Text Generation
Transformers
Safetensors
qwen3
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Instructions to use DCAgent2/nl2bash-stackshuffle with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use DCAgent2/nl2bash-stackshuffle with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="DCAgent2/nl2bash-stackshuffle") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("DCAgent2/nl2bash-stackshuffle") model = AutoModelForCausalLM.from_pretrained("DCAgent2/nl2bash-stackshuffle", device_map="auto") 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 Settings
- vLLM
How to use DCAgent2/nl2bash-stackshuffle with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "DCAgent2/nl2bash-stackshuffle" # 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/nl2bash-stackshuffle", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/DCAgent2/nl2bash-stackshuffle
- SGLang
How to use DCAgent2/nl2bash-stackshuffle 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/nl2bash-stackshuffle" \ --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/nl2bash-stackshuffle", "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/nl2bash-stackshuffle" \ --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/nl2bash-stackshuffle", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use DCAgent2/nl2bash-stackshuffle with Docker Model Runner:
docker model run hf.co/DCAgent2/nl2bash-stackshuffle
| { | |
| "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": 12.87344450799192, | |
| "learning_rate": 2.39880059970015e-07, | |
| "loss": 0.7749, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7653565406799316, | |
| "step": 5, | |
| "valid_targets_mean": 2484.8, | |
| "valid_targets_min": 784 | |
| }, | |
| { | |
| "epoch": 0.01050420168067227, | |
| "grad_norm": 12.707723147793125, | |
| "learning_rate": 5.397301349325338e-07, | |
| "loss": 0.7606, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7621961236000061, | |
| "step": 10, | |
| "valid_targets_mean": 2830.7, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 0.015756302521008403, | |
| "grad_norm": 11.204400117593265, | |
| "learning_rate": 8.395802098950526e-07, | |
| "loss": 0.7482, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7473458051681519, | |
| "step": 15, | |
| "valid_targets_mean": 3091.0, | |
| "valid_targets_min": 1146 | |
| }, | |
| { | |
| "epoch": 0.02100840336134454, | |
| "grad_norm": 9.688069489628813, | |
| "learning_rate": 1.1394302848575713e-06, | |
| "loss": 0.7466, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7422138452529907, | |
| "step": 20, | |
| "valid_targets_mean": 2413.2, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 0.026260504201680673, | |
| "grad_norm": 7.62742888265603, | |
| "learning_rate": 1.43928035982009e-06, | |
| "loss": 0.7287, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7463569045066833, | |
| "step": 25, | |
| "valid_targets_mean": 2688.5, | |
| "valid_targets_min": 1011 | |
| }, | |
| { | |
| "epoch": 0.031512605042016806, | |
| "grad_norm": 5.508936794357446, | |
| "learning_rate": 1.7391304347826088e-06, | |
| "loss": 0.6518, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6350823640823364, | |
| "step": 30, | |
| "valid_targets_mean": 2500.6, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 0.03676470588235294, | |
| "grad_norm": 4.810314643948938, | |
| "learning_rate": 2.0389805097451275e-06, | |
| "loss": 0.6558, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6311067938804626, | |
| "step": 35, | |
| "valid_targets_mean": 2686.3, | |
| "valid_targets_min": 893 | |
| }, | |
| { | |
| "epoch": 0.04201680672268908, | |
| "grad_norm": 4.8557060388176545, | |
| "learning_rate": 2.3388305847076464e-06, | |
| "loss": 0.6284, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6066815257072449, | |
| "step": 40, | |
| "valid_targets_mean": 2177.2, | |
| "valid_targets_min": 1080 | |
| }, | |
| { | |
| "epoch": 0.04726890756302521, | |
| "grad_norm": 2.9119019293559685, | |
| "learning_rate": 2.6386806596701653e-06, | |
| "loss": 0.6094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6400288343429565, | |
| "step": 45, | |
| "valid_targets_mean": 2633.9, | |
| "valid_targets_min": 963 | |
| }, | |
| { | |
| "epoch": 0.052521008403361345, | |
| "grad_norm": 2.5325988257181504, | |
| "learning_rate": 2.9385307346326843e-06, | |
| "loss": 0.5741, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5878869891166687, | |
| "step": 50, | |
| "valid_targets_mean": 2203.5, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 0.05777310924369748, | |
| "grad_norm": 1.6332369639610234, | |
| "learning_rate": 3.2383808095952024e-06, | |
| "loss": 0.5072, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.49458950757980347, | |
| "step": 55, | |
| "valid_targets_mean": 3035.3, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 0.06302521008403361, | |
| "grad_norm": 1.2034100946296309, | |
| "learning_rate": 3.5382308845577213e-06, | |
| "loss": 0.5137, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4873230755329132, | |
| "step": 60, | |
| "valid_targets_mean": 2940.2, | |
| "valid_targets_min": 1009 | |
| }, | |
| { | |
| "epoch": 0.06827731092436974, | |
| "grad_norm": 1.1067968813199072, | |
| "learning_rate": 3.83808095952024e-06, | |
| "loss": 0.511, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4636586606502533, | |
| "step": 65, | |
| "valid_targets_mean": 2532.8, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 0.07352941176470588, | |
| "grad_norm": 1.2850520360535964, | |
| "learning_rate": 4.137931034482759e-06, | |
| "loss": 0.4961, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4967666566371918, | |
