm": 0.6419175889567692, | |
| "learning_rate": 3.370788824032441e-05, | |
| "loss": 0.2706, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27997952699661255, | |
| "step": 2225, | |
| "valid_targets_mean": 2992.5, | |
| "valid_targets_min": 1255 | |
| }, | |
| { | |
| "epoch": 2.342436974789916, | |
| "grad_norm": 0.6429288935922298, | |
| "learning_rate": 3.366969518958746e-05, | |
| "loss": 0.2797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29206526279449463, | |
| "step": 2230, | |
| "valid_targets_mean": 3205.4, | |
| "valid_targets_min": 1586 | |
| }, | |
| { | |
| "epoch": 2.347689075630252, | |
| "grad_norm": 0.5926026217237049, | |
| "learning_rate": 3.363140835452945e-05, | |
| "loss": 0.2777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2542131543159485, | |
| "step": 2235, | |
| "valid_targets_mean": 3410.4, | |
| "valid_targets_min": 1181 | |
| }, | |
| { | |
| "epoch": 2.3529411764705883, | |
| "grad_norm": 0.6805026667731596, | |
| "learning_rate": 3.3593027997826654e-05, | |
| "loss": 0.2666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26993298530578613, | |
| "step": 2240, | |
| "valid_targets_mean": 2876.6, | |
| "valid_targets_min": 1428 | |
| }, | |
| { | |
| "epoch": 2.3581932773109244, | |
| "grad_norm": 0.6149224206718514, | |
| "learning_rate": 3.355455438279702e-05, | |
| "loss": 0.2674, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27489912509918213, | |
| "step": 2245, | |
| "valid_targets_mean": 3850.8, | |
| "valid_targets_min": 1708 | |
| }, | |
| { | |
| "epoch": 2.3634453781512605, | |
| "grad_norm": 0.7148223911895215, | |
| "learning_rate": 3.351598777339827e-05, | |
| "loss": 0.2817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31252795457839966, | |
| "step": 2250, | |
| "valid_targets_mean": 3311.1, | |
| "valid_targets_min": 1171 | |
| }, | |
| { | |
| "epoch": 2.3686974789915967, | |
| "grad_norm": 0.6343777416548871, | |
| "learning_rate": 3.3477328434226164e-05, | |
| "loss": 0.2884, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2894600033760071, | |
| "step": 2255, | |
| "valid_targets_mean": 3490.8, | |
| "valid_targets_min": 1834 | |
| }, | |
| { | |
| "epoch": 2.3739495798319328, | |
| "grad_norm": 0.5713389821309428, | |
| "learning_rate": 3.343857663051264e-05, | |
| "loss": 0.2728, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28061312437057495, | |
| "step": 2260, | |
| "valid_targets_mean": 4200.2, | |
| "valid_targets_min": 1529 | |
| }, | |
| { | |
| "epoch": 2.379201680672269, | |
| "grad_norm": 0.6593813910051031, | |
| "learning_rate": 3.339973262812403e-05, | |
| "loss": 0.2949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28659868240356445, | |
| "step": 2265, | |
| "valid_targets_mean": 3408.4, | |
| "valid_targets_min": 845 | |
| }, | |
| { | |
| "epoch": 2.384453781512605, | |
| "grad_norm": 0.5688545517719381, | |
| "learning_rate": 3.336079669355921e-05, | |
| "loss": 0.2671, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2584436535835266, | |
| "step": 2270, | |
| "valid_targets_mean": 3692.9, | |
| "valid_targets_min": 2113 | |
| }, | |
| { | |
| "epoch": 2.389705882352941, | |
| "grad_norm": 0.6802358237019724, | |
| "learning_rate": 3.332176909394777e-05, | |
| "loss": 0.2938, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3029017448425293, | |
| "step": 2275, | |
| "valid_targets_mean": 2837.8, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 2.3949579831932772, | |
| "grad_norm": 0.6451243325404382, | |
| "learning_rate": 3.3282650097048206e-05, | |
| "loss": 0.2884, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26884186267852783, | |
| "step": 2280, | |
| "valid_targets_mean": 3303.2, | |
| "valid_targets_min": 1433 | |
| }, | |
| { | |
| "epoch": 2.4002100840336134, | |
| "grad_norm": 0.7202674984435038, | |
| "learning_rate": 3.324343997124606e-05, | |
| "loss": 0.2868, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3011503219604492, | |
| "step": 2285, | |
| "valid_targets_mean": 3398.4, | |
| "valid_targets_min": 1035 | |
| }, | |
| { | |
| "epoch": 2.4054621848739495, | |
| "grad_norm": 0.6862578318316971, | |
| "learning_rate": 3.3204138985552094e-05, | |
| "loss": 0.281, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28322312235832214, | |
| "step": 2290, | |
| "valid_targets_mean": 2917.6, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 2.4107142857142856, | |
| "grad_norm": 0.6876766550885656, | |
| "learning_rate": 3.316474740960041e-05, | |
| "loss": 0.2872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2967104911804199, | |
| "step": 2295, | |
| "valid_targets_mean": 3035.1, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 2.4159663865546217, | |
| "grad_norm": 0.603593965602333, | |
| "learning_rate": 3.312526551364667e-05, | |
| "loss": 0.2821, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26389676332473755, | |
| "step": 2300, | |
| "valid_targets_mean": 3494.5, | |
| "valid_targets_min": 1913 | |
| }, | |
| { | |
| "epoch": 2.421218487394958, | |
| "grad_norm": 0.6258585295206295, | |
| "learning_rate": 3.308569356856616e-05, | |
| "loss": 0.2672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2644697427749634, | |
| "step": 2305, | |
| "valid_targets_mean": 3577.7, | |
| "valid_targets_min": 1410 | |
| }, | |
| { | |
| "epoch": 2.426470588235294, | |
| "grad_norm": 0.6298839340403605, | |
| "learning_rate": 3.3046031845851993e-05, | |
| "loss": 0.2848, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26435354351997375, | |
| "step": 2310, | |
| "valid_targets_mean": 3195.6, | |
| "valid_targets_min": 1390 | |
| }, | |
| { | |
| "epoch": 2.43172268907563, | |
| "grad_norm": 0.6686960769778729, | |
| "learning_rate": 3.3006280617613214e-05, | |
| "loss": 0.2852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.333891898393631, | |
| "step": 2315, | |
| "valid_targets_mean": 3242.2, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 2.4369747899159666, | |
| "grad_norm": 0.6994882884638276, | |
| "learning_rate": 3.296644015657295e-05, | |
| "loss": 0.28, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29009950160980225, | |
| "step": 2320, | |
| "valid_targets_mean": 3322.3, | |
| "valid_targets_min": 1451 | |
| }, | |
| { | |
| "epoch": 2.4422268907563023, | |
| "grad_norm": 0.6022542746468339, | |
| "learning_rate": 3.292651073606652e-05, | |
| "loss": 0.2766, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24634996056556702, | |
| "step": 2325, | |
| "valid_targets_mean": 3026.8, | |
| "valid_targets_min": 1459 | |
| }, | |
| { | |
| "epoch": 2.447478991596639, | |
| "grad_norm": 0.7597076620942451, | |
| "learning_rate": 3.288649263003958e-05, | |
| "loss": 0.2712, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30025792121887207, | |
| "step": 2330, | |
| "valid_targets_mean": 2535.8, | |
| "valid_targets_min": 999 | |
| }, | |
| { | |
| "epoch": 2.452731092436975, | |
| "grad_norm": 0.5901929232154103, | |
| "learning_rate": 3.284638611304623e-05, | |
| "loss": 0.2667, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24954357743263245, | |
| "step": 2335, | |
| "valid_targets_mean": 3463.1, | |
| "valid_targets_min": 1564 | |
| }, | |
| { | |
| "epoch": 2.457983193277311, | |
| "grad_norm": 0.6305958469820415, | |
| "learning_rate": 3.280619146024714e-05, | |
| "loss": 0.2732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2825808525085449, | |
| "step": 2340, | |
| "valid_targets_mean": 3291.3, | |
| "valid_targets_min": 1470 | |
| }, | |
| { | |
| "epoch": 2.463235294117647, | |
| "grad_norm": 0.6407328134623678, | |
| "learning_rate": 3.2765908947407665e-05, | |
| "loss": 0.2567, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26442772150039673, | |
| "step": 2345, | |
| "valid_targets_mean": 3654.3, | |
| "valid_targets_min": 1935 | |
| }, | |
| { | |
| "epoch": 2.4684873949579833, | |
| "grad_norm": 0.6192241477472046, | |
| "learning_rate": 3.2725538850895933e-05, | |
| "loss": 0.2602, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25553327798843384, | |
| "step": 2350, | |
| "valid_targets_mean": 3334.8, | |
| "valid_targets_min": 1375 | |
| }, | |
| { | |
| "epoch": 2.4737394957983194, | |
| "grad_norm": 0.5958070120631284, | |
| "learning_rate": 3.268508144768096e-05, | |
| "loss": 0.2704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25967782735824585, | |
| "step": 2355, | |
| "valid_targets_mean": 3422.6, | |
| "valid_targets_min": 1294 | |
| }, | |
| { | |
| "epoch": 2.4789915966386555, | |
| "grad_norm": 0.5693719485778825, | |
| "learning_rate": 3.264453701533075e-05, | |
| "loss": 0.2583, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25220775604248047, | |
| "step": 2360, | |
| "valid_targets_mean": 3406.5, | |
| "valid_targets_min": 1496 | |
| }, | |
| { | |
| "epoch": 2.4842436974789917, | |
| "grad_norm": 0.6509477999261548, | |
| "learning_rate": 3.260390583201041e-05, | |
| "loss": 0.2596, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27943873405456543, | |
| "step": 2365, | |
| "valid_targets_mean": 3359.4, | |
| "valid_targets_min": 1333 | |
| }, | |
| { | |
| "epoch": 2.4894957983193278, | |
| "grad_norm": 0.6327828789713742, | |
| "learning_rate": 3.2563188176480217e-05, | |
| "loss": 0.2618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27448466420173645, | |
| "step": 2370, | |
| "valid_targets_mean": 3461.3, | |
| "valid_targets_min": 2264 | |
| }, | |
| { | |
| "epoch": 2.494747899159664, | |
| "grad_norm": 0.6536523986307107, | |
| "learning_rate": 3.252238432809368e-05, | |
| "loss": 0.2801, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2812950015068054, | |
| "step": 2375, | |
| "valid_targets_mean": 3145.4, | |
| "valid_targets_min": 1454 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "grad_norm": 0.6139150491194192, | |
| "learning_rate": 3.2481494566795716e-05, | |
| "loss": 0.2761, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27129268646240234, | |
| "step": 2380, | |
| "valid_targets_mean": 3347.1, | |
| "valid_targets_min": 1712 | |
| }, | |
| { | |
| "epoch": 2.505252100840336, | |
| "grad_norm": 0.6255364740679421, | |
| "learning_rate": 3.244051917312061e-05, | |
| "loss": 0.2653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2420213669538498, | |
| "step": 2385, | |
| "valid_targets_mean": 3450.1, | |
| "valid_targets_min": 1433 | |
| }, | |
| { | |
| "epoch": 2.5105042016806722, | |
| "grad_norm": 0.5931305720132205, | |
| "learning_rate": 3.239945842819018e-05, | |
| "loss": 0.2664, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2606050372123718, | |
| "step": 2390, | |
| "valid_targets_mean": 3711.4, | |
| "valid_targets_min": 1943 | |
| }, | |
| { | |
| "epoch": 2.5157563025210083, | |
| "grad_norm": 0.5818743103427734, | |
| "learning_rate": 3.2358312613711804e-05, | |
| "loss": 0.2649, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24495342373847961, | |
| "step": 2395, | |
| "valid_targets_mean": 3649.8, | |
| "valid_targets_min": 1945 | |
| }, | |
| { | |
| "epoch": 2.5210084033613445, | |
| "grad_norm": 0.6430898542162261, | |
| "learning_rate": 3.231708201197653e-05, | |
| "loss": 0.2809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2672806978225708, | |
| "step": 2400, | |
| "valid_targets_mean": 2710.3, | |
| "valid_targets_min": 815 | |
| }, | |
| { | |
| "epoch": 2.5262605042016806, | |
| "grad_norm": 0.6238779450079466, | |
| "learning_rate": 3.227576690585706e-05, | |
| "loss": 0.2769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29193007946014404, | |
| "step": 2405, | |
| "valid_targets_mean": 3416.9, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 2.5315126050420167, | |
| "grad_norm": 0.5995241914800916, | |
| "learning_rate": 3.223436757880591e-05, | |
| "loss": 0.2674, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2675004005432129, | |
| "step": 2410, | |
| "valid_targets_mean": 3632.6, | |
| "valid_targets_min": 1727 | |
| }, | |
| { | |
| "epoch": 2.536764705882353, | |
| "grad_norm": 0.6158294085854414, | |
| "learning_rate": 3.219288431485337e-05, | |
| "loss": 0.2642, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2477218508720398, | |
| "step": 2415, | |
| "valid_targets_mean": 2918.1, | |
| "valid_targets_min": 1622 | |
| }, | |
| { | |
| "epoch": 2.542016806722689, | |
| "grad_norm": 0.8470134075805191, | |
| "learning_rate": 3.2151317398605646e-05, | |
| "loss": 0.2375, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1993420273065567, | |
| "step": 2420, | |
| "valid_targets_mean": 1337.4, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 2.5472689075630255, | |
| "grad_norm": 0.7459112191461049, | |
| "learning_rate": 3.2109667115242815e-05, | |
| "loss": 0.1872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18858204782009125, | |
| "step": 2425, | |
| "valid_targets_mean": 1537.9, | |
| "valid_targets_min": 957 | |
| }, | |
| { | |
| "epoch": 2.552521008403361, | |
| "grad_norm": 0.764955976635607, | |
| "learning_rate": 3.206793375051694e-05, | |
| "loss": 0.1806, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17688506841659546, | |
| "step": 2430, | |
| "valid_targets_mean": 1386.6, | |
| "valid_targets_min": 796 | |
| }, | |
| { | |
| "epoch": 2.5577731092436977, | |
| "grad_norm": 0.9461176823521338, | |
| "learning_rate": 3.2026117590750086e-05, | |
| "loss": 0.1832, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18542060256004333, | |
| "step": 2435, | |
| "valid_targets_mean": 1313.4, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 2.5630252100840334, | |
| "grad_norm": 0.7925028804553308, | |
| "learning_rate": 3.198421892283235e-05, | |
| "loss": 0.1811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18615254759788513, | |
| "step": 2440, | |
| "valid_targets_mean": 1576.1, | |
| "valid_targets_min": 638 | |
| }, | |
| { | |
| "epoch": 2.56827731092437, | |
| "grad_norm": 0.7519556440109215, | |
| "learning_rate": 3.1942238034219885e-05, | |
| "loss": 0.1794, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18748995661735535, | |
| "step": 2445, | |
| "valid_targets_mean": 1508.0, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 2.5735294117647056, | |
| "grad_norm": 0.7720688633693402, | |
| "learning_rate": 3.190017521293296e-05, | |
| "loss": 0.1859, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16558489203453064, | |
| "step": 2450, | |
| "valid_targets_mean": 1276.2, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 2.578781512605042, | |
| "grad_norm": 0.8326618926243746, | |
| "learning_rate": 3.185803074755395e-05, | |
| "loss": 0.17, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1751636564731598, | |
| "step": 2455, | |
| "valid_targets_mean": 1297.8, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 2.5840336134453783, | |
| "grad_norm": 0.7347796399611563, | |
| "learning_rate": 3.1815804927225363e-05, | |
| "loss": 0.1901, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1736968755722046, | |
| "step": 2460, | |
| "valid_targets_mean": 1447.1, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 2.5892857142857144, | |
| "grad_norm": 0.7982508928947285, | |
| "learning_rate": 3.1773498041647876e-05, | |
| "loss": 0.1796, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20142167806625366, | |
| "step": 2465, | |
| "valid_targets_mean": 1720.9, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 2.5945378151260505, | |
| "grad_norm": 0.7855892482872033, | |
| "learning_rate": 3.173111038107834e-05, | |
| "loss": 0.1727, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18200182914733887, | |
| "step": 2470, | |
| "valid_targets_mean": 1435.6, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 2.5997899159663866, | |
| "grad_norm": 0.7759266996305735, | |
| "learning_rate": 3.168864223632776e-05, | |
| "loss": 0.1783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17558130621910095, | |
| "step": 2475, | |
| "valid_targets_mean": 1619.9, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 2.6050420168067228, | |
| "grad_norm": 0.6783776091632051, | |
| "learning_rate": 3.1646093898759334e-05, | |
| "loss": 0.1783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16645342111587524, | |
| "step": 2480, | |
| "valid_targets_mean": 1491.7, | |
| "valid_targets_min": 717 | |
| }, | |
| { | |
| "epoch": 2.610294117647059, | |
| "grad_norm": 0.7711910637914073, | |
| "learning_rate": 3.160346566028645e-05, | |
| "loss": 0.1673, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16639794409275055, | |
| "step": 2485, | |
| "valid_targets_mean": 1278.4, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 2.615546218487395, | |
| "grad_norm": 0.7139740512846204, | |
| "learning_rate": 3.156075781337068e-05, | |
| "loss": 0.1657, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15931686758995056, | |
| "step": 2490, | |
| "valid_targets_mean": 1346.2, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 2.620798319327731, | |
| "grad_norm": 0.7301516156278756, | |
| "learning_rate": 3.1517970651019735e-05, | |
| "loss": 0.1709, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1589939445257187, | |
| "step": 2495, | |
| "valid_targets_mean": 1324.8, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 2.6260504201680672, | |
| "grad_norm": 0.8269901402636304, | |
| "learning_rate": 3.1475104466785536e-05, | |
| "loss": 0.1823, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1824565976858139, | |
| "step": 2500, | |
| "valid_targets_mean": 1293.7, | |
| "valid_targets_min": 631 | |
| }, | |
| { | |
| "epoch": 2.6313025210084033, | |
| "grad_norm": 0.8200023743718566, | |
| "learning_rate": 3.143215955476211e-05, | |
| "loss": 0.1829, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19661253690719604, | |
| "step": 2505, | |
| "valid_targets_mean": 1490.3, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 2.6365546218487395, | |
| "grad_norm": 0.7270060792013762, | |
| "learning_rate": 3.138913620958365e-05, | |
| "loss": 0.1796, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1958092451095581, | |
| "step": 2510, | |
| "valid_targets_mean": 1748.8, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 2.6418067226890756, | |
| "grad_norm": 0.728759015109214, | |
| "learning_rate": 3.134603472642244e-05, | |
| "loss": 0.1742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16739603877067566, | |
| "step": 2515, | |
| "valid_targets_mean": 1434.9, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 2.6470588235294117, | |
| "grad_norm": 0.8025983924576464, | |
| "learning_rate": 3.1302855400986866e-05, | |
| "loss": 0.1765, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17958450317382812, | |
| "step": 2520, | |
| "valid_targets_mean": 1402.4, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 2.652310924369748, | |
| "grad_norm": 0.7842835266150214, | |
| "learning_rate": 3.1259598529519336e-05, | |
| "loss": 0.1771, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18177050352096558, | |
| "step": 2525, | |
| "valid_targets_mean": 1400.9, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 2.657563025210084, | |
| "grad_norm": 0.7753249562445936, | |
| "learning_rate": 3.121626440879432e-05, | |
| "loss": 0.1705, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1475217640399933, | |
| "step": 2530, | |
| "valid_targets_mean": 1109.4, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 2.66281512605042, | |
| "grad_norm": 0.7722564104049819, | |
| "learning_rate": 3.117285333611625e-05, | |
| "loss": 0.1691, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18021240830421448, | |
| "step": 2535, | |
| "valid_targets_mean": 1552.2, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 2.668067226890756, | |
| "grad_norm": 0.8324884710244702, | |
| "learning_rate": 3.112936560931752e-05, | |
| "loss": 0.1678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1621646136045456, | |
| "step": 2540, | |
| "valid_targets_mean": 1393.4, | |
| "valid_targets_min": 563 | |
| }, | |
| { | |
| "epoch": 2.6733193277310923, | |
| "grad_norm": 0.745079772832585, | |
| "learning_rate": 3.108580152675642e-05, | |
| "loss": 0.1715, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16466446220874786, | |
| "step": 2545, | |
| "valid_targets_mean": 1476.5, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 2.678571428571429, | |
| "grad_norm": 0.7667224190162166, | |
| "learning_rate": 3.104216138731511e-05, | |
| "loss": 0.1689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17364142835140228, | |
| "step": 2550, | |
| "valid_targets_mean": 1660.7, | |
| "valid_targets_min": 1107 | |
| }, | |
| { | |
| "epoch": 2.6838235294117645, | |
| "grad_norm": 0.8705202254035619, | |
| "learning_rate": 3.0998445490397546e-05, | |
| "loss": 0.1821, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17687436938285828, | |
| "step": 2555, | |
| "valid_targets_mean": 1498.9, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 2.689075630252101, | |
| "grad_norm": 0.7422198641689959, | |
| "learning_rate": 3.095465413592743e-05, | |
| "loss": 0.1637, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1798504889011383, | |
| "step": 2560, | |
| "valid_targets_mean": 1571.8, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 2.6943277310924367, | |
| "grad_norm": 0.8053104805546935, | |
| "learning_rate": 3.091078762434616e-05, | |
| "loss": 0.1717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.175202876329422, | |
| "step": 2565, | |
| "valid_targets_mean": 1496.6, | |
| "valid_targets_min": 631 | |
| }, | |
| { | |
| "epoch": 2.6995798319327733, | |
| "grad_norm": 0.7483069663873031, | |
| "learning_rate": 3.086684625661081e-05, | |
| "loss": 0.1719, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16679367423057556, | |
| "step": 2570, | |
| "valid_targets_mean": 1476.0, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 2.7048319327731094, | |
| "grad_norm": 0.7649264183537576, | |
| "learning_rate": 3.082283033419197e-05, | |
| "loss": 0.1764, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16266639530658722, | |
| "step": 2575, | |
| "valid_targets_mean": 1448.8, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 2.7100840336134455, | |
| "grad_norm": 0.7515149499265241, | |
| "learning_rate": 3.077874015907174e-05, | |
| "loss": 0.1657, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18091100454330444, | |
| "step": 2580, | |
| "valid_targets_mean": 1305.0, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 2.7153361344537816, | |
| "grad_norm": 0.7377534312107921, | |
| "learning_rate": 3.073457603374167e-05, | |
| "loss": 0.1655, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1628911793231964, | |
| "step": 2585, | |
| "valid_targets_mean": 1487.6, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 2.7205882352941178, | |
| "grad_norm": 0.7717568619128692, | |
| "learning_rate": 3.069033826120066e-05, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1856994479894638, | |
| "step": 2590, | |
| "valid_targets_mean": 1619.1, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 2.725840336134454, | |
| "grad_norm": 0.9267036067403248, | |
| "learning_rate": 3.064602714495286e-05, | |
| "loss": 0.1696, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19266122579574585, | |
| "step": 2595, | |
| "valid_targets_mean": 1382.1, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 2.73109243697479, | |
| "grad_norm": 0.7500546233055486, | |
| "learning_rate": 3.060164298900565e-05, | |
| "loss": 0.1689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1605774164199829, | |
| "step": 2600, | |
| "valid_targets_mean": 1260.2, | |
| "valid_targets_min": 794 | |
| }, | |
| { | |
| "epoch": 2.736344537815126, | |
| "grad_norm": 0.9445550576448264, | |
| "learning_rate": 3.055718609786749e-05, | |
| "loss": 0.17, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16394124925136566, | |
| "step": 2605, | |
| "valid_targets_mean": 1414.1, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 2.741596638655462, | |
| "grad_norm": 0.7880924622108275, | |
| "learning_rate": 3.051265677654585e-05, | |
| "loss": 0.1683, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1700436919927597, | |
| "step": 2610, | |
| "valid_targets_mean": 1450.9, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 2.7468487394957983, | |
| "grad_norm": 0.7995324958770467, | |
| "learning_rate": 3.0468055330545162e-05, | |
| "loss": 0.1769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18648771941661835, | |
| "step": 2615, | |
| "valid_targets_mean": 1626.9, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 2.7521008403361344, | |
| "grad_norm": 0.776585890510735, | |
| "learning_rate": 3.0423382065864653e-05, | |
| "loss": 0.1587, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16372038424015045, | |
| "step": 2620, | |
| "valid_targets_mean": 1370.7, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 2.7573529411764706, | |
| "grad_norm": 0.8233797828888295, | |
| "learning_rate": 3.0378637288996304e-05, | |
| "loss": 0.1754, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1707250475883484, | |
| "step": 2625, | |
| "valid_targets_mean": 1327.7, | |
| "valid_targets_min": 690 | |
| }, | |
| { | |
| "epoch": 2.7626050420168067, | |
| "grad_norm": 0.7313593178890555, | |
| "learning_rate": 3.033382130692269e-05, | |
| "loss": 0.1718, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16939744353294373, | |
| "step": 2630, | |
| "valid_targets_mean": 1433.4, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 2.767857142857143, | |
| "grad_norm": 0.7983600909257519, | |
| "learning_rate": 3.0288934427114955e-05, | |
| "loss": 0.1794, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16010600328445435, | |
| "step": 2635, | |
| "valid_targets_mean": 1398.2, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 2.773109243697479, | |
| "grad_norm": 0.7280785441425935, | |
| "learning_rate": 3.0243976957530602e-05, | |
| "loss": 0.1629, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16162815690040588, | |
| "step": 2640, | |
| "valid_targets_mean": 1479.6, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 2.778361344537815, | |
| "grad_norm": 0.7839303867427334, | |
| "learning_rate": 3.0198949206611475e-05, | |
| "loss": 0.1695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18356014788150787, | |
| "step": 2645, | |
| "valid_targets_mean": 1549.6, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 2.783613445378151, | |
| "grad_norm": 0.7391305503246356, | |
| "learning_rate": 3.0153851483281582e-05, | |
| "loss": 0.1702, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16755986213684082, | |
| "step": 2650, | |
| "valid_targets_mean": 1363.6, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 2.7888655462184873, | |
| "grad_norm": 0.7900413678108272, | |
| "learning_rate": 3.010868409694499e-05, | |
| "loss": 0.1672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1623220145702362, | |
| "step": 2655, | |
| "valid_targets_mean": 1367.9, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 2.7941176470588234, | |
| "grad_norm": 0.7222759869118193, | |
| "learning_rate": 3.0063447357483712e-05, | |
| "loss": 0.165, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16432425379753113, | |
| "step": 2660, | |
| "valid_targets_mean": 1393.1, | |
| "valid_targets_min": 504 | |
| }, | |
| { | |
| "epoch": 2.79936974789916, | |
| "grad_norm": 0.7458633773608461, | |
| "learning_rate": 3.0018141575255587e-05, | |
| "loss": 0.167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16822689771652222, | |
| "step": 2665, | |
| "valid_targets_mean": 1577.6, | |
| "valid_targets_min": 1006 | |
| }, | |
| { | |
| "epoch": 2.8046218487394956, | |
| "grad_norm": 0.7555827595039865, | |
| "learning_rate": 2.9972767061092107e-05, | |
| "loss": 0.1725, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15489238500595093, | |
| "step": 2670, | |
| "valid_targets_mean": 1338.3, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 2.809873949579832, | |
| "grad_norm": 0.6713741267521114, | |
| "learning_rate": 2.9927324126296352e-05, | |
| "loss": 0.1752, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1501934975385666, | |
| "step": 2675, | |
| "valid_targets_mean": 1335.4, | |
| "valid_targets_min": 886 | |
| }, | |
| { | |
| "epoch": 2.815126050420168, | |
| "grad_norm": 0.677528061579685, | |
| "learning_rate": 2.98818130826408e-05, | |
| "loss": 0.1694, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1680719256401062, | |
| "step": 2680, | |
| "valid_targets_mean": 1633.0, | |
| "valid_targets_min": 909 | |
| }, | |
| { | |
| "epoch": 2.8203781512605044, | |
| "grad_norm": 0.8220892091600075, | |
| "learning_rate": 2.9836234242365204e-05, | |
| "loss": 0.1728, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17720499634742737, | |
| "step": 2685, | |
| "valid_targets_mean": 1308.2, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 2.82563025210084, | |
| "grad_norm": 0.8315780706625598, | |
| "learning_rate": 2.979058791817447e-05, | |
| "loss": 0.1678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1801975965499878, | |
| "step": 2690, | |
| "valid_targets_mean": 1479.8, | |
| "valid_targets_min": 861 | |
| }, | |
| { | |
| "epoch": 2.8308823529411766, | |
| "grad_norm": 0.7948697249631214, | |
| "learning_rate": 2.974487442323647e-05, | |
| "loss": 0.1653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16297297179698944, | |
| "step": 2695, | |
| "valid_targets_mean": 1284.6, | |
| "valid_targets_min": 752 | |
| }, | |
| { | |
| "epoch": 2.8361344537815127, | |
| "grad_norm": 0.7446282359364188, | |
| "learning_rate": 2.9699094071179938e-05, | |
| "loss": 0.1677, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18304650485515594, | |
| "step": 2700, | |
| "valid_targets_mean": 1584.1, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 2.841386554621849, | |
| "grad_norm": 0.718619587429062, | |
| "learning_rate": 2.9653247176092286e-05, | |
| "loss": 0.1528, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1484600156545639, | |
| "step": 2705, | |
| "valid_targets_mean": 1445.1, | |
| "valid_targets_min": 638 | |
| }, | |
| { | |
| "epoch": 2.846638655462185, | |
| "grad_norm": 0.8381765419273387, | |
| "learning_rate": 2.9607334052517466e-05, | |
| "loss": 0.1681, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1818123310804367, | |
| "step": 2710, | |
| "valid_targets_mean": 1449.8, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 2.851890756302521, | |
| "grad_norm": 0.7250086940420657, | |
| "learning_rate": 2.9561355015453803e-05, | |
| "loss": 0.1685, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16832154989242554, | |
| "step": 2715, | |
| "valid_targets_mean": 1551.5, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 2.857142857142857, | |
| "grad_norm": 0.7099316939262316, | |
| "learning_rate": 2.9515310380351847e-05, | |
| "loss": 0.1605, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16392531991004944, | |
| "step": 2720, | |
| "valid_targets_mean": 1306.3, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 2.8623949579831933, | |
| "grad_norm": 0.7472866347997246, | |
| "learning_rate": 2.9469200463112186e-05, | |
| "loss": 0.1714, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17414012551307678, | |
| "step": 2725, | |
| "valid_targets_mean": 1487.9, | |
| "valid_targets_min": 855 | |
| }, | |
| { | |
| "epoch": 2.8676470588235294, | |
| "grad_norm": 0.7126081269612909, | |
| "learning_rate": 2.9423025580083298e-05, | |
| "loss": 0.168, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15881696343421936, | |
| "step": 2730, | |
| "valid_targets_mean": 1503.1, | |
| "valid_targets_min": 902 | |
| }, | |
| { | |
| "epoch": 2.8728991596638656, | |
| "grad_norm": 0.7172891409501757, | |
| "learning_rate": 2.937678604805938e-05, | |
| "loss": 0.1724, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16258196532726288, | |
| "step": 2735, | |
| "valid_targets_mean": 1501.8, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 2.8781512605042017, | |
| "grad_norm": 0.7828132309340979, | |
| "learning_rate": 2.9330482184278168e-05, | |
| "loss": 0.1667, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1718115508556366, | |
| "step": 2740, | |
| "valid_targets_mean": 1378.9, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 2.883403361344538, | |
| "grad_norm": 0.7837917980983125, | |
| "learning_rate": 2.9284114306418758e-05, | |
| "loss": 0.1614, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1576087772846222, | |
| "step": 2745, | |
| "valid_targets_mean": 1222.1, | |
| "valid_targets_min": 244 | |
| }, | |
| { | |
| "epoch": 2.888655462184874, | |
| "grad_norm": 0.7472309237344245, | |
| "learning_rate": 2.923768273259942e-05, | |
| "loss": 0.1743, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15255385637283325, | |
| "step": 2750, | |
| "valid_targets_mean": 1328.9, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 2.89390756302521, | |
| "grad_norm": 0.796880354826496, | |
| "learning_rate": 2.919118778137546e-05, | |
| "loss": 0.1691, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15773728489875793, | |
| "step": 2755, | |
| "valid_targets_mean": 1268.8, | |
| "valid_targets_min": 585 | |
| }, | |
| { | |
| "epoch": 2.899159663865546, | |
| "grad_norm": 0.7916812281947033, | |
| "learning_rate": 2.9144629771736954e-05, | |
| "loss": 0.1816, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18590059876441956, | |
| "step": 2760, | |
| "valid_targets_mean": 1596.2, | |
| "valid_targets_min": 1041 | |
| }, | |
| { | |
| "epoch": 2.9044117647058822, | |
| "grad_norm": 0.7189169579851691, | |
| "learning_rate": 2.9098009023106645e-05, | |
| "loss": 0.1706, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15632124245166779, | |
| "step": 2765, | |
| "valid_targets_mean": 1373.1, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 2.9096638655462184, | |
| "grad_norm": 0.7049554537703587, | |
| "learning_rate": 2.9051325855337684e-05, | |
| "loss": 0.1717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15910044312477112, | |
| "step": 2770, | |
| "valid_targets_mean": 1468.1, | |
| "valid_targets_min": 959 | |
| }, | |
| { | |
| "epoch": 2.9149159663865545, | |
| "grad_norm": 0.7054773517266925, | |
| "learning_rate": 2.90045805887115e-05, | |
| "loss": 0.166, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1691712886095047, | |
| "step": 2775, | |
| "valid_targets_mean": 1426.8, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 2.9201680672268906, | |
| "grad_norm": 0.8510450051223593, | |
| "learning_rate": 2.8957773543935518e-05, | |