| "step": 70, | |
| "valid_targets_mean": 1951.5, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 0.07878151260504201, | |
| "grad_norm": 0.8971762854941586, | |
| "learning_rate": 4.437781109445278e-06, | |
| "loss": 0.4949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.43605613708496094, | |
| "step": 75, | |
| "valid_targets_mean": 2723.1, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 0.08403361344537816, | |
| "grad_norm": 0.9430146035301179, | |
| "learning_rate": 4.737631184407796e-06, | |
| "loss": 0.4615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4186355173587799, | |
| "step": 80, | |
| "valid_targets_mean": 2138.4, | |
| "valid_targets_min": 619 | |
| }, | |
| { | |
| "epoch": 0.08928571428571429, | |
| "grad_norm": 0.9397684354473573, | |
| "learning_rate": 5.037481259370315e-06, | |
| "loss": 0.4629, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.447378933429718, | |
| "step": 85, | |
| "valid_targets_mean": 2195.4, | |
| "valid_targets_min": 985 | |
| }, | |
| { | |
| "epoch": 0.09453781512605042, | |
| "grad_norm": 0.9254373058058498, | |
| "learning_rate": 5.337331334332834e-06, | |
| "loss": 0.4579, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.44608554244041443, | |
| "step": 90, | |
| "valid_targets_mean": 2305.1, | |
| "valid_targets_min": 1115 | |
| }, | |
| { | |
| "epoch": 0.09978991596638656, | |
| "grad_norm": 0.8895449710201235, | |
| "learning_rate": 5.6371814092953526e-06, | |
| "loss": 0.4429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4577831029891968, | |
| "step": 95, | |
| "valid_targets_mean": 2365.8, | |
| "valid_targets_min": 1000 | |
| }, | |
| { | |
| "epoch": 0.10504201680672269, | |
| "grad_norm": 0.8582249839693272, | |
| "learning_rate": 5.937031484257871e-06, | |
| "loss": 0.4524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.43060487508773804, | |
| "step": 100, | |
| "valid_targets_mean": 2525.8, | |
| "valid_targets_min": 691 | |
| }, | |
| { | |
| "epoch": 0.11029411764705882, | |
| "grad_norm": 0.8328052594807136, | |
| "learning_rate": 6.2368815592203904e-06, | |
| "loss": 0.4351, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.43569284677505493, | |
| "step": 105, | |
| "valid_targets_mean": 2433.1, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 0.11554621848739496, | |
| "grad_norm": 0.8004894157898372, | |
| "learning_rate": 6.536731634182909e-06, | |
| "loss": 0.435, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4115639328956604, | |
| "step": 110, | |
| "valid_targets_mean": 2809.9, | |
| "valid_targets_min": 1462 | |
| }, | |
| { | |
| "epoch": 0.1207983193277311, | |
| "grad_norm": 0.8370123985640117, | |
| "learning_rate": 6.8365817091454274e-06, | |
| "loss": 0.4329, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.40281569957733154, | |
| "step": 115, | |
| "valid_targets_mean": 2199.2, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 0.12605042016806722, | |
| "grad_norm": 0.8695717739921266, | |
| "learning_rate": 7.136431784107947e-06, | |
| "loss": 0.4223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4021134674549103, | |
| "step": 120, | |
| "valid_targets_mean": 2273.1, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 0.13130252100840337, | |
| "grad_norm": 0.8006986968807501, | |
| "learning_rate": 7.436281859070465e-06, | |
| "loss": 0.4335, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4061618447303772, | |
| "step": 125, | |
| "valid_targets_mean": 2860.1, | |
| "valid_targets_min": 1215 | |
| }, | |
| { | |
| "epoch": 0.13655462184873948, | |
| "grad_norm": 0.8969789811481348, | |
| "learning_rate": 7.736131934032984e-06, | |
| "loss": 0.4152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.47167330980300903, | |
| "step": 130, | |
| "valid_targets_mean": 2499.8, | |
| "valid_targets_min": 1218 | |
| }, | |
| { | |
| "epoch": 0.14180672268907563, | |
| "grad_norm": 0.8068868280254967, | |
| "learning_rate": 8.035982008995503e-06, | |
| "loss": 0.4163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4264189898967743, | |
| "step": 135, | |
| "valid_targets_mean": 2426.6, | |
| "valid_targets_min": 985 | |
| }, | |
| { | |
| "epoch": 0.14705882352941177, | |
| "grad_norm": 0.9031868308784958, | |
| "learning_rate": 8.335832083958023e-06, | |
| "loss": 0.4126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4020998775959015, | |
| "step": 140, | |
| "valid_targets_mean": 2240.5, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 0.15231092436974789, | |
| "grad_norm": 0.7811330017492059, | |
| "learning_rate": 8.63568215892054e-06, | |
| "loss": 0.3964, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4105680286884308, | |
| "step": 145, | |
| "valid_targets_mean": 2559.5, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 0.15756302521008403, | |
| "grad_norm": 0.8087924248265574, | |
| "learning_rate": 8.93553223388306e-06, | |
| "loss": 0.4058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4330301880836487, | |
| "step": 150, | |
| "valid_targets_mean": 2632.9, | |
| "valid_targets_min": 1231 | |
| }, | |
| { | |
| "epoch": 0.16281512605042017, | |
| "grad_norm": 0.7719589146187298, | |
| "learning_rate": 9.235382308845579e-06, | |
| "loss": 0.3981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3908872604370117, | |
| "step": 155, | |
| "valid_targets_mean": 2874.4, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 0.16806722689075632, | |
| "grad_norm": 0.8718120536448756, | |
| "learning_rate": 9.535232383808097e-06, | |
| "loss": 0.3981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.41889429092407227, | |