| "loss": 0.1674, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1627202183008194, | |
| "step": 2780, | |
| "valid_targets_mean": 1360.6, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 2.9254201680672267, | |
| "grad_norm": 0.7255441493487386, | |
| "learning_rate": 2.8910905042141046e-05, | |
| "loss": 0.1715, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17565231025218964, | |
| "step": 2785, | |
| "valid_targets_mean": 1550.9, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 2.9306722689075633, | |
| "grad_norm": 0.7448004871056576, | |
| "learning_rate": 2.8863975404881022e-05, | |
| "loss": 0.166, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.170585036277771, | |
| "step": 2790, | |
| "valid_targets_mean": 1635.2, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 2.935924369747899, | |
| "grad_norm": 0.8125236767167314, | |
| "learning_rate": 2.881698495412781e-05, | |
| "loss": 0.1626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16882717609405518, | |
| "step": 2795, | |
| "valid_targets_mean": 1338.4, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 2.9411764705882355, | |
| "grad_norm": 0.7100556725806012, | |
| "learning_rate": 2.8769934012271004e-05, | |
| "loss": 0.16, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1684436947107315, | |
| "step": 2800, | |
| "valid_targets_mean": 1609.4, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 2.946428571428571, | |
| "grad_norm": 0.742789781874144, | |
| "learning_rate": 2.8722822902115228e-05, | |
| "loss": 0.1778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1685582548379898, | |
| "step": 2805, | |
| "valid_targets_mean": 1568.2, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 2.9516806722689077, | |
| "grad_norm": 0.7211288410702199, | |
| "learning_rate": 2.8675651946877884e-05, | |
| "loss": 0.1657, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17647072672843933, | |
| "step": 2810, | |
| "valid_targets_mean": 1670.1, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 2.956932773109244, | |
| "grad_norm": 0.7463660696176379, | |
| "learning_rate": 2.862842147018696e-05, | |
| "loss": 0.16, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17189790308475494, | |
| "step": 2815, | |
| "valid_targets_mean": 1533.6, | |
| "valid_targets_min": 1047 | |
| }, | |
| { | |
| "epoch": 2.96218487394958, | |
| "grad_norm": 0.7064510545820726, | |
| "learning_rate": 2.858113179607881e-05, | |
| "loss": 0.1556, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1601008176803589, | |
| "step": 2820, | |
| "valid_targets_mean": 1528.5, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 2.967436974789916, | |
| "grad_norm": 0.7179205966287691, | |
| "learning_rate": 2.8533783248995938e-05, | |
| "loss": 0.1522, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15484890341758728, | |
| "step": 2825, | |
| "valid_targets_mean": 1404.0, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.972689075630252, | |
| "grad_norm": 0.7844547172494378, | |
| "learning_rate": 2.848637615378475e-05, | |
| "loss": 0.1693, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16542135179042816, | |
| "step": 2830, | |
| "valid_targets_mean": 1366.3, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 2.9779411764705883, | |
| "grad_norm": 0.7331691954809272, | |
| "learning_rate": 2.8438910835693314e-05, | |
| "loss": 0.165, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17777270078659058, | |
| "step": 2835, | |
| "valid_targets_mean": 1436.8, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 2.9831932773109244, | |
| "grad_norm": 0.6822641232454941, | |
| "learning_rate": 2.8391387620369192e-05, | |
| "loss": 0.1587, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1580207496881485, | |
| "step": 2840, | |
| "valid_targets_mean": 1425.5, | |
| "valid_targets_min": 874 | |
| }, | |
| { | |
| "epoch": 2.9884453781512605, | |
| "grad_norm": 0.7108579727414756, | |
| "learning_rate": 2.834380683385714e-05, | |
| "loss": 0.1616, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16222992539405823, | |
| "step": 2845, | |
| "valid_targets_mean": 1604.8, | |
| "valid_targets_min": 1280 | |
| }, | |
| { | |
| "epoch": 2.9936974789915967, | |
| "grad_norm": 0.7199485862107549, | |
| "learning_rate": 2.8296168802596892e-05, | |
| "loss": 0.1667, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16997018456459045, | |
| "step": 2850, | |
| "valid_targets_mean": 1522.8, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 2.9989495798319328, | |
| "grad_norm": 0.7245690379476749, | |
| "learning_rate": 2.8248473853420933e-05, | |
| "loss": 0.171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18465590476989746, | |
| "step": 2855, | |
| "valid_targets_mean": 1654.8, | |
| "valid_targets_min": 918 | |
| }, | |
| { | |
| "epoch": 3.004201680672269, | |
| "grad_norm": 0.6536431805707318, | |
| "learning_rate": 2.8200722313552247e-05, | |
| "loss": 0.2481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2750670313835144, | |
| "step": 2860, | |
| "valid_targets_mean": 3678.0, | |
| "valid_targets_min": 702 | |
| }, | |
| { | |
| "epoch": 3.009453781512605, | |
| "grad_norm": 0.5616581452948703, | |
| "learning_rate": 2.8152914510602073e-05, | |
| "loss": 0.2663, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2694254219532013, | |
| "step": 2865, | |
| "valid_targets_mean": 4133.2, | |
| "valid_targets_min": 1550 | |
| }, | |
| { | |
| "epoch": 3.014705882352941, | |
| "grad_norm": 0.5914552324016616, | |
| "learning_rate": 2.810505077256765e-05, | |
| "loss": 0.2531, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26886335015296936, | |
| "step": 2870, | |
| "valid_targets_mean": 3929.9, | |
| "valid_targets_min": 1296 | |
| }, | |
| { | |
| "epoch": 3.0199579831932772, | |
| "grad_norm": 0.5617545432111057, | |
| "learning_rate": 2.8057131427829983e-05, | |
| "loss": 0.2488, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23543497920036316, | |
| "step": 2875, | |
| "valid_targets_mean": 4362.6, | |
| "valid_targets_min": 1681 | |
| }, | |
| { | |
| "epoch": 3.0252100840336134, | |
| "grad_norm": 0.5909970139213782, | |
| "learning_rate": 2.800915680515157e-05, | |
| "loss": 0.2512, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.269046425819397, | |
| "step": 2880, | |
| "valid_targets_mean": 3591.2, | |
| "valid_targets_min": 1044 | |
| }, | |
| { | |
| "epoch": 3.0304621848739495, | |
| "grad_norm": 0.548113977574793, | |
| "learning_rate": 2.7961127233674167e-05, | |
| "loss": 0.2608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2423868477344513, | |
| "step": 2885, | |
| "valid_targets_mean": 3697.8, | |
| "valid_targets_min": 920 | |
| }, | |
| { | |
| "epoch": 3.0357142857142856, | |
| "grad_norm": 0.5485874887437441, | |
| "learning_rate": 2.7913043042916513e-05, | |
| "loss": 0.2389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25993612408638, | |
| "step": 2890, | |
| "valid_targets_mean": 5826.4, | |
| "valid_targets_min": 2060 | |
| }, | |
| { | |
| "epoch": 3.0409663865546217, | |
| "grad_norm": 0.5329423434447529, | |
| "learning_rate": 2.786490456277208e-05, | |
| "loss": 0.2544, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24909579753875732, | |
| "step": 2895, | |
| "valid_targets_mean": 4033.6, | |
| "valid_targets_min": 467 | |
| }, | |
| { | |
| "epoch": 3.046218487394958, | |
| "grad_norm": 0.5983501432535179, | |
| "learning_rate": 2.7816712123506796e-05, | |
| "loss": 0.2631, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2781432271003723, | |
| "step": 2900, | |
| "valid_targets_mean": 3789.9, | |
| "valid_targets_min": 1689 | |
| }, | |
| { | |
| "epoch": 3.051470588235294, | |
| "grad_norm": 0.5971903589113992, | |
| "learning_rate": 2.776846605575681e-05, | |
| "loss": 0.2551, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2563186287879944, | |
| "step": 2905, | |
| "valid_targets_mean": 3695.6, | |
| "valid_targets_min": 1477 | |
| }, | |
| { | |
| "epoch": 3.05672268907563, | |
| "grad_norm": 0.5806312781845833, | |
| "learning_rate": 2.772016669052618e-05, | |
| "loss": 0.2649, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2883797883987427, | |
| "step": 2910, | |
| "valid_targets_mean": 4019.2, | |
| "valid_targets_min": 1223 | |
| }, | |
| { | |
| "epoch": 3.0619747899159666, | |
| "grad_norm": 0.5300497777002529, | |
| "learning_rate": 2.767181435918464e-05, | |
| "loss": 0.2572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27011632919311523, | |
| "step": 2915, | |
| "valid_targets_mean": 4429.1, | |
| "valid_targets_min": 1552 | |
| }, | |
| { | |
| "epoch": 3.0672268907563027, | |
| "grad_norm": 0.6074007571412762, | |
| "learning_rate": 2.7623409393465298e-05, | |
| "loss": 0.2513, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2498617321252823, | |
| "step": 2920, | |
| "valid_targets_mean": 3706.6, | |
| "valid_targets_min": 1439 | |
| }, | |
| { | |
| "epoch": 3.072478991596639, | |
| "grad_norm": 0.626168733306276, | |
| "learning_rate": 2.7574952125462386e-05, | |
| "loss": 0.2522, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23380611836910248, | |
| "step": 2925, | |
| "valid_targets_mean": 3199.5, | |
| "valid_targets_min": 1402 | |
| }, | |
| { | |
| "epoch": 3.077731092436975, | |
| "grad_norm": 0.6268290865377733, | |
| "learning_rate": 2.7526442887628962e-05, | |
| "loss": 0.2704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27413225173950195, | |
| "step": 2930, | |
| "valid_targets_mean": 3583.0, | |
| "valid_targets_min": 1586 | |
| }, | |
| { | |
| "epoch": 3.082983193277311, | |
| "grad_norm": 0.6176358259961854, | |
| "learning_rate": 2.7477882012774633e-05, | |
| "loss": 0.2567, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2532740831375122, | |
| "step": 2935, | |
| "valid_targets_mean": 3411.6, | |
| "valid_targets_min": 1538 | |
| }, | |
| { | |
| "epoch": 3.088235294117647, | |
| "grad_norm": 0.6568380568558851, | |
| "learning_rate": 2.742926983406328e-05, | |
| "loss": 0.2605, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2723920941352844, | |
| "step": 2940, | |
| "valid_targets_mean": 2824.4, | |
| "valid_targets_min": 1058 | |
| }, | |
| { | |
| "epoch": 3.0934873949579833, | |
| "grad_norm": 0.6596581392165033, | |
| "learning_rate": 2.7380606685010767e-05, | |
| "loss": 0.2743, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2707256078720093, | |
| "step": 2945, | |
| "valid_targets_mean": 2777.6, | |
| "valid_targets_min": 1205 | |
| }, | |
| { | |
| "epoch": 3.0987394957983194, | |
| "grad_norm": 0.6126808354592166, | |
| "learning_rate": 2.7331892899482654e-05, | |
| "loss": 0.2574, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.266759991645813, | |
| "step": 2950, | |
| "valid_targets_mean": 3421.2, | |
| "valid_targets_min": 1165 | |
| }, | |
| { | |
| "epoch": 3.1039915966386555, | |
| "grad_norm": 0.6307476273476691, | |
| "learning_rate": 2.7283128811691893e-05, | |
| "loss": 0.2605, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2549319565296173, | |
| "step": 2955, | |
| "valid_targets_mean": 3220.3, | |
| "valid_targets_min": 1267 | |
| }, | |
| { | |
| "epoch": 3.1092436974789917, | |
| "grad_norm": 0.6279599813034801, | |
| "learning_rate": 2.7234314756196564e-05, | |
| "loss": 0.255, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24072325229644775, | |
| "step": 2960, | |
| "valid_targets_mean": 3117.6, | |
| "valid_targets_min": 1026 | |
| }, | |
| { | |
| "epoch": 3.1144957983193278, | |
| "grad_norm": 0.612911115813002, | |
| "learning_rate": 2.718545106789755e-05, | |
| "loss": 0.2471, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25921279191970825, | |
| "step": 2965, | |
| "valid_targets_mean": 3350.1, | |
| "valid_targets_min": 1345 | |
| }, | |
| { | |
| "epoch": 3.119747899159664, | |
| "grad_norm": 0.6086001253575483, | |
| "learning_rate": 2.713653808203626e-05, | |
| "loss": 0.2625, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24595654010772705, | |
| "step": 2970, | |
| "valid_targets_mean": 3563.4, | |
| "valid_targets_min": 2050 | |
| }, | |
| { | |
| "epoch": 3.125, | |
| "grad_norm": 0.6022736229918767, | |
| "learning_rate": 2.7087576134192305e-05, | |
| "loss": 0.263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23543179035186768, | |
| "step": 2975, | |
| "valid_targets_mean": 3142.4, | |
| "valid_targets_min": 1166 | |
| }, | |
| { | |
| "epoch": 3.130252100840336, | |
| "grad_norm": 0.6154494544856133, | |
| "learning_rate": 2.7038565560281252e-05, | |
| "loss": 0.2626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2742067575454712, | |
| "step": 2980, | |
| "valid_targets_mean": 3220.9, | |
| "valid_targets_min": 997 | |
| }, | |
| { | |
| "epoch": 3.1355042016806722, | |
| "grad_norm": 0.6742689274132734, | |
| "learning_rate": 2.6989506696552224e-05, | |
| "loss": 0.2457, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2767985463142395, | |
| "step": 2985, | |
| "valid_targets_mean": 2980.7, | |
| "valid_targets_min": 1293 | |
| }, | |
| { | |
| "epoch": 3.1407563025210083, | |
| "grad_norm": 0.6039332698980796, | |
| "learning_rate": 2.6940399879585685e-05, | |
| "loss": 0.255, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23629191517829895, | |
| "step": 2990, | |
| "valid_targets_mean": 3027.2, | |
| "valid_targets_min": 1732 | |
| }, | |
| { | |
| "epoch": 3.1460084033613445, | |
| "grad_norm": 0.6818227952934203, | |
| "learning_rate": 2.6891245446291093e-05, | |
| "loss": 0.2566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27924174070358276, | |
| "step": 2995, | |
| "valid_targets_mean": 3159.9, | |
| "valid_targets_min": 1108 | |
| }, | |
| { | |
| "epoch": 3.1512605042016806, | |
| "grad_norm": 0.6924339005528242, | |
| "learning_rate": 2.6842043733904578e-05, | |
| "loss": 0.2517, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25699329376220703, | |
| "step": 3000, | |
| "valid_targets_mean": 3364.2, | |
| "valid_targets_min": 571 | |
| }, | |
| { | |
| "epoch": 3.1565126050420167, | |
| "grad_norm": 0.675775810675891, | |
| "learning_rate": 2.6792795079986634e-05, | |
| "loss": 0.258, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25223487615585327, | |
| "step": 3005, | |
| "valid_targets_mean": 2891.1, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 3.161764705882353, | |
| "grad_norm": 0.7118892851934508, | |
| "learning_rate": 2.6743499822419833e-05, | |
| "loss": 0.246, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26765280961990356, | |
| "step": 3010, | |
| "valid_targets_mean": 2982.2, | |
| "valid_targets_min": 1798 | |
| }, | |
| { | |
| "epoch": 3.167016806722689, | |
| "grad_norm": 0.6135998796483827, | |
| "learning_rate": 2.669415829940646e-05, | |
| "loss": 0.2528, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24437305331230164, | |
| "step": 3015, | |
| "valid_targets_mean": 3367.0, | |
| "valid_targets_min": 1338 | |
| }, | |
| { | |
| "epoch": 3.172268907563025, | |
| "grad_norm": 0.6308992296328928, | |
| "learning_rate": 2.6644770849466226e-05, | |
| "loss": 0.2566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26935499906539917, | |
| "step": 3020, | |
| "valid_targets_mean": 3564.4, | |
| "valid_targets_min": 1276 | |
| }, | |
| { | |
| "epoch": 3.177521008403361, | |
| "grad_norm": 0.6604476871808482, | |
| "learning_rate": 2.6595337811433925e-05, | |
| "loss": 0.2546, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25287342071533203, | |
| "step": 3025, | |
| "valid_targets_mean": 3842.9, | |
| "valid_targets_min": 1668 | |
| }, | |
| { | |
| "epoch": 3.1827731092436973, | |
| "grad_norm": 0.5864716374259419, | |
| "learning_rate": 2.654585952445714e-05, | |
| "loss": 0.2538, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2597090005874634, | |
| "step": 3030, | |
| "valid_targets_mean": 3920.2, | |
| "valid_targets_min": 1373 | |
| }, | |
| { | |
| "epoch": 3.1880252100840334, | |
| "grad_norm": 0.6138428018636582, | |
| "learning_rate": 2.6496336327993882e-05, | |
| "loss": 0.2572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2661425769329071, | |
| "step": 3035, | |
| "valid_targets_mean": 3629.1, | |
| "valid_targets_min": 1858 | |
| }, | |
| { | |
| "epoch": 3.19327731092437, | |
| "grad_norm": 0.5951194931237926, | |
| "learning_rate": 2.6446768561810264e-05, | |
| "loss": 0.2559, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2323005348443985, | |
| "step": 3040, | |
| "valid_targets_mean": 3384.1, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 3.198529411764706, | |
| "grad_norm": 0.6617141663860238, | |
| "learning_rate": 2.6397156565978198e-05, | |
| "loss": 0.2537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28470370173454285, | |
| "step": 3045, | |
| "valid_targets_mean": 3499.2, | |
| "valid_targets_min": 2483 | |
| }, | |
| { | |
| "epoch": 3.203781512605042, | |
| "grad_norm": 0.6472312935786627, | |
| "learning_rate": 2.6347500680873043e-05, | |
| "loss": 0.2491, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24719394743442535, | |
| "step": 3050, | |
| "valid_targets_mean": 2993.0, | |
| "valid_targets_min": 977 | |
| }, | |
| { | |
| "epoch": 3.2090336134453783, | |
| "grad_norm": 0.5796753467452805, | |
| "learning_rate": 2.6297801247171258e-05, | |
| "loss": 0.2403, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2180900275707245, | |
| "step": 3055, | |
| "valid_targets_mean": 3620.9, | |
| "valid_targets_min": 1856 | |
| }, | |
| { | |
| "epoch": 3.2142857142857144, | |
| "grad_norm": 0.663197194188445, | |
| "learning_rate": 2.6248058605848097e-05, | |
| "loss": 0.2469, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24358363449573517, | |
| "step": 3060, | |
| "valid_targets_mean": 3015.3, | |
| "valid_targets_min": 1200 | |
| }, | |
| { | |
| "epoch": 3.2195378151260505, | |
| "grad_norm": 0.5921044967379778, | |
| "learning_rate": 2.6198273098175234e-05, | |
| "loss": 0.2468, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21505384147167206, | |
| "step": 3065, | |
| "valid_targets_mean": 3572.4, | |
| "valid_targets_min": 1607 | |
| }, | |
| { | |
| "epoch": 3.2247899159663866, | |
| "grad_norm": 0.656610014798293, | |
| "learning_rate": 2.614844506571844e-05, | |
| "loss": 0.2586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24926620721817017, | |
| "step": 3070, | |
| "valid_targets_mean": 2880.9, | |
| "valid_targets_min": 1358 | |
| }, | |
| { | |
| "epoch": 3.2300420168067228, | |
| "grad_norm": 0.6222197525250114, | |
| "learning_rate": 2.609857485033525e-05, | |
| "loss": 0.2518, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2442103624343872, | |
| "step": 3075, | |
| "valid_targets_mean": 3209.0, | |
| "valid_targets_min": 1219 | |
| }, | |
| { | |
| "epoch": 3.235294117647059, | |
| "grad_norm": 0.587521968482715, | |
| "learning_rate": 2.6048662794172595e-05, | |
| "loss": 0.2545, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23662276566028595, | |
| "step": 3080, | |
| "valid_targets_mean": 3512.0, | |
| "valid_targets_min": 1168 | |
| }, | |
| { | |
| "epoch": 3.240546218487395, | |
| "grad_norm": 0.6685246266555611, | |
| "learning_rate": 2.599870923966447e-05, | |
| "loss": 0.2345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2525354027748108, | |
| "step": 3085, | |
| "valid_targets_mean": 2764.2, | |
| "valid_targets_min": 474 | |
| }, | |
| { | |
| "epoch": 3.245798319327731, | |
| "grad_norm": 0.6202106667084994, | |
| "learning_rate": 2.5948714529529565e-05, | |
| "loss": 0.2481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24517540633678436, | |
| "step": 3090, | |
| "valid_targets_mean": 3665.3, | |
| "valid_targets_min": 1140 | |
| }, | |
| { | |
| "epoch": 3.2510504201680672, | |
| "grad_norm": 0.5785793477278848, | |
| "learning_rate": 2.5898679006768958e-05, | |
| "loss": 0.2418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22179822623729706, | |
| "step": 3095, | |
| "valid_targets_mean": 3486.1, | |
| "valid_targets_min": 1482 | |
| }, | |
| { | |
| "epoch": 3.2563025210084033, | |
| "grad_norm": 0.6033278556859933, | |
| "learning_rate": 2.5848603014663702e-05, | |
| "loss": 0.2375, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21563810110092163, | |
| "step": 3100, | |
| "valid_targets_mean": 3097.8, | |
| "valid_targets_min": 904 | |
| }, | |
| { | |
| "epoch": 3.2615546218487395, | |
| "grad_norm": 0.6455004631708245, | |
| "learning_rate": 2.579848689677252e-05, | |
| "loss": 0.2513, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29056334495544434, | |
| "step": 3105, | |
| "valid_targets_mean": 3727.6, | |
| "valid_targets_min": 1194 | |
| }, | |
| { | |
| "epoch": 3.2668067226890756, | |
| "grad_norm": 0.5925539085498962, | |
| "learning_rate": 2.5748330996929414e-05, | |
| "loss": 0.2471, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2517462372779846, | |
| "step": 3110, | |
| "valid_targets_mean": 3901.4, | |
| "valid_targets_min": 1527 | |
| }, | |
| { | |
| "epoch": 3.2720588235294117, | |
| "grad_norm": 0.6625252347137165, | |
| "learning_rate": 2.5698135659241334e-05, | |
| "loss": 0.2458, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2597498893737793, | |
| "step": 3115, | |
| "valid_targets_mean": 3953.8, | |
| "valid_targets_min": 1428 | |
| }, | |
| { | |
| "epoch": 3.277310924369748, | |
| "grad_norm": 0.6869236027507247, | |
| "learning_rate": 2.564790122808579e-05, | |
| "loss": 0.2531, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27155032753944397, | |
| "step": 3120, | |
| "valid_targets_mean": 3239.1, | |
| "valid_targets_min": 1460 | |
| }, | |
| { | |
| "epoch": 3.282563025210084, | |
| "grad_norm": 0.641846245505168, | |
| "learning_rate": 2.5597628048108504e-05, | |
| "loss": 0.2653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26535165309906006, | |
| "step": 3125, | |
| "valid_targets_mean": 3494.8, | |
| "valid_targets_min": 1172 | |
| }, | |
| { | |
| "epoch": 3.28781512605042, | |
| "grad_norm": 0.6112401795648174, | |
| "learning_rate": 2.5547316464221058e-05, | |
| "loss": 0.2422, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2558062672615051, | |
| "step": 3130, | |
| "valid_targets_mean": 3518.0, | |
| "valid_targets_min": 1229 | |
| }, | |
| { | |
| "epoch": 3.293067226890756, | |
| "grad_norm": 0.5712486962165064, | |
| "learning_rate": 2.5496966821598493e-05, | |
| "loss": 0.24, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23176561295986176, | |
| "step": 3135, | |
| "valid_targets_mean": 3590.5, | |
| "valid_targets_min": 1523 | |
| }, | |
| { | |
| "epoch": 3.2983193277310923, | |
| "grad_norm": 0.6085435476900142, | |
| "learning_rate": 2.5446579465676977e-05, | |
| "loss": 0.2376, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23727688193321228, | |
| "step": 3140, | |
| "valid_targets_mean": 3541.0, | |
| "valid_targets_min": 997 | |
| }, | |
| { | |
| "epoch": 3.3035714285714284, | |
| "grad_norm": 0.6531904825025399, | |
| "learning_rate": 2.5396154742151407e-05, | |
| "loss": 0.2383, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2473251223564148, | |
| "step": 3145, | |
| "valid_targets_mean": 3417.0, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 3.3088235294117645, | |
| "grad_norm": 0.6509222958342396, | |
| "learning_rate": 2.5345692996973075e-05, | |
| "loss": 0.2445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24736976623535156, | |
| "step": 3150, | |
| "valid_targets_mean": 3183.2, | |
| "valid_targets_min": 1282 | |
| }, | |
| { | |
| "epoch": 3.314075630252101, | |
| "grad_norm": 0.5707923779013074, | |
| "learning_rate": 2.5295194576347233e-05, | |
| "loss": 0.229, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21255888044834137, | |
| "step": 3155, | |
| "valid_targets_mean": 3807.4, | |
| "valid_targets_min": 1143 | |
| }, | |
| { | |
| "epoch": 3.3193277310924367, | |
| "grad_norm": 0.638283494349549, | |
| "learning_rate": 2.5244659826730774e-05, | |
| "loss": 0.2369, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23158839344978333, | |
| "step": 3160, | |
| "valid_targets_mean": 3191.6, | |
| "valid_targets_min": 1438 | |
| }, | |
| { | |
| "epoch": 3.3245798319327733, | |
| "grad_norm": 0.5983879302794481, | |
| "learning_rate": 2.519408909482984e-05, | |
| "loss": 0.2587, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23230604827404022, | |
| "step": 3165, | |
| "valid_targets_mean": 3549.1, | |
| "valid_targets_min": 2091 | |
| }, | |
| { | |
| "epoch": 3.3298319327731094, | |
| "grad_norm": 0.6356003504281118, | |
| "learning_rate": 2.5143482727597427e-05, | |
| "loss": 0.2408, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23692654073238373, | |
| "step": 3170, | |
| "valid_targets_mean": 3175.4, | |
| "valid_targets_min": 1305 | |
| }, | |
| { | |
| "epoch": 3.3350840336134455, | |
| "grad_norm": 0.5822669278425938, | |
| "learning_rate": 2.509284107223102e-05, | |
| "loss": 0.2441, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2323712259531021, | |
| "step": 3175, | |
| "valid_targets_mean": 3699.6, | |
| "valid_targets_min": 1428 | |
| }, | |
| { | |
| "epoch": 3.3403361344537816, | |
| "grad_norm": 0.6034461853939603, | |
| "learning_rate": 2.50421644761702e-05, | |
| "loss": 0.242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2210870236158371, | |
| "step": 3180, | |
| "valid_targets_mean": 3405.7, | |
| "valid_targets_min": 1388 | |
| }, | |
| { | |
| "epoch": 3.3455882352941178, | |
| "grad_norm": 0.6142007275077153, | |
| "learning_rate": 2.499145328709429e-05, | |
| "loss": 0.2497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25747138261795044, | |
| "step": 3185, | |
| "valid_targets_mean": 3644.9, | |
| "valid_targets_min": 1823 | |
| }, | |
| { | |
| "epoch": 3.350840336134454, | |
| "grad_norm": 0.6119837836741985, | |
| "learning_rate": 2.4940707852919924e-05, | |
| "loss": 0.2296, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2111436128616333, | |
| "step": 3190, | |
| "valid_targets_mean": 3011.4, | |
| "valid_targets_min": 1109 | |
| }, | |
| { | |
| "epoch": 3.35609243697479, | |
| "grad_norm": 0.6533534419454151, | |
| "learning_rate": 2.488992852179868e-05, | |
| "loss": 0.2436, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23850302398204803, | |
| "step": 3195, | |
| "valid_targets_mean": 3141.6, | |
| "valid_targets_min": 1649 | |
| }, | |
| { | |
| "epoch": 3.361344537815126, | |
| "grad_norm": 0.7140300662865305, | |
| "learning_rate": 2.483911564211472e-05, | |
| "loss": 0.2366, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26096731424331665, | |
| "step": 3200, | |
| "valid_targets_mean": 2904.2, | |
| "valid_targets_min": 1249 | |
| }, | |
| { | |
| "epoch": 3.366596638655462, | |
| "grad_norm": 0.7251927942844403, | |
| "learning_rate": 2.4788269562482355e-05, | |
| "loss": 0.2531, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26130279898643494, | |
| "step": 3205, | |
| "valid_targets_mean": 3641.2, | |
| "valid_targets_min": 1793 | |
| }, | |
| { | |
| "epoch": 3.3718487394957983, | |
| "grad_norm": 0.6421393556947993, | |
| "learning_rate": 2.4737390631743675e-05, | |
| "loss": 0.2463, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2507364749908447, | |
| "step": 3210, | |
| "valid_targets_mean": 3154.0, | |
| "valid_targets_min": 1054 | |
| }, | |
| { | |
| "epoch": 3.3771008403361344, | |
| "grad_norm": 0.6437737473605846, | |
| "learning_rate": 2.4686479198966146e-05, | |
| "loss": 0.2458, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24308231472969055, | |
| "step": 3215, | |
| "valid_targets_mean": 3276.6, | |
| "valid_targets_min": 1055 | |
| }, | |
| { | |
| "epoch": 3.3823529411764706, | |
| "grad_norm": 0.6571399883695811, | |
| "learning_rate": 2.463553561344024e-05, | |
| "loss": 0.2507, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24305208027362823, | |
| "step": 3220, | |
| "valid_targets_mean": 3750.0, | |
| "valid_targets_min": 1241 | |
| }, | |
| { | |
| "epoch": 3.3876050420168067, | |
| "grad_norm": 0.650161760454692, | |
| "learning_rate": 2.4584560224677013e-05, | |
| "loss": 0.2453, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24910904467105865, | |
| "step": 3225, | |
| "valid_targets_mean": 3649.5, | |
| "valid_targets_min": 1174 | |
| }, | |
| { | |
| "epoch": 3.392857142857143, | |
| "grad_norm": 0.645357588744853, | |
| "learning_rate": 2.4533553382405702e-05, | |
| "loss": 0.2565, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2722882926464081, | |
| "step": 3230, | |
| "valid_targets_mean": 3284.9, | |
| "valid_targets_min": 992 | |
| }, | |
| { | |
| "epoch": 3.398109243697479, | |
| "grad_norm": 0.6297523245044137, | |
| "learning_rate": 2.4482515436571365e-05, | |
| "loss": 0.2473, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25592392683029175, | |
| "step": 3235, | |
| "valid_targets_mean": 3817.9, | |
| "valid_targets_min": 1133 | |
| }, | |
| { | |
| "epoch": 3.403361344537815, | |
| "grad_norm": 0.6382149094735594, | |
| "learning_rate": 2.4431446737332437e-05, | |
| "loss": 0.2478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2542281150817871, | |
| "step": 3240, | |
| "valid_targets_mean": 3242.6, | |
| "valid_targets_min": 1409 | |
| }, | |
| { | |
| "epoch": 3.408613445378151, | |
| "grad_norm": 0.726502424494956, | |
| "learning_rate": 2.438034763505833e-05, | |
| "loss": 0.2499, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25886449217796326, | |
| "step": 3245, | |
| "valid_targets_mean": 3062.4, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 3.4138655462184873, | |
| "grad_norm": 0.659719001774402, | |
| "learning_rate": 2.432921848032707e-05, | |
| "loss": 0.2515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25948482751846313, | |
| "step": 3250, | |
| "valid_targets_mean": 3307.1, | |
| "valid_targets_min": 1239 | |
| }, | |
| { | |
| "epoch": 3.4191176470588234, | |
| "grad_norm": 0.6125599642876945, | |
| "learning_rate": 2.427805962392286e-05, | |
| "loss": 0.2416, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24624377489089966, | |
| "step": 3255, | |
| "valid_targets_mean": 3635.1, | |
| "valid_targets_min": 1349 | |
| }, | |
| { | |
| "epoch": 3.4243697478991595, | |
| "grad_norm": 0.6755501860373633, | |
| "learning_rate": 2.4226871416833683e-05, | |
| "loss": 0.2453, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2604214549064636, | |
| "step": 3260, | |
| "valid_targets_mean": 3250.2, | |
| "valid_targets_min": 1434 | |
| }, | |
| { | |
| "epoch": 3.4296218487394956, | |
| "grad_norm": 0.6523430203235168, | |
| "learning_rate": 2.4175654210248848e-05, | |
| "loss": 0.2408, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23800113797187805, | |
| "step": 3265, | |
| "valid_targets_mean": 3233.2, | |
| "valid_targets_min": 521 | |
| }, | |
| { | |
| "epoch": 3.434873949579832, | |
| "grad_norm": 0.6487475032548123, | |
| "learning_rate": 2.4124408355556677e-05, | |
| "loss": 0.2548, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2616475224494934, | |
| "step": 3270, | |
| "valid_targets_mean": 3477.9, | |
| "valid_targets_min": 937 | |
| }, | |
| { | |
| "epoch": 3.440126050420168, | |
| "grad_norm": 0.7183856650514495, | |
| "learning_rate": 2.4073134204342014e-05, | |
| "loss": 0.254, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25091075897216797, | |
| "step": 3275, | |
| "valid_targets_mean": 3096.8, | |
| "valid_targets_min": 1239 | |
| }, | |
| { | |
| "epoch": 3.4453781512605044, | |
| "grad_norm": 0.6435718093662096, | |
| "learning_rate": 2.4021832108383837e-05, | |
| "loss": 0.2332, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23243436217308044, | |
| "step": 3280, | |
| "valid_targets_mean": 3552.1, | |
| "valid_targets_min": 1457 | |
| }, | |
| { | |
| "epoch": 3.4506302521008405, | |
| "grad_norm": 0.6794408128931803, | |
| "learning_rate": 2.3970502419652854e-05, | |
| "loss": 0.2358, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24406090378761292, | |
| "step": 3285, | |
| "valid_targets_mean": 3231.3, | |
| "valid_targets_min": 1307 | |
| }, | |
| { | |
| "epoch": 3.4558823529411766, | |
| "grad_norm": 0.6369525387600705, | |
| "learning_rate": 2.3919145490309076e-05, | |
| "loss": 0.2397, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24544468522071838, | |
| "step": 3290, | |
| "valid_targets_mean": 3674.0, | |
| "valid_targets_min": 1773 | |
| }, | |
| { | |
| "epoch": 3.4611344537815127, | |
| "grad_norm": 0.6864257084128846, | |
| "learning_rate": 2.38677616726994e-05, | |
| "loss": 0.2314, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23525570333003998, | |
| "step": 3295, | |
| "valid_targets_mean": 3274.8, | |
| "valid_targets_min": 1035 | |
| }, | |
| { | |
| "epoch": 3.466386554621849, | |
| "grad_norm": 0.6249370596202156, | |
| "learning_rate": 2.381635131935521e-05, | |
| "loss": 0.2307, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23440296947956085, | |
| "step": 3300, | |
| "valid_targets_mean": 3696.1, | |
| "valid_targets_min": 1333 | |
| }, | |
| { | |
| "epoch": 3.471638655462185, | |
| "grad_norm": 0.609366608375307, | |
| "learning_rate": 2.3764914782989926e-05, | |
| "loss": 0.235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2386169135570526, | |
| "step": 3305, | |
| "valid_targets_mean": 3717.7, | |
| "valid_targets_min": 1062 | |
| }, | |
| { | |
| "epoch": 3.476890756302521, | |
| "grad_norm": 0.6957649137605235, | |
| "learning_rate": 2.3713452416496625e-05, | |
| "loss": 0.2263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2374548614025116, | |
| "step": 3310, | |
| "valid_targets_mean": 2991.9, | |
| "valid_targets_min": 993 | |
| }, | |
| { | |
| "epoch": 3.482142857142857, | |
| "grad_norm": 0.6360187725166639, | |
| "learning_rate": 2.366196457294558e-05, | |
| "loss": 0.2258, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23413576185703278, | |
| "step": 3315, | |
| "valid_targets_mean": 3540.1, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 3.4873949579831933, | |
| "grad_norm": 0.6151473063134504, | |
| "learning_rate": 2.3610451605581858e-05, | |
| "loss": 0.2278, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21768996119499207, | |
| "step": 3320, | |
| "valid_targets_mean": 3565.8, | |
| "valid_targets_min": 1152 | |
| }, | |
| { | |
| "epoch": 3.4926470588235294, | |
| "grad_norm": 0.6627058497672066, | |
| "learning_rate": 2.3558913867822905e-05, | |
| "loss": 0.238, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24642956256866455, | |
| "step": 3325, | |
| "valid_targets_mean": 3437.2, | |
| "valid_targets_min": 1128 | |
| }, | |
| { | |
| "epoch": 3.4978991596638656, | |
| "grad_norm": 0.6487038665095798, | |
| "learning_rate": 2.3507351713256105e-05, | |