| "step": 160, | |
| "valid_targets_mean": 2393.1, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 0.17331932773109243, | |
| "grad_norm": 0.8830904007290984, | |
| "learning_rate": 9.835082458770614e-06, | |
| "loss": 0.4037, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38836172223091125, | |
| "step": 165, | |
| "valid_targets_mean": 2164.8, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 0.17857142857142858, | |
| "grad_norm": 0.8311418026284046, | |
| "learning_rate": 1.0134932533733135e-05, | |
| "loss": 0.3887, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3843948543071747, | |
| "step": 170, | |
| "valid_targets_mean": 2470.9, | |
| "valid_targets_min": 1080 | |
| }, | |
| { | |
| "epoch": 0.18382352941176472, | |
| "grad_norm": 0.8211185772584042, | |
| "learning_rate": 1.0434782608695653e-05, | |
| "loss": 0.4169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4080219268798828, | |
| "step": 175, | |
| "valid_targets_mean": 2487.4, | |
| "valid_targets_min": 688 | |
| }, | |
| { | |
| "epoch": 0.18907563025210083, | |
| "grad_norm": 0.8630859100580258, | |
| "learning_rate": 1.0734632683658172e-05, | |
| "loss": 0.3844, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.42095762491226196, | |
| "step": 180, | |
| "valid_targets_mean": 2252.2, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 0.19432773109243698, | |
| "grad_norm": 0.7978994945119684, | |
| "learning_rate": 1.103448275862069e-05, | |
| "loss": 0.3844, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.41790223121643066, | |
| "step": 185, | |
| "valid_targets_mean": 2912.1, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 0.19957983193277312, | |
| "grad_norm": 0.8619777697515183, | |
| "learning_rate": 1.1334332833583211e-05, | |
| "loss": 0.3922, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4195566773414612, | |
| "step": 190, | |
| "valid_targets_mean": 2451.5, | |
| "valid_targets_min": 923 | |
| }, | |
| { | |
| "epoch": 0.20483193277310924, | |
| "grad_norm": 0.8075469823787889, | |
| "learning_rate": 1.1634182908545729e-05, | |
| "loss": 0.381, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3835638463497162, | |
| "step": 195, | |
| "valid_targets_mean": 2334.6, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 0.21008403361344538, | |
| "grad_norm": 0.7886022090561562, | |
| "learning_rate": 1.1934032983508246e-05, | |
| "loss": 0.3732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.41375166177749634, | |
| "step": 200, | |
| "valid_targets_mean": 2972.5, | |
| "valid_targets_min": 1031 | |
| }, | |
| { | |
| "epoch": 0.21533613445378152, | |
| "grad_norm": 0.8061255296748036, | |
| "learning_rate": 1.2233883058470766e-05, | |
| "loss": 0.3888, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4238708019256592, | |
| "step": 205, | |
| "valid_targets_mean": 2525.8, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 0.22058823529411764, | |
| "grad_norm": 0.8001397668970225, | |
| "learning_rate": 1.2533733133433283e-05, | |
| "loss": 0.3628, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3657309412956238, | |
| "step": 210, | |
| "valid_targets_mean": 2680.4, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 0.22584033613445378, | |
| "grad_norm": 0.7848361664620995, | |
| "learning_rate": 1.2833583208395803e-05, | |
| "loss": 0.3762, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.37307411432266235, | |
| "step": 215, | |
| "valid_targets_mean": 2759.8, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 0.23109243697478993, | |
| "grad_norm": 0.9451943941703829, | |
| "learning_rate": 1.313343328335832e-05, | |
| "loss": 0.3721, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35972413420677185, | |
| "step": 220, | |
| "valid_targets_mean": 1986.0, | |
| "valid_targets_min": 1143 | |
| }, | |
| { | |
| "epoch": 0.23634453781512604, | |
| "grad_norm": 1.0257525942300505, | |
| "learning_rate": 1.3433283358320841e-05, | |
| "loss": 0.3903, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.37196558713912964, | |
| "step": 225, | |
| "valid_targets_mean": 2237.6, | |
| "valid_targets_min": 1023 | |
| }, | |
| { | |
| "epoch": 0.2415966386554622, | |
| "grad_norm": 0.8450544047029999, | |
| "learning_rate": 1.3733133433283359e-05, | |
| "loss": 0.3774, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.400920033454895, | |
| "step": 230, | |
| "valid_targets_mean": 2558.1, | |
| "valid_targets_min": 843 | |
| }, | |
| { | |
| "epoch": 0.24684873949579833, | |
| "grad_norm": 0.8081787234697447, | |
| "learning_rate": 1.4032983508245878e-05, | |
| "loss": 0.3585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35195618867874146, | |
| "step": 235, | |
| "valid_targets_mean": 2621.5, | |
| "valid_targets_min": 749 | |
| }, | |
| { | |
| "epoch": 0.25210084033613445, | |
| "grad_norm": 0.8046828946061387, | |
| "learning_rate": 1.4332833583208396e-05, | |
| "loss": 0.3612, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.340299516916275, | |
| "step": 240, | |
| "valid_targets_mean": 2374.2, | |
| "valid_targets_min": 1079 | |
| }, | |
| { | |
| "epoch": 0.25735294117647056, | |
| "grad_norm": 0.6971318909820771, | |
| "learning_rate": 1.4632683658170917e-05, | |
| "loss": 0.357, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3586806356906891, | |
| "step": 245, | |
| "valid_targets_mean": 3350.4, | |
| "valid_targets_min": 707 | |
| }, | |
| { | |
| "epoch": 0.26260504201680673, | |
| "grad_norm": 0.8055301071475804, | |
| "learning_rate": 1.4932533733133435e-05, | |
| "loss": 0.3433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3433215320110321, | |
| "step": 250, | |