| "loss": 0.2494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22559967637062073, | |
| "step": 3330, | |
| "valid_targets_mean": 3062.5, | |
| "valid_targets_min": 1123 | |
| }, | |
| { | |
| "epoch": 3.5031512605042017, | |
| "grad_norm": 0.7141894926568165, | |
| "learning_rate": 2.3455765495636353e-05, | |
| "loss": 0.2404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2572299838066101, | |
| "step": 3335, | |
| "valid_targets_mean": 2870.6, | |
| "valid_targets_min": 1278 | |
| }, | |
| { | |
| "epoch": 3.508403361344538, | |
| "grad_norm": 0.7021679105572891, | |
| "learning_rate": 2.3404155568883643e-05, | |
| "loss": 0.224, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23877769708633423, | |
| "step": 3340, | |
| "valid_targets_mean": 3168.8, | |
| "valid_targets_min": 1834 | |
| }, | |
| { | |
| "epoch": 3.513655462184874, | |
| "grad_norm": 0.597320613139139, | |
| "learning_rate": 2.3352522287080625e-05, | |
| "loss": 0.2369, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21225066483020782, | |
| "step": 3345, | |
| "valid_targets_mean": 3446.2, | |
| "valid_targets_min": 420 | |
| }, | |
| { | |
| "epoch": 3.51890756302521, | |
| "grad_norm": 0.6380276235544516, | |
| "learning_rate": 2.3300866004470182e-05, | |
| "loss": 0.2432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24665868282318115, | |
| "step": 3350, | |
| "valid_targets_mean": 3379.9, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 3.524159663865546, | |
| "grad_norm": 0.7068358500231999, | |
| "learning_rate": 2.324918707545302e-05, | |
| "loss": 0.241, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24660983681678772, | |
| "step": 3355, | |
| "valid_targets_mean": 2930.9, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 3.5294117647058822, | |
| "grad_norm": 0.6415894870301354, | |
| "learning_rate": 2.3197485854585183e-05, | |
| "loss": 0.2365, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2307543158531189, | |
| "step": 3360, | |
| "valid_targets_mean": 3399.6, | |
| "valid_targets_min": 929 | |
| }, | |
| { | |
| "epoch": 3.5346638655462184, | |
| "grad_norm": 0.6658926618185477, | |
| "learning_rate": 2.3145762696575675e-05, | |
| "loss": 0.2382, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25282666087150574, | |
| "step": 3365, | |
| "valid_targets_mean": 3603.7, | |
| "valid_targets_min": 1217 | |
| }, | |
| { | |
| "epoch": 3.5399159663865545, | |
| "grad_norm": 0.9220749340704859, | |
| "learning_rate": 2.3094017956283995e-05, | |
| "loss": 0.223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17968519032001495, | |
| "step": 3370, | |
| "valid_targets_mean": 1437.6, | |
| "valid_targets_min": 544 | |
| }, | |
| { | |
| "epoch": 3.5451680672268906, | |
| "grad_norm": 0.8146754285295436, | |
| "learning_rate": 2.3042251988717728e-05, | |
| "loss": 0.1748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1631278544664383, | |
| "step": 3375, | |
| "valid_targets_mean": 1329.9, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 3.5504201680672267, | |
| "grad_norm": 0.8079169271047991, | |
| "learning_rate": 2.2990465149030077e-05, | |
| "loss": 0.1602, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14737124741077423, | |
| "step": 3380, | |
| "valid_targets_mean": 1353.5, | |
| "valid_targets_min": 514 | |
| }, | |
| { | |
| "epoch": 3.5556722689075633, | |
| "grad_norm": 0.8535871657760787, | |
| "learning_rate": 2.2938657792517446e-05, | |
| "loss": 0.164, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16740497946739197, | |
| "step": 3385, | |
| "valid_targets_mean": 1316.6, | |
| "valid_targets_min": 701 | |
| }, | |
| { | |
| "epoch": 3.560924369747899, | |
| "grad_norm": 0.7719202998888485, | |
| "learning_rate": 2.2886830274617017e-05, | |
| "loss": 0.1537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1611393243074417, | |
| "step": 3390, | |
| "valid_targets_mean": 1592.3, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 3.5661764705882355, | |
| "grad_norm": 0.7958529561306797, | |
| "learning_rate": 2.2834982950904267e-05, | |
| "loss": 0.1582, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14342686533927917, | |
| "step": 3395, | |
| "valid_targets_mean": 1302.9, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.571428571428571, | |
| "grad_norm": 0.8182753300383552, | |
| "learning_rate": 2.2783116177090573e-05, | |
| "loss": 0.1669, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16086168587207794, | |
| "step": 3400, | |
| "valid_targets_mean": 1477.3, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 3.5766806722689077, | |
| "grad_norm": 0.8440198093915002, | |
| "learning_rate": 2.2731230309020762e-05, | |
| "loss": 0.1445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14826279878616333, | |
| "step": 3405, | |
| "valid_targets_mean": 1277.8, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 3.581932773109244, | |
| "grad_norm": 0.7376313701958287, | |
| "learning_rate": 2.2679325702670642e-05, | |
| "loss": 0.1665, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1509246826171875, | |
| "step": 3410, | |
| "valid_targets_mean": 1484.6, | |
| "valid_targets_min": 708 | |
| }, | |
| { | |
| "epoch": 3.58718487394958, | |
| "grad_norm": 0.7318924656750266, | |
| "learning_rate": 2.2627402714144586e-05, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13782326877117157, | |
| "step": 3415, | |
| "valid_targets_mean": 1430.1, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 3.592436974789916, | |
| "grad_norm": 0.7765160826126889, | |
| "learning_rate": 2.2575461699673085e-05, | |
| "loss": 0.1547, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13097913563251495, | |
| "step": 3420, | |
| "valid_targets_mean": 1201.6, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 3.597689075630252, | |
| "grad_norm": 0.7638002189556976, | |
| "learning_rate": 2.2523503015610303e-05, | |
| "loss": 0.1527, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1455516219139099, | |
| "step": 3425, | |
| "valid_targets_mean": 1363.5, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 3.6029411764705883, | |
| "grad_norm": 1.2976616643941823, | |
| "learning_rate": 2.2471527018431637e-05, | |
| "loss": 0.1559, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16458860039710999, | |
| "step": 3430, | |
| "valid_targets_mean": 1451.6, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 3.6081932773109244, | |
| "grad_norm": 0.7946926976962566, | |
| "learning_rate": 2.2419534064731238e-05, | |
| "loss": 0.145, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15298821032047272, | |
| "step": 3435, | |
| "valid_targets_mean": 1393.4, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 3.6134453781512605, | |
| "grad_norm": 0.7749793995924292, | |
| "learning_rate": 2.236752451121963e-05, | |
| "loss": 0.1465, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1496553122997284, | |
| "step": 3440, | |
| "valid_targets_mean": 1498.2, | |
| "valid_targets_min": 821 | |
| }, | |
| { | |
| "epoch": 3.6186974789915967, | |
| "grad_norm": 0.8079854962978418, | |
| "learning_rate": 2.2315498714721198e-05, | |
| "loss": 0.1432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1507023572921753, | |
| "step": 3445, | |
| "valid_targets_mean": 1411.3, | |
| "valid_targets_min": 691 | |
| }, | |
| { | |
| "epoch": 3.6239495798319328, | |
| "grad_norm": 0.7810771746182552, | |
| "learning_rate": 2.2263457032171762e-05, | |
| "loss": 0.1525, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1364845335483551, | |
| "step": 3450, | |
| "valid_targets_mean": 1252.6, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 3.629201680672269, | |
| "grad_norm": 0.7854800292698711, | |
| "learning_rate": 2.2211399820616154e-05, | |
| "loss": 0.1548, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15126636624336243, | |
| "step": 3455, | |
| "valid_targets_mean": 1422.4, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 3.634453781512605, | |
| "grad_norm": 0.7420584859304036, | |
| "learning_rate": 2.2159327437205733e-05, | |
| "loss": 0.1575, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14271925389766693, | |
| "step": 3460, | |
| "valid_targets_mean": 1345.2, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 3.639705882352941, | |
| "grad_norm": 0.7964739958712432, | |
| "learning_rate": 2.210724023919595e-05, | |
| "loss": 0.1527, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16524961590766907, | |
| "step": 3465, | |
| "valid_targets_mean": 1557.1, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 3.6449579831932772, | |
| "grad_norm": 0.737091402225577, | |
| "learning_rate": 2.205513858394389e-05, | |
| "loss": 0.1525, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16383059322834015, | |
| "step": 3470, | |
| "valid_targets_mean": 1626.4, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 3.6502100840336134, | |
| "grad_norm": 0.7022804917492798, | |
| "learning_rate": 2.2003022828905832e-05, | |
| "loss": 0.1504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14327232539653778, | |
| "step": 3475, | |
| "valid_targets_mean": 1605.8, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 3.6554621848739495, | |
| "grad_norm": 0.9525205693478359, | |
| "learning_rate": 2.19508933316348e-05, | |
| "loss": 0.1504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15683533251285553, | |
| "step": 3480, | |
| "valid_targets_mean": 1267.2, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 3.6607142857142856, | |
| "grad_norm": 0.76863394644955, | |
| "learning_rate": 2.189875044977808e-05, | |
| "loss": 0.1386, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12910358607769012, | |
| "step": 3485, | |
| "valid_targets_mean": 1270.9, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 3.6659663865546217, | |
| "grad_norm": 0.8107277676634275, | |
| "learning_rate": 2.1846594541074796e-05, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15735816955566406, | |
| "step": 3490, | |
| "valid_targets_mean": 1564.4, | |
| "valid_targets_min": 872 | |
| }, | |
| { | |
| "epoch": 3.671218487394958, | |
| "grad_norm": 0.8063598539952364, | |
| "learning_rate": 2.179442596335345e-05, | |
| "loss": 0.1422, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13597534596920013, | |
| "step": 3495, | |
| "valid_targets_mean": 1323.2, | |
| "valid_targets_min": 584 | |
| }, | |
| { | |
| "epoch": 3.6764705882352944, | |
| "grad_norm": 0.7114116421864716, | |
| "learning_rate": 2.174224507452945e-05, | |
| "loss": 0.149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1412769854068756, | |
| "step": 3500, | |
| "valid_targets_mean": 1568.7, | |
| "valid_targets_min": 999 | |
| }, | |
| { | |
| "epoch": 3.68172268907563, | |
| "grad_norm": 0.9027372835508958, | |
| "learning_rate": 2.1690052232602677e-05, | |
| "loss": 0.1554, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17291805148124695, | |
| "step": 3505, | |
| "valid_targets_mean": 1610.1, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 3.6869747899159666, | |
| "grad_norm": 0.7921785974495511, | |
| "learning_rate": 2.1637847795655022e-05, | |
| "loss": 0.1454, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13400542736053467, | |
| "step": 3510, | |
| "valid_targets_mean": 1269.1, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 3.6922268907563023, | |
| "grad_norm": 0.7595580625561714, | |
| "learning_rate": 2.1585632121847925e-05, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14459824562072754, | |
| "step": 3515, | |
| "valid_targets_mean": 1640.6, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 3.697478991596639, | |
| "grad_norm": 0.8461026963697874, | |
| "learning_rate": 2.153340556941991e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13768309354782104, | |
| "step": 3520, | |
| "valid_targets_mean": 1270.5, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 3.7027310924369745, | |
| "grad_norm": 0.8139291786249493, | |
| "learning_rate": 2.1481168496684145e-05, | |
| "loss": 0.153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13502518832683563, | |
| "step": 3525, | |
| "valid_targets_mean": 1266.8, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 3.707983193277311, | |
| "grad_norm": 0.7773733165903327, | |
| "learning_rate": 2.142892126202597e-05, | |
| "loss": 0.1434, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14710843563079834, | |
| "step": 3530, | |
| "valid_targets_mean": 1562.7, | |
| "valid_targets_min": 843 | |
| }, | |
| { | |
| "epoch": 3.713235294117647, | |
| "grad_norm": 0.8540100911584824, | |
| "learning_rate": 2.137666422390045e-05, | |
| "loss": 0.143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14058175683021545, | |
| "step": 3535, | |
| "valid_targets_mean": 1169.9, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.7184873949579833, | |
| "grad_norm": 0.8192520728614282, | |
| "learning_rate": 2.132439774082991e-05, | |
| "loss": 0.1351, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.136886328458786, | |
| "step": 3540, | |
| "valid_targets_mean": 1270.4, | |
| "valid_targets_min": 772 | |
| }, | |
| { | |
| "epoch": 3.7237394957983194, | |
| "grad_norm": 0.7591394569870296, | |
| "learning_rate": 2.1272122171401467e-05, | |
| "loss": 0.1438, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12614212930202484, | |
| "step": 3545, | |
| "valid_targets_mean": 1218.7, | |
| "valid_targets_min": 619 | |
| }, | |
| { | |
| "epoch": 3.7289915966386555, | |
| "grad_norm": 0.8033788971182879, | |
| "learning_rate": 2.1219837874264573e-05, | |
| "loss": 0.1497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14868700504302979, | |
| "step": 3550, | |
| "valid_targets_mean": 1416.0, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 3.7342436974789917, | |
| "grad_norm": 0.7186280622481173, | |
| "learning_rate": 2.116754520812857e-05, | |
| "loss": 0.1436, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14419475197792053, | |
| "step": 3555, | |
| "valid_targets_mean": 1500.3, | |
| "valid_targets_min": 831 | |
| }, | |
| { | |
| "epoch": 3.7394957983193278, | |
| "grad_norm": 0.8001515867898759, | |
| "learning_rate": 2.111524453176022e-05, | |
| "loss": 0.1483, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14432606101036072, | |
| "step": 3560, | |
| "valid_targets_mean": 1469.2, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 3.744747899159664, | |
| "grad_norm": 0.7937436890534154, | |
| "learning_rate": 2.1062936203981227e-05, | |
| "loss": 0.1449, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1394154131412506, | |
| "step": 3565, | |
| "valid_targets_mean": 1317.1, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "grad_norm": 0.768116550112638, | |
| "learning_rate": 2.10106205836658e-05, | |
| "loss": 0.1398, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13312450051307678, | |
| "step": 3570, | |
| "valid_targets_mean": 1313.2, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 3.755252100840336, | |
| "grad_norm": 0.8277161948796012, | |
| "learning_rate": 2.095829802973817e-05, | |
| "loss": 0.1449, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15327024459838867, | |
| "step": 3575, | |
| "valid_targets_mean": 1500.1, | |
| "valid_targets_min": 931 | |
| }, | |
| { | |
| "epoch": 3.7605042016806722, | |
| "grad_norm": 0.7538438394237286, | |
| "learning_rate": 2.090596890117016e-05, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14686134457588196, | |
| "step": 3580, | |
| "valid_targets_mean": 1631.1, | |
| "valid_targets_min": 1011 | |
| }, | |
| { | |
| "epoch": 3.7657563025210083, | |
| "grad_norm": 0.8173261031285455, | |
| "learning_rate": 2.0853633556978674e-05, | |
| "loss": 0.1542, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14837123453617096, | |
| "step": 3585, | |
| "valid_targets_mean": 1393.4, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 3.7710084033613445, | |
| "grad_norm": 0.8094111718458211, | |
| "learning_rate": 2.0801292356223265e-05, | |
| "loss": 0.1419, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14551788568496704, | |
| "step": 3590, | |
| "valid_targets_mean": 1533.6, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 3.7762605042016806, | |
| "grad_norm": 0.9683838865909208, | |
| "learning_rate": 2.0748945658003695e-05, | |
| "loss": 0.1363, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13133487105369568, | |
| "step": 3595, | |
| "valid_targets_mean": 1311.3, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 3.7815126050420167, | |
| "grad_norm": 0.7858949321807766, | |
| "learning_rate": 2.0696593821457392e-05, | |
| "loss": 0.149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13803954422473907, | |
| "step": 3600, | |
| "valid_targets_mean": 1410.7, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 3.786764705882353, | |
| "grad_norm": 0.7682170107004075, | |
| "learning_rate": 2.064423720575709e-05, | |
| "loss": 0.1444, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1322372555732727, | |
| "step": 3605, | |
| "valid_targets_mean": 1287.4, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 3.792016806722689, | |
| "grad_norm": 0.785944883857896, | |
| "learning_rate": 2.0591876170108257e-05, | |
| "loss": 0.1397, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13903062045574188, | |
| "step": 3610, | |
| "valid_targets_mean": 1356.6, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 3.7972689075630255, | |
| "grad_norm": 1.0766948059764874, | |
| "learning_rate": 2.0539511073746743e-05, | |
| "loss": 0.1395, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1596774160861969, | |
| "step": 3615, | |
| "valid_targets_mean": 1579.2, | |
| "valid_targets_min": 852 | |
| }, | |
| { | |
| "epoch": 3.802521008403361, | |
| "grad_norm": 0.8802148487961788, | |
| "learning_rate": 2.0487142275936213e-05, | |
| "loss": 0.1496, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1571769416332245, | |
| "step": 3620, | |
| "valid_targets_mean": 1478.4, | |
| "valid_targets_min": 978 | |
| }, | |
| { | |
| "epoch": 3.8077731092436977, | |
| "grad_norm": 0.9017272636806979, | |
| "learning_rate": 2.0434770135965743e-05, | |
| "loss": 0.1523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1524774432182312, | |
| "step": 3625, | |
| "valid_targets_mean": 1463.6, | |
| "valid_targets_min": 545 | |
| }, | |
| { | |
| "epoch": 3.8130252100840334, | |
| "grad_norm": 0.794244937697772, | |
| "learning_rate": 2.0382395013147347e-05, | |
| "loss": 0.1458, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15117822587490082, | |
| "step": 3630, | |
| "valid_targets_mean": 1314.1, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 3.81827731092437, | |
| "grad_norm": 0.683828437827683, | |
| "learning_rate": 2.033001726681349e-05, | |
| "loss": 0.1479, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17605149745941162, | |
| "step": 3635, | |
| "valid_targets_mean": 1785.9, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 3.8235294117647056, | |
| "grad_norm": 0.8701621923894823, | |
| "learning_rate": 2.027763725631463e-05, | |
| "loss": 0.1445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1420620083808899, | |
| "step": 3640, | |
| "valid_targets_mean": 1337.8, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 3.828781512605042, | |
| "grad_norm": 0.8336992892361698, | |
| "learning_rate": 2.0225255341016794e-05, | |
| "loss": 0.1437, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13892850279808044, | |
| "step": 3645, | |
| "valid_targets_mean": 1443.2, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 3.8340336134453783, | |
| "grad_norm": 0.8497904024804793, | |
| "learning_rate": 2.017287188029904e-05, | |
| "loss": 0.14, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14770525693893433, | |
| "step": 3650, | |
| "valid_targets_mean": 1358.2, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.8392857142857144, | |
| "grad_norm": 0.7320787153441641, | |
| "learning_rate": 2.0120487233551035e-05, | |
| "loss": 0.1391, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13209758698940277, | |
| "step": 3655, | |
| "valid_targets_mean": 1293.9, | |
| "valid_targets_min": 672 | |
| }, | |
| { | |
| "epoch": 3.8445378151260505, | |
| "grad_norm": 0.8025485910991912, | |
| "learning_rate": 2.006810176017059e-05, | |
| "loss": 0.1343, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1370171755552292, | |
| "step": 3660, | |
| "valid_targets_mean": 1362.8, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 3.8497899159663866, | |
| "grad_norm": 0.7870212464053592, | |
| "learning_rate": 2.0015715819561205e-05, | |
| "loss": 0.147, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1440301537513733, | |
| "step": 3665, | |
| "valid_targets_mean": 1569.8, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 3.8550420168067228, | |
| "grad_norm": 0.7590451152844765, | |
| "learning_rate": 1.9963329771129558e-05, | |
| "loss": 0.1425, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12877610325813293, | |
| "step": 3670, | |
| "valid_targets_mean": 1300.3, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 3.860294117647059, | |
| "grad_norm": 0.8086227373232483, | |
| "learning_rate": 1.9910943974283066e-05, | |
| "loss": 0.1416, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13503482937812805, | |
| "step": 3675, | |
| "valid_targets_mean": 1306.6, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.865546218487395, | |
| "grad_norm": 0.7360884641253568, | |
| "learning_rate": 1.9858558788427447e-05, | |
| "loss": 0.1456, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1362096667289734, | |
| "step": 3680, | |
| "valid_targets_mean": 1580.3, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 3.870798319327731, | |
| "grad_norm": 0.8447821850921354, | |
| "learning_rate": 1.9806174572964205e-05, | |
| "loss": 0.1497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15582120418548584, | |
| "step": 3685, | |
| "valid_targets_mean": 1388.9, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 3.8760504201680672, | |
| "grad_norm": 0.7698208399307699, | |
| "learning_rate": 1.975379168728819e-05, | |
| "loss": 0.1404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1433495581150055, | |
| "step": 3690, | |
| "valid_targets_mean": 1522.7, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 3.8813025210084033, | |
| "grad_norm": 0.8559947536872534, | |
| "learning_rate": 1.9701410490785128e-05, | |
| "loss": 0.1434, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15050233900547028, | |
| "step": 3695, | |
| "valid_targets_mean": 1404.7, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 3.8865546218487395, | |
| "grad_norm": 0.8384924745171093, | |
| "learning_rate": 1.9649031342829178e-05, | |
| "loss": 0.1449, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14834997057914734, | |
| "step": 3700, | |
| "valid_targets_mean": 1505.6, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 3.8918067226890756, | |
| "grad_norm": 0.904316575159625, | |
| "learning_rate": 1.9596654602780398e-05, | |
| "loss": 0.1418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14717188477516174, | |
| "step": 3705, | |
| "valid_targets_mean": 1389.2, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 3.8970588235294117, | |
| "grad_norm": 0.8089593091562093, | |
| "learning_rate": 1.9544280629982364e-05, | |
| "loss": 0.1507, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16768889129161835, | |
| "step": 3710, | |
| "valid_targets_mean": 1636.1, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 3.902310924369748, | |
| "grad_norm": 1.1385386317837363, | |
| "learning_rate": 1.949190978375966e-05, | |
| "loss": 0.151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1377800554037094, | |
| "step": 3715, | |
| "valid_targets_mean": 1567.6, | |
| "valid_targets_min": 908 | |
| }, | |
| { | |
| "epoch": 3.907563025210084, | |
| "grad_norm": 0.7625175757891974, | |
| "learning_rate": 1.9439542423415413e-05, | |
| "loss": 0.1456, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1343471109867096, | |
| "step": 3720, | |
| "valid_targets_mean": 1259.7, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 3.91281512605042, | |
| "grad_norm": 0.7881066764674685, | |
| "learning_rate": 1.9387178908228838e-05, | |
| "loss": 0.1505, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1399286389350891, | |
| "step": 3725, | |
| "valid_targets_mean": 1342.3, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 3.918067226890756, | |
| "grad_norm": 0.7975143538194005, | |
| "learning_rate": 1.933481959745276e-05, | |
| "loss": 0.1378, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13079893589019775, | |
| "step": 3730, | |
| "valid_targets_mean": 1350.1, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 3.9233193277310923, | |
| "grad_norm": 0.7958077340565655, | |
| "learning_rate": 1.9282464850311184e-05, | |
| "loss": 0.1418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14043010771274567, | |
| "step": 3735, | |
| "valid_targets_mean": 1477.9, | |
| "valid_targets_min": 769 | |
| }, | |
| { | |
| "epoch": 3.928571428571429, | |
| "grad_norm": 0.8357581182064178, | |
| "learning_rate": 1.9230115025996764e-05, | |
| "loss": 0.1489, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13817790150642395, | |
| "step": 3740, | |
| "valid_targets_mean": 1340.2, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 3.9338235294117645, | |
| "grad_norm": 0.7898072090831375, | |
| "learning_rate": 1.9177770483668416e-05, | |
| "loss": 0.1404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13853862881660461, | |
| "step": 3745, | |
| "valid_targets_mean": 1375.1, | |
| "valid_targets_min": 707 | |
| }, | |
| { | |
| "epoch": 3.939075630252101, | |
| "grad_norm": 0.8156265634718887, | |
| "learning_rate": 1.912543158244881e-05, | |
| "loss": 0.1368, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13628491759300232, | |
| "step": 3750, | |
| "valid_targets_mean": 1429.2, | |
| "valid_targets_min": 964 | |
| }, | |
| { | |
| "epoch": 3.9443277310924367, | |
| "grad_norm": 0.8498119635524579, | |
| "learning_rate": 1.9073098681421895e-05, | |
| "loss": 0.1494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17182615399360657, | |
| "step": 3755, | |
| "valid_targets_mean": 1610.7, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 3.9495798319327733, | |
| "grad_norm": 0.7982935292914652, | |
| "learning_rate": 1.902077213963048e-05, | |
| "loss": 0.1466, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1300593912601471, | |
| "step": 3760, | |
| "valid_targets_mean": 1399.4, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 3.9548319327731094, | |
| "grad_norm": 0.7854022510245628, | |
| "learning_rate": 1.896845231607372e-05, | |
| "loss": 0.1389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13969582319259644, | |
| "step": 3765, | |
| "valid_targets_mean": 1396.2, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 3.9600840336134455, | |
| "grad_norm": 0.5894392686560074, | |
| "learning_rate": 1.8916139569704704e-05, | |
| "loss": 0.1356, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10743817687034607, | |
| "step": 3770, | |
| "valid_targets_mean": 2173.6, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 3.9653361344537816, | |
| "grad_norm": 0.7395136244392762, | |
| "learning_rate": 1.886383425942795e-05, | |
| "loss": 0.1304, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1325417160987854, | |
| "step": 3775, | |
| "valid_targets_mean": 1583.9, | |
| "valid_targets_min": 930 | |
| }, | |
| { | |
| "epoch": 3.9705882352941178, | |
| "grad_norm": 0.7865246777578737, | |
| "learning_rate": 1.8811536744096956e-05, | |
| "loss": 0.1425, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13856448233127594, | |
| "step": 3780, | |
| "valid_targets_mean": 1454.1, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 3.975840336134454, | |
| "grad_norm": 0.7942309792805228, | |
| "learning_rate": 1.8759247382511748e-05, | |
| "loss": 0.1395, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14219339191913605, | |
| "step": 3785, | |
| "valid_targets_mean": 1557.5, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 3.98109243697479, | |
| "grad_norm": 0.8125974018026644, | |
| "learning_rate": 1.8706966533416413e-05, | |
| "loss": 0.1386, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14180970191955566, | |
| "step": 3790, | |
| "valid_targets_mean": 1599.9, | |
| "valid_targets_min": 1070 | |
| }, | |
| { | |
| "epoch": 3.986344537815126, | |
| "grad_norm": 0.7872308732821972, | |
| "learning_rate": 1.8654694555496624e-05, | |
| "loss": 0.138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15511301159858704, | |
| "step": 3795, | |
| "valid_targets_mean": 1596.4, | |
| "valid_targets_min": 626 | |
| }, | |
| { | |
| "epoch": 3.991596638655462, | |
| "grad_norm": 0.8264202266884269, | |
| "learning_rate": 1.860243180737721e-05, | |
| "loss": 0.1378, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13887375593185425, | |
| "step": 3800, | |
| "valid_targets_mean": 1423.5, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 3.9968487394957983, | |
| "grad_norm": 0.8547785642731142, | |
| "learning_rate": 1.8550178647619664e-05, | |
| "loss": 0.1445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14218434691429138, | |
| "step": 3805, | |
| "valid_targets_mean": 1343.2, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 4.0021008403361344, | |
| "grad_norm": 0.9794746439498689, | |
| "learning_rate": 1.84979354347197e-05, | |
| "loss": 0.1849, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2445564717054367, | |
| "step": 3810, | |
| "valid_targets_mean": 3945.9, | |
| "valid_targets_min": 1698 | |
| }, | |
| { | |
| "epoch": 4.007352941176471, | |
| "grad_norm": 0.6295029416310093, | |
| "learning_rate": 1.8445702527104782e-05, | |
| "loss": 0.2336, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21728411316871643, | |
| "step": 3815, | |
| "valid_targets_mean": 3601.0, | |
| "valid_targets_min": 981 | |
| }, | |
| { | |
| "epoch": 4.012605042016807, | |
| "grad_norm": 0.6261229756084788, | |
| "learning_rate": 1.8393480283131677e-05, | |
| "loss": 0.2318, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2355230748653412, | |
| "step": 3820, | |
| "valid_targets_mean": 3562.5, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 4.017857142857143, | |
| "grad_norm": 0.5498474690761935, | |
| "learning_rate": 1.8341269061083996e-05, | |
| "loss": 0.2298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23167270421981812, | |
| "step": 3825, | |
| "valid_targets_mean": 4304.1, | |
| "valid_targets_min": 1201 | |
| }, | |
| { | |
| "epoch": 4.023109243697479, | |
| "grad_norm": 0.593694354846493, | |
| "learning_rate": 1.8289069219169717e-05, | |
| "loss": 0.216, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22287964820861816, | |
| "step": 3830, | |
| "valid_targets_mean": 3669.4, | |
| "valid_targets_min": 980 | |
| }, | |
| { | |
| "epoch": 4.0283613445378155, | |
| "grad_norm": 0.6618298739900584, | |
| "learning_rate": 1.8236881115518766e-05, | |
| "loss": 0.2384, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2557893693447113, | |
| "step": 3835, | |
| "valid_targets_mean": 4120.9, | |
| "valid_targets_min": 1512 | |
| }, | |
| { | |
| "epoch": 4.033613445378151, | |
| "grad_norm": 0.6142046700506524, | |
| "learning_rate": 1.8184705108180516e-05, | |
| "loss": 0.2077, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2090051770210266, | |
| "step": 3840, | |
| "valid_targets_mean": 3758.4, | |
| "valid_targets_min": 1734 | |
| }, | |
| { | |
| "epoch": 4.038865546218488, | |
| "grad_norm": 0.5794721395589425, | |
| "learning_rate": 1.8132541555121353e-05, | |
| "loss": 0.2284, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2171669602394104, | |
| "step": 3845, | |
| "valid_targets_mean": 4416.4, | |
| "valid_targets_min": 1381 | |
| }, | |
| { | |
| "epoch": 4.044117647058823, | |
| "grad_norm": 0.5873970760320343, | |
| "learning_rate": 1.808039081422223e-05, | |
| "loss": 0.2267, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23693841695785522, | |
| "step": 3850, | |
| "valid_targets_mean": 3650.9, | |
| "valid_targets_min": 1024 | |
| }, | |
| { | |
| "epoch": 4.04936974789916, | |
| "grad_norm": 0.6544469337267305, | |
| "learning_rate": 1.802825324327618e-05, | |
| "loss": 0.2364, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22341692447662354, | |
| "step": 3855, | |
| "valid_targets_mean": 3792.1, | |
| "valid_targets_min": 2107 | |
| }, | |
| { | |
| "epoch": 4.054621848739496, | |
| "grad_norm": 0.6368703549214123, | |
| "learning_rate": 1.7976129199985886e-05, | |
| "loss": 0.2273, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.216568723320961, | |
| "step": 3860, | |
| "valid_targets_mean": 3349.4, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 4.059873949579832, | |
| "grad_norm": 0.6024180533305669, | |
| "learning_rate": 1.7924019041961228e-05, | |
| "loss": 0.2282, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22188133001327515, | |
| "step": 3865, | |
| "valid_targets_mean": 3829.7, | |
| "valid_targets_min": 1597 | |
| }, | |
| { | |
| "epoch": 4.065126050420168, | |
| "grad_norm": 0.6323374001658106, | |
| "learning_rate": 1.7871923126716827e-05, | |
| "loss": 0.2319, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21390947699546814, | |
| "step": 3870, | |
| "valid_targets_mean": 2841.6, | |
| "valid_targets_min": 1535 | |
| }, | |
| { | |
| "epoch": 4.070378151260504, | |
| "grad_norm": 0.5926408570492498, | |
| "learning_rate": 1.7819841811669573e-05, | |
| "loss": 0.2226, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23295366764068604, | |
| "step": 3875, | |
| "valid_targets_mean": 4229.0, | |
| "valid_targets_min": 1385 | |
| }, | |
| { | |
| "epoch": 4.07563025210084, | |
| "grad_norm": 0.6456267570500575, | |
| "learning_rate": 1.7767775454136194e-05, | |
| "loss": 0.2314, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24474671483039856, | |
| "step": 3880, | |
| "valid_targets_mean": 3543.5, | |
| "valid_targets_min": 1451 | |
| }, | |
| { | |
| "epoch": 4.080882352941177, | |
| "grad_norm": 0.7628071822726314, | |
| "learning_rate": 1.7715724411330806e-05, | |
| "loss": 0.2372, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2485944777727127, | |
| "step": 3885, | |
| "valid_targets_mean": 2809.7, | |
| "valid_targets_min": 1269 | |
| }, | |
| { | |
| "epoch": 4.086134453781512, | |
| "grad_norm": 0.6699354098606171, | |
| "learning_rate": 1.7663689040362446e-05, | |
| "loss": 0.2262, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23240669071674347, | |
| "step": 3890, | |
| "valid_targets_mean": 2924.8, | |
| "valid_targets_min": 935 | |
| }, | |
| { | |
| "epoch": 4.091386554621849, | |
| "grad_norm": 0.6582143455705185, | |
| "learning_rate": 1.761166969823262e-05, | |
| "loss": 0.2376, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2368423342704773, | |
| "step": 3895, | |
| "valid_targets_mean": 3029.1, | |
| "valid_targets_min": 1695 | |
| }, | |
| { | |
| "epoch": 4.0966386554621845, | |
| "grad_norm": 0.7189637917203952, | |
| "learning_rate": 1.7559666741832896e-05, | |
| "loss": 0.2321, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2396736443042755, | |
| "step": 3900, | |
| "valid_targets_mean": 3217.1, | |
| "valid_targets_min": 671 | |
| }, | |
| { | |
| "epoch": 4.101890756302521, | |
| "grad_norm": 0.651225210749888, | |
| "learning_rate": 1.750768052794239e-05, | |
| "loss": 0.2289, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23836606740951538, | |
| "step": 3905, | |
| "valid_targets_mean": 3716.9, | |
| "valid_targets_min": 1020 | |
| }, | |
| { | |
| "epoch": 4.107142857142857, | |
| "grad_norm": 0.6518633827861082, | |
| "learning_rate": 1.7455711413225372e-05, | |
| "loss": 0.2251, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.233098566532135, | |
| "step": 3910, | |
| "valid_targets_mean": 3237.8, | |
| "valid_targets_min": 1889 | |
| }, | |
| { | |
| "epoch": 4.112394957983193, | |
| "grad_norm": 0.6876543636905016, | |
| "learning_rate": 1.7403759754228794e-05, | |
| "loss": 0.2211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22473348677158356, | |
| "step": 3915, | |
| "valid_targets_mean": 2959.4, | |
| "valid_targets_min": 1385 | |
| }, | |
| { | |
| "epoch": 4.117647058823529, | |
| "grad_norm": 0.7187642767454995, | |
| "learning_rate": 1.7351825907379837e-05, | |
| "loss": 0.2268, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25105470418930054, | |
| "step": 3920, | |
| "valid_targets_mean": 3390.4, | |
| "valid_targets_min": 1588 | |
| }, | |
| { | |
| "epoch": 4.1228991596638656, | |
| "grad_norm": 0.6768076603434932, | |
| "learning_rate": 1.7299910228983513e-05, | |
| "loss": 0.2292, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23110558092594147, | |
| "step": 3925, | |
| "valid_targets_mean": 3135.2, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 4.128151260504202, | |
| "grad_norm": 0.6753396109242978, | |
| "learning_rate": 1.7248013075220128e-05, | |
| "loss": 0.2302, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22585970163345337, | |
| "step": 3930, | |
| "valid_targets_mean": 2970.0, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 4.133403361344538, | |
| "grad_norm": 0.6444337615377012, | |
| "learning_rate": 1.7196134802142945e-05, | |
| "loss": 0.225, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2224642038345337, | |
| "step": 3935, | |
| "valid_targets_mean": 3744.2, | |
| "valid_targets_min": 1934 | |
| }, | |
| { | |
| "epoch": 4.138655462184874, | |
| "grad_norm": 0.6792362480524322, | |
| "learning_rate": 1.7144275765675673e-05, | |
| "loss": 0.2255, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23104789853096008, | |
| "step": 3940, | |
| "valid_targets_mean": 3753.4, | |
| "valid_targets_min": 863 | |
| }, | |
| { | |
| "epoch": 4.14390756302521, | |
| "grad_norm": 0.6240116424894049, | |
| "learning_rate": 1.7092436321610042e-05, | |
| "loss": 0.2203, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2116166055202484, | |
| "step": 3945, | |
| "valid_targets_mean": 3309.5, | |
| "valid_targets_min": 1410 | |
| }, | |
| { | |
| "epoch": 4.149159663865547, | |
| "grad_norm": 0.6706279698577615, | |
| "learning_rate": 1.704061682560336e-05, | |
| "loss": 0.23, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25118085741996765, | |
| "step": 3950, | |
| "valid_targets_mean": 4311.5, | |
| "valid_targets_min": 2040 | |
| }, | |
| { | |
| "epoch": 4.154411764705882, | |
| "grad_norm": 0.7103731685784668, | |
| "learning_rate": 1.698881763317609e-05, | |
| "loss": 0.2221, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23094293475151062, | |
| "step": 3955, | |
| "valid_targets_mean": 2688.4, | |
| "valid_targets_min": 1458 | |
| }, | |
| { | |
| "epoch": 4.159663865546219, | |
| "grad_norm": 0.6553558995778831, | |
| "learning_rate": 1.69370390997094e-05, | |
| "loss": 0.2234, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2209138721227646, | |
| "step": 3960, | |
| "valid_targets_mean": 3155.8, | |
| "valid_targets_min": 1110 | |
| }, | |
| { | |
| "epoch": 4.1649159663865545, | |
| "grad_norm": 0.6997001621889637, | |
| "learning_rate": 1.6885281580442677e-05, | |
| "loss": 0.2202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22238752245903015, | |
| "step": 3965, | |
| "valid_targets_mean": 2572.8, | |
| "valid_targets_min": 513 | |
| }, | |
| { | |
| "epoch": 4.170168067226891, | |
| "grad_norm": 2.385982583274398, | |
| "learning_rate": 1.6833545430471194e-05, | |
| "loss": 0.225, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2642630636692047, | |
| "step": 3970, | |
| "valid_targets_mean": 2760.3, | |
| "valid_targets_min": 1636 | |
| }, | |
| { | |
| "epoch": 4.175420168067227, | |
| "grad_norm": 0.6281184165319054, | |
| "learning_rate": 1.678183100474359e-05, | |
| "loss": 0.226, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23373137414455414, | |
| "step": 3975, | |
| "valid_targets_mean": 3912.6, | |
| "valid_targets_min": 1801 | |
| }, | |
| { | |
| "epoch": 4.180672268907563, | |
| "grad_norm": 0.7646111653644774, | |
| "learning_rate": 1.6730138658059448e-05, | |
| "loss": 0.2232, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2311634123325348, | |
| "step": 3980, | |
| "valid_targets_mean": 3197.1, | |
| "valid_targets_min": 924 | |
| }, | |
| { | |
| "epoch": 4.185924369747899, | |
| "grad_norm": 0.7085277743946552, | |
| "learning_rate": 1.667846874506689e-05, | |
| "loss": 0.2264, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2161899358034134, | |
| "step": 3985, | |
| "valid_targets_mean": 2886.2, | |
| "valid_targets_min": 1401 | |
| }, | |
| { | |
| "epoch": 4.1911764705882355, | |
| "grad_norm": 0.7229719956908336, | |
| "learning_rate": 1.662682162026012e-05, | |
| "loss": 0.2322, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24561478197574615, | |
| "step": 3990, | |
| "valid_targets_mean": 3041.9, | |
| "valid_targets_min": 1265 | |
| }, | |
| { | |
| "epoch": 4.196428571428571, | |
| "grad_norm": 0.6233236849228904, | |
| "learning_rate": 1.6575197637976995e-05, | |
| "loss": 0.2102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20357422530651093, | |
| "step": 3995, | |
| "valid_targets_mean": 3318.8, | |
| "valid_targets_min": 1648 | |
| }, | |
| { | |
| "epoch": 4.201680672268908, | |
| "grad_norm": 0.684658245418384, | |
| "learning_rate": 1.652359715239661e-05, | |
| "loss": 0.2252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21792471408843994, | |
| "step": 4000, | |
| "valid_targets_mean": 3989.8, | |
| "valid_targets_min": 1844 | |
| }, | |
| { | |
| "epoch": 4.206932773109243, | |
| "grad_norm": 0.6398769831531746, | |
| "learning_rate": 1.6472020517536828e-05, | |
| "loss": 0.2161, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2144005000591278, | |
| "step": 4005, | |
| "valid_targets_mean": 3636.7, | |
| "valid_targets_min": 1261 | |
| }, | |
| { | |
| "epoch": 4.21218487394958, | |
| "grad_norm": 0.7341420848535544, | |
| "learning_rate": 1.6420468087251907e-05, | |
| "loss": 0.2205, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20942559838294983, | |
| "step": 4010, | |
| "valid_targets_mean": 3165.4, | |
| "valid_targets_min": 1294 | |
| }, | |
| { | |
| "epoch": 4.217436974789916, | |
| "grad_norm": 0.7202011687730974, | |
| "learning_rate": 1.6368940215230026e-05, | |
| "loss": 0.2235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24036064743995667, | |
| "step": 4015, | |
| "valid_targets_mean": 3169.3, | |
| "valid_targets_min": 1432 | |
| }, | |
| { | |
| "epoch": 4.222689075630252, | |
| "grad_norm": 0.6828224156387892, | |
| "learning_rate": 1.6317437254990875e-05, | |
| "loss": 0.218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23701542615890503, | |
| "step": 4020, | |
| "valid_targets_mean": 3107.8, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 4.227941176470588, | |
| "grad_norm": 0.6787140516821543, | |
| "learning_rate": 1.626595955988325e-05, | |
| "loss": 0.2258, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21809794008731842, | |
| "step": 4025, | |
| "valid_targets_mean": 3005.8, | |
| "valid_targets_min": 988 | |
| }, | |
| { | |
| "epoch": 4.233193277310924, | |
| "grad_norm": 0.7166611868537497, | |
| "learning_rate": 1.6214507483082593e-05, | |
| "loss": 0.2252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.233636736869812, | |
| "step": 4030, | |
| "valid_targets_mean": 3192.8, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 4.23844537815126, | |
| "grad_norm": 0.6756135280787572, | |
| "learning_rate": 1.6163081377588588e-05, | |
| "loss": 0.209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22536295652389526, | |
| "step": 4035, | |
| "valid_targets_mean": 3349.0, | |
| "valid_targets_min": 1328 | |
| }, | |
| { | |
| "epoch": 4.243697478991597, | |
| "grad_norm": 0.6733677254055387, | |
| "learning_rate": 1.6111681596222732e-05, | |
| "loss": 0.2174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2062094807624817, | |
| "step": 4040, | |
| "valid_targets_mean": 3037.0, | |
| "valid_targets_min": 1302 | |
| }, | |
| { | |
| "epoch": 4.248949579831933, | |
| "grad_norm": 0.7009721818052325, | |
| "learning_rate": 1.606030849162593e-05, | |
| "loss": 0.2173, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2229950726032257, | |
| "step": 4045, | |
| "valid_targets_mean": 3635.6, | |
| "valid_targets_min": 1701 | |
| }, | |
| { | |
| "epoch": 4.254201680672269, | |
| "grad_norm": 0.6661284889755613, | |
| "learning_rate": 1.600896241625605e-05, | |
| "loss": 0.2136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22747060656547546, | |
| "step": 4050, | |
| "valid_targets_mean": 3422.1, | |
| "valid_targets_min": 1319 | |
| }, | |
| { | |
| "epoch": 4.2594537815126055, | |
| "grad_norm": 0.6094686088609381, | |
| "learning_rate": 1.5957643722385532e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21023216843605042, | |
| "step": 4055, | |
| "valid_targets_mean": 3806.1, | |
| "valid_targets_min": 1194 | |
| }, | |
| { | |
| "epoch": 4.264705882352941, | |
| "grad_norm": 0.6259994730524294, | |
| "learning_rate": 1.590635276209896e-05, | |
| "loss": 0.2292, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20072561502456665, | |
| "step": 4060, | |
| "valid_targets_mean": 3004.6, | |
| "valid_targets_min": 1453 | |
| }, | |
| { | |
| "epoch": 4.269957983193278, | |
| "grad_norm": 0.5928071676270973, | |
| "learning_rate": 1.5855089887290634e-05, | |
| "loss": 0.2154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2144814431667328, | |
| "step": 4065, | |
| "valid_targets_mean": 3975.1, | |
| "valid_targets_min": 1743 | |
| }, | |
| { | |
| "epoch": 4.275210084033613, | |
| "grad_norm": 0.7034766602053386, | |
| "learning_rate": 1.5803855449662175e-05, | |
| "loss": 0.2244, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23000982403755188, | |
| "step": 4070, | |
| "valid_targets_mean": 3192.5, | |
| "valid_targets_min": 1234 | |
| }, | |
| { | |
| "epoch": 4.28046218487395, | |
| "grad_norm": 0.6797161663877974, | |
| "learning_rate": 1.5752649800720096e-05, | |
| "loss": 0.232, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2336946427822113, | |
| "step": 4075, | |
| "valid_targets_mean": 3073.8, | |
| "valid_targets_min": 1188 | |
| }, | |
| { | |
| "epoch": 4.285714285714286, | |
| "grad_norm": 0.5845573966663017, | |
| "learning_rate": 1.5701473291773395e-05, | |
| "loss": 0.2252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2029433399438858, | |
| "step": 4080, | |
| "valid_targets_mean": 3894.5, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 4.290966386554622, | |
| "grad_norm": 0.9566677599879087, | |
| "learning_rate": 1.5650326273931166e-05, | |
| "loss": 0.2134, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21042492985725403, | |
| "step": 4085, | |
| "valid_targets_mean": 3189.6, | |
| "valid_targets_min": 1364 | |
| }, | |
| { | |
| "epoch": 4.296218487394958, | |
| "grad_norm": 0.6929467596600297, | |
| "learning_rate": 1.559920909810016e-05, | |
| "loss": 0.2129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22422900795936584, | |
| "step": 4090, | |
| "valid_targets_mean": 3735.1, | |
| "valid_targets_min": 1514 | |
| }, | |
| { | |
| "epoch": 4.301470588235294, | |
| "grad_norm": 0.7889834685287505, | |
| "learning_rate": 1.5548122114982393e-05, | |
| "loss": 0.2079, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22729304432868958, | |
| "step": 4095, | |
| "valid_targets_mean": 2972.1, | |
| "valid_targets_min": 1073 | |
| }, | |
| { | |
| "epoch": 4.30672268907563, | |
| "grad_norm": 0.7145190034824895, | |
| "learning_rate": 1.5497065675072728e-05, | |
| "loss": 0.2131, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22352570295333862, | |
| "step": 4100, | |
| "valid_targets_mean": 2926.5, | |
| "valid_targets_min": 1026 | |
| }, | |
| { | |
| "epoch": 4.311974789915967, | |
| "grad_norm": 0.6561593136868893, | |
| "learning_rate": 1.5446040128656485e-05, | |
| "loss": 0.2064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19251301884651184, | |
| "step": 4105, | |
| "valid_targets_mean": 2859.4, | |
| "valid_targets_min": 857 | |
| }, | |
| { | |
| "epoch": 4.317226890756302, | |
| "grad_norm": 0.6862200086122758, | |
| "learning_rate": 1.539504582580704e-05, | |
| "loss": 0.204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21139201521873474, | |
| "step": 4110, | |
| "valid_targets_mean": 3083.8, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 4.322478991596639, | |
| "grad_norm": 0.776255577609975, | |
| "learning_rate": 1.5344083116383395e-05, | |
| "loss": 0.2201, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24367398023605347, | |
| "step": 4115, | |
| "valid_targets_mean": 3220.5, | |
| "valid_targets_min": 1175 | |
| }, | |
| { | |
| "epoch": 4.3277310924369745, | |
| "grad_norm": 0.6825477266343268, | |
| "learning_rate": 1.529315235002781e-05, | |
| "loss": 0.2168, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20516598224639893, | |
| "step": 4120, | |
| "valid_targets_mean": 3000.1, | |
| "valid_targets_min": 1411 | |
| }, | |
| { | |
| "epoch": 4.332983193277311, | |
| "grad_norm": 0.6569828485036016, | |
| "learning_rate": 1.5242253876163399e-05, | |
| "loss": 0.2167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2060314565896988, | |
| "step": 4125, | |
| "valid_targets_mean": 3553.2, | |
| "valid_targets_min": 1743 | |
| }, | |
| { | |
| "epoch": 4.338235294117647, | |
| "grad_norm": 0.7319933985960059, | |
| "learning_rate": 1.5191388043991712e-05, | |
| "loss": 0.2178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2330959588289261, | |
| "step": 4130, | |
| "valid_targets_mean": 3180.8, | |
| "valid_targets_min": 1398 | |
| }, | |
| { | |
| "epoch": 4.343487394957983, | |
| "grad_norm": 0.6718430936496146, | |
| "learning_rate": 1.5140555202490359e-05, | |
| "loss": 0.2108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20311537384986877, | |
| "step": 4135, | |
| "valid_targets_mean": 3596.6, | |
| "valid_targets_min": 1767 | |
| }, | |
| { | |
| "epoch": 4.348739495798319, | |
| "grad_norm": 0.6541737718826137, | |
| "learning_rate": 1.5089755700410602e-05, | |
| "loss": 0.218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20354878902435303, | |
| "step": 4140, | |
| "valid_targets_mean": 3805.3, | |
| "valid_targets_min": 1145 | |
| }, | |
| { | |
| "epoch": 4.3539915966386555, | |
| "grad_norm": 0.6642189302688583, | |
| "learning_rate": 1.5038989886274992e-05, | |
| "loss": 0.2104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2232103943824768, | |
| "step": 4145, | |
| "valid_targets_mean": 3490.6, | |
| "valid_targets_min": 1722 | |
| }, | |
| { | |
| "epoch": 4.359243697478991, | |
| "grad_norm": 0.6521008246137644, | |
| "learning_rate": 1.498825810837492e-05, | |
| "loss": 0.2014, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19541659951210022, | |
| "step": 4150, | |
| "valid_targets_mean": 3157.6, | |
| "valid_targets_min": 1423 | |
| }, | |
| { | |
| "epoch": 4.364495798319328, | |
| "grad_norm": 0.6924321128175812, | |
| "learning_rate": 1.4937560714768283e-05, | |
| "loss": 0.2236, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22630862891674042, | |
| "step": 4155, | |
| "valid_targets_mean": 3276.0, | |
| "valid_targets_min": 1426 | |
| }, | |
| { | |
| "epoch": 4.369747899159664, | |
| "grad_norm": 0.7066936740974944, | |
| "learning_rate": 1.4886898053277086e-05, | |
| "loss": 0.225, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2139957845211029, | |
| "step": 4160, | |
| "valid_targets_mean": 3190.3, | |
| "valid_targets_min": 538 | |
| }, | |
| { | |
| "epoch": 4.375, | |
| "grad_norm": 0.7759687498686932, | |
| "learning_rate": 1.4836270471485029e-05, | |
| "loss": 0.2155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2279035598039627, | |
| "step": 4165, | |
| "valid_targets_mean": 2498.0, | |
| "valid_targets_min": 537 | |
| }, | |
| { | |
| "epoch": 4.380252100840336, | |
| "grad_norm": 0.6772887664667172, | |
| "learning_rate": 1.4785678316735144e-05, | |
| "loss": 0.2255, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21676117181777954, | |
| "step": 4170, | |
| "valid_targets_mean": 3068.9, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 4.385504201680672, | |
| "grad_norm": 0.8980937779941081, | |
| "learning_rate": 1.4735121936127406e-05, | |
| "loss": 0.2121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24256817996501923, | |
| "step": 4175, | |
| "valid_targets_mean": 3478.7, | |
| "valid_targets_min": 1625 | |
| }, | |
| { | |
| "epoch": 4.390756302521009, | |
| "grad_norm": 0.7329255290185969, | |
| "learning_rate": 1.4684601676516366e-05, | |
| "loss": 0.2311, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2386029064655304, | |
| "step": 4180, | |
| "valid_targets_mean": 3130.2, | |
| "valid_targets_min": 1400 | |
| }, | |
| { | |
| "epoch": 4.3960084033613445, | |
| "grad_norm": 0.7786336588807071, | |
| "learning_rate": 1.4634117884508738e-05, | |
| "loss": 0.214, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19800731539726257, | |
| "step": 4185, | |
| "valid_targets_mean": 3135.8, | |
| "valid_targets_min": 1648 | |
| }, | |
| { | |
| "epoch": 4.401260504201681, | |
| "grad_norm": 0.7631967058787871, | |
| "learning_rate": 1.458367090646105e-05, | |
| "loss": 0.2248, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2008107602596283, | |
| "step": 4190, | |
| "valid_targets_mean": 2437.4, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 4.406512605042017, | |
| "grad_norm": 0.7191244352959169, | |
| "learning_rate": 1.4533261088477257e-05, | |
| "loss": 0.2154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20372244715690613, | |
| "step": 4195, | |
| "valid_targets_mean": 2783.8, | |
| "valid_targets_min": 885 | |
| }, | |
| { | |
| "epoch": 4.411764705882353, | |
| "grad_norm": 0.7150862191687413, | |
| "learning_rate": 1.448288877640637e-05, | |
| "loss": 0.222, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20869606733322144, | |
| "step": 4200, | |
| "valid_targets_mean": 3689.0, | |
| "valid_targets_min": 1325 | |
| }, | |
| { | |
| "epoch": 4.417016806722689, | |
| "grad_norm": 0.716489559808733, | |
| "learning_rate": 1.4432554315840082e-05, | |
| "loss": 0.219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20727670192718506, | |
| "step": 4205, | |
| "valid_targets_mean": 3362.6, | |
| "valid_targets_min": 953 | |
| }, | |
| { | |
| "epoch": 4.4222689075630255, | |
| "grad_norm": 0.7087446848068543, | |
| "learning_rate": 1.438225805211039e-05, | |
| "loss": 0.2114, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.224325031042099, | |
| "step": 4210, | |
| "valid_targets_mean": 3481.2, | |
| "valid_targets_min": 1592 | |
| }, | |
| { | |
| "epoch": 4.427521008403361, | |
| "grad_norm": 0.6500807963798068, | |
| "learning_rate": 1.4332000330287248e-05, | |
| "loss": 0.2204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20948271453380585, | |
| "step": 4215, | |
| "valid_targets_mean": 3169.1, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 4.432773109243698, | |
| "grad_norm": 0.6730778830395577, | |
| "learning_rate": 1.428178149517617e-05, | |
| "loss": 0.2244, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2134353220462799, | |
| "step": 4220, | |
| "valid_targets_mean": 3418.1, | |
| "valid_targets_min": 1066 | |
| }, | |
| { | |
| "epoch": 4.438025210084033, | |
| "grad_norm": 0.6727213761188319, | |
| "learning_rate": 1.4231601891315876e-05, | |
| "loss": 0.2218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2243449091911316, | |
| "step": 4225, | |
| "valid_targets_mean": 4071.0, | |
| "valid_targets_min": 1885 | |
| }, | |
| { | |
| "epoch": 4.44327731092437, | |
| "grad_norm": 0.7011377195102847, | |
| "learning_rate": 1.418146186297594e-05, | |
| "loss": 0.2064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19244351983070374, | |
| "step": 4230, | |
| "valid_targets_mean": 3914.1, | |
| "valid_targets_min": 852 | |
| }, | |
| { | |
| "epoch": 4.448529411764706, | |
| "grad_norm": 0.6685868237134905, | |
| "learning_rate": 1.413136175415441e-05, | |
| "loss": 0.2199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22170649468898773, | |
| "step": 4235, | |
| "valid_targets_mean": 2790.1, | |
| "valid_targets_min": 1025 | |
| }, | |
| { | |
| "epoch": 4.453781512605042, | |
| "grad_norm": 0.6475753723469297, | |
| "learning_rate": 1.4081301908575457e-05, | |
| "loss": 0.2077, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2239975482225418, | |
| "step": 4240, | |
| "valid_targets_mean": 4196.2, | |
| "valid_targets_min": 1397 | |
| }, | |
| { | |
| "epoch": 4.459033613445378, | |
| "grad_norm": 0.8141488892094502, | |
| "learning_rate": 1.4031282669687035e-05, | |
| "loss": 0.207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19613388180732727, | |
| "step": 4245, | |
| "valid_targets_mean": 3816.1, | |
| "valid_targets_min": 1920 | |
| }, | |
| { | |
| "epoch": 4.464285714285714, | |
| "grad_norm": 0.6806187299183513, | |
| "learning_rate": 1.3981304380658474e-05, | |
| "loss": 0.2048, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20587432384490967, | |
| "step": 4250, | |
| "valid_targets_mean": 3625.5, | |
| "valid_targets_min": 1687 | |
| }, | |
| { | |
| "epoch": 4.46953781512605, | |
| "grad_norm": 0.7502387892882345, | |
| "learning_rate": 1.3931367384378183e-05, | |
| "loss": 0.2101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2327464073896408, | |
| "step": 4255, | |
| "valid_targets_mean": 3404.2, | |
| "valid_targets_min": 1210 | |
| }, | |
| { | |
| "epoch": 4.474789915966387, | |
| "grad_norm": 0.6570388636908775, | |
| "learning_rate": 1.3881472023451262e-05, | |
| "loss": 0.201, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1919768750667572, | |
| "step": 4260, | |
| "valid_targets_mean": 3334.3, | |
| "valid_targets_min": 1360 | |
| }, | |
| { | |
| "epoch": 4.480042016806722, | |
| "grad_norm": 0.6969625105218161, | |
| "learning_rate": 1.3831618640197153e-05, | |
| "loss": 0.2051, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21205592155456543, | |
| "step": 4265, | |
| "valid_targets_mean": 3091.5, | |
| "valid_targets_min": 1447 | |
| }, | |
| { | |
| "epoch": 4.485294117647059, | |
| "grad_norm": 0.7307640724228113, | |
| "learning_rate": 1.3781807576647315e-05, | |
| "loss": 0.1982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.204860657453537, | |
| "step": 4270, | |
| "valid_targets_mean": 2715.1, | |
| "valid_targets_min": 1243 | |
| }, | |
| { | |
| "epoch": 4.4905462184873945, | |
| "grad_norm": 0.7293687135782331, | |
| "learning_rate": 1.373203917454285e-05, | |
| "loss": 0.207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21605542302131653, | |
| "step": 4275, | |
| "valid_targets_mean": 2944.4, | |
| "valid_targets_min": 897 | |
| }, | |
| { | |
| "epoch": 4.495798319327731, | |
| "grad_norm": 0.7287294839891085, | |
| "learning_rate": 1.3682313775332186e-05, | |
| "loss": 0.2254, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2306896150112152, | |
| "step": 4280, | |
| "valid_targets_mean": 3085.3, | |
| "valid_targets_min": 1312 | |
| }, | |
| { | |
| "epoch": 4.501050420168067, | |
| "grad_norm": 0.6982921140070419, | |
| "learning_rate": 1.3632631720168705e-05, | |
| "loss": 0.2065, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20665548741817474, | |
| "step": 4285, | |
| "valid_targets_mean": 3413.9, | |
| "valid_targets_min": 1469 | |
| }, | |
| { | |
| "epoch": 4.506302521008403, | |
| "grad_norm": 0.7131094152829368, | |
| "learning_rate": 1.358299334990842e-05, | |
| "loss": 0.2055, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2063906341791153, | |
| "step": 4290, | |
| "valid_targets_mean": 2992.4, | |
| "valid_targets_min": 1593 | |
| }, | |
| { | |
| "epoch": 4.51155462184874, | |
| "grad_norm": 0.7372869299957252, | |
| "learning_rate": 1.3533399005107635e-05, | |
| "loss": 0.2104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22824421525001526, | |
| "step": 4295, | |
| "valid_targets_mean": 3131.6, | |
| "valid_targets_min": 1664 | |
| }, | |
| { | |
| "epoch": 4.516806722689076, | |
| "grad_norm": 0.6346433949828432, | |
| "learning_rate": 1.3483849026020603e-05, | |
| "loss": 0.1992, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19013050198554993, | |
| "step": 4300, | |
| "valid_targets_mean": 3709.5, | |
| "valid_targets_min": 1605 | |
| }, | |
| { | |
| "epoch": 4.522058823529412, | |
| "grad_norm": 0.6754439443280119, | |
| "learning_rate": 1.3434343752597199e-05, | |
| "loss": 0.2258, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22575265169143677, | |
| "step": 4305, | |
| "valid_targets_mean": 3649.6, | |
| "valid_targets_min": 1737 | |
| }, | |
| { | |
| "epoch": 4.527310924369748, | |
| "grad_norm": 0.6967616700028672, | |
| "learning_rate": 1.3384883524480576e-05, | |
| "loss": 0.2156, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22875870764255524, | |
| "step": 4310, | |
| "valid_targets_mean": 3512.8, | |
| "valid_targets_min": 1817 | |
| }, | |
| { | |
| "epoch": 4.532563025210084, | |
| "grad_norm": 0.6886827979081415, | |
| "learning_rate": 1.333546868100486e-05, | |
| "loss": 0.2043, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20903798937797546, | |
| "step": 4315, | |
| "valid_targets_mean": 3413.6, | |
| "valid_targets_min": 2026 | |
| }, | |
| { | |
| "epoch": 4.53781512605042, | |
| "grad_norm": 0.660189404042565, | |
| "learning_rate": 1.3286099561192784e-05, | |
| "loss": 0.207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21688057482242584, | |
| "step": 4320, | |
| "valid_targets_mean": 3905.8, | |
| "valid_targets_min": 1214 | |
| }, | |
| { | |
| "epoch": 4.543067226890757, | |
| "grad_norm": 0.8043966125939357, | |
| "learning_rate": 1.3236776503753397e-05, | |
| "loss": 0.1721, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14316143095493317, | |
| "step": 4325, | |
| "valid_targets_mean": 1467.6, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 4.548319327731092, | |
| "grad_norm": 0.7841054688198438, | |
| "learning_rate": 1.3187499847079713e-05, | |
| "loss": 0.1451, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14375540614128113, | |
| "step": 4330, | |
| "valid_targets_mean": 1573.2, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 4.553571428571429, | |
| "grad_norm": 0.8186335723421362, | |
| "learning_rate": 1.3138269929246427e-05, | |
| "loss": 0.1396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15070709586143494, | |
| "step": 4335, | |
| "valid_targets_mean": 1469.3, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 4.5588235294117645, | |
| "grad_norm": 0.8679170015389952, | |
| "learning_rate": 1.308908708800753e-05, | |
| "loss": 0.1393, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14169779419898987, | |
| "step": 4340, | |
| "valid_targets_mean": 1390.9, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 4.564075630252101, | |
| "grad_norm": 0.9114994147726516, | |
| "learning_rate": 1.3039951660794063e-05, | |
| "loss": 0.1401, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14317017793655396, | |
| "step": 4345, | |
| "valid_targets_mean": 1436.4, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 4.569327731092437, | |
| "grad_norm": 0.8898038163908437, | |
| "learning_rate": 1.2990863984711773e-05, | |
| "loss": 0.1421, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17335724830627441, | |
| "step": 4350, | |
| "valid_targets_mean": 1576.6, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 4.574579831932773, | |
| "grad_norm": 0.785064729278413, | |
| "learning_rate": 1.294182439653878e-05, | |
| "loss": 0.1329, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12577390670776367, | |
| "step": 4355, | |
| "valid_targets_mean": 1400.9, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 4.579831932773109, | |
| "grad_norm": 0.7745818948625097, | |
| "learning_rate": 1.2892833232723297e-05, | |
| "loss": 0.1444, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21761628985404968, | |
| "step": 4360, | |
| "valid_targets_mean": 1609.2, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 4.5850840336134455, | |
| "grad_norm": 0.9711741060616659, | |
| "learning_rate": 1.2843890829381303e-05, | |
| "loss": 0.1307, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14271649718284607, | |
| "step": 4365, | |
| "valid_targets_mean": 1564.4, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 4.590336134453781, | |
| "grad_norm": 0.8866160542903732, | |
| "learning_rate": 1.2794997522294263e-05, | |
| "loss": 0.1366, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13775603473186493, | |
| "step": 4370, | |
| "valid_targets_mean": 1521.2, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 4.595588235294118, | |
| "grad_norm": 0.8231923650364629, | |
| "learning_rate": 1.2746153646906766e-05, | |
| "loss": 0.1282, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13401484489440918, | |
| "step": 4375, | |
| "valid_targets_mean": 1525.2, | |
| "valid_targets_min": 815 | |
| }, | |
| { | |
| "epoch": 4.600840336134453, | |
| "grad_norm": 0.8550257487811143, | |
| "learning_rate": 1.2697359538324303e-05, | |
| "loss": 0.1316, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13395115733146667, | |
| "step": 4380, | |
| "valid_targets_mean": 1584.4, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 4.60609243697479, | |
| "grad_norm": 0.8426859205924878, | |
| "learning_rate": 1.2648615531310907e-05, | |
| "loss": 0.1317, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12041187286376953, | |
| "step": 4385, | |
| "valid_targets_mean": 1384.5, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 4.6113445378151265, | |
| "grad_norm": 0.786905654669723, | |
| "learning_rate": 1.259992196028688e-05, | |
| "loss": 0.1232, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11873074620962143, | |
| "step": 4390, | |
| "valid_targets_mean": 1481.8, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 4.616596638655462, | |
| "grad_norm": 0.8303948199867104, | |
| "learning_rate": 1.2551279159326495e-05, | |
| "loss": 0.1208, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11647927761077881, | |
| "step": 4395, | |
| "valid_targets_mean": 1247.1, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 4.621848739495798, | |
| "grad_norm": 0.8753106179981709, | |
| "learning_rate": 1.2502687462155709e-05, | |
| "loss": 0.1312, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1494443118572235, | |
| "step": 4400, | |
| "valid_targets_mean": 1686.6, | |
| "valid_targets_min": 896 | |
| }, | |
| { | |
| "epoch": 4.6271008403361344, | |
| "grad_norm": 0.8050150714693087, | |
| "learning_rate": 1.2454147202149865e-05, | |
| "loss": 0.1312, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12719589471817017, | |
| "step": 4405, | |
| "valid_targets_mean": 1415.3, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 4.632352941176471, | |
| "grad_norm": 0.8009830063907718, | |
| "learning_rate": 1.2405658712331409e-05, | |
| "loss": 0.1342, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12356242537498474, | |
| "step": 4410, | |
| "valid_targets_mean": 1527.1, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 4.637605042016807, | |
| "grad_norm": 0.7347334443196466, | |
| "learning_rate": 1.2357222325367604e-05, | |
| "loss": 0.1311, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11749774217605591, | |
| "step": 4415, | |
| "valid_targets_mean": 1439.2, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 4.642857142857143, | |
| "grad_norm": 0.8064940468859054, | |
| "learning_rate": 1.2308838373568249e-05, | |
| "loss": 0.1308, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12089011818170547, | |
| "step": 4420, | |
| "valid_targets_mean": 1403.5, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 4.648109243697479, | |
| "grad_norm": 0.7972098191885254, | |
| "learning_rate": 1.2260507188883395e-05, | |
| "loss": 0.1317, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12170597910881042, | |
| "step": 4425, | |
| "valid_targets_mean": 1499.6, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 4.6533613445378155, | |
| "grad_norm": 0.975037389066893, | |
| "learning_rate": 1.2212229102901077e-05, | |
| "loss": 0.1311, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13412654399871826, | |
| "step": 4430, | |
| "valid_targets_mean": 1218.8, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 4.658613445378151, | |
| "grad_norm": 0.8641754817962347, | |
| "learning_rate": 1.2164004446845037e-05, | |
| "loss": 0.1226, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13030359148979187, | |
| "step": 4435, | |
| "valid_targets_mean": 1603.4, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 4.663865546218488, | |
| "grad_norm": 0.9180121006760533, | |
| "learning_rate": 1.2115833551572438e-05, | |
| "loss": 0.1216, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12268111109733582, | |
| "step": 4440, | |