| "valid_targets_mean": 2368.8, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 0.26785714285714285, | |
| "grad_norm": 0.8154286910342443, | |
| "learning_rate": 1.5232383808095954e-05, | |
| "loss": 0.358, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33199673891067505, | |
| "step": 255, | |
| "valid_targets_mean": 2458.3, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 0.27310924369747897, | |
| "grad_norm": 0.8416573919590596, | |
| "learning_rate": 1.5532233883058472e-05, | |
| "loss": 0.3717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3616263270378113, | |
| "step": 260, | |
| "valid_targets_mean": 2318.0, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 0.27836134453781514, | |
| "grad_norm": 0.7678040770406672, | |
| "learning_rate": 1.5832083958020993e-05, | |
| "loss": 0.3621, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32790547609329224, | |
| "step": 265, | |
| "valid_targets_mean": 2317.4, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 0.28361344537815125, | |
| "grad_norm": 0.8164391899177017, | |
| "learning_rate": 1.613193403298351e-05, | |
| "loss": 0.3392, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2930743098258972, | |
| "step": 270, | |
| "valid_targets_mean": 2118.6, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 0.28886554621848737, | |
| "grad_norm": 0.8248688372806788, | |
| "learning_rate": 1.6431784107946028e-05, | |
| "loss": 0.3646, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3935636579990387, | |
| "step": 275, | |
| "valid_targets_mean": 2466.6, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 0.29411764705882354, | |
| "grad_norm": 0.8270276632057298, | |
| "learning_rate": 1.6731634182908546e-05, | |
| "loss": 0.3683, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.36888253688812256, | |
| "step": 280, | |
| "valid_targets_mean": 2239.0, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 0.29936974789915966, | |
| "grad_norm": 0.8394101421073619, | |
| "learning_rate": 1.7031484257871064e-05, | |
| "loss": 0.3563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3346269130706787, | |
| "step": 285, | |
| "valid_targets_mean": 2003.6, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 0.30462184873949577, | |
| "grad_norm": 0.7147367181058835, | |
| "learning_rate": 1.7331334332833585e-05, | |
| "loss": 0.3588, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3740352988243103, | |
| "step": 290, | |
| "valid_targets_mean": 3159.9, | |
| "valid_targets_min": 1551 | |
| }, | |
| { | |
| "epoch": 0.30987394957983194, | |
| "grad_norm": 0.8491621210785679, | |
| "learning_rate": 1.7631184407796102e-05, | |
| "loss": 0.3563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.357749342918396, | |
| "step": 295, | |
| "valid_targets_mean": 2103.2, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 0.31512605042016806, | |
| "grad_norm": 0.8318921687266874, | |
| "learning_rate": 1.7931034482758623e-05, | |
| "loss": 0.3319, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3408452868461609, | |
| "step": 300, | |
| "valid_targets_mean": 2045.0, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 0.32037815126050423, | |
| "grad_norm": 0.8396347924881318, | |
| "learning_rate": 1.823088455772114e-05, | |
| "loss": 0.3421, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3563663959503174, | |
| "step": 305, | |
| "valid_targets_mean": 2256.4, | |
| "valid_targets_min": 854 | |
| }, | |
| { | |
| "epoch": 0.32563025210084034, | |
| "grad_norm": 0.9205468135189474, | |
| "learning_rate": 1.853073463268366e-05, | |
| "loss": 0.3732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.39193105697631836, | |
| "step": 310, | |
| "valid_targets_mean": 2059.1, | |
| "valid_targets_min": 1011 | |
| }, | |
| { | |
| "epoch": 0.33088235294117646, | |
| "grad_norm": 0.9135710271321404, | |
| "learning_rate": 1.8830584707646176e-05, | |
| "loss": 0.3372, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3061055541038513, | |
| "step": 315, | |
| "valid_targets_mean": 1836.5, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 0.33613445378151263, | |
| "grad_norm": 0.913185572728451, | |
| "learning_rate": 1.9130434782608697e-05, | |
| "loss": 0.3472, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3164549171924591, | |
| "step": 320, | |
| "valid_targets_mean": 2192.6, | |
| "valid_targets_min": 594 | |
| }, | |
| { | |
| "epoch": 0.34138655462184875, | |
| "grad_norm": 0.7411966725143038, | |
| "learning_rate": 1.9430284857571215e-05, | |
| "loss": 0.3551, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34439602494239807, | |
| "step": 325, | |
| "valid_targets_mean": 2740.4, | |
| "valid_targets_min": 926 | |
| }, | |
| { | |
| "epoch": 0.34663865546218486, | |
| "grad_norm": 0.7141106760577947, | |
| "learning_rate": 1.9730134932533736e-05, | |
| "loss": 0.3467, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3687133193016052, | |
| "step": 330, | |
| "valid_targets_mean": 2977.1, | |
| "valid_targets_min": 885 | |
| }, | |
| { | |
| "epoch": 0.35189075630252103, | |
| "grad_norm": 0.8006543236123854, | |
| "learning_rate": 2.0029985007496254e-05, | |
| "loss": 0.3435, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3490038514137268, | |
| "step": 335, | |
| "valid_targets_mean": 2345.1, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 0.35714285714285715, | |
| "grad_norm": 0.7847814844972669, | |
| "learning_rate": 2.0329835082458775e-05, | |
| "loss": 0.3439, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3197982609272003, | |
| "step": 340, | |
| "valid_targets_mean": 2583.3, | |
| "valid_targets_min": 758 | |