| "valid_targets_mean": 1407.5, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 4.669117647058823, | |
| "grad_norm": 0.8547903927275693, | |
| "learning_rate": 1.2067716747571616e-05, | |
| "loss": 0.1261, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13655883073806763, | |
| "step": 4445, | |
| "valid_targets_mean": 1627.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 4.67436974789916, | |
| "grad_norm": 0.881631394330736, | |
| "learning_rate": 1.201965436495978e-05, | |
| "loss": 0.1273, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12907208502292633, | |
| "step": 4450, | |
| "valid_targets_mean": 1424.6, | |
| "valid_targets_min": 1004 | |
| }, | |
| { | |
| "epoch": 4.679621848739496, | |
| "grad_norm": 0.8063717130180919, | |
| "learning_rate": 1.1971646733480784e-05, | |
| "loss": 0.1288, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13420622050762177, | |
| "step": 4455, | |
| "valid_targets_mean": 1478.4, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 4.684873949579832, | |
| "grad_norm": 0.8260967863457012, | |
| "learning_rate": 1.1923694182502848e-05, | |
| "loss": 0.1348, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11544696986675262, | |
| "step": 4460, | |
| "valid_targets_mean": 1341.6, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 4.690126050420168, | |
| "grad_norm": 0.8029075297918149, | |
| "learning_rate": 1.1875797041016286e-05, | |
| "loss": 0.1247, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12873606383800507, | |
| "step": 4465, | |
| "valid_targets_mean": 1627.8, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 4.695378151260504, | |
| "grad_norm": 0.8998520655105815, | |
| "learning_rate": 1.1827955637631283e-05, | |
| "loss": 0.1311, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14146029949188232, | |
| "step": 4470, | |
| "valid_targets_mean": 1465.2, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 4.70063025210084, | |
| "grad_norm": 0.9673884965286654, | |
| "learning_rate": 1.1780170300575602e-05, | |
| "loss": 0.1295, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15864944458007812, | |
| "step": 4475, | |
| "valid_targets_mean": 1623.8, | |
| "valid_targets_min": 796 | |
| }, | |
| { | |
| "epoch": 4.705882352941177, | |
| "grad_norm": 0.8028753858678307, | |
| "learning_rate": 1.1732441357692353e-05, | |
| "loss": 0.1197, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11006323993206024, | |
| "step": 4480, | |
| "valid_targets_mean": 1318.8, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 4.711134453781512, | |
| "grad_norm": 0.7757517539215945, | |
| "learning_rate": 1.1684769136437742e-05, | |
| "loss": 0.1238, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1143508329987526, | |
| "step": 4485, | |
| "valid_targets_mean": 1281.6, | |
| "valid_targets_min": 665 | |
| }, | |
| { | |
| "epoch": 4.716386554621849, | |
| "grad_norm": 0.8198556072030991, | |
| "learning_rate": 1.1637153963878815e-05, | |
| "loss": 0.1189, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10272709280252457, | |
| "step": 4490, | |
| "valid_targets_mean": 1267.4, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 4.7216386554621845, | |
| "grad_norm": 0.993622221837871, | |
| "learning_rate": 1.1589596166691231e-05, | |
| "loss": 0.1259, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14048606157302856, | |
| "step": 4495, | |
| "valid_targets_mean": 1505.4, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 4.726890756302521, | |
| "grad_norm": 0.9175391246008294, | |
| "learning_rate": 1.1542096071157012e-05, | |
| "loss": 0.1224, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12400451302528381, | |
| "step": 4500, | |
| "valid_targets_mean": 1243.3, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 4.732142857142857, | |
| "grad_norm": 0.8780528943864113, | |
| "learning_rate": 1.1494654003162285e-05, | |
| "loss": 0.1243, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12478488683700562, | |
| "step": 4505, | |
| "valid_targets_mean": 1371.4, | |
| "valid_targets_min": 924 | |
| }, | |
| { | |
| "epoch": 4.737394957983193, | |
| "grad_norm": 1.0957377833225224, | |
| "learning_rate": 1.1447270288195089e-05, | |
| "loss": 0.1287, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1323080062866211, | |
| "step": 4510, | |
| "valid_targets_mean": 1497.8, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 4.742647058823529, | |
| "grad_norm": 0.9093558038524169, | |
| "learning_rate": 1.1399945251343114e-05, | |
| "loss": 0.1237, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14012199640274048, | |
| "step": 4515, | |
| "valid_targets_mean": 1446.2, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 4.7478991596638656, | |
| "grad_norm": 0.8399340996586211, | |
| "learning_rate": 1.1352679217291457e-05, | |
| "loss": 0.1251, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10750658810138702, | |
| "step": 4520, | |
| "valid_targets_mean": 1165.6, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 4.753151260504202, | |
| "grad_norm": 0.8065610789721998, | |
| "learning_rate": 1.1305472510320419e-05, | |
| "loss": 0.1192, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12666700780391693, | |
| "step": 4525, | |
| "valid_targets_mean": 1593.9, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 4.758403361344538, | |
| "grad_norm": 0.825002414616717, | |
| "learning_rate": 1.1258325454303286e-05, | |
| "loss": 0.13, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13277189433574677, | |
| "step": 4530, | |
| "valid_targets_mean": 1652.5, | |
| "valid_targets_min": 573 | |
| }, | |
| { | |
| "epoch": 4.7636554621848735, | |
| "grad_norm": 0.8052034847245988, | |
| "learning_rate": 1.1211238372704073e-05, | |
| "loss": 0.1265, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12504729628562927, | |
| "step": 4535, | |
| "valid_targets_mean": 1694.2, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 4.76890756302521, | |
| "grad_norm": 0.7944186688073329, | |
| "learning_rate": 1.1164211588575339e-05, | |
| "loss": 0.1295, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11452824622392654, | |
| "step": 4540, | |
| "valid_targets_mean": 1335.4, | |
| "valid_targets_min": 763 | |
| }, | |
| { | |
| "epoch": 4.774159663865547, | |
| "grad_norm": 0.9867441295496705, | |
| "learning_rate": 1.1117245424555967e-05, | |
| "loss": 0.1213, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12596900761127472, | |
| "step": 4545, | |
| "valid_targets_mean": 1470.9, | |
| "valid_targets_min": 956 | |
| }, | |
| { | |
| "epoch": 4.779411764705882, | |
| "grad_norm": 0.9111318412419092, | |
| "learning_rate": 1.1070340202868915e-05, | |
| "loss": 0.1244, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12238766252994537, | |
| "step": 4550, | |
| "valid_targets_mean": 1322.6, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 4.784663865546219, | |
| "grad_norm": 0.9026375196081591, | |
| "learning_rate": 1.1023496245319056e-05, | |
| "loss": 0.1254, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12969665229320526, | |
| "step": 4555, | |
| "valid_targets_mean": 1267.3, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 4.7899159663865545, | |
| "grad_norm": 0.8044017655815434, | |
| "learning_rate": 1.0976713873290925e-05, | |
| "loss": 0.1178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11323502659797668, | |
| "step": 4560, | |
| "valid_targets_mean": 1303.4, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 4.795168067226891, | |
| "grad_norm": 0.8200442636224358, | |
| "learning_rate": 1.0929993407746543e-05, | |
| "loss": 0.1178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10392217338085175, | |
| "step": 4565, | |
| "valid_targets_mean": 1200.9, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 4.800420168067227, | |
| "grad_norm": 0.839175470474291, | |
| "learning_rate": 1.0883335169223212e-05, | |
| "loss": 0.1287, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13048666715621948, | |
| "step": 4570, | |
| "valid_targets_mean": 1514.3, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 4.805672268907563, | |
| "grad_norm": 0.8992567495172576, | |
| "learning_rate": 1.0836739477831297e-05, | |
| "loss": 0.1282, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13388502597808838, | |
| "step": 4575, | |
| "valid_targets_mean": 1504.9, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 4.810924369747899, | |
| "grad_norm": 0.8667203247346105, | |
| "learning_rate": 1.0790206653252055e-05, | |
| "loss": 0.1295, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12979502975940704, | |
| "step": 4580, | |
| "valid_targets_mean": 1499.6, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 4.8161764705882355, | |
| "grad_norm": 0.8236209984599626, | |
| "learning_rate": 1.0743737014735434e-05, | |
| "loss": 0.1251, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11989112198352814, | |
| "step": 4585, | |
| "valid_targets_mean": 1397.2, | |
| "valid_targets_min": 736 | |
| }, | |
| { | |
| "epoch": 4.821428571428571, | |
| "grad_norm": 0.8603157391015415, | |
| "learning_rate": 1.0697330881097857e-05, | |
| "loss": 0.1304, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1339968591928482, | |
| "step": 4590, | |
| "valid_targets_mean": 1373.4, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 4.826680672268908, | |
| "grad_norm": 0.8846127818913729, | |
| "learning_rate": 1.0650988570720077e-05, | |
| "loss": 0.1246, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13660481572151184, | |
| "step": 4595, | |
| "valid_targets_mean": 1546.9, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 4.831932773109243, | |
| "grad_norm": 0.8465612563420826, | |
| "learning_rate": 1.0604710401544963e-05, | |
| "loss": 0.1193, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11273416876792908, | |
| "step": 4600, | |
| "valid_targets_mean": 1265.4, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 4.83718487394958, | |
| "grad_norm": 0.8108208139117962, | |
| "learning_rate": 1.0558496691075318e-05, | |
| "loss": 0.123, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11026652157306671, | |
| "step": 4605, | |
| "valid_targets_mean": 1368.5, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 4.842436974789916, | |
| "grad_norm": 0.8597461483981371, | |
| "learning_rate": 1.0512347756371726e-05, | |
| "loss": 0.1148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12328451871871948, | |
| "step": 4610, | |
| "valid_targets_mean": 1360.0, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 4.847689075630252, | |
| "grad_norm": 0.8960809913992954, | |
| "learning_rate": 1.0466263914050362e-05, | |
| "loss": 0.1252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1297706663608551, | |
| "step": 4615, | |
| "valid_targets_mean": 1508.2, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 4.852941176470588, | |
| "grad_norm": 0.8624681465534092, | |
| "learning_rate": 1.0420245480280804e-05, | |
| "loss": 0.124, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13054832816123962, | |
| "step": 4620, | |
| "valid_targets_mean": 1582.8, | |
| "valid_targets_min": 665 | |
| }, | |
| { | |
| "epoch": 4.858193277310924, | |
| "grad_norm": 0.9437920037528255, | |
| "learning_rate": 1.0374292770783891e-05, | |
| "loss": 0.1205, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14127075672149658, | |
| "step": 4625, | |
| "valid_targets_mean": 1475.8, | |
| "valid_targets_min": 1093 | |
| }, | |
| { | |
| "epoch": 4.86344537815126, | |
| "grad_norm": 0.8557418798703278, | |
| "learning_rate": 1.0328406100829542e-05, | |
| "loss": 0.126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1389474719762802, | |
| "step": 4630, | |
| "valid_targets_mean": 1493.8, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 4.868697478991597, | |
| "grad_norm": 0.87445714339707, | |
| "learning_rate": 1.0282585785234578e-05, | |
| "loss": 0.1263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13523709774017334, | |
| "step": 4635, | |
| "valid_targets_mean": 1507.4, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 4.873949579831933, | |
| "grad_norm": 0.8287344010980091, | |
| "learning_rate": 1.023683213836061e-05, | |
| "loss": 0.122, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11704841256141663, | |
| "step": 4640, | |
| "valid_targets_mean": 1353.1, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 4.879201680672269, | |
| "grad_norm": 0.8509384511271177, | |
| "learning_rate": 1.0191145474111823e-05, | |
| "loss": 0.1225, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11140401661396027, | |
| "step": 4645, | |
| "valid_targets_mean": 1293.3, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 4.884453781512605, | |
| "grad_norm": 0.9759630950705781, | |
| "learning_rate": 1.0145526105932872e-05, | |
| "loss": 0.1235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13949152827262878, | |
| "step": 4650, | |
| "valid_targets_mean": 1433.0, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 4.889705882352941, | |
| "grad_norm": 0.9316341415558835, | |
| "learning_rate": 1.0099974346806714e-05, | |
| "loss": 0.1242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12794744968414307, | |
| "step": 4655, | |
| "valid_targets_mean": 1385.1, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 4.894957983193278, | |
| "grad_norm": 0.8801743318163907, | |
| "learning_rate": 1.0054490509252423e-05, | |
| "loss": 0.1221, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1252344250679016, | |
| "step": 4660, | |
| "valid_targets_mean": 1543.5, | |
| "valid_targets_min": 701 | |
| }, | |
| { | |
| "epoch": 4.900210084033613, | |
| "grad_norm": 0.8895967821752813, | |
| "learning_rate": 1.0009074905323118e-05, | |
| "loss": 0.1364, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12817296385765076, | |
| "step": 4665, | |
| "valid_targets_mean": 1708.4, | |
| "valid_targets_min": 1202 | |
| }, | |
| { | |
| "epoch": 4.90546218487395, | |
| "grad_norm": 0.803668498672539, | |
| "learning_rate": 9.963727846603784e-06, | |
| "loss": 0.1265, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12417686730623245, | |
| "step": 4670, | |
| "valid_targets_mean": 1497.2, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 4.910714285714286, | |
| "grad_norm": 0.8209572965038717, | |
| "learning_rate": 9.918449644209087e-06, | |
| "loss": 0.125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11642634868621826, | |
| "step": 4675, | |
| "valid_targets_mean": 1333.4, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 4.915966386554622, | |
| "grad_norm": 0.9106589305140345, | |
| "learning_rate": 9.873240608781341e-06, | |
| "loss": 0.1266, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.126824289560318, | |
| "step": 4680, | |
| "valid_targets_mean": 1260.1, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 4.921218487394958, | |
| "grad_norm": 0.8246969179034082, | |
| "learning_rate": 9.828101050488308e-06, | |
| "loss": 0.1202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11583910137414932, | |
| "step": 4685, | |
| "valid_targets_mean": 1421.8, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 4.926470588235294, | |
| "grad_norm": 0.9467345871410094, | |
| "learning_rate": 9.783031279021063e-06, | |
| "loss": 0.1272, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.121206134557724, | |
| "step": 4690, | |
| "valid_targets_mean": 1255.4, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 4.93172268907563, | |
| "grad_norm": 0.8946786916723871, | |
| "learning_rate": 9.738031603591926e-06, | |
| "loss": 0.1217, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12146471440792084, | |
| "step": 4695, | |
| "valid_targets_mean": 1334.6, | |
| "valid_targets_min": 821 | |
| }, | |
| { | |
| "epoch": 4.936974789915967, | |
| "grad_norm": 0.8415482635056759, | |
| "learning_rate": 9.69310233293227e-06, | |
| "loss": 0.1194, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12440873682498932, | |
| "step": 4700, | |
| "valid_targets_mean": 1562.7, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 4.942226890756302, | |
| "grad_norm": 0.8253084558385871, | |
| "learning_rate": 9.648243775290476e-06, | |
| "loss": 0.1203, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12124086171388626, | |
| "step": 4705, | |
| "valid_targets_mean": 1435.9, | |
| "valid_targets_min": 781 | |
| }, | |
| { | |
| "epoch": 4.947478991596639, | |
| "grad_norm": 0.8212423721287204, | |
| "learning_rate": 9.60345623842974e-06, | |
| "loss": 0.1296, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10891197621822357, | |
| "step": 4710, | |
| "valid_targets_mean": 1341.2, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 4.9527310924369745, | |
| "grad_norm": 0.7441257737363668, | |
| "learning_rate": 9.558740029626046e-06, | |
| "loss": 0.12, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10063184797763824, | |
| "step": 4715, | |
| "valid_targets_mean": 1319.9, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 4.957983193277311, | |
| "grad_norm": 0.8222604138587376, | |
| "learning_rate": 9.51409545566597e-06, | |
| "loss": 0.1236, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12508775293827057, | |
| "step": 4720, | |
| "valid_targets_mean": 1657.6, | |
| "valid_targets_min": 868 | |
| }, | |
| { | |
| "epoch": 4.963235294117647, | |
| "grad_norm": 0.8441572887401301, | |
| "learning_rate": 9.469522822844663e-06, | |
| "loss": 0.1134, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12001735717058182, | |
| "step": 4725, | |
| "valid_targets_mean": 1566.1, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 4.968487394957983, | |
| "grad_norm": 0.8622442317465762, | |
| "learning_rate": 9.425022436963664e-06, | |
| "loss": 0.1153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13704422116279602, | |
| "step": 4730, | |
| "valid_targets_mean": 1575.5, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 4.973739495798319, | |
| "grad_norm": 0.8300300731222955, | |
| "learning_rate": 9.380594603328875e-06, | |
| "loss": 0.1228, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11850959062576294, | |
| "step": 4735, | |
| "valid_targets_mean": 1568.6, | |
| "valid_targets_min": 639 | |
| }, | |
| { | |
| "epoch": 4.9789915966386555, | |
| "grad_norm": 0.8820976325913817, | |
| "learning_rate": 9.336239626748432e-06, | |
| "loss": 0.1199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12084612995386124, | |
| "step": 4740, | |
| "valid_targets_mean": 1482.9, | |
| "valid_targets_min": 920 | |
| }, | |
| { | |
| "epoch": 4.984243697478991, | |
| "grad_norm": 0.8235056111823619, | |
| "learning_rate": 9.291957811530602e-06, | |
| "loss": 0.1149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11309683322906494, | |
| "step": 4745, | |
| "valid_targets_mean": 1354.8, | |
| "valid_targets_min": 847 | |
| }, | |
| { | |
| "epoch": 4.989495798319328, | |
| "grad_norm": 0.8380994731336534, | |
| "learning_rate": 9.247749461481712e-06, | |
| "loss": 0.1207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12259923666715622, | |
| "step": 4750, | |
| "valid_targets_mean": 1451.1, | |
| "valid_targets_min": 557 | |
| }, | |
| { | |
| "epoch": 4.994747899159664, | |
| "grad_norm": 0.9271264619980075, | |
| "learning_rate": 9.203614879904083e-06, | |
| "loss": 0.1212, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12632964551448822, | |
| "step": 4755, | |
| "valid_targets_mean": 1332.4, | |
| "valid_targets_min": 645 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.7766606123712497, | |
| "learning_rate": 9.159554369593901e-06, | |
| "loss": 0.1228, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11304584890604019, | |
| "step": 4760, | |
| "valid_targets_mean": 1721.1, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 5.005252100840337, | |
| "grad_norm": 1.0368311360471962, | |
| "learning_rate": 9.115568232839193e-06, | |
| "loss": 0.2281, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2315981090068817, | |
| "step": 4765, | |
| "valid_targets_mean": 3398.8, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 5.010504201680672, | |
| "grad_norm": 0.6460571203150131, | |
| "learning_rate": 9.071656771417721e-06, | |
| "loss": 0.207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17781823873519897, | |
| "step": 4770, | |
| "valid_targets_mean": 4347.4, | |
| "valid_targets_min": 1651 | |
| }, | |
| { | |
| "epoch": 5.015756302521009, | |
| "grad_norm": 0.542133417895063, | |
| "learning_rate": 9.027820286594902e-06, | |
| "loss": 0.2101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1896631270647049, | |
| "step": 4775, | |
| "valid_targets_mean": 4633.2, | |
| "valid_targets_min": 1560 | |
| }, | |
| { | |
| "epoch": 5.0210084033613445, | |
| "grad_norm": 0.6608510247217927, | |
| "learning_rate": 8.984059079121785e-06, | |
| "loss": 0.2035, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19018647074699402, | |
| "step": 4780, | |
| "valid_targets_mean": 3938.1, | |
| "valid_targets_min": 1841 | |
| }, | |
| { | |
| "epoch": 5.026260504201681, | |
| "grad_norm": 0.6358924043276708, | |
| "learning_rate": 8.940373449232935e-06, | |
| "loss": 0.2111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22464792430400848, | |
| "step": 4785, | |
| "valid_targets_mean": 3724.8, | |
| "valid_targets_min": 1374 | |
| }, | |
| { | |
| "epoch": 5.031512605042017, | |
| "grad_norm": 0.6599880706584261, | |
| "learning_rate": 8.896763696644405e-06, | |
| "loss": 0.2016, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18435746431350708, | |
| "step": 4790, | |
| "valid_targets_mean": 3264.8, | |
| "valid_targets_min": 1089 | |
| }, | |
| { | |
| "epoch": 5.036764705882353, | |
| "grad_norm": 0.7096826802832755, | |
| "learning_rate": 8.853230120551693e-06, | |
| "loss": 0.2038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21627312898635864, | |
| "step": 4795, | |
| "valid_targets_mean": 3374.9, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 5.042016806722689, | |
| "grad_norm": 0.6612340433792787, | |
| "learning_rate": 8.809773019627635e-06, | |
| "loss": 0.2061, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21441474556922913, | |
| "step": 4800, | |
| "valid_targets_mean": 3514.9, | |
| "valid_targets_min": 1574 | |
| }, | |
| { | |
| "epoch": 5.0472689075630255, | |
| "grad_norm": 0.6216659402671705, | |
| "learning_rate": 8.766392692020413e-06, | |
| "loss": 0.2138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21736544370651245, | |
| "step": 4805, | |
| "valid_targets_mean": 3956.7, | |
| "valid_targets_min": 1899 | |
| }, | |
| { | |
| "epoch": 5.052521008403361, | |
| "grad_norm": 0.6397951889136652, | |
| "learning_rate": 8.723089435351497e-06, | |
| "loss": 0.2059, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.213152214884758, | |
| "step": 4810, | |
| "valid_targets_mean": 4397.6, | |
| "valid_targets_min": 2261 | |
| }, | |
| { | |
| "epoch": 5.057773109243698, | |
| "grad_norm": 0.6827468645370026, | |
| "learning_rate": 8.679863546713559e-06, | |
| "loss": 0.2105, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19350934028625488, | |
| "step": 4815, | |
| "valid_targets_mean": 3263.8, | |
| "valid_targets_min": 1873 | |
| }, | |
| { | |
| "epoch": 5.063025210084033, | |
| "grad_norm": 0.6432078260343645, | |
| "learning_rate": 8.636715322668498e-06, | |
| "loss": 0.212, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2112460732460022, | |
| "step": 4820, | |
| "valid_targets_mean": 3430.2, | |
| "valid_targets_min": 1013 | |
| }, | |
| { | |
| "epoch": 5.06827731092437, | |
| "grad_norm": 0.5998462380114775, | |
| "learning_rate": 8.593645059245365e-06, | |
| "loss": 0.1989, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1965964436531067, | |
| "step": 4825, | |
| "valid_targets_mean": 4041.6, | |
| "valid_targets_min": 1366 | |
| }, | |
| { | |
| "epoch": 5.073529411764706, | |
| "grad_norm": 0.6552876974701827, | |
| "learning_rate": 8.550653051938333e-06, | |
| "loss": 0.2088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2239168882369995, | |
| "step": 4830, | |
| "valid_targets_mean": 3515.2, | |
| "valid_targets_min": 2050 | |
| }, | |
| { | |
| "epoch": 5.078781512605042, | |
| "grad_norm": 0.7164022528196604, | |
| "learning_rate": 8.507739595704695e-06, | |
| "loss": 0.2149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21687167882919312, | |
| "step": 4835, | |
| "valid_targets_mean": 3048.9, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 5.084033613445378, | |
| "grad_norm": 0.7024986732808598, | |
| "learning_rate": 8.464904984962832e-06, | |
| "loss": 0.2057, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2093815803527832, | |
| "step": 4840, | |
| "valid_targets_mean": 2982.6, | |
| "valid_targets_min": 855 | |
| }, | |
| { | |
| "epoch": 5.089285714285714, | |
| "grad_norm": 0.6486210855345914, | |
| "learning_rate": 8.422149513590151e-06, | |
| "loss": 0.2096, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22854307293891907, | |
| "step": 4845, | |
| "valid_targets_mean": 3753.4, | |
| "valid_targets_min": 2125 | |
| }, | |
| { | |
| "epoch": 5.09453781512605, | |
| "grad_norm": 0.7106124392970365, | |
| "learning_rate": 8.37947347492115e-06, | |
| "loss": 0.2146, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20508921146392822, | |
| "step": 4850, | |
| "valid_targets_mean": 2946.8, | |
| "valid_targets_min": 1173 | |
| }, | |
| { | |
| "epoch": 5.099789915966387, | |
| "grad_norm": 0.7291320506136616, | |
| "learning_rate": 8.33687716174532e-06, | |
| "loss": 0.2051, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20715506374835968, | |
| "step": 4855, | |
| "valid_targets_mean": 3284.2, | |
| "valid_targets_min": 1438 | |
| }, | |
| { | |
| "epoch": 5.105042016806722, | |
| "grad_norm": 0.667621729812997, | |
| "learning_rate": 8.294360866305192e-06, | |
| "loss": 0.2038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18001650273799896, | |
| "step": 4860, | |
| "valid_targets_mean": 3818.4, | |
| "valid_targets_min": 1133 | |
| }, | |
| { | |
| "epoch": 5.110294117647059, | |
| "grad_norm": 0.6904135635931854, | |
| "learning_rate": 8.251924880294317e-06, | |
| "loss": 0.2068, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1863369643688202, | |
| "step": 4865, | |
| "valid_targets_mean": 3339.8, | |
| "valid_targets_min": 1051 | |
| }, | |
| { | |
| "epoch": 5.1155462184873945, | |
| "grad_norm": 0.708134359007726, | |
| "learning_rate": 8.20956949485527e-06, | |
| "loss": 0.202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2200697362422943, | |
| "step": 4870, | |
| "valid_targets_mean": 3285.1, | |
| "valid_targets_min": 1642 | |
| }, | |
| { | |
| "epoch": 5.120798319327731, | |
| "grad_norm": 0.6714863526693571, | |
| "learning_rate": 8.167295000577622e-06, | |
| "loss": 0.2061, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19819751381874084, | |
| "step": 4875, | |
| "valid_targets_mean": 3340.8, | |
| "valid_targets_min": 1747 | |
| }, | |
| { | |
| "epoch": 5.126050420168067, | |
| "grad_norm": 0.6991162613090433, | |
| "learning_rate": 8.125101687496e-06, | |
| "loss": 0.2163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2069099098443985, | |
| "step": 4880, | |
| "valid_targets_mean": 3218.2, | |
| "valid_targets_min": 1765 | |
| }, | |
| { | |
| "epoch": 5.131302521008403, | |
| "grad_norm": 0.6933394799250305, | |
| "learning_rate": 8.082989845088038e-06, | |
| "loss": 0.2094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20274317264556885, | |
| "step": 4885, | |
| "valid_targets_mean": 3182.9, | |
| "valid_targets_min": 1233 | |
| }, | |
| { | |
| "epoch": 5.13655462184874, | |
| "grad_norm": 0.7022583842802075, | |
| "learning_rate": 8.040959762272441e-06, | |
| "loss": 0.195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19732248783111572, | |
| "step": 4890, | |
| "valid_targets_mean": 3020.8, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 5.141806722689076, | |
| "grad_norm": 0.6578222367990266, | |
| "learning_rate": 7.999011727406995e-06, | |
| "loss": 0.2022, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1947716623544693, | |
| "step": 4895, | |
| "valid_targets_mean": 3269.2, | |
| "valid_targets_min": 926 | |
| }, | |
| { | |
| "epoch": 5.147058823529412, | |
| "grad_norm": 0.6729734257141221, | |
| "learning_rate": 7.957146028286524e-06, | |
| "loss": 0.2027, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1842532753944397, | |
| "step": 4900, | |
| "valid_targets_mean": 3256.4, | |
| "valid_targets_min": 1016 | |
| }, | |
| { | |
| "epoch": 5.152310924369748, | |
| "grad_norm": 0.6998982376570767, | |
| "learning_rate": 7.915362952141017e-06, | |
| "loss": 0.2048, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19443146884441376, | |
| "step": 4905, | |
| "valid_targets_mean": 3058.4, | |
| "valid_targets_min": 1367 | |
| }, | |
| { | |
| "epoch": 5.157563025210084, | |
| "grad_norm": 0.6159690568204116, | |
| "learning_rate": 7.873662785633594e-06, | |
| "loss": 0.2023, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1794825792312622, | |
| "step": 4910, | |
| "valid_targets_mean": 3513.8, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 5.16281512605042, | |
| "grad_norm": 0.718929827088859, | |
| "learning_rate": 7.832045814858538e-06, | |
| "loss": 0.1975, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19105938076972961, | |
| "step": 4915, | |
| "valid_targets_mean": 2834.2, | |
| "valid_targets_min": 1480 | |
| }, | |
| { | |
| "epoch": 5.168067226890757, | |
| "grad_norm": 0.6775963512647579, | |
| "learning_rate": 7.790512325339367e-06, | |
| "loss": 0.1982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18776269257068634, | |
| "step": 4920, | |
| "valid_targets_mean": 2854.1, | |
| "valid_targets_min": 1229 | |
| }, | |
| { | |
| "epoch": 5.173319327731092, | |
| "grad_norm": 0.7289317859752761, | |
| "learning_rate": 7.74906260202685e-06, | |
| "loss": 0.209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19116216897964478, | |
| "step": 4925, | |
| "valid_targets_mean": 3215.8, | |
| "valid_targets_min": 1727 | |
| }, | |
| { | |
| "epoch": 5.178571428571429, | |
| "grad_norm": 0.6500227003936777, | |
| "learning_rate": 7.707696929297053e-06, | |
| "loss": 0.204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1965217888355255, | |
| "step": 4930, | |
| "valid_targets_mean": 3578.9, | |
| "valid_targets_min": 1860 | |
| }, | |
| { | |
| "epoch": 5.1838235294117645, | |
| "grad_norm": 0.7034986881633343, | |
| "learning_rate": 7.666415590949382e-06, | |
| "loss": 0.2052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20505331456661224, | |
| "step": 4935, | |
| "valid_targets_mean": 3121.1, | |
| "valid_targets_min": 1145 | |
| }, | |
| { | |
| "epoch": 5.189075630252101, | |
| "grad_norm": 0.8145007649387389, | |
| "learning_rate": 7.625218870204676e-06, | |
| "loss": 0.2074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20866146683692932, | |
| "step": 4940, | |
| "valid_targets_mean": 2910.1, | |
| "valid_targets_min": 1195 | |
| }, | |
| { | |
| "epoch": 5.194327731092437, | |
| "grad_norm": 0.6372684720958994, | |
| "learning_rate": 7.584107049703195e-06, | |
| "loss": 0.1956, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17585617303848267, | |
| "step": 4945, | |
| "valid_targets_mean": 3215.2, | |
| "valid_targets_min": 1087 | |
| }, | |
| { | |
| "epoch": 5.199579831932773, | |
| "grad_norm": 0.6712839739955457, | |
| "learning_rate": 7.5430804115027615e-06, | |
| "loss": 0.2045, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19442957639694214, | |
| "step": 4950, | |
| "valid_targets_mean": 3842.3, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 5.204831932773109, | |
| "grad_norm": 0.7139371019585867, | |
| "learning_rate": 7.502139237076744e-06, | |
| "loss": 0.1986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18818247318267822, | |
| "step": 4955, | |
| "valid_targets_mean": 3112.6, | |
| "valid_targets_min": 1097 | |
| }, | |
| { | |
| "epoch": 5.2100840336134455, | |
| "grad_norm": 0.6441054449323667, | |
| "learning_rate": 7.461283807312194e-06, | |
| "loss": 0.1961, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20336642861366272, | |
| "step": 4960, | |
| "valid_targets_mean": 3587.6, | |
| "valid_targets_min": 1833 | |
| }, | |
| { | |
| "epoch": 5.215336134453781, | |
| "grad_norm": 0.7139397353122559, | |
| "learning_rate": 7.420514402507886e-06, | |
| "loss": 0.2008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21995452046394348, | |
| "step": 4965, | |
| "valid_targets_mean": 3660.6, | |
| "valid_targets_min": 1182 | |
| }, | |
| { | |
| "epoch": 5.220588235294118, | |
| "grad_norm": 0.6921935226344126, | |
| "learning_rate": 7.379831302372389e-06, | |
| "loss": 0.2009, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2377263605594635, | |
| "step": 4970, | |
| "valid_targets_mean": 3839.4, | |
| "valid_targets_min": 1586 | |
| }, | |
| { | |
| "epoch": 5.225840336134453, | |
| "grad_norm": 0.78550496353629, | |
| "learning_rate": 7.3392347860221556e-06, | |
| "loss": 0.2021, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2115703821182251, | |
| "step": 4975, | |
| "valid_targets_mean": 3062.7, | |
| "valid_targets_min": 1714 | |
| }, | |
| { | |
| "epoch": 5.23109243697479, | |
| "grad_norm": 0.717261721079846, | |
| "learning_rate": 7.298725131979629e-06, | |
| "loss": 0.1999, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2131146639585495, | |
| "step": 4980, | |
| "valid_targets_mean": 3611.1, | |
| "valid_targets_min": 1504 | |
| }, | |
| { | |
| "epoch": 5.236344537815126, | |
| "grad_norm": 0.6319115820709261, | |
| "learning_rate": 7.258302618171287e-06, | |
| "loss": 0.1956, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1644587367773056, | |
| "step": 4985, | |
| "valid_targets_mean": 3457.1, | |
| "valid_targets_min": 1170 | |
| }, | |
| { | |
| "epoch": 5.241596638655462, | |
| "grad_norm": 0.760031676694523, | |
| "learning_rate": 7.217967521925775e-06, | |
| "loss": 0.1937, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20226861536502838, | |
| "step": 4990, | |
| "valid_targets_mean": 3089.3, | |
| "valid_targets_min": 1436 | |
| }, | |
| { | |
| "epoch": 5.246848739495798, | |
| "grad_norm": 0.9470898190916125, | |
| "learning_rate": 7.177720119971998e-06, | |
| "loss": 0.1967, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19645410776138306, | |
| "step": 4995, | |
| "valid_targets_mean": 3257.2, | |
| "valid_targets_min": 1302 | |
| }, | |
| { | |
| "epoch": 5.2521008403361344, | |
| "grad_norm": 0.706018495456836, | |