| }, | |
| { | |
| "epoch": 0.36239495798319327, | |
| "grad_norm": 0.9492857105812816, | |
| "learning_rate": 2.0629685157421292e-05, | |
| "loss": 0.3519, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3854501247406006, | |
| "step": 345, | |
| "valid_targets_mean": 2491.8, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 0.36764705882352944, | |
| "grad_norm": 0.9063284452571975, | |
| "learning_rate": 2.092953523238381e-05, | |
| "loss": 0.3585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3553091883659363, | |
| "step": 350, | |
| "valid_targets_mean": 1893.2, | |
| "valid_targets_min": 638 | |
| }, | |
| { | |
| "epoch": 0.37289915966386555, | |
| "grad_norm": 0.7830522409215324, | |
| "learning_rate": 2.1229385307346328e-05, | |
| "loss": 0.3371, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35570859909057617, | |
| "step": 355, | |
| "valid_targets_mean": 2596.5, | |
| "valid_targets_min": 571 | |
| }, | |
| { | |
| "epoch": 0.37815126050420167, | |
| "grad_norm": 0.9034654728534351, | |
| "learning_rate": 2.152923538230885e-05, | |
| "loss": 0.3633, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3349594175815582, | |
| "step": 360, | |
| "valid_targets_mean": 1895.6, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 0.38340336134453784, | |
| "grad_norm": 0.7718514232416079, | |
| "learning_rate": 2.1829085457271363e-05, | |
| "loss": 0.3499, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3692238926887512, | |
| "step": 365, | |
| "valid_targets_mean": 3021.2, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 0.38865546218487396, | |
| "grad_norm": 0.7100879122429138, | |
| "learning_rate": 2.2128935532233884e-05, | |
| "loss": 0.3618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.38098081946372986, | |
| "step": 370, | |
| "valid_targets_mean": 3113.8, | |
| "valid_targets_min": 1343 | |
| }, | |
| { | |
| "epoch": 0.3939075630252101, | |
| "grad_norm": 0.9674664797223852, | |
| "learning_rate": 2.2428785607196405e-05, | |
| "loss": 0.3563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.335666298866272, | |
| "step": 375, | |
| "valid_targets_mean": 1930.9, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 0.39915966386554624, | |
| "grad_norm": 0.9243345478210326, | |
| "learning_rate": 2.2728635682158923e-05, | |
| "loss": 0.3336, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33842527866363525, | |
| "step": 380, | |
| "valid_targets_mean": 2132.2, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 0.40441176470588236, | |
| "grad_norm": 0.8789172880338919, | |
| "learning_rate": 2.302848575712144e-05, | |
| "loss": 0.3538, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3681184947490692, | |
| "step": 385, | |
| "valid_targets_mean": 2242.4, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 0.4096638655462185, | |
| "grad_norm": 0.8586032021607275, | |
| "learning_rate": 2.3328335832083958e-05, | |
| "loss": 0.3611, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.40162259340286255, | |
| "step": 390, | |
| "valid_targets_mean": 2104.2, | |
| "valid_targets_min": 1341 | |
| }, | |
| { | |
| "epoch": 0.41491596638655465, | |
| "grad_norm": 0.7759128620985057, | |
| "learning_rate": 2.362818590704648e-05, | |
| "loss": 0.3534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30948907136917114, | |
| "step": 395, | |
| "valid_targets_mean": 2376.9, | |
| "valid_targets_min": 1305 | |
| }, | |
| { | |
| "epoch": 0.42016806722689076, | |
| "grad_norm": 0.8511116280560821, | |
| "learning_rate": 2.3928035982009e-05, | |
| "loss": 0.3428, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.335288405418396, | |
| "step": 400, | |
| "valid_targets_mean": 2221.9, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 0.4254201680672269, | |
| "grad_norm": 0.8282588416488064, | |
| "learning_rate": 2.4227886056971515e-05, | |
| "loss": 0.3246, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2992272973060608, | |
| "step": 405, | |
| "valid_targets_mean": 2456.8, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 0.43067226890756305, | |
| "grad_norm": 0.8894895733438389, | |
| "learning_rate": 2.4527736131934036e-05, | |
| "loss": 0.3417, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3676615357398987, | |
| "step": 410, | |
| "valid_targets_mean": 2224.9, | |
| "valid_targets_min": 1024 | |
| }, | |
| { | |
| "epoch": 0.43592436974789917, | |
| "grad_norm": 0.6776249586308555, | |
| "learning_rate": 2.4827586206896553e-05, | |
| "loss": 0.3381, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3416297435760498, | |
| "step": 415, | |
| "valid_targets_mean": 3067.1, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 0.4411764705882353, | |
| "grad_norm": 0.775573641251283, | |
| "learning_rate": 2.5127436281859074e-05, | |
| "loss": 0.3254, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33019617199897766, | |
| "step": 420, | |
| "valid_targets_mean": 2569.1, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 0.44642857142857145, | |
| "grad_norm": 0.8373072328217271, | |
| "learning_rate": 2.542728635682159e-05, | |
| "loss": 0.3402, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33131086826324463, | |
| "step": 425, | |
| "valid_targets_mean": 2256.8, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 0.45168067226890757, | |
| "grad_norm": 0.8453545100732337, | |
| "learning_rate": 2.572713643178411e-05, | |
| "loss": 0.3243, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3512518107891083, | |
| "step": 430, | |
| "valid_targets_mean": 2501.5, | |
| "valid_targets_min": 544 | |
| }, | |
| { | |
| "epoch": 0.4569327731092437, | |
| "grad_norm": 0.8253450678943121, | |
| "learning_rate": 2.602698650674663e-05, | |
| "loss": 0.3365, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3878710865974426, | |
| "step": 435, | |
| "valid_targets_mean": 2262.6, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 0.46218487394957986, | |
| "grad_norm": 0.6830047934435176, | |
| "learning_rate": 2.632683658170915e-05, | |
| "loss": 0.3246, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32986515760421753, | |
| "step": 440, | |
| "valid_targets_mean": 2881.2, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 0.46743697478991597, | |
| "grad_norm": 0.7176824990335685, | |
| "learning_rate": 2.6626686656671666e-05, | |
| "loss": 0.3396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3406386971473694, | |
| "step": 445, | |
| "valid_targets_mean": 2875.5, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 0.4726890756302521, | |
| "grad_norm": 0.8474215992592125, | |
| "learning_rate": 2.6926536731634184e-05, | |
| "loss": 0.3537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3648040294647217, | |
| "step": 450, | |
| "valid_targets_mean": 2027.9, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 0.47794117647058826, | |
| "grad_norm": 1.095403929365044, | |
| "learning_rate": 2.7226386806596705e-05, | |
| "loss": 0.3518, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33743467926979065, | |
| "step": 455, | |
| "valid_targets_mean": 2896.9, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 0.4831932773109244, | |
| "grad_norm": 0.7470324162963815, | |
| "learning_rate": 2.752623688155922e-05, | |
| "loss": 0.3492, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.322024405002594, | |
| "step": 460, | |
| "valid_targets_mean": 2677.2, | |
| "valid_targets_min": 1139 | |
| }, | |
| { | |
| "epoch": 0.4884453781512605, | |
| "grad_norm": 0.7287288163080067, | |
| "learning_rate": 2.782608695652174e-05, | |
| "loss": 0.3337, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3294079601764679, | |
| "step": 465, | |
| "valid_targets_mean": 2836.6, | |
| "valid_targets_min": 668 | |
| }, | |
| { | |
| "epoch": 0.49369747899159666, | |
| "grad_norm": 0.7571889256497794, | |
| "learning_rate": 2.812593703148426e-05, | |
| "loss": 0.3408, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3153488337993622, | |
| "step": 470, | |
| "valid_targets_mean": 2510.1, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 0.4989495798319328, | |
| "grad_norm": 0.7390238171828771, | |
| "learning_rate": 2.8425787106446782e-05, | |
| "loss": 0.3346, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3535560965538025, | |
| "step": 475, | |
| "valid_targets_mean": 2723.5, | |
| "valid_targets_min": 1336 | |
| }, | |
| { | |
| "epoch": 0.5042016806722689, | |
| "grad_norm": 0.7519915733907849, | |
| "learning_rate": 2.8725637181409296e-05, | |
| "loss": 0.3336, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35906684398651123, | |
| "step": 480, | |
| "valid_targets_mean": 2768.1, | |
| "valid_targets_min": 906 | |
| }, | |
| { | |
| "epoch": 0.509453781512605, | |
| "grad_norm": 0.8835497372225222, | |
| "learning_rate": 2.9025487256371818e-05, | |
| "loss": 0.3071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30021750926971436, | |
| "step": 485, | |
| "valid_targets_mean": 1918.7, | |
| "valid_targets_min": 595 | |
| }, | |
| { | |
| "epoch": 0.5147058823529411, | |
| "grad_norm": 0.7720200309244449, | |
| "learning_rate": 2.9325337331334335e-05, | |
| "loss": 0.3371, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2973299026489258, | |
| "step": 490, | |
| "valid_targets_mean": 2384.1, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 0.5199579831932774, | |
| "grad_norm": 0.7920180853215862, | |
| "learning_rate": 2.9625187406296856e-05, | |
| "loss": 0.3357, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31775835156440735, | |
| "step": 495, | |
| "valid_targets_mean": 2716.2, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 0.5252100840336135, | |
| "grad_norm": 0.8442161626685477, | |
| "learning_rate": 2.992503748125937e-05, | |
| "loss": 0.3404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.304614782333374, | |
| "step": 500, | |
| "valid_targets_mean": 1944.6, | |
| "valid_targets_min": 898 | |
| }, | |
| { | |
| "epoch": 0.5304621848739496, | |
| "grad_norm": 0.8063704786094186, | |
| "learning_rate": 3.022488755622189e-05, | |
| "loss": 0.324, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3063597083091736, | |
| "step": 505, | |
| "valid_targets_mean": 2314.1, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 0.5357142857142857, | |
| "grad_norm": 0.8885170741957255, | |
| "learning_rate": 3.052473763118441e-05, | |
| "loss": 0.3186, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.301100492477417, | |
| "step": 510, | |
| "valid_targets_mean": 1699.1, | |
| "valid_targets_min": 649 | |
| }, | |
| { | |
| "epoch": 0.5409663865546218, | |
| "grad_norm": 0.7707180348154409, | |
| "learning_rate": 3.0824587706146934e-05, | |
| "loss": 0.3444, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33862438797950745, | |
| "step": 515, | |
| "valid_targets_mean": 2648.4, | |
| "valid_targets_min": 850 | |
| }, | |
| { | |
| "epoch": 0.5462184873949579, | |
| "grad_norm": 0.7331236147987067, | |
| "learning_rate": 3.112443778110945e-05, | |
| "loss": 0.3572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3474993109703064, | |
| "step": 520, | |
| "valid_targets_mean": 2870.6, | |
| "valid_targets_min": 843 | |
| }, | |
| { | |
| "epoch": 0.5514705882352942, | |
| "grad_norm": 0.9814672284053637, | |