| "learning_rate": 7.137560688437184e-06, | |
| "loss": 0.1976, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21298207342624664, | |
| "step": 5000, | |
| "valid_targets_mean": 3425.1, | |
| "valid_targets_min": 2023 | |
| }, | |
| { | |
| "epoch": 5.257352941176471, | |
| "grad_norm": 0.6165263612026762, | |
| "learning_rate": 7.097489502845047e-06, | |
| "loss": 0.1802, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1659245491027832, | |
| "step": 5005, | |
| "valid_targets_mean": 3196.4, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 5.262605042016807, | |
| "grad_norm": 0.7663728544419008, | |
| "learning_rate": 7.0575068381138525e-06, | |
| "loss": 0.211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19875086843967438, | |
| "step": 5010, | |
| "valid_targets_mean": 3197.8, | |
| "valid_targets_min": 1059 | |
| }, | |
| { | |
| "epoch": 5.267857142857143, | |
| "grad_norm": 0.6844105981294532, | |
| "learning_rate": 7.0176129685545414e-06, | |
| "loss": 0.1926, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16719815135002136, | |
| "step": 5015, | |
| "valid_targets_mean": 3301.7, | |
| "valid_targets_min": 964 | |
| }, | |
| { | |
| "epoch": 5.273109243697479, | |
| "grad_norm": 0.6796898164555453, | |
| "learning_rate": 6.977808167868867e-06, | |
| "loss": 0.2043, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1949189305305481, | |
| "step": 5020, | |
| "valid_targets_mean": 3694.2, | |
| "valid_targets_min": 1105 | |
| }, | |
| { | |
| "epoch": 5.2783613445378155, | |
| "grad_norm": 0.8023813457851872, | |
| "learning_rate": 6.9380927091475085e-06, | |
| "loss": 0.2102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20866572856903076, | |
| "step": 5025, | |
| "valid_targets_mean": 2612.6, | |
| "valid_targets_min": 1491 | |
| }, | |
| { | |
| "epoch": 5.283613445378151, | |
| "grad_norm": 0.762817989942347, | |
| "learning_rate": 6.898466864868165e-06, | |
| "loss": 0.2144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20279455184936523, | |
| "step": 5030, | |
| "valid_targets_mean": 3110.5, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 5.288865546218488, | |
| "grad_norm": 0.7866878344731686, | |
| "learning_rate": 6.858930906893751e-06, | |
| "loss": 0.1934, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19141477346420288, | |
| "step": 5035, | |
| "valid_targets_mean": 3259.2, | |
| "valid_targets_min": 1822 | |
| }, | |
| { | |
| "epoch": 5.294117647058823, | |
| "grad_norm": 0.7702679497523474, | |
| "learning_rate": 6.819485106470454e-06, | |
| "loss": 0.1944, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1862347275018692, | |
| "step": 5040, | |
| "valid_targets_mean": 3547.6, | |
| "valid_targets_min": 1625 | |
| }, | |
| { | |
| "epoch": 5.29936974789916, | |
| "grad_norm": 0.6629392322076868, | |
| "learning_rate": 6.780129734225949e-06, | |
| "loss": 0.1864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17328760027885437, | |
| "step": 5045, | |
| "valid_targets_mean": 3628.2, | |
| "valid_targets_min": 1415 | |
| }, | |
| { | |
| "epoch": 5.304621848739496, | |
| "grad_norm": 0.7042102260307026, | |
| "learning_rate": 6.740865060167483e-06, | |
| "loss": 0.1917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19156163930892944, | |
| "step": 5050, | |
| "valid_targets_mean": 3398.9, | |
| "valid_targets_min": 1460 | |
| }, | |
| { | |
| "epoch": 5.309873949579832, | |
| "grad_norm": 0.7519060990832099, | |
| "learning_rate": 6.701691353680049e-06, | |
| "loss": 0.1975, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19511482119560242, | |
| "step": 5055, | |
| "valid_targets_mean": 2947.4, | |
| "valid_targets_min": 1079 | |
| }, | |
| { | |
| "epoch": 5.315126050420168, | |
| "grad_norm": 0.727969914435581, | |
| "learning_rate": 6.662608883524542e-06, | |
| "loss": 0.1769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18351078033447266, | |
| "step": 5060, | |
| "valid_targets_mean": 3003.0, | |
| "valid_targets_min": 1274 | |
| }, | |
| { | |
| "epoch": 5.320378151260504, | |
| "grad_norm": 0.8817302867844496, | |
| "learning_rate": 6.6236179178359115e-06, | |
| "loss": 0.1928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21124529838562012, | |
| "step": 5065, | |
| "valid_targets_mean": 3306.3, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 5.32563025210084, | |
| "grad_norm": 0.713024621019574, | |
| "learning_rate": 6.5847187241213e-06, | |
| "loss": 0.2055, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19047285616397858, | |
| "step": 5070, | |
| "valid_targets_mean": 3332.2, | |
| "valid_targets_min": 1312 | |
| }, | |
| { | |
| "epoch": 5.330882352941177, | |
| "grad_norm": 0.752933122434432, | |
| "learning_rate": 6.5459115692582386e-06, | |
| "loss": 0.1924, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1971365213394165, | |
| "step": 5075, | |
| "valid_targets_mean": 2601.3, | |
| "valid_targets_min": 987 | |
| }, | |
| { | |
| "epoch": 5.336134453781512, | |
| "grad_norm": 0.624699265138433, | |
| "learning_rate": 6.507196719492819e-06, | |
| "loss": 0.1949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1911473274230957, | |
| "step": 5080, | |
| "valid_targets_mean": 4341.5, | |
| "valid_targets_min": 1856 | |
| }, | |
| { | |
| "epoch": 5.341386554621849, | |
| "grad_norm": 0.7283950240407708, | |
| "learning_rate": 6.468574440437801e-06, | |
| "loss": 0.1966, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19926682114601135, | |
| "step": 5085, | |
| "valid_targets_mean": 3290.9, | |
| "valid_targets_min": 1601 | |
| }, | |
| { | |
| "epoch": 5.3466386554621845, | |
| "grad_norm": 0.7710715895807497, | |
| "learning_rate": 6.43004499707089e-06, | |
| "loss": 0.204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20834487676620483, | |
| "step": 5090, | |
| "valid_targets_mean": 3119.4, | |
| "valid_targets_min": 1986 | |
| }, | |
| { | |
| "epoch": 5.351890756302521, | |
| "grad_norm": 0.7170439306426001, | |
| "learning_rate": 6.3916086537328545e-06, | |
| "loss": 0.1858, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21489310264587402, | |
| "step": 5095, | |
| "valid_targets_mean": 3777.1, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 5.357142857142857, | |
| "grad_norm": 0.6744006460741591, | |
| "learning_rate": 6.3532656741257145e-06, | |
| "loss": 0.1852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16385085880756378, | |
| "step": 5100, | |
| "valid_targets_mean": 2878.0, | |
| "valid_targets_min": 512 | |
| }, | |
| { | |
| "epoch": 5.362394957983193, | |
| "grad_norm": 0.673275766985947, | |
| "learning_rate": 6.315016321310965e-06, | |
| "loss": 0.1902, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1753840446472168, | |
| "step": 5105, | |
| "valid_targets_mean": 3215.4, | |
| "valid_targets_min": 1202 | |
| }, | |
| { | |
| "epoch": 5.367647058823529, | |
| "grad_norm": 0.7266719247932443, | |
| "learning_rate": 6.276860857707727e-06, | |
| "loss": 0.2117, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20921765267848969, | |
| "step": 5110, | |
| "valid_targets_mean": 3693.1, | |
| "valid_targets_min": 1455 | |
| }, | |
| { | |
| "epoch": 5.3728991596638656, | |
| "grad_norm": 0.6960543396247455, | |
| "learning_rate": 6.238799545090994e-06, | |
| "loss": 0.1953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18564675748348236, | |
| "step": 5115, | |
| "valid_targets_mean": 3477.5, | |
| "valid_targets_min": 1571 | |
| }, | |
| { | |
| "epoch": 5.378151260504202, | |
| "grad_norm": 0.789570094298452, | |
| "learning_rate": 6.20083264458981e-06, | |
| "loss": 0.2082, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22905242443084717, | |
| "step": 5120, | |
| "valid_targets_mean": 2912.1, | |
| "valid_targets_min": 1450 | |
| }, | |
| { | |
| "epoch": 5.383403361344538, | |
| "grad_norm": 0.687911349078074, | |
| "learning_rate": 6.162960416685469e-06, | |
| "loss": 0.1928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18090611696243286, | |
| "step": 5125, | |
| "valid_targets_mean": 3323.0, | |
| "valid_targets_min": 1431 | |
| }, | |
| { | |
| "epoch": 5.388655462184874, | |
| "grad_norm": 0.8129192543706223, | |
| "learning_rate": 6.125183121209741e-06, | |
| "loss": 0.2089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2075553834438324, | |
| "step": 5130, | |
| "valid_targets_mean": 3461.4, | |
| "valid_targets_min": 1147 | |
| }, | |
| { | |
| "epoch": 5.39390756302521, | |
| "grad_norm": 0.7873625003498075, | |
| "learning_rate": 6.087501017343107e-06, | |
| "loss": 0.2065, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2097611427307129, | |
| "step": 5135, | |
| "valid_targets_mean": 2641.6, | |
| "valid_targets_min": 998 | |
| }, | |
| { | |
| "epoch": 5.399159663865547, | |
| "grad_norm": 0.7237167630580533, | |
| "learning_rate": 6.0499143636129384e-06, | |
| "loss": 0.199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20370076596736908, | |
| "step": 5140, | |
| "valid_targets_mean": 3260.9, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 5.404411764705882, | |
| "grad_norm": 0.85938348372471, | |
| "learning_rate": 6.012423417891766e-06, | |
| "loss": 0.1987, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20711749792099, | |
| "step": 5145, | |
| "valid_targets_mean": 2796.2, | |
| "valid_targets_min": 1005 | |
| }, | |
| { | |
| "epoch": 5.409663865546219, | |
| "grad_norm": 0.6968015629432648, | |
| "learning_rate": 5.975028437395493e-06, | |
| "loss": 0.1985, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2024056762456894, | |
| "step": 5150, | |
| "valid_targets_mean": 3183.6, | |
| "valid_targets_min": 2298 | |
| }, | |
| { | |
| "epoch": 5.4149159663865545, | |
| "grad_norm": 0.7775633759987758, | |
| "learning_rate": 5.937729678681612e-06, | |
| "loss": 0.207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20395274460315704, | |
| "step": 5155, | |
| "valid_targets_mean": 2821.2, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 5.420168067226891, | |
| "grad_norm": 0.7494510775854252, | |
| "learning_rate": 5.9005273976474684e-06, | |
| "loss": 0.1891, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18208947777748108, | |
| "step": 5160, | |
| "valid_targets_mean": 3265.2, | |
| "valid_targets_min": 1665 | |
| }, | |
| { | |
| "epoch": 5.425420168067227, | |
| "grad_norm": 0.824725420965561, | |
| "learning_rate": 5.8634218495285126e-06, | |
| "loss": 0.2046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20658603310585022, | |
| "step": 5165, | |
| "valid_targets_mean": 3019.3, | |
| "valid_targets_min": 1076 | |
| }, | |
| { | |
| "epoch": 5.430672268907563, | |
| "grad_norm": 0.706863386587082, | |
| "learning_rate": 5.8264132888965155e-06, | |
| "loss": 0.193, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19833385944366455, | |
| "step": 5170, | |
| "valid_targets_mean": 3610.8, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 5.435924369747899, | |
| "grad_norm": 0.7103390006031945, | |
| "learning_rate": 5.789501969657853e-06, | |
| "loss": 0.2088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1992931365966797, | |
| "step": 5175, | |
| "valid_targets_mean": 3255.8, | |
| "valid_targets_min": 1728 | |
| }, | |
| { | |
| "epoch": 5.4411764705882355, | |
| "grad_norm": 0.6848028235034054, | |
| "learning_rate": 5.752688145051761e-06, | |
| "loss": 0.2018, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18287193775177002, | |
| "step": 5180, | |
| "valid_targets_mean": 3377.2, | |
| "valid_targets_min": 981 | |
| }, | |
| { | |
| "epoch": 5.446428571428571, | |
| "grad_norm": 0.6806365974211747, | |
| "learning_rate": 5.715972067648561e-06, | |
| "loss": 0.1865, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1667267084121704, | |
| "step": 5185, | |
| "valid_targets_mean": 2839.1, | |
| "valid_targets_min": 1351 | |
| }, | |
| { | |
| "epoch": 5.451680672268908, | |
| "grad_norm": 0.6810171841346309, | |
| "learning_rate": 5.679353989347989e-06, | |
| "loss": 0.1968, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20622935891151428, | |
| "step": 5190, | |
| "valid_targets_mean": 4396.7, | |
| "valid_targets_min": 2416 | |
| }, | |
| { | |
| "epoch": 5.456932773109243, | |
| "grad_norm": 0.6886243059430002, | |
| "learning_rate": 5.642834161377408e-06, | |
| "loss": 0.1913, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1882258951663971, | |
| "step": 5195, | |
| "valid_targets_mean": 4668.6, | |
| "valid_targets_min": 1233 | |
| }, | |
| { | |
| "epoch": 5.46218487394958, | |
| "grad_norm": 0.660869505715319, | |
| "learning_rate": 5.606412834290121e-06, | |
| "loss": 0.1875, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1740989089012146, | |
| "step": 5200, | |
| "valid_targets_mean": 3936.4, | |
| "valid_targets_min": 1524 | |
| }, | |
| { | |
| "epoch": 5.467436974789916, | |
| "grad_norm": 0.6389923753551053, | |
| "learning_rate": 5.570090257963645e-06, | |
| "loss": 0.1911, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18670515716075897, | |
| "step": 5205, | |
| "valid_targets_mean": 4389.3, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 5.472689075630252, | |
| "grad_norm": 0.7185606904022306, | |
| "learning_rate": 5.533866681597981e-06, | |
| "loss": 0.1893, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18405769765377045, | |
| "step": 5210, | |
| "valid_targets_mean": 3241.3, | |
| "valid_targets_min": 906 | |
| }, | |
| { | |
| "epoch": 5.477941176470588, | |
| "grad_norm": 0.796159259324853, | |
| "learning_rate": 5.497742353713928e-06, | |
| "loss": 0.1852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20212048292160034, | |
| "step": 5215, | |
| "valid_targets_mean": 3680.4, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 5.483193277310924, | |
| "grad_norm": 0.6843887461039875, | |
| "learning_rate": 5.461717522151371e-06, | |
| "loss": 0.1767, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1704184114933014, | |
| "step": 5220, | |
| "valid_targets_mean": 3168.9, | |
| "valid_targets_min": 1484 | |
| }, | |
| { | |
| "epoch": 5.48844537815126, | |
| "grad_norm": 0.81830636371663, | |
| "learning_rate": 5.4257924340675535e-06, | |
| "loss": 0.188, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18880800902843475, | |
| "step": 5225, | |
| "valid_targets_mean": 3667.5, | |
| "valid_targets_min": 1440 | |
| }, | |
| { | |
| "epoch": 5.493697478991597, | |
| "grad_norm": 0.761192266055787, | |
| "learning_rate": 5.389967335935429e-06, | |
| "loss": 0.2039, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2186485230922699, | |
| "step": 5230, | |
| "valid_targets_mean": 3025.6, | |
| "valid_targets_min": 1349 | |
| }, | |
| { | |
| "epoch": 5.498949579831933, | |
| "grad_norm": 0.7532532316921411, | |
| "learning_rate": 5.3542424735419306e-06, | |
| "loss": 0.1948, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1849045306444168, | |
| "step": 5235, | |
| "valid_targets_mean": 2935.8, | |
| "valid_targets_min": 1156 | |
| }, | |
| { | |
| "epoch": 5.504201680672269, | |
| "grad_norm": 0.7266362371210848, | |
| "learning_rate": 5.318618091986294e-06, | |
| "loss": 0.1904, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17571350932121277, | |
| "step": 5240, | |
| "valid_targets_mean": 3190.4, | |
| "valid_targets_min": 993 | |
| }, | |
| { | |
| "epoch": 5.509453781512605, | |
| "grad_norm": 0.7883604048005873, | |
| "learning_rate": 5.283094435678398e-06, | |
| "loss": 0.1845, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20258891582489014, | |
| "step": 5245, | |
| "valid_targets_mean": 2785.6, | |
| "valid_targets_min": 1275 | |
| }, | |
| { | |
| "epoch": 5.514705882352941, | |
| "grad_norm": 0.738824710435114, | |
| "learning_rate": 5.2476717483370685e-06, | |
| "loss": 0.1907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1991308033466339, | |
| "step": 5250, | |
| "valid_targets_mean": 3266.8, | |
| "valid_targets_min": 1669 | |
| }, | |
| { | |
| "epoch": 5.519957983193278, | |
| "grad_norm": 0.7925166684746748, | |
| "learning_rate": 5.212350272988394e-06, | |
| "loss": 0.1974, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1969340294599533, | |
| "step": 5255, | |
| "valid_targets_mean": 3357.5, | |
| "valid_targets_min": 1710 | |
| }, | |
| { | |
| "epoch": 5.525210084033613, | |
| "grad_norm": 0.6825177393092328, | |
| "learning_rate": 5.177130251964088e-06, | |
| "loss": 0.1917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18405650556087494, | |
| "step": 5260, | |
| "valid_targets_mean": 3468.1, | |
| "valid_targets_min": 1685 | |
| }, | |
| { | |
| "epoch": 5.53046218487395, | |
| "grad_norm": 0.6591391403288339, | |
| "learning_rate": 5.1420119268998085e-06, | |
| "loss": 0.1951, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19749081134796143, | |
| "step": 5265, | |
| "valid_targets_mean": 3659.6, | |
| "valid_targets_min": 1931 | |
| }, | |
| { | |
| "epoch": 5.535714285714286, | |
| "grad_norm": 0.6736986111232942, | |
| "learning_rate": 5.106995538733488e-06, | |
| "loss": 0.1919, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18770131468772888, | |
| "step": 5270, | |
| "valid_targets_mean": 3862.3, | |
| "valid_targets_min": 2009 | |
| }, | |
| { | |
| "epoch": 5.540966386554622, | |
| "grad_norm": 0.8741377267243634, | |
| "learning_rate": 5.07208132770372e-06, | |
| "loss": 0.1753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16808158159255981, | |
| "step": 5275, | |
| "valid_targets_mean": 1792.6, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 5.546218487394958, | |
| "grad_norm": 0.8019745661839314, | |
| "learning_rate": 5.037269533348075e-06, | |
| "loss": 0.1325, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12899786233901978, | |
| "step": 5280, | |
| "valid_targets_mean": 1354.7, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 5.551470588235294, | |
| "grad_norm": 0.9281156697299153, | |
| "learning_rate": 5.002560394501463e-06, | |
| "loss": 0.1286, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13236215710639954, | |
| "step": 5285, | |
| "valid_targets_mean": 1556.9, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 5.55672268907563, | |
| "grad_norm": 0.823501068446433, | |
| "learning_rate": 4.9679541492945135e-06, | |
| "loss": 0.1263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11238055676221848, | |
| "step": 5290, | |
| "valid_targets_mean": 1397.1, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 5.561974789915967, | |
| "grad_norm": 0.8106452499925574, | |
| "learning_rate": 4.933451035151913e-06, | |
| "loss": 0.1248, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12909407913684845, | |
| "step": 5295, | |
| "valid_targets_mean": 1698.1, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 5.567226890756302, | |
| "grad_norm": 0.8991107502754927, | |
| "learning_rate": 4.899051288790806e-06, | |
| "loss": 0.1232, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12105128914117813, | |
| "step": 5300, | |
| "valid_targets_mean": 1393.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 5.572478991596639, | |
| "grad_norm": 0.8395119193730466, | |
| "learning_rate": 4.86475514621916e-06, | |
| "loss": 0.1317, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11696095764636993, | |
| "step": 5305, | |
| "valid_targets_mean": 1443.1, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 5.5777310924369745, | |
| "grad_norm": 1.0738907935762767, | |
| "learning_rate": 4.830562842734112e-06, | |
| "loss": 0.1115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11815018951892853, | |
| "step": 5310, | |
| "valid_targets_mean": 1392.2, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 5.582983193277311, | |
| "grad_norm": 0.8292783339611027, | |
| "learning_rate": 4.796474612920421e-06, | |
| "loss": 0.1324, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11230309307575226, | |
| "step": 5315, | |
| "valid_targets_mean": 1386.6, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 5.588235294117647, | |
| "grad_norm": 0.9357218955051299, | |
| "learning_rate": 4.762490690648813e-06, | |
| "loss": 0.1191, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12718072533607483, | |
| "step": 5320, | |
| "valid_targets_mean": 1442.9, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 5.593487394957983, | |
| "grad_norm": 0.8519534800000159, | |
| "learning_rate": 4.728611309074374e-06, | |
| "loss": 0.119, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11761420965194702, | |
| "step": 5325, | |
| "valid_targets_mean": 1536.1, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 5.598739495798319, | |
| "grad_norm": 0.8547416140469031, | |
| "learning_rate": 4.694836700634975e-06, | |
| "loss": 0.1162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1196620911359787, | |
| "step": 5330, | |
| "valid_targets_mean": 1463.1, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 5.6039915966386555, | |
| "grad_norm": 1.0901335325216264, | |
| "learning_rate": 4.661167097049668e-06, | |
| "loss": 0.1227, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12231962382793427, | |
| "step": 5335, | |
| "valid_targets_mean": 1292.8, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 5.609243697478991, | |
| "grad_norm": 0.9123241183930774, | |
| "learning_rate": 4.62760272931708e-06, | |
| "loss": 0.1098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11801913380622864, | |
| "step": 5340, | |
| "valid_targets_mean": 1542.4, | |
| "valid_targets_min": 878 | |
| }, | |
| { | |
| "epoch": 5.614495798319328, | |
| "grad_norm": 0.8570929790691967, | |
| "learning_rate": 4.594143827713844e-06, | |
| "loss": 0.1095, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10810776799917221, | |
| "step": 5345, | |
| "valid_targets_mean": 1506.8, | |
| "valid_targets_min": 942 | |
| }, | |
| { | |
| "epoch": 5.619747899159664, | |
| "grad_norm": 0.9539604016186798, | |
| "learning_rate": 4.56079062179303e-06, | |
| "loss": 0.1107, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11601479351520538, | |
| "step": 5350, | |
| "valid_targets_mean": 1401.9, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 5.625, | |
| "grad_norm": 0.9114757769971337, | |
| "learning_rate": 4.527543340382538e-06, | |
| "loss": 0.1171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11921755969524384, | |
| "step": 5355, | |
| "valid_targets_mean": 1483.8, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 5.630252100840336, | |
| "grad_norm": 0.9360433208438923, | |
| "learning_rate": 4.494402211583569e-06, | |
| "loss": 0.1168, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11960253119468689, | |
| "step": 5360, | |
| "valid_targets_mean": 1356.3, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 5.635504201680672, | |
| "grad_norm": 0.8687564177093906, | |
| "learning_rate": 4.461367462769009e-06, | |
| "loss": 0.1185, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10865601897239685, | |
| "step": 5365, | |
| "valid_targets_mean": 1340.6, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 5.640756302521009, | |
| "grad_norm": 0.8626601637005205, | |
| "learning_rate": 4.428439320581923e-06, | |
| "loss": 0.1201, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.119922935962677, | |
| "step": 5370, | |
| "valid_targets_mean": 1528.6, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 5.6460084033613445, | |
| "grad_norm": 1.2286926999590546, | |
| "learning_rate": 4.3956180109339705e-06, | |
| "loss": 0.1161, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10904563963413239, | |
| "step": 5375, | |
| "valid_targets_mean": 1268.9, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 5.651260504201681, | |
| "grad_norm": 0.8729368734062463, | |
| "learning_rate": 4.362903759003849e-06, | |
| "loss": 0.1132, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1106632798910141, | |
| "step": 5380, | |
| "valid_targets_mean": 1360.1, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 5.656512605042017, | |
| "grad_norm": 0.8066508995038737, | |
| "learning_rate": 4.33029678923576e-06, | |
| "loss": 0.1136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11501973122358322, | |
| "step": 5385, | |
| "valid_targets_mean": 1452.3, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 5.661764705882353, | |
| "grad_norm": 0.8482374901140713, | |
| "learning_rate": 4.297797325337889e-06, | |
| "loss": 0.1051, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11828213930130005, | |
| "step": 5390, | |
| "valid_targets_mean": 1391.3, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 5.667016806722689, | |
| "grad_norm": 0.826557032894644, | |
| "learning_rate": 4.265405590280822e-06, | |
| "loss": 0.1104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09589506685733795, | |
| "step": 5395, | |
| "valid_targets_mean": 1344.4, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 5.6722689075630255, | |
| "grad_norm": 0.941561782597229, | |
| "learning_rate": 4.233121806296072e-06, | |
| "loss": 0.1139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12858964502811432, | |
| "step": 5400, | |
| "valid_targets_mean": 1545.2, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.677521008403361, | |
| "grad_norm": 0.912093405528764, | |
| "learning_rate": 4.200946194874514e-06, | |
| "loss": 0.1109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10562796890735626, | |
| "step": 5405, | |
| "valid_targets_mean": 1354.6, | |
| "valid_targets_min": 779 | |
| }, | |
| { | |
| "epoch": 5.682773109243698, | |
| "grad_norm": 0.9260849926734132, | |
| "learning_rate": 4.168878976764872e-06, | |
| "loss": 0.125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11829014122486115, | |
| "step": 5410, | |
| "valid_targets_mean": 1434.1, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 5.688025210084033, | |
| "grad_norm": 0.826771685553631, | |
| "learning_rate": 4.136920371972228e-06, | |
| "loss": 0.1094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1030585840344429, | |
| "step": 5415, | |
| "valid_targets_mean": 1388.2, | |
| "valid_targets_min": 868 | |
| }, | |
| { | |
| "epoch": 5.69327731092437, | |
| "grad_norm": 1.6999684647726199, | |
| "learning_rate": 4.105070599756482e-06, | |
| "loss": 0.1163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10433146357536316, | |
| "step": 5420, | |
| "valid_targets_mean": 1282.3, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.698529411764706, | |
| "grad_norm": 0.8817697258847107, | |
| "learning_rate": 4.073329878630856e-06, | |
| "loss": 0.1127, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11964145302772522, | |
| "step": 5425, | |
| "valid_targets_mean": 1457.2, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 5.703781512605042, | |
| "grad_norm": 0.8518914261076463, | |
| "learning_rate": 4.041698426360425e-06, | |
| "loss": 0.1157, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11360062658786774, | |
| "step": 5430, | |
| "valid_targets_mean": 1448.5, | |
| "valid_targets_min": 806 | |
| }, | |
| { | |
| "epoch": 5.709033613445378, | |
| "grad_norm": 0.8301646651006851, | |
| "learning_rate": 4.010176459960569e-06, | |
| "loss": 0.1063, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09808893501758575, | |
| "step": 5435, | |
| "valid_targets_mean": 1380.7, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 5.714285714285714, | |
| "grad_norm": 0.8473401775683042, | |
| "learning_rate": 3.978764195695528e-06, | |
| "loss": 0.1111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11080893129110336, | |
| "step": 5440, | |
| "valid_targets_mean": 1515.7, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 5.71953781512605, | |
| "grad_norm": 0.9100300218409771, | |
| "learning_rate": 3.947461849076912e-06, | |
| "loss": 0.1007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1048063337802887, | |
| "step": 5445, | |
| "valid_targets_mean": 1402.0, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 5.724789915966387, | |
| "grad_norm": 0.8889294607662074, | |
| "learning_rate": 3.916269634862193e-06, | |
| "loss": 0.1113, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10396813601255417, | |
| "step": 5450, | |
| "valid_targets_mean": 1574.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 5.730042016806722, | |
| "grad_norm": 0.8903144486882467, | |
| "learning_rate": 3.885187767053269e-06, | |
| "loss": 0.1153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10951384902000427, | |
| "step": 5455, | |
| "valid_targets_mean": 1430.1, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.735294117647059, | |
| "grad_norm": 0.9031317882555354, | |
| "learning_rate": 3.854216458894973e-06, | |
| "loss": 0.1117, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12100645154714584, | |
| "step": 5460, | |
| "valid_targets_mean": 1372.3, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 5.740546218487395, | |
| "grad_norm": 0.8606806374553407, | |
| "learning_rate": 3.823355922873606e-06, | |
| "loss": 0.1084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09598094969987869, | |
| "step": 5465, | |
| "valid_targets_mean": 1192.8, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 5.745798319327731, | |
| "grad_norm": 0.9625073380564342, | |
| "learning_rate": 3.792606370715508e-06, | |
| "loss": 0.1138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10911808162927628, | |
| "step": 5470, | |
| "valid_targets_mean": 1508.0, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 5.751050420168067, | |
| "grad_norm": 0.8921081945325963, | |
| "learning_rate": 3.761968013385584e-06, | |
| "loss": 0.1069, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10961748659610748, | |
| "step": 5475, | |
| "valid_targets_mean": 1430.2, | |
| "valid_targets_min": 1046 | |
| }, | |
| { | |
| "epoch": 5.756302521008403, | |
| "grad_norm": 0.9534472437981408, | |
| "learning_rate": 3.731441061085843e-06, | |
| "loss": 0.114, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12638525664806366, | |
| "step": 5480, | |
| "valid_targets_mean": 1651.8, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 5.76155462184874, | |
| "grad_norm": 0.8796809522933133, | |
| "learning_rate": 3.701025723253988e-06, | |
| "loss": 0.1137, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11247321218252182, | |
| "step": 5485, | |
| "valid_targets_mean": 1540.4, | |
| "valid_targets_min": 846 | |
| }, | |
| { | |
| "epoch": 5.766806722689076, | |
| "grad_norm": 0.845414134902536, | |
| "learning_rate": 3.67072220856197e-06, | |
| "loss": 0.1172, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10926534980535507, | |
| "step": 5490, | |
| "valid_targets_mean": 1455.1, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 5.772058823529412, | |
| "grad_norm": 0.8126469766696464, | |
| "learning_rate": 3.640530724914517e-06, | |
| "loss": 0.1056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10028908401727676, | |
| "step": 5495, | |
| "valid_targets_mean": 1497.4, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 5.777310924369748, | |
| "grad_norm": 0.8406164976266153, | |
| "learning_rate": 3.61045147944777e-06, | |
| "loss": 0.1071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11252301931381226, | |
| "step": 5500, | |
| "valid_targets_mean": 1543.8, | |
| "valid_targets_min": 982 | |
| }, | |
| { | |
| "epoch": 5.782563025210084, | |
| "grad_norm": 0.9415916945254411, | |
| "learning_rate": 3.5804846785278315e-06, | |
| "loss": 0.1119, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10923096537590027, | |
| "step": 5505, | |
| "valid_targets_mean": 1346.3, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 5.78781512605042, | |
| "grad_norm": 0.9260499456205905, | |
| "learning_rate": 3.5506305277493326e-06, | |
| "loss": 0.1071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10536296665668488, | |
| "step": 5510, | |
| "valid_targets_mean": 1395.7, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 5.793067226890757, | |
| "grad_norm": 0.8774386593252896, | |
| "learning_rate": 3.520889231934057e-06, | |
| "loss": 0.1056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10436506569385529, | |
| "step": 5515, | |
| "valid_targets_mean": 1290.6, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 5.798319327731092, | |
| "grad_norm": 0.8896778034184114, | |
| "learning_rate": 3.4912609951295084e-06, | |
| "loss": 0.1088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1191120520234108, | |
| "step": 5520, | |
| "valid_targets_mean": 1568.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 5.803571428571429, | |
| "grad_norm": 0.9421881494917845, | |
| "learning_rate": 3.461746020607526e-06, | |
| "loss": 0.1162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13356809318065643, | |
| "step": 5525, | |
| "valid_targets_mean": 1703.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 5.8088235294117645, | |
| "grad_norm": 0.9745264130361654, | |
| "learning_rate": 3.432344510862895e-06, | |
| "loss": 0.1158, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12390495836734772, | |
| "step": 5530, | |
| "valid_targets_mean": 1652.8, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 5.814075630252101, | |
| "grad_norm": 0.8607153880999272, | |
| "learning_rate": 3.403056667611917e-06, | |
| "loss": 0.1093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10633715987205505, | |
| "step": 5535, | |
| "valid_targets_mean": 1481.8, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 5.819327731092437, | |
| "grad_norm": 0.9758519266902133, | |
| "learning_rate": 3.373882691791088e-06, | |
| "loss": 0.1153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10976162552833557, | |
| "step": 5540, | |
| "valid_targets_mean": 1686.5, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 5.824579831932773, | |
| "grad_norm": 0.8967263497321722, | |
| "learning_rate": 3.344822783555679e-06, | |
| "loss": 0.1078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10004294663667679, | |
| "step": 5545, | |
| "valid_targets_mean": 1345.1, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 5.829831932773109, | |
| "grad_norm": 1.0391240197314284, | |
| "learning_rate": 3.315877142278363e-06, | |
| "loss": 0.1144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11110391467809677, | |
| "step": 5550, | |
| "valid_targets_mean": 1619.1, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 5.8350840336134455, | |
| "grad_norm": 0.8413197113632568, | |
| "learning_rate": 3.287045966547866e-06, | |
| "loss": 0.1064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11233742535114288, | |
| "step": 5555, | |
| "valid_targets_mean": 1423.2, | |
| "valid_targets_min": 674 | |
| }, | |