| "learning_rate": 3.142428785607197e-05, | |
| "loss": 0.3307, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.334958016872406, | |
| "step": 525, | |
| "valid_targets_mean": 2415.3, | |
| "valid_targets_min": 904 | |
| }, | |
| { | |
| "epoch": 0.5567226890756303, | |
| "grad_norm": 0.7379782030619627, | |
| "learning_rate": 3.172413793103448e-05, | |
| "loss": 0.3268, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29642948508262634, | |
| "step": 530, | |
| "valid_targets_mean": 2672.1, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 0.5619747899159664, | |
| "grad_norm": 0.7091140893245311, | |
| "learning_rate": 3.2023988005997004e-05, | |
| "loss": 0.3153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28314292430877686, | |
| "step": 535, | |
| "valid_targets_mean": 2483.3, | |
| "valid_targets_min": 845 | |
| }, | |
| { | |
| "epoch": 0.5672268907563025, | |
| "grad_norm": 0.7575431351825522, | |
| "learning_rate": 3.2323838080959525e-05, | |
| "loss": 0.324, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33548930287361145, | |
| "step": 540, | |
| "valid_targets_mean": 2529.8, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 0.5724789915966386, | |
| "grad_norm": 0.7600622983828266, | |
| "learning_rate": 3.262368815592204e-05, | |
| "loss": 0.3395, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3454827666282654, | |
| "step": 545, | |
| "valid_targets_mean": 2808.9, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 0.5777310924369747, | |
| "grad_norm": 0.8183170811821894, | |
| "learning_rate": 3.292353823088456e-05, | |
| "loss": 0.3224, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35002923011779785, | |
| "step": 550, | |
| "valid_targets_mean": 2794.5, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 0.582983193277311, | |
| "grad_norm": 0.8057355550873148, | |
| "learning_rate": 3.3223388305847075e-05, | |
| "loss": 0.3333, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3431859612464905, | |
| "step": 555, | |
| "valid_targets_mean": 2285.7, | |
| "valid_targets_min": 557 | |
| }, | |
| { | |
| "epoch": 0.5882352941176471, | |
| "grad_norm": 0.6311718548224762, | |
| "learning_rate": 3.3523238380809596e-05, | |
| "loss": 0.3372, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3158794641494751, | |
| "step": 560, | |
| "valid_targets_mean": 3329.5, | |
| "valid_targets_min": 1114 | |
| }, | |
| { | |
| "epoch": 0.5934873949579832, | |
| "grad_norm": 0.7483509818624398, | |
| "learning_rate": 3.382308845577212e-05, | |
| "loss": 0.3101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.312336266040802, | |
| "step": 565, | |
| "valid_targets_mean": 2451.1, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 0.5987394957983193, | |
| "grad_norm": 0.8180747042153188, | |
| "learning_rate": 3.412293853073464e-05, | |
| "loss": 0.3288, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32389289140701294, | |
| "step": 570, | |
| "valid_targets_mean": 2601.9, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 0.6039915966386554, | |
| "grad_norm": 0.7501508192177901, | |
| "learning_rate": 3.442278860569715e-05, | |
| "loss": 0.3178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3347775936126709, | |
| "step": 575, | |
| "valid_targets_mean": 2365.9, | |
| "valid_targets_min": 795 | |
| }, | |
| { | |
| "epoch": 0.6092436974789915, | |
| "grad_norm": 0.8986962858297536, | |
| "learning_rate": 3.4722638680659673e-05, | |
| "loss": 0.3175, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2883639931678772, | |
| "step": 580, | |
| "valid_targets_mean": 2002.2, | |
| "valid_targets_min": 1167 | |
| }, | |
| { | |
| "epoch": 0.6144957983193278, | |
| "grad_norm": 0.7802444960510159, | |
| "learning_rate": 3.5022488755622194e-05, | |
| "loss": 0.3346, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32042810320854187, | |
| "step": 585, | |
| "valid_targets_mean": 2517.3, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 0.6197478991596639, | |
| "grad_norm": 0.8123611572368629, | |
| "learning_rate": 3.5322338830584716e-05, | |
| "loss": 0.3421, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3365861773490906, | |
| "step": 590, | |
| "valid_targets_mean": 2191.0, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 0.625, | |
| "grad_norm": 0.7007108414732764, | |
| "learning_rate": 3.562218890554723e-05, | |
| "loss": 0.3405, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3454342484474182, | |
| "step": 595, | |
| "valid_targets_mean": 2924.4, | |
| "valid_targets_min": 1165 | |
| }, | |
| { | |
| "epoch": 0.6302521008403361, | |
| "grad_norm": 0.730987978325978, | |
| "learning_rate": 3.592203898050975e-05, | |
| "loss": 0.3132, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29325562715530396, | |
| "step": 600, | |
| "valid_targets_mean": 2913.4, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 0.6355042016806722, | |
| "grad_norm": 0.7733724038568615, | |
| "learning_rate": 3.6221889055472265e-05, | |
| "loss": 0.3195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3013491928577423, | |
| "step": 605, | |
| "valid_targets_mean": 2295.1, | |
| "valid_targets_min": 850 | |
| }, | |
| { | |
| "epoch": 0.6407563025210085, | |
| "grad_norm": 0.8343650173831091, | |
| "learning_rate": 3.6521739130434786e-05, | |
| "loss": 0.3194, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29321059584617615, | |
| "step": 610, | |
| "valid_targets_mean": 1838.8, | |
| "valid_targets_min": 1024 | |
| }, | |
| { | |
| "epoch": 0.6460084033613446, | |
| "grad_norm": 0.8241043537661473, | |
| "learning_rate": 3.68215892053973e-05, | |