| { | |
| "epoch": 5.840336134453781, | |
| "grad_norm": 0.8505366043709849, | |
| "learning_rate": 3.258329454167599e-06, | |
| "loss": 0.1055, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10158196091651917, | |
| "step": 5560, | |
| "valid_targets_mean": 1207.4, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 5.845588235294118, | |
| "grad_norm": 0.8995630860603766, | |
| "learning_rate": 3.2297278021542833e-06, | |
| "loss": 0.1047, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11643253266811371, | |
| "step": 5565, | |
| "valid_targets_mean": 1542.4, | |
| "valid_targets_min": 912 | |
| }, | |
| { | |
| "epoch": 5.850840336134453, | |
| "grad_norm": 0.8887960162735108, | |
| "learning_rate": 3.201241206736614e-06, | |
| "loss": 0.1131, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11594435572624207, | |
| "step": 5570, | |
| "valid_targets_mean": 1591.0, | |
| "valid_targets_min": 897 | |
| }, | |
| { | |
| "epoch": 5.85609243697479, | |
| "grad_norm": 1.1842033013727196, | |
| "learning_rate": 3.1728698633539266e-06, | |
| "loss": 0.1073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1004495844244957, | |
| "step": 5575, | |
| "valid_targets_mean": 1290.1, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 5.8613445378151265, | |
| "grad_norm": 0.8528216924870448, | |
| "learning_rate": 3.1446139666548212e-06, | |
| "loss": 0.1097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09786748886108398, | |
| "step": 5580, | |
| "valid_targets_mean": 1407.2, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 5.866596638655462, | |
| "grad_norm": 0.9063160104966684, | |
| "learning_rate": 3.1164737104958707e-06, | |
| "loss": 0.1159, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11985988914966583, | |
| "step": 5585, | |
| "valid_targets_mean": 1600.2, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 5.871848739495798, | |
| "grad_norm": 0.8331809940510251, | |
| "learning_rate": 3.0884492879402452e-06, | |
| "loss": 0.1122, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10528066009283066, | |
| "step": 5590, | |
| "valid_targets_mean": 1442.4, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 5.8771008403361344, | |
| "grad_norm": 0.8662258381631335, | |
| "learning_rate": 3.0605408912564184e-06, | |
| "loss": 0.1078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11069589108228683, | |
| "step": 5595, | |
| "valid_targets_mean": 1426.6, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 5.882352941176471, | |
| "grad_norm": 0.8059447911663884, | |
| "learning_rate": 3.032748711916851e-06, | |
| "loss": 0.1073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1017690896987915, | |
| "step": 5600, | |
| "valid_targets_mean": 1430.9, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 5.887605042016807, | |
| "grad_norm": 0.8787623158989677, | |
| "learning_rate": 3.0050729405966495e-06, | |
| "loss": 0.1123, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11519140005111694, | |
| "step": 5605, | |
| "valid_targets_mean": 1412.6, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 5.892857142857143, | |
| "grad_norm": 1.0058669337260462, | |
| "learning_rate": 2.9775137671722754e-06, | |
| "loss": 0.1074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12139453738927841, | |
| "step": 5610, | |
| "valid_targets_mean": 1372.1, | |
| "valid_targets_min": 667 | |
| }, | |
| { | |
| "epoch": 5.898109243697479, | |
| "grad_norm": 0.9818342504773482, | |
| "learning_rate": 2.9500713807202565e-06, | |
| "loss": 0.1165, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1251143366098404, | |
| "step": 5615, | |
| "valid_targets_mean": 1628.1, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 5.9033613445378155, | |
| "grad_norm": 1.0048701808002658, | |
| "learning_rate": 2.9227459695158543e-06, | |
| "loss": 0.1195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12856556475162506, | |
| "step": 5620, | |
| "valid_targets_mean": 1464.7, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 5.908613445378151, | |
| "grad_norm": 0.9524777797952462, | |
| "learning_rate": 2.895537721031809e-06, | |
| "loss": 0.1112, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13392376899719238, | |
| "step": 5625, | |
| "valid_targets_mean": 1745.9, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 5.913865546218488, | |
| "grad_norm": 0.8360052998799868, | |
| "learning_rate": 2.8684468219370364e-06, | |
| "loss": 0.1099, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09906312078237534, | |
| "step": 5630, | |
| "valid_targets_mean": 1129.1, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.919117647058823, | |
| "grad_norm": 0.9259328855340732, | |
| "learning_rate": 2.841473458095334e-06, | |
| "loss": 0.1092, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12108924984931946, | |
| "step": 5635, | |
| "valid_targets_mean": 1624.6, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 5.92436974789916, | |
| "grad_norm": 0.9476207176974106, | |
| "learning_rate": 2.8146178145641355e-06, | |
| "loss": 0.1111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1358291506767273, | |
| "step": 5640, | |
| "valid_targets_mean": 1755.8, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.929621848739496, | |
| "grad_norm": 0.8733881430652086, | |
| "learning_rate": 2.787880075593217e-06, | |
| "loss": 0.1096, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11107189953327179, | |
| "step": 5645, | |
| "valid_targets_mean": 1516.4, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 5.934873949579832, | |
| "grad_norm": 0.8803395218820865, | |
| "learning_rate": 2.7612604246234355e-06, | |
| "loss": 0.1058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10555368661880493, | |
| "step": 5650, | |
| "valid_targets_mean": 1391.4, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 5.940126050420168, | |
| "grad_norm": 0.8940529152418495, | |
| "learning_rate": 2.7347590442854867e-06, | |
| "loss": 0.1075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11712536960840225, | |
| "step": 5655, | |
| "valid_targets_mean": 1412.4, | |
| "valid_targets_min": 778 | |
| }, | |
| { | |
| "epoch": 5.945378151260504, | |
| "grad_norm": 0.9009479204017542, | |
| "learning_rate": 2.7083761163986433e-06, | |
| "loss": 0.1185, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12473637610673904, | |
| "step": 5660, | |
| "valid_targets_mean": 1599.6, | |
| "valid_targets_min": 851 | |
| }, | |
| { | |
| "epoch": 5.95063025210084, | |
| "grad_norm": 0.8299195293692447, | |
| "learning_rate": 2.682111821969493e-06, | |
| "loss": 0.1072, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10053186863660812, | |
| "step": 5665, | |
| "valid_targets_mean": 1398.4, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 5.955882352941177, | |
| "grad_norm": 0.8224894370809741, | |
| "learning_rate": 2.6559663411907233e-06, | |
| "loss": 0.1068, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09880638122558594, | |
| "step": 5670, | |
| "valid_targets_mean": 1477.9, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 5.961134453781512, | |
| "grad_norm": 0.8446943544157789, | |
| "learning_rate": 2.629939853439856e-06, | |
| "loss": 0.1043, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09644494950771332, | |
| "step": 5675, | |
| "valid_targets_mean": 1299.4, | |
| "valid_targets_min": 639 | |
| }, | |
| { | |
| "epoch": 5.966386554621849, | |
| "grad_norm": 0.8887196740407839, | |
| "learning_rate": 2.604032537278052e-06, | |
| "loss": 0.1009, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10244621336460114, | |
| "step": 5680, | |
| "valid_targets_mean": 1277.9, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 5.9716386554621845, | |
| "grad_norm": 0.8494276928184287, | |
| "learning_rate": 2.5782445704488447e-06, | |
| "loss": 0.1103, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10847003012895584, | |
| "step": 5685, | |
| "valid_targets_mean": 1529.9, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 5.976890756302521, | |
| "grad_norm": 0.8658895151511194, | |
| "learning_rate": 2.5525761298769515e-06, | |
| "loss": 0.1061, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10970406234264374, | |
| "step": 5690, | |
| "valid_targets_mean": 1331.6, | |
| "valid_targets_min": 695 | |
| }, | |
| { | |
| "epoch": 5.982142857142857, | |
| "grad_norm": 0.9065356041104413, | |
| "learning_rate": 2.527027391667054e-06, | |
| "loss": 0.1048, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10520265996456146, | |
| "step": 5695, | |
| "valid_targets_mean": 1581.9, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 5.987394957983193, | |
| "grad_norm": 0.9469535446284602, | |
| "learning_rate": 2.501598531102587e-06, | |
| "loss": 0.1059, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10947225987911224, | |
| "step": 5700, | |
| "valid_targets_mean": 1327.6, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.992647058823529, | |
| "grad_norm": 0.9476175324156466, | |
| "learning_rate": 2.47628972264452e-06, | |
| "loss": 0.1063, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11112868785858154, | |
| "step": 5705, | |
| "valid_targets_mean": 1346.4, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 5.9978991596638656, | |
| "grad_norm": 0.9233409415833587, | |
| "learning_rate": 2.4511011399301944e-06, | |
| "loss": 0.1098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10038024187088013, | |
| "step": 5710, | |
| "valid_targets_mean": 1483.2, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.003151260504202, | |
| "grad_norm": 1.1538573337023281, | |
| "learning_rate": 2.4260329557721107e-06, | |
| "loss": 0.1738, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20704257488250732, | |
| "step": 5715, | |
| "valid_targets_mean": 4404.5, | |
| "valid_targets_min": 1323 | |
| }, | |
| { | |
| "epoch": 6.008403361344538, | |
| "grad_norm": 1.1897027864306338, | |
| "learning_rate": 2.4010853421567237e-06, | |
| "loss": 0.2143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22920134663581848, | |
| "step": 5720, | |
| "valid_targets_mean": 4483.4, | |
| "valid_targets_min": 1005 | |
| }, | |
| { | |
| "epoch": 6.013655462184874, | |
| "grad_norm": 0.8684521464937454, | |
| "learning_rate": 2.376258470243309e-06, | |
| "loss": 0.2006, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19509011507034302, | |
| "step": 5725, | |
| "valid_targets_mean": 4502.1, | |
| "valid_targets_min": 1509 | |
| }, | |
| { | |
| "epoch": 6.01890756302521, | |
| "grad_norm": 0.8037614431305709, | |
| "learning_rate": 2.351552510362758e-06, | |
| "loss": 0.2021, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1814684271812439, | |
| "step": 5730, | |
| "valid_targets_mean": 3430.2, | |
| "valid_targets_min": 1561 | |
| }, | |
| { | |
| "epoch": 6.024159663865547, | |
| "grad_norm": 0.7816962283075021, | |
| "learning_rate": 2.326967632016406e-06, | |
| "loss": 0.1892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18898342549800873, | |
| "step": 5735, | |
| "valid_targets_mean": 3385.4, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 6.029411764705882, | |
| "grad_norm": 0.6426902354697276, | |
| "learning_rate": 2.302504003874888e-06, | |
| "loss": 0.204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1760241985321045, | |
| "step": 5740, | |
| "valid_targets_mean": 3982.6, | |
| "valid_targets_min": 1413 | |
| }, | |
| { | |
| "epoch": 6.034663865546219, | |
| "grad_norm": 0.632043071580013, | |
| "learning_rate": 2.2781617937769675e-06, | |
| "loss": 0.1805, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19574542343616486, | |
| "step": 5745, | |
| "valid_targets_mean": 4911.7, | |
| "valid_targets_min": 2160 | |
| }, | |
| { | |
| "epoch": 6.0399159663865545, | |
| "grad_norm": 0.7493769564211563, | |
| "learning_rate": 2.2539411687283907e-06, | |
| "loss": 0.2004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19801229238510132, | |
| "step": 5750, | |
| "valid_targets_mean": 3782.6, | |
| "valid_targets_min": 1511 | |
| }, | |
| { | |
| "epoch": 6.045168067226891, | |
| "grad_norm": 0.7505509884796049, | |
| "learning_rate": 2.229842294900744e-06, | |
| "loss": 0.1974, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20454156398773193, | |
| "step": 5755, | |
| "valid_targets_mean": 3030.0, | |
| "valid_targets_min": 1462 | |
| }, | |
| { | |
| "epoch": 6.050420168067227, | |
| "grad_norm": 0.6430262170455331, | |
| "learning_rate": 2.205865337630302e-06, | |
| "loss": 0.2006, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18196547031402588, | |
| "step": 5760, | |
| "valid_targets_mean": 3772.4, | |
| "valid_targets_min": 874 | |
| }, | |
| { | |
| "epoch": 6.055672268907563, | |
| "grad_norm": 0.7098617069387976, | |
| "learning_rate": 2.182010461416899e-06, | |
| "loss": 0.1983, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19023019075393677, | |
| "step": 5765, | |
| "valid_targets_mean": 3337.1, | |
| "valid_targets_min": 953 | |
| }, | |
| { | |
| "epoch": 6.060924369747899, | |
| "grad_norm": 0.6552026128807007, | |
| "learning_rate": 2.158277829922817e-06, | |
| "loss": 0.2004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2095913290977478, | |
| "step": 5770, | |
| "valid_targets_mean": 4284.9, | |
| "valid_targets_min": 1265 | |
| }, | |
| { | |
| "epoch": 6.0661764705882355, | |
| "grad_norm": 0.5640222204672125, | |
| "learning_rate": 2.1346676059716253e-06, | |
| "loss": 0.1961, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19262853264808655, | |
| "step": 5775, | |
| "valid_targets_mean": 5010.6, | |
| "valid_targets_min": 1596 | |
| }, | |
| { | |
| "epoch": 6.071428571428571, | |
| "grad_norm": 0.6700501160751591, | |
| "learning_rate": 2.1111799515471066e-06, | |
| "loss": 0.1951, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2048107534646988, | |
| "step": 5780, | |
| "valid_targets_mean": 3490.4, | |
| "valid_targets_min": 1759 | |
| }, | |
| { | |
| "epoch": 6.076680672268908, | |
| "grad_norm": 1.192594245087854, | |
| "learning_rate": 2.0878150277921195e-06, | |
| "loss": 0.1997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20699971914291382, | |
| "step": 5785, | |
| "valid_targets_mean": 3264.3, | |
| "valid_targets_min": 1353 | |
| }, | |
| { | |
| "epoch": 6.081932773109243, | |
| "grad_norm": 0.6402893347468779, | |
| "learning_rate": 2.064572995007492e-06, | |
| "loss": 0.1987, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17887401580810547, | |
| "step": 5790, | |
| "valid_targets_mean": 3814.6, | |
| "valid_targets_min": 1643 | |
| }, | |
| { | |
| "epoch": 6.08718487394958, | |
| "grad_norm": 0.6836188652244888, | |
| "learning_rate": 2.0414540126509274e-06, | |
| "loss": 0.1928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17512783408164978, | |
| "step": 5795, | |
| "valid_targets_mean": 3166.1, | |
| "valid_targets_min": 1269 | |
| }, | |
| { | |
| "epoch": 6.092436974789916, | |
| "grad_norm": 0.6778220747996802, | |
| "learning_rate": 2.0184582393359253e-06, | |
| "loss": 0.209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20338422060012817, | |
| "step": 5800, | |
| "valid_targets_mean": 3617.4, | |
| "valid_targets_min": 1681 | |
| }, | |
| { | |
| "epoch": 6.097689075630252, | |
| "grad_norm": 0.6622080466063627, | |
| "learning_rate": 1.99558583283066e-06, | |
| "loss": 0.1926, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17835170030593872, | |
| "step": 5805, | |
| "valid_targets_mean": 3919.8, | |
| "valid_targets_min": 1477 | |
| }, | |
| { | |
| "epoch": 6.102941176470588, | |
| "grad_norm": 0.7482662124846838, | |
| "learning_rate": 1.9728369500569333e-06, | |
| "loss": 0.2007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1978331208229065, | |
| "step": 5810, | |
| "valid_targets_mean": 3418.5, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 6.108193277310924, | |
| "grad_norm": 1.0158385440208195, | |
| "learning_rate": 1.950211747089077e-06, | |
| "loss": 0.1963, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2101389467716217, | |
| "step": 5815, | |
| "valid_targets_mean": 3706.7, | |
| "valid_targets_min": 2077 | |
| }, | |
| { | |
| "epoch": 6.11344537815126, | |
| "grad_norm": 0.7058981750753525, | |
| "learning_rate": 1.927710379152881e-06, | |
| "loss": 0.1819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1703696846961975, | |
| "step": 5820, | |
| "valid_targets_mean": 2878.1, | |
| "valid_targets_min": 1505 | |
| }, | |
| { | |
| "epoch": 6.118697478991597, | |
| "grad_norm": 0.7508480855114319, | |
| "learning_rate": 1.9053330006245452e-06, | |
| "loss": 0.2016, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21035335958003998, | |
| "step": 5825, | |
| "valid_targets_mean": 3418.0, | |
| "valid_targets_min": 1957 | |
| }, | |
| { | |
| "epoch": 6.123949579831933, | |
| "grad_norm": 0.8433194225905101, | |
| "learning_rate": 1.8830797650296006e-06, | |
| "loss": 0.2058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25868695974349976, | |
| "step": 5830, | |
| "valid_targets_mean": 4106.4, | |
| "valid_targets_min": 1704 | |
| }, | |
| { | |
| "epoch": 6.129201680672269, | |
| "grad_norm": 0.7590435573654869, | |
| "learning_rate": 1.8609508250418628e-06, | |
| "loss": 0.1922, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2122974842786789, | |
| "step": 5835, | |
| "valid_targets_mean": 3046.9, | |
| "valid_targets_min": 1422 | |
| }, | |
| { | |
| "epoch": 6.1344537815126055, | |
| "grad_norm": 0.685478517649954, | |
| "learning_rate": 1.8389463324823986e-06, | |
| "loss": 0.186, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1737407147884369, | |
| "step": 5840, | |
| "valid_targets_mean": 3447.4, | |
| "valid_targets_min": 1603 | |
| }, | |
| { | |
| "epoch": 6.139705882352941, | |
| "grad_norm": 0.8161452603910854, | |
| "learning_rate": 1.81706643831846e-06, | |
| "loss": 0.1991, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20304900407791138, | |
| "step": 5845, | |
| "valid_targets_mean": 2597.8, | |
| "valid_targets_min": 1685 | |
| }, | |
| { | |
| "epoch": 6.144957983193278, | |
| "grad_norm": 0.7114292866383795, | |
| "learning_rate": 1.7953112926624715e-06, | |
| "loss": 0.1864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19642803072929382, | |
| "step": 5850, | |
| "valid_targets_mean": 3290.6, | |
| "valid_targets_min": 1208 | |
| }, | |
| { | |
| "epoch": 6.150210084033613, | |
| "grad_norm": 0.7279693187875432, | |
| "learning_rate": 1.7736810447709852e-06, | |
| "loss": 0.1954, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1854780614376068, | |
| "step": 5855, | |
| "valid_targets_mean": 2661.8, | |
| "valid_targets_min": 1501 | |
| }, | |
| { | |
| "epoch": 6.15546218487395, | |
| "grad_norm": 0.7284805186121608, | |
| "learning_rate": 1.7521758430436553e-06, | |
| "loss": 0.1961, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22248005867004395, | |
| "step": 5860, | |
| "valid_targets_mean": 3849.0, | |
| "valid_targets_min": 1866 | |
| }, | |
| { | |
| "epoch": 6.160714285714286, | |
| "grad_norm": 0.641586434100809, | |
| "learning_rate": 1.7307958350222366e-06, | |
| "loss": 0.1819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1723538041114807, | |
| "step": 5865, | |
| "valid_targets_mean": 3397.9, | |
| "valid_targets_min": 1239 | |
| }, | |
| { | |
| "epoch": 6.165966386554622, | |
| "grad_norm": 0.7494400065396022, | |
| "learning_rate": 1.7095411673895545e-06, | |
| "loss": 0.1915, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20020216703414917, | |
| "step": 5870, | |
| "valid_targets_mean": 3155.5, | |
| "valid_targets_min": 1177 | |
| }, | |
| { | |
| "epoch": 6.171218487394958, | |
| "grad_norm": 0.7120710111731775, | |
| "learning_rate": 1.6884119859685034e-06, | |
| "loss": 0.1922, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18129676580429077, | |
| "step": 5875, | |
| "valid_targets_mean": 3594.5, | |
| "valid_targets_min": 1270 | |
| }, | |
| { | |
| "epoch": 6.176470588235294, | |
| "grad_norm": 0.7012711548565949, | |
| "learning_rate": 1.6674084357210562e-06, | |
| "loss": 0.1943, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18761314451694489, | |
| "step": 5880, | |
| "valid_targets_mean": 3571.2, | |
| "valid_targets_min": 1216 | |
| }, | |
| { | |
| "epoch": 6.18172268907563, | |
| "grad_norm": 0.6937981892604655, | |
| "learning_rate": 1.6465306607472608e-06, | |
| "loss": 0.1927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18734653294086456, | |
| "step": 5885, | |
| "valid_targets_mean": 3290.8, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 6.186974789915967, | |
| "grad_norm": 0.7223230217430263, | |
| "learning_rate": 1.625778804284246e-06, | |
| "loss": 0.1955, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18657483160495758, | |
| "step": 5890, | |
| "valid_targets_mean": 2986.1, | |
| "valid_targets_min": 1292 | |
| }, | |
| { | |
| "epoch": 6.192226890756302, | |
| "grad_norm": 0.703895451312637, | |
| "learning_rate": 1.6051530087052558e-06, | |
| "loss": 0.1986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18239927291870117, | |
| "step": 5895, | |
| "valid_targets_mean": 3485.4, | |
| "valid_targets_min": 1639 | |
| }, | |
| { | |
| "epoch": 6.197478991596639, | |
| "grad_norm": 0.7220190947994646, | |
| "learning_rate": 1.5846534155186511e-06, | |
| "loss": 0.1811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19022542238235474, | |
| "step": 5900, | |
| "valid_targets_mean": 2959.4, | |
| "valid_targets_min": 1394 | |
| }, | |
| { | |
| "epoch": 6.2027310924369745, | |
| "grad_norm": 0.7192311642483811, | |
| "learning_rate": 1.564280165366956e-06, | |
| "loss": 0.1979, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20613990724086761, | |
| "step": 5905, | |
| "valid_targets_mean": 3428.9, | |
| "valid_targets_min": 2280 | |
| }, | |
| { | |
| "epoch": 6.207983193277311, | |
| "grad_norm": 0.7691711026249906, | |
| "learning_rate": 1.544033398025886e-06, | |
| "loss": 0.1868, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21603600680828094, | |
| "step": 5910, | |
| "valid_targets_mean": 3233.9, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 6.213235294117647, | |
| "grad_norm": 0.8449315953755364, | |
| "learning_rate": 1.5239132524033951e-06, | |
| "loss": 0.1834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18134553730487823, | |
| "step": 5915, | |
| "valid_targets_mean": 3211.6, | |
| "valid_targets_min": 1654 | |
| }, | |
| { | |
| "epoch": 6.218487394957983, | |
| "grad_norm": 0.7395008703366813, | |
| "learning_rate": 1.50391986653871e-06, | |
| "loss": 0.1916, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18029049038887024, | |
| "step": 5920, | |
| "valid_targets_mean": 3057.8, | |
| "valid_targets_min": 584 | |
| }, | |
| { | |
| "epoch": 6.223739495798319, | |
| "grad_norm": 0.7120133644780183, | |
| "learning_rate": 1.4840533776013954e-06, | |
| "loss": 0.1916, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20296408236026764, | |
| "step": 5925, | |
| "valid_targets_mean": 3572.6, | |
| "valid_targets_min": 2498 | |
| }, | |
| { | |
| "epoch": 6.2289915966386555, | |
| "grad_norm": 0.752548834195417, | |
| "learning_rate": 1.4643139218904035e-06, | |
| "loss": 0.1897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18757691979408264, | |
| "step": 5930, | |
| "valid_targets_mean": 3161.1, | |
| "valid_targets_min": 1350 | |
| }, | |
| { | |
| "epoch": 6.234243697478991, | |
| "grad_norm": 0.7837418288101413, | |
| "learning_rate": 1.4447016348331477e-06, | |
| "loss": 0.1975, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19451862573623657, | |
| "step": 5935, | |
| "valid_targets_mean": 3486.1, | |
| "valid_targets_min": 1624 | |
| }, | |
| { | |
| "epoch": 6.239495798319328, | |
| "grad_norm": 0.7679677373950752, | |
| "learning_rate": 1.4252166509845777e-06, | |
| "loss": 0.1762, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16891759634017944, | |
| "step": 5940, | |
| "valid_targets_mean": 2769.6, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 6.244747899159663, | |
| "grad_norm": 0.6545085160769695, | |
| "learning_rate": 1.4058591040262259e-06, | |
| "loss": 0.1878, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1811821609735489, | |
| "step": 5945, | |
| "valid_targets_mean": 3734.1, | |
| "valid_targets_min": 2209 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "grad_norm": 0.7272522502400294, | |
| "learning_rate": 1.3866291267653309e-06, | |
| "loss": 0.1889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18539613485336304, | |
| "step": 5950, | |
| "valid_targets_mean": 3107.0, | |
| "valid_targets_min": 1259 | |
| }, | |
| { | |
| "epoch": 6.255252100840336, | |
| "grad_norm": 0.6965968942209451, | |
| "learning_rate": 1.3675268511339067e-06, | |
| "loss": 0.1829, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17434164881706238, | |
| "step": 5955, | |
| "valid_targets_mean": 3685.6, | |
| "valid_targets_min": 1183 | |
| }, | |
| { | |
| "epoch": 6.260504201680672, | |
| "grad_norm": 0.7307769179047728, | |
| "learning_rate": 1.3485524081878288e-06, | |
| "loss": 0.1806, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20974670350551605, | |
| "step": 5960, | |
| "valid_targets_mean": 3891.3, | |
| "valid_targets_min": 1101 | |
| }, | |
| { | |
| "epoch": 6.265756302521009, | |
| "grad_norm": 0.7157157431884991, | |
| "learning_rate": 1.3297059281059533e-06, | |
| "loss": 0.1975, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19651910662651062, | |
| "step": 5965, | |
| "valid_targets_mean": 3877.6, | |
| "valid_targets_min": 1658 | |
| }, | |
| { | |
| "epoch": 6.2710084033613445, | |
| "grad_norm": 0.8654778291967979, | |
| "learning_rate": 1.3109875401892146e-06, | |
| "loss": 0.187, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19856885075569153, | |
| "step": 5970, | |
| "valid_targets_mean": 3050.1, | |
| "valid_targets_min": 1423 | |
| }, | |
| { | |
| "epoch": 6.276260504201681, | |
| "grad_norm": 0.7591061487662136, | |
| "learning_rate": 1.2923973728597373e-06, | |
| "loss": 0.1957, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18958264589309692, | |
| "step": 5975, | |
| "valid_targets_mean": 3202.4, | |
| "valid_targets_min": 1185 | |
| }, | |
| { | |
| "epoch": 6.281512605042017, | |
| "grad_norm": 0.7845162142528881, | |
| "learning_rate": 1.2739355536599508e-06, | |
| "loss": 0.2075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2176717221736908, | |
| "step": 5980, | |
| "valid_targets_mean": 3779.1, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 6.286764705882353, | |
| "grad_norm": 0.7274113285487317, | |
| "learning_rate": 1.2556022092517362e-06, | |
| "loss": 0.1891, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1754409372806549, | |
| "step": 5985, | |
| "valid_targets_mean": 3628.9, | |
| "valid_targets_min": 1352 | |
| }, | |
| { | |
| "epoch": 6.292016806722689, | |
| "grad_norm": 0.6322394750569781, | |
| "learning_rate": 1.237397465415524e-06, | |
| "loss": 0.19, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19616565108299255, | |
| "step": 5990, | |
| "valid_targets_mean": 4061.2, | |
| "valid_targets_min": 1535 | |
| }, | |
| { | |
| "epoch": 6.2972689075630255, | |
| "grad_norm": 0.7257557508034214, | |
| "learning_rate": 1.2193214470494664e-06, | |
| "loss": 0.18, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.167373925447464, | |
| "step": 5995, | |
| "valid_targets_mean": 2753.1, | |
| "valid_targets_min": 854 | |
| }, | |
| { | |
| "epoch": 6.302521008403361, | |
| "grad_norm": 0.7431922528740501, | |
| "learning_rate": 1.201374278168541e-06, | |
| "loss": 0.1768, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16021127998828888, | |
| "step": 6000, | |
| "valid_targets_mean": 2942.5, | |
| "valid_targets_min": 1179 | |
| }, | |
| { | |
| "epoch": 6.307773109243698, | |
| "grad_norm": 0.7197510596784839, | |
| "learning_rate": 1.1835560819037429e-06, | |
| "loss": 0.1872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1804109513759613, | |
| "step": 6005, | |
| "valid_targets_mean": 3598.9, | |
| "valid_targets_min": 1290 | |
| }, | |
| { | |
| "epoch": 6.313025210084033, | |
| "grad_norm": 0.8045418304005293, | |
| "learning_rate": 1.1658669805012112e-06, | |
| "loss": 0.1761, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1749718338251114, | |
| "step": 6010, | |
| "valid_targets_mean": 2332.6, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 6.31827731092437, | |
| "grad_norm": 0.7127205041558864, | |
| "learning_rate": 1.1483070953213904e-06, | |
| "loss": 0.1745, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18035557866096497, | |
| "step": 6015, | |
| "valid_targets_mean": 3670.2, | |
| "valid_targets_min": 1764 | |
| }, | |
| { | |
| "epoch": 6.323529411764706, | |
| "grad_norm": 0.8051860130698884, | |
| "learning_rate": 1.130876546838211e-06, | |
| "loss": 0.2005, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2282528579235077, | |
| "step": 6020, | |
| "valid_targets_mean": 3528.6, | |
| "valid_targets_min": 1410 | |
| }, | |
| { | |
| "epoch": 6.328781512605042, | |
| "grad_norm": 0.7625429310294065, | |
| "learning_rate": 1.1135754546382627e-06, | |
| "loss": 0.1813, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19633427262306213, | |
| "step": 6025, | |
| "valid_targets_mean": 3147.7, | |
| "valid_targets_min": 1683 | |
| }, | |
| { | |
| "epoch": 6.334033613445378, | |
| "grad_norm": 0.7138024696131663, | |
| "learning_rate": 1.0964039374199564e-06, | |
| "loss": 0.1872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19841980934143066, | |
| "step": 6030, | |
| "valid_targets_mean": 3632.5, | |
| "valid_targets_min": 996 | |
| }, | |
| { | |
| "epoch": 6.339285714285714, | |
| "grad_norm": 0.7688382202900614, | |
| "learning_rate": 1.07936211299273e-06, | |
| "loss": 0.188, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18963909149169922, | |
| "step": 6035, | |
| "valid_targets_mean": 3135.3, | |
| "valid_targets_min": 1349 | |
| }, | |
| { | |
| "epoch": 6.34453781512605, | |
| "grad_norm": 0.7341808809524797, | |
| "learning_rate": 1.0624500982762376e-06, | |
| "loss": 0.1859, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2044352889060974, | |
| "step": 6040, | |
| "valid_targets_mean": 3391.6, | |
| "valid_targets_min": 1248 | |
| }, | |
| { | |
| "epoch": 6.349789915966387, | |
| "grad_norm": 0.7221599296550733, | |
| "learning_rate": 1.045668009299523e-06, | |
| "loss": 0.1844, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17875009775161743, | |
| "step": 6045, | |
| "valid_targets_mean": 3323.4, | |
| "valid_targets_min": 1445 | |
| }, | |
| { | |
| "epoch": 6.355042016806722, | |
| "grad_norm": 0.7155618590130363, | |
| "learning_rate": 1.029015961200266e-06, | |
| "loss": 0.181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17386895418167114, | |
| "step": 6050, | |
| "valid_targets_mean": 3134.3, | |
| "valid_targets_min": 2000 | |
| }, | |
| { | |
| "epoch": 6.360294117647059, | |
| "grad_norm": 0.8696023819163883, | |
| "learning_rate": 1.012494068223946e-06, | |
| "loss": 0.1759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18593569099903107, | |
| "step": 6055, | |
| "valid_targets_mean": 2828.3, | |
| "valid_targets_min": 1197 | |
| }, | |
| { | |
| "epoch": 6.3655462184873945, | |
| "grad_norm": 0.7145375703144178, | |
| "learning_rate": 9.961024437230948e-07, | |
| "loss": 0.1948, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18828941881656647, | |
| "step": 6060, | |
| "valid_targets_mean": 3219.0, | |
| "valid_targets_min": 1223 | |
| }, | |
| { | |
| "epoch": 6.370798319327731, | |
| "grad_norm": 0.7434045401240451, | |
| "learning_rate": 9.798412001565038e-07, | |
| "loss": 0.1947, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17213355004787445, | |
| "step": 6065, | |
| "valid_targets_mean": 3199.2, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 6.376050420168067, | |
| "grad_norm": 0.760275705548783, | |
| "learning_rate": 9.63710449088453e-07, | |
| "loss": 0.1912, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19253242015838623, | |
| "step": 6070, | |
| "valid_targets_mean": 2817.8, | |
| "valid_targets_min": 1603 | |
| }, | |
| { | |
| "epoch": 6.381302521008403, | |
| "grad_norm": 0.729204870471661, | |
| "learning_rate": 9.477103011879408e-07, | |
| "loss": 0.1931, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17323127388954163, | |
| "step": 6075, | |
| "valid_targets_mean": 2942.2, | |
| "valid_targets_min": 1135 | |
| }, | |
| { | |
| "epoch": 6.38655462184874, | |
| "grad_norm": 0.7069419195201837, | |
| "learning_rate": 9.318408662279399e-07, | |
| "loss": 0.1945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21406994760036469, | |
| "step": 6080, | |
| "valid_targets_mean": 4077.0, | |
| "valid_targets_min": 1994 | |
| }, | |
| { | |
| "epoch": 6.391806722689076, | |
| "grad_norm": 0.7466628023501196, | |
| "learning_rate": 9.161022530846253e-07, | |
| "loss": 0.1942, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16813525557518005, | |
| "step": 6085, | |
| "valid_targets_mean": 3060.1, | |
| "valid_targets_min": 1066 | |
| }, | |
| { | |
| "epoch": 6.397058823529412, | |
| "grad_norm": 0.7366467317633626, | |
| "learning_rate": 9.004945697366452e-07, | |
| "loss": 0.1932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20294198393821716, | |
| "step": 6090, | |
| "valid_targets_mean": 3687.4, | |
| "valid_targets_min": 1211 | |
| }, | |
| { | |
| "epoch": 6.402310924369748, | |
| "grad_norm": 0.7660213872137742, | |
| "learning_rate": 8.850179232643641e-07, | |
| "loss": 0.1903, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1757078468799591, | |
| "step": 6095, | |
| "valid_targets_mean": 2972.2, | |
| "valid_targets_min": 1195 | |
| }, | |
| { | |
| "epoch": 6.407563025210084, | |
| "grad_norm": 0.8448238623784213, | |
| "learning_rate": 8.696724198491369e-07, | |
| "loss": 0.1894, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18741391599178314, | |
| "step": 6100, | |
| "valid_targets_mean": 2429.5, | |
| "valid_targets_min": 484 | |
| }, | |
| { | |
| "epoch": 6.41281512605042, | |
| "grad_norm": 0.7208665843891413, | |
| "learning_rate": 8.544581647725825e-07, | |
| "loss": 0.1976, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2011285126209259, | |
| "step": 6105, | |
| "valid_targets_mean": 3851.9, | |
| "valid_targets_min": 1257 | |
| }, | |
| { | |
| "epoch": 6.418067226890757, | |
| "grad_norm": 0.7164569848209156, | |
| "learning_rate": 8.393752624158603e-07, | |
| "loss": 0.1867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17893822491168976, | |