| "loss": 0.3207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32989972829818726, | |
| "step": 615, | |
| "valid_targets_mean": 2260.1, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 0.6512605042016807, | |
| "grad_norm": 0.8204056966498567, | |
| "learning_rate": 3.712143928035982e-05, | |
| "loss": 0.3318, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32474619150161743, | |
| "step": 620, | |
| "valid_targets_mean": 2229.6, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 0.6565126050420168, | |
| "grad_norm": 0.7119267457996982, | |
| "learning_rate": 3.742128935532234e-05, | |
| "loss": 0.3374, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3413497507572174, | |
| "step": 625, | |
| "valid_targets_mean": 2606.0, | |
| "valid_targets_min": 886 | |
| }, | |
| { | |
| "epoch": 0.6617647058823529, | |
| "grad_norm": 0.8282736254561877, | |
| "learning_rate": 3.772113943028486e-05, | |
| "loss": 0.3145, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30755430459976196, | |
| "step": 630, | |
| "valid_targets_mean": 1902.6, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 0.667016806722689, | |
| "grad_norm": 0.7523816591845006, | |
| "learning_rate": 3.802098950524738e-05, | |
| "loss": 0.3373, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34018948674201965, | |
| "step": 635, | |
| "valid_targets_mean": 2412.2, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 0.6722689075630253, | |
| "grad_norm": 0.762541870874189, | |
| "learning_rate": 3.83208395802099e-05, | |
| "loss": 0.3223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3180159330368042, | |
| "step": 640, | |
| "valid_targets_mean": 2287.2, | |
| "valid_targets_min": 994 | |
| }, | |
| { | |
| "epoch": 0.6775210084033614, | |
| "grad_norm": 1.0029207465299363, | |
| "learning_rate": 3.862068965517242e-05, | |
| "loss": 0.3572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33629217743873596, | |
| "step": 645, | |
| "valid_targets_mean": 2997.9, | |
| "valid_targets_min": 1170 | |
| }, | |
| { | |
| "epoch": 0.6827731092436975, | |
| "grad_norm": 0.7213964566380182, | |
| "learning_rate": 3.8920539730134934e-05, | |
| "loss": 0.3128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2845936417579651, | |
| "step": 650, | |
| "valid_targets_mean": 2431.8, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 0.6880252100840336, | |
| "grad_norm": 0.8519735795162455, | |
| "learning_rate": 3.9220389805097455e-05, | |
| "loss": 0.3433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3487856984138489, | |
| "step": 655, | |
| "valid_targets_mean": 2218.6, | |
| "valid_targets_min": 954 | |
| }, | |
| { | |
| "epoch": 0.6932773109243697, | |
| "grad_norm": 0.7046220956008892, | |
| "learning_rate": 3.9520239880059976e-05, | |
| "loss": 0.3199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3212737441062927, | |
| "step": 660, | |
| "valid_targets_mean": 2970.2, | |
| "valid_targets_min": 1285 | |
| }, | |
| { | |
| "epoch": 0.6985294117647058, | |
| "grad_norm": 0.7336185220602566, | |
| "learning_rate": 3.982008995502249e-05, | |
| "loss": 0.3157, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3486784100532532, | |
| "step": 665, | |
| "valid_targets_mean": 2758.5, | |
| "valid_targets_min": 924 | |
| }, | |
| { | |
| "epoch": 0.7037815126050421, | |
| "grad_norm": 0.8287435711662308, | |
| "learning_rate": 3.9999989022799437e-05, | |
| "loss": 0.3279, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3096350431442261, | |
| "step": 670, | |
| "valid_targets_mean": 1968.1, | |
| "valid_targets_min": 890 | |
| }, | |
| { | |
| "epoch": 0.7090336134453782, | |
| "grad_norm": 0.8202847564434639, | |
| "learning_rate": 3.9999865529431466e-05, | |
| "loss": 0.3302, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3207584023475647, | |
| "step": 675, | |
| "valid_targets_mean": 2397.3, | |
| "valid_targets_min": 792 | |
| }, | |
| { | |
| "epoch": 0.7142857142857143, | |
| "grad_norm": 0.6219543558575774, | |
| "learning_rate": 3.9999604822044886e-05, | |
| "loss": 0.3056, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2984098196029663, | |
| "step": 680, | |
| "valid_targets_mean": 3388.2, | |
| "valid_targets_min": 1117 | |
| }, | |
| { | |
| "epoch": 0.7195378151260504, | |
| "grad_norm": 0.7173660798726208, | |
| "learning_rate": 3.999920690242835e-05, | |
| "loss": 0.304, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29944783449172974, | |
| "step": 685, | |
| "valid_targets_mean": 2623.8, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 0.7247899159663865, | |
| "grad_norm": 0.6960902757404932, | |
| "learning_rate": 3.999867177331189e-05, | |
| "loss": 0.3217, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3305845856666565, | |
| "step": 690, | |
| "valid_targets_mean": 3031.8, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 0.7300420168067226, | |
| "grad_norm": 0.9054223785892286, | |
| "learning_rate": 3.9997999438366895e-05, | |
| "loss": 0.3396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33356064558029175, | |
| "step": 695, | |
| "valid_targets_mean": 2308.3, | |
| "valid_targets_min": 989 | |
| }, | |
| { | |
| "epoch": 0.7352941176470589, | |
| "grad_norm": 0.7885273248848218, | |
| "learning_rate": 3.9997189902206065e-05, | |
| "loss": 0.3376, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35361307859420776, | |
| "step": 700, | |
| "valid_targets_mean": 2145.7, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 0.740546218487395, | |
| "grad_norm": 0.7007173394110273, | |
| "learning_rate": 3.999624317038344e-05, | |
| "loss": 0.3054, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3023601770401001