| "step": 6110, | |
| "valid_targets_mean": 3483.2, | |
| "valid_targets_min": 1372 | |
| }, | |
| { | |
| "epoch": 6.423319327731092, | |
| "grad_norm": 0.7706889942916458, | |
| "learning_rate": 8.244238162589435e-07, | |
| "loss": 0.1886, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19799582660198212, | |
| "step": 6115, | |
| "valid_targets_mean": 3186.8, | |
| "valid_targets_min": 1769 | |
| }, | |
| { | |
| "epoch": 6.428571428571429, | |
| "grad_norm": 0.7885866860059115, | |
| "learning_rate": 8.096039288799251e-07, | |
| "loss": 0.1906, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18775229156017303, | |
| "step": 6120, | |
| "valid_targets_mean": 3407.6, | |
| "valid_targets_min": 1074 | |
| }, | |
| { | |
| "epoch": 6.4338235294117645, | |
| "grad_norm": 0.6919907540962501, | |
| "learning_rate": 7.949157019543064e-07, | |
| "loss": 0.1957, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1812029927968979, | |
| "step": 6125, | |
| "valid_targets_mean": 3796.4, | |
| "valid_targets_min": 980 | |
| }, | |
| { | |
| "epoch": 6.439075630252101, | |
| "grad_norm": 0.7918341376555602, | |
| "learning_rate": 7.803592362542911e-07, | |
| "loss": 0.1995, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20236995816230774, | |
| "step": 6130, | |
| "valid_targets_mean": 2808.1, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 6.444327731092437, | |
| "grad_norm": 0.7965898165350053, | |
| "learning_rate": 7.659346316481086e-07, | |
| "loss": 0.1842, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22702500224113464, | |
| "step": 6135, | |
| "valid_targets_mean": 3677.5, | |
| "valid_targets_min": 1471 | |
| }, | |
| { | |
| "epoch": 6.449579831932773, | |
| "grad_norm": 0.7266460626729361, | |
| "learning_rate": 7.516419870993296e-07, | |
| "loss": 0.1783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16017696261405945, | |
| "step": 6140, | |
| "valid_targets_mean": 3032.6, | |
| "valid_targets_min": 1353 | |
| }, | |
| { | |
| "epoch": 6.454831932773109, | |
| "grad_norm": 0.7762725308287282, | |
| "learning_rate": 7.374814006661667e-07, | |
| "loss": 0.1851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17704874277114868, | |
| "step": 6145, | |
| "valid_targets_mean": 2601.9, | |
| "valid_targets_min": 1084 | |
| }, | |
| { | |
| "epoch": 6.4600840336134455, | |
| "grad_norm": 0.6779697306222039, | |
| "learning_rate": 7.234529695008241e-07, | |
| "loss": 0.1838, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17760631442070007, | |
| "step": 6150, | |
| "valid_targets_mean": 3280.8, | |
| "valid_targets_min": 1428 | |
| }, | |
| { | |
| "epoch": 6.465336134453781, | |
| "grad_norm": 0.8121257910719274, | |
| "learning_rate": 7.095567898488175e-07, | |
| "loss": 0.1811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1856568455696106, | |
| "step": 6155, | |
| "valid_targets_mean": 3229.4, | |
| "valid_targets_min": 1193 | |
| }, | |
| { | |
| "epoch": 6.470588235294118, | |
| "grad_norm": 0.7371997769488223, | |
| "learning_rate": 6.957929570483224e-07, | |
| "loss": 0.1817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17043937742710114, | |
| "step": 6160, | |
| "valid_targets_mean": 3519.2, | |
| "valid_targets_min": 1614 | |
| }, | |
| { | |
| "epoch": 6.475840336134453, | |
| "grad_norm": 0.6638062829293753, | |
| "learning_rate": 6.821615655295155e-07, | |
| "loss": 0.174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16736310720443726, | |
| "step": 6165, | |
| "valid_targets_mean": 3699.1, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 6.48109243697479, | |
| "grad_norm": 0.7220475008819989, | |
| "learning_rate": 6.686627088139253e-07, | |
| "loss": 0.1747, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14820286631584167, | |
| "step": 6170, | |
| "valid_targets_mean": 2730.4, | |
| "valid_targets_min": 1575 | |
| }, | |
| { | |
| "epoch": 6.486344537815126, | |
| "grad_norm": 0.6962914540227392, | |
| "learning_rate": 6.552964795137895e-07, | |
| "loss": 0.1796, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18003645539283752, | |
| "step": 6175, | |
| "valid_targets_mean": 3345.0, | |
| "valid_targets_min": 1697 | |
| }, | |
| { | |
| "epoch": 6.491596638655462, | |
| "grad_norm": 0.7440882846707011, | |
| "learning_rate": 6.420629693314317e-07, | |
| "loss": 0.1846, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19298087060451508, | |
| "step": 6180, | |
| "valid_targets_mean": 3602.2, | |
| "valid_targets_min": 1970 | |
| }, | |
| { | |
| "epoch": 6.496848739495798, | |
| "grad_norm": 0.7715353825685302, | |
| "learning_rate": 6.289622690586151e-07, | |
| "loss": 0.1967, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18994097411632538, | |
| "step": 6185, | |
| "valid_targets_mean": 3128.4, | |
| "valid_targets_min": 887 | |
| }, | |
| { | |
| "epoch": 6.5021008403361344, | |
| "grad_norm": 0.8887946663567943, | |
| "learning_rate": 6.159944685759289e-07, | |
| "loss": 0.1788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1865498125553131, | |
| "step": 6190, | |
| "valid_targets_mean": 2712.2, | |
| "valid_targets_min": 1279 | |
| }, | |
| { | |
| "epoch": 6.507352941176471, | |
| "grad_norm": 0.6867564122848154, | |
| "learning_rate": 6.031596568521792e-07, | |
| "loss": 0.1742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1672389805316925, | |
| "step": 6195, | |
| "valid_targets_mean": 3052.6, | |
| "valid_targets_min": 1021 | |
| }, | |
| { | |
| "epoch": 6.512605042016807, | |
| "grad_norm": 0.8081151613494477, | |
| "learning_rate": 5.904579219437567e-07, | |
| "loss": 0.1874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1853436529636383, | |
| "step": 6200, | |
| "valid_targets_mean": 3352.1, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 6.517857142857143, | |
| "grad_norm": 0.8297017876682581, | |
| "learning_rate": 5.778893509940497e-07, | |
| "loss": 0.1855, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2378537356853485, | |
| "step": 6205, | |
| "valid_targets_mean": 2860.6, | |
| "valid_targets_min": 1039 | |
| }, | |
| { | |
| "epoch": 6.523109243697479, | |
| "grad_norm": 0.7037062721004519, | |
| "learning_rate": 5.654540302328526e-07, | |
| "loss": 0.1874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17788448929786682, | |
| "step": 6210, | |
| "valid_targets_mean": 3742.9, | |
| "valid_targets_min": 2459 | |
| }, | |
| { | |
| "epoch": 6.5283613445378155, | |
| "grad_norm": 0.6751043734055584, | |
| "learning_rate": 5.531520449757465e-07, | |
| "loss": 0.1865, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16557300090789795, | |
| "step": 6215, | |
| "valid_targets_mean": 3435.8, | |
| "valid_targets_min": 1276 | |
| }, | |
| { | |
| "epoch": 6.533613445378151, | |
| "grad_norm": 0.6332717965969442, | |
| "learning_rate": 5.409834796235447e-07, | |
| "loss": 0.1827, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17843201756477356, | |
| "step": 6220, | |
| "valid_targets_mean": 4335.8, | |
| "valid_targets_min": 1572 | |
| }, | |
| { | |
| "epoch": 6.538865546218488, | |
| "grad_norm": 0.7540653616743905, | |
| "learning_rate": 5.28948417661701e-07, | |
| "loss": 0.1847, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1931922882795334, | |
| "step": 6225, | |
| "valid_targets_mean": 2975.5, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.544117647058823, | |
| "grad_norm": 0.945825539132947, | |
| "learning_rate": 5.170469416597223e-07, | |
| "loss": 0.1348, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1272469162940979, | |
| "step": 6230, | |
| "valid_targets_mean": 1444.9, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 6.54936974789916, | |
| "grad_norm": 0.9201038959571372, | |
| "learning_rate": 5.05279133270633e-07, | |
| "loss": 0.1261, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12720605731010437, | |
| "step": 6235, | |
| "valid_targets_mean": 1456.5, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 6.554621848739496, | |
| "grad_norm": 0.9294548695403332, | |
| "learning_rate": 4.936450732303866e-07, | |
| "loss": 0.1212, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12623611092567444, | |
| "step": 6240, | |
| "valid_targets_mean": 1432.6, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 6.559873949579832, | |
| "grad_norm": 0.81095763836755, | |
| "learning_rate": 4.821448413573237e-07, | |
| "loss": 0.1169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11129063367843628, | |
| "step": 6245, | |
| "valid_targets_mean": 1432.6, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 6.565126050420168, | |
| "grad_norm": 0.8336053499869773, | |
| "learning_rate": 4.70778516551631e-07, | |
| "loss": 0.123, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11743263155221939, | |
| "step": 6250, | |
| "valid_targets_mean": 1387.0, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 6.570378151260504, | |
| "grad_norm": 0.8189714574359003, | |
| "learning_rate": 4.595461767947829e-07, | |
| "loss": 0.1256, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12419360131025314, | |
| "step": 6255, | |
| "valid_targets_mean": 1527.4, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 6.57563025210084, | |
| "grad_norm": 0.8726128195536536, | |
| "learning_rate": 4.484478991490249e-07, | |
| "loss": 0.108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10007977485656738, | |
| "step": 6260, | |
| "valid_targets_mean": 1353.6, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 6.580882352941177, | |
| "grad_norm": 0.9528237066239539, | |
| "learning_rate": 4.374837597568338e-07, | |
| "loss": 0.1271, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10588499903678894, | |
| "step": 6265, | |
| "valid_targets_mean": 1309.9, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 6.586134453781512, | |
| "grad_norm": 0.773375907639893, | |
| "learning_rate": 4.266538338403936e-07, | |
| "loss": 0.1134, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1066712811589241, | |
| "step": 6270, | |
| "valid_targets_mean": 1465.4, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 6.591386554621849, | |
| "grad_norm": 0.7447038454026255, | |
| "learning_rate": 4.159581957010894e-07, | |
| "loss": 0.1176, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10983015596866608, | |
| "step": 6275, | |
| "valid_targets_mean": 1379.4, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 6.5966386554621845, | |
| "grad_norm": 0.8705541079832045, | |
| "learning_rate": 4.0539691871898545e-07, | |
| "loss": 0.1092, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10850921273231506, | |
| "step": 6280, | |
| "valid_targets_mean": 1158.2, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 6.601890756302521, | |
| "grad_norm": 0.9416604661089228, | |
| "learning_rate": 3.949700753523344e-07, | |
| "loss": 0.1118, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1087367981672287, | |
| "step": 6285, | |
| "valid_targets_mean": 1543.1, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 6.607142857142857, | |
| "grad_norm": 0.7499389879872029, | |
| "learning_rate": 3.8467773713707134e-07, | |
| "loss": 0.1101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0929175540804863, | |
| "step": 6290, | |
| "valid_targets_mean": 1300.1, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 6.612394957983193, | |
| "grad_norm": 0.7928586147639457, | |
| "learning_rate": 3.7451997468632486e-07, | |
| "loss": 0.1076, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10093091428279877, | |
| "step": 6295, | |
| "valid_targets_mean": 1401.3, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 6.617647058823529, | |
| "grad_norm": 0.9017571609664792, | |
| "learning_rate": 3.6449685768993327e-07, | |
| "loss": 0.1058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11259915679693222, | |
| "step": 6300, | |
| "valid_targets_mean": 1436.8, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 6.6228991596638656, | |
| "grad_norm": 0.8365779847032611, | |
| "learning_rate": 3.5460845491396944e-07, | |
| "loss": 0.1145, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12376931309700012, | |
| "step": 6305, | |
| "valid_targets_mean": 1531.6, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 6.628151260504202, | |
| "grad_norm": 0.889104983252716, | |
| "learning_rate": 3.448548342002589e-07, | |
| "loss": 0.1106, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1163712590932846, | |
| "step": 6310, | |
| "valid_targets_mean": 1469.6, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 6.633403361344538, | |
| "grad_norm": 0.8740557078182234, | |
| "learning_rate": 3.352360624659312e-07, | |
| "loss": 0.1176, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12843522429466248, | |
| "step": 6315, | |
| "valid_targets_mean": 1437.8, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 6.6386554621848735, | |
| "grad_norm": 0.8792645901441288, | |
| "learning_rate": 3.2575220570294276e-07, | |
| "loss": 0.1102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11173952370882034, | |
| "step": 6320, | |
| "valid_targets_mean": 1327.3, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 6.64390756302521, | |
| "grad_norm": 0.8115595326058354, | |
| "learning_rate": 3.164033289776369e-07, | |
| "loss": 0.1138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12014767527580261, | |
| "step": 6325, | |
| "valid_targets_mean": 1685.2, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 6.649159663865547, | |
| "grad_norm": 0.9521276969718082, | |
| "learning_rate": 3.071894964302935e-07, | |
| "loss": 0.113, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12036529183387756, | |
| "step": 6330, | |
| "valid_targets_mean": 1595.3, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 6.654411764705882, | |
| "grad_norm": 0.7903280323515541, | |
| "learning_rate": 2.981107712746867e-07, | |
| "loss": 0.1071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09706208854913712, | |
| "step": 6335, | |
| "valid_targets_mean": 1381.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 6.659663865546219, | |
| "grad_norm": 0.8355498112239595, | |
| "learning_rate": 2.891672157976522e-07, | |
| "loss": 0.1063, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10185357928276062, | |
| "step": 6340, | |
| "valid_targets_mean": 1345.7, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 6.6649159663865545, | |
| "grad_norm": 0.8948565870788178, | |
| "learning_rate": 2.803588913586608e-07, | |
| "loss": 0.1052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10712474584579468, | |
| "step": 6345, | |
| "valid_targets_mean": 1342.1, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 6.670168067226891, | |
| "grad_norm": 0.8221625640757199, | |
| "learning_rate": 2.716858583894033e-07, | |
| "loss": 0.1075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10669823735952377, | |
| "step": 6350, | |
| "valid_targets_mean": 1450.6, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 6.675420168067227, | |
| "grad_norm": 0.8413623877263701, | |
| "learning_rate": 2.6314817639335964e-07, | |
| "loss": 0.109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11155045032501221, | |
| "step": 6355, | |
| "valid_targets_mean": 1461.3, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 6.680672268907563, | |
| "grad_norm": 0.8587976871867034, | |
| "learning_rate": 2.547459039454103e-07, | |
| "loss": 0.1128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.120185486972332, | |
| "step": 6360, | |
| "valid_targets_mean": 1723.1, | |
| "valid_targets_min": 1075 | |
| }, | |
| { | |
| "epoch": 6.685924369747899, | |
| "grad_norm": 0.92862789354394, | |
| "learning_rate": 2.4647909869142117e-07, | |
| "loss": 0.1139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1129080206155777, | |
| "step": 6365, | |
| "valid_targets_mean": 1669.1, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 6.6911764705882355, | |
| "grad_norm": 0.8365500928151788, | |
| "learning_rate": 2.3834781734784817e-07, | |
| "loss": 0.1094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11981870979070663, | |
| "step": 6370, | |
| "valid_targets_mean": 1521.4, | |
| "valid_targets_min": 957 | |
| }, | |
| { | |
| "epoch": 6.696428571428571, | |
| "grad_norm": 0.8689676540452689, | |
| "learning_rate": 2.3035211570135995e-07, | |
| "loss": 0.1085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09756392240524292, | |
| "step": 6375, | |
| "valid_targets_mean": 1282.1, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 6.701680672268908, | |
| "grad_norm": 0.9369896336428621, | |
| "learning_rate": 2.224920486084403e-07, | |
| "loss": 0.1135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10868951678276062, | |
| "step": 6380, | |
| "valid_targets_mean": 1499.7, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 6.706932773109243, | |
| "grad_norm": 0.8236205446292434, | |
| "learning_rate": 2.1476766999502408e-07, | |
| "loss": 0.1034, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11456996202468872, | |
| "step": 6385, | |
| "valid_targets_mean": 1621.1, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 6.71218487394958, | |
| "grad_norm": 0.8914772064174625, | |
| "learning_rate": 2.071790328561152e-07, | |
| "loss": 0.1056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11084552109241486, | |
| "step": 6390, | |
| "valid_targets_mean": 1390.9, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.717436974789916, | |
| "grad_norm": 0.847420104124652, | |
| "learning_rate": 1.997261892554403e-07, | |
| "loss": 0.0992, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09959986060857773, | |
| "step": 6395, | |
| "valid_targets_mean": 1254.4, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 6.722689075630252, | |
| "grad_norm": 0.874119271608596, | |
| "learning_rate": 1.9240919032506688e-07, | |
| "loss": 0.1089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10325537621974945, | |
| "step": 6400, | |
| "valid_targets_mean": 1479.2, | |
| "valid_targets_min": 817 | |
| }, | |
| { | |
| "epoch": 6.727941176470588, | |
| "grad_norm": 0.8485613033201405, | |
| "learning_rate": 1.8522808626507683e-07, | |
| "loss": 0.1064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11246100813150406, | |
| "step": 6405, | |
| "valid_targets_mean": 1464.4, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 6.733193277310924, | |
| "grad_norm": 0.8500768031572533, | |
| "learning_rate": 1.781829263432111e-07, | |
| "loss": 0.1057, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10820727050304413, | |
| "step": 6410, | |
| "valid_targets_mean": 1381.2, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 6.73844537815126, | |
| "grad_norm": 1.0070862147853497, | |
| "learning_rate": 1.7127375889452569e-07, | |
| "loss": 0.1097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10374858230352402, | |
| "step": 6415, | |
| "valid_targets_mean": 1486.0, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 6.743697478991597, | |
| "grad_norm": 0.9783876731250509, | |
| "learning_rate": 1.6450063132107396e-07, | |
| "loss": 0.1081, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11519736051559448, | |
| "step": 6420, | |
| "valid_targets_mean": 1459.7, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 6.748949579831933, | |
| "grad_norm": 0.8259991898073508, | |
| "learning_rate": 1.5786359009156928e-07, | |
| "loss": 0.1044, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09908154606819153, | |
| "step": 6425, | |
| "valid_targets_mean": 1397.0, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 6.754201680672269, | |
| "grad_norm": 0.8592177771547324, | |
| "learning_rate": 1.5136268074107398e-07, | |
| "loss": 0.1029, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10556816309690475, | |
| "step": 6430, | |
| "valid_targets_mean": 1288.8, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 6.759453781512605, | |
| "grad_norm": 0.876887343291053, | |
| "learning_rate": 1.44997947870682e-07, | |
| "loss": 0.1128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11252713203430176, | |
| "step": 6435, | |
| "valid_targets_mean": 1446.1, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 6.764705882352941, | |
| "grad_norm": 0.9653593171457929, | |
| "learning_rate": 1.3876943514721465e-07, | |
| "loss": 0.1151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14397484064102173, | |
| "step": 6440, | |
| "valid_targets_mean": 1673.2, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 6.769957983193278, | |
| "grad_norm": 0.8493841465052333, | |
| "learning_rate": 1.3267718530292296e-07, | |
| "loss": 0.1042, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10939159989356995, | |
| "step": 6445, | |
| "valid_targets_mean": 1598.7, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 6.775210084033613, | |
| "grad_norm": 0.821915166356728, | |
| "learning_rate": 1.267212401351925e-07, | |
| "loss": 0.1029, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10267669707536697, | |
| "step": 6450, | |
| "valid_targets_mean": 1445.2, | |
| "valid_targets_min": 551 | |
| }, | |
| { | |
| "epoch": 6.78046218487395, | |
| "grad_norm": 0.9162688760398915, | |
| "learning_rate": 1.2090164050625907e-07, | |
| "loss": 0.1092, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11637894809246063, | |
| "step": 6455, | |
| "valid_targets_mean": 1552.2, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.785714285714286, | |
| "grad_norm": 0.8967489554798673, | |
| "learning_rate": 1.1521842634292013e-07, | |
| "loss": 0.1055, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10807375609874725, | |
| "step": 6460, | |
| "valid_targets_mean": 1511.6, | |
| "valid_targets_min": 825 | |
| }, | |
| { | |
| "epoch": 6.790966386554622, | |
| "grad_norm": 0.8909809447006225, | |
| "learning_rate": 1.0967163663627044e-07, | |
| "loss": 0.1008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10812873393297195, | |
| "step": 6465, | |
| "valid_targets_mean": 1489.3, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 6.796218487394958, | |
| "grad_norm": 0.8597923475845829, | |
| "learning_rate": 1.0426130944143353e-07, | |
| "loss": 0.0978, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09637298434972763, | |
| "step": 6470, | |
| "valid_targets_mean": 1356.9, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 6.801470588235294, | |
| "grad_norm": 0.8605974789325253, | |
| "learning_rate": 9.898748187729513e-08, | |
| "loss": 0.111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09894619882106781, | |
| "step": 6475, | |
| "valid_targets_mean": 1359.9, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.80672268907563, | |
| "grad_norm": 0.9100875928994199, | |
| "learning_rate": 9.385019012625007e-08, | |
| "loss": 0.1135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11625594645738602, | |
| "step": 6480, | |
| "valid_targets_mean": 1443.4, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 6.811974789915967, | |
| "grad_norm": 0.8583853259041662, | |
| "learning_rate": 8.884946943395811e-08, | |
| "loss": 0.1104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11173100769519806, | |
| "step": 6485, | |
| "valid_targets_mean": 1376.3, | |
| "valid_targets_min": 944 | |
| }, | |
| { | |
| "epoch": 6.817226890756302, | |
| "grad_norm": 0.7696906815936799, | |
| "learning_rate": 8.398535410910402e-08, | |
| "loss": 0.1038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09793375432491302, | |
| "step": 6490, | |
| "valid_targets_mean": 1597.7, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 6.822478991596639, | |
| "grad_norm": 0.7990711794579531, | |
| "learning_rate": 7.925787752314674e-08, | |
| "loss": 0.1124, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09811538457870483, | |
| "step": 6495, | |
| "valid_targets_mean": 1419.2, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 6.8277310924369745, | |
| "grad_norm": 0.7717945193131686, | |
| "learning_rate": 7.466707211010838e-08, | |
| "loss": 0.108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10290801525115967, | |
| "step": 6500, | |
| "valid_targets_mean": 1570.5, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 6.832983193277311, | |
| "grad_norm": 0.9330044417782807, | |
| "learning_rate": 7.02129693663478e-08, | |
| "loss": 0.102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10135114192962646, | |
| "step": 6505, | |
| "valid_targets_mean": 1177.1, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 6.838235294117647, | |
| "grad_norm": 0.885799823289419, | |
| "learning_rate": 6.589559985033189e-08, | |
| "loss": 0.1051, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10052669793367386, | |
| "step": 6510, | |
| "valid_targets_mean": 1280.1, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 6.843487394957983, | |
| "grad_norm": 0.8895050648007775, | |
| "learning_rate": 6.171499318244234e-08, | |
| "loss": 0.0972, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0963287502527237, | |
| "step": 6515, | |
| "valid_targets_mean": 1253.6, | |
| "valid_targets_min": 604 | |
| }, | |
| { | |
| "epoch": 6.848739495798319, | |
| "grad_norm": 0.8626286788780626, | |
| "learning_rate": 5.767117804476696e-08, | |
| "loss": 0.108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0979156345129013, | |
| "step": 6520, | |
| "valid_targets_mean": 1232.8, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 6.8539915966386555, | |
| "grad_norm": 0.8311072710033767, | |
| "learning_rate": 5.376418218089541e-08, | |
| "loss": 0.109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1046762764453888, | |
| "step": 6525, | |
| "valid_targets_mean": 1484.8, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 6.859243697478991, | |
| "grad_norm": 0.9079288548187959, | |
| "learning_rate": 4.99940323957393e-08, | |
| "loss": 0.1056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11257081478834152, | |
| "step": 6530, | |
| "valid_targets_mean": 1383.6, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.864495798319328, | |
| "grad_norm": 0.8918440819640617, | |
| "learning_rate": 4.63607545553435e-08, | |
| "loss": 0.1065, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10769190639257431, | |
| "step": 6535, | |
| "valid_targets_mean": 1386.2, | |
| "valid_targets_min": 840 | |
| }, | |
| { | |
| "epoch": 6.869747899159664, | |
| "grad_norm": 0.9081288698653714, | |
| "learning_rate": 4.2864373586706254e-08, | |
| "loss": 0.1084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10562418401241302, | |
| "step": 6540, | |
| "valid_targets_mean": 1375.6, | |
| "valid_targets_min": 772 | |
| }, | |
| { | |
| "epoch": 6.875, | |
| "grad_norm": 0.8331522601558252, | |
| "learning_rate": 3.950491347761487e-08, | |
| "loss": 0.1045, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10698384046554565, | |
| "step": 6545, | |
| "valid_targets_mean": 1612.4, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 6.880252100840336, | |
| "grad_norm": 0.8223352044912656, | |
| "learning_rate": 3.628239727647254e-08, | |
| "loss": 0.1039, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10564574599266052, | |
| "step": 6550, | |
| "valid_targets_mean": 1427.0, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 6.885504201680672, | |
| "grad_norm": 1.2026759527578401, | |
| "learning_rate": 3.319684709215176e-08, | |
| "loss": 0.1063, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10793040692806244, | |
| "step": 6555, | |
| "valid_targets_mean": 1479.3, | |
| "valid_targets_min": 766 | |
| }, | |
| { | |
| "epoch": 6.890756302521009, | |
| "grad_norm": 0.8039951616902973, | |
| "learning_rate": 3.024828409383007e-08, | |
| "loss": 0.1018, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09126448631286621, | |
| "step": 6560, | |
| "valid_targets_mean": 1241.5, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 6.8960084033613445, | |
| "grad_norm": 0.8938242214668102, | |
| "learning_rate": 2.743672851085233e-08, | |
| "loss": 0.1082, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11531618982553482, | |
| "step": 6565, | |
| "valid_targets_mean": 1410.9, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 6.901260504201681, | |
| "grad_norm": 0.891140302112469, | |
| "learning_rate": 2.4762199632588634e-08, | |
| "loss": 0.1181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11160510778427124, | |
| "step": 6570, | |
| "valid_targets_mean": 1569.1, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 6.906512605042017, | |
| "grad_norm": 0.8886137842875292, | |
| "learning_rate": 2.2224715808309983e-08, | |
| "loss": 0.1078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10890690982341766, | |
| "step": 6575, | |
| "valid_targets_mean": 1378.0, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 6.911764705882353, | |
| "grad_norm": 0.9367326463088681, | |
| "learning_rate": 1.9824294447043923e-08, | |
| "loss": 0.1115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13127145171165466, | |
| "step": 6580, | |
| "valid_targets_mean": 1620.0, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 6.917016806722689, | |
| "grad_norm": 0.838793196574803, | |
| "learning_rate": 1.7560952017481313e-08, | |
| "loss": 0.1025, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09811349958181381, | |
| "step": 6585, | |
| "valid_targets_mean": 1355.1, | |
| "valid_targets_min": 564 | |
| }, | |
| { | |
| "epoch": 6.9222689075630255, | |
| "grad_norm": 0.8347467550105157, | |
| "learning_rate": 1.5434704047836424e-08, | |
| "loss": 0.1027, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09899817407131195, | |
| "step": 6590, | |
| "valid_targets_mean": 1419.1, | |
| "valid_targets_min": 651 | |
| }, | |
| { | |
| "epoch": 6.927521008403361, | |
| "grad_norm": 0.8905919808272537, | |
| "learning_rate": 1.344556512576256e-08, | |
| "loss": 0.1108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10358569025993347, | |
| "step": 6595, | |
| "valid_targets_mean": 1291.7, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 6.932773109243698, | |
| "grad_norm": 0.8193427692652885, | |
| "learning_rate": 1.1593548898236606e-08, | |
| "loss": 0.1027, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09667745232582092, | |
| "step": 6600, | |
| "valid_targets_mean": 1236.8, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 6.938025210084033, | |
| "grad_norm": 0.8054209308544474, | |
| "learning_rate": 9.878668071474639e-09, | |
| "loss": 0.101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08912120014429092, | |
| "step": 6605, | |
| "valid_targets_mean": 1258.9, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 6.94327731092437, | |
| "grad_norm": 0.8379632670512256, | |
| "learning_rate": 8.3009344108409e-09, | |
| "loss": 0.1071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10500848293304443, | |
| "step": 6610, | |
| "valid_targets_mean": 1495.5, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 6.948529411764706, | |
| "grad_norm": 0.9047510484875828, | |
| "learning_rate": 6.860358740763406e-09, | |
| "loss": 0.1143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12173473089933395, | |
| "step": 6615, | |
| "valid_targets_mean": 1421.0, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 6.953781512605042, | |
| "grad_norm": 0.8386152042305247, | |
| "learning_rate": 5.55695094467179e-09, | |
| "loss": 0.1013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11426173895597458, | |
| "step": 6620, | |
| "valid_targets_mean": 1487.2, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 6.959033613445378, | |
| "grad_norm": 0.8298195930802489, | |
| "learning_rate": 4.3907199649151355e-09, | |
| "loss": 0.105, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10774379968643188, | |
| "step": 6625, | |
| "valid_targets_mean": 1512.4, | |
| "valid_targets_min": 823 | |
| }, | |
| { | |
| "epoch": 6.964285714285714, | |
| "grad_norm": 0.7758294188223108, | |
| "learning_rate": 3.361673802708687e-09, | |
| "loss": 0.0926, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08553959429264069, | |
| "step": 6630, | |
| "valid_targets_mean": 1375.4, | |
| "valid_targets_min": 595 | |
| }, | |
| { | |
| "epoch": 6.96953781512605, | |
| "grad_norm": 0.9086042968946347, | |
| "learning_rate": 2.469819518080563e-09, | |
| "loss": 0.1032, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11077478528022766, | |
| "step": 6635, | |
| "valid_targets_mean": 1401.2, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 6.974789915966387, | |
| "grad_norm": 1.0872068857891495, | |
| "learning_rate": 1.7151632298140209e-09, | |
| "loss": 0.1005, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09122096002101898, | |
| "step": 6640, | |
| "valid_targets_mean": 1185.6, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 6.980042016806722, | |
| "grad_norm": 0.9264996959156637, | |
| "learning_rate": 1.0977101154163727e-09, | |
| "loss": 0.102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08891530334949493, | |
| "step": 6645, | |
| "valid_targets_mean": 1364.2, | |
| "valid_targets_min": 685 | |
| }, | |
| { | |
| "epoch": 6.985294117647059, | |
| "grad_norm": 0.8282065868459698, | |
| "learning_rate": 6.174644110767958e-10, | |
| "loss": 0.0982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0922815352678299, | |
| "step": 6650, | |
| "valid_targets_mean": 1345.2, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 6.990546218487395, | |
| "grad_norm": 0.7963997364292627, | |
| "learning_rate": 2.744294116419077e-10, | |
| "loss": 0.1045, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09688857197761536, | |
| "step": 6655, | |
| "valid_targets_mean": 1431.4, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 6.995798319327731, | |
| "grad_norm": 0.8692235718392078, | |
| "learning_rate": 6.860747058468064e-11, | |
| "loss": 0.1056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11209136247634888, | |
| "step": 6660, | |
| "valid_targets_mean": 1505.8, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0974183976650238, | |
| "step": 6664, | |
| "total_flos": 1189064928198656.0, | |
| "train_loss": 0.2180497848695996, | |
| "train_runtime": 25724.8893, | |
| "train_samples_per_second": 4.145, | |
| "train_steps_per_second": 0.259, | |
| "valid_targets_mean": 1721.1, | |
| "valid_targets_min": 760 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 6664, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 7, | |
| "save_steps": 200, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 1189064928198656.0, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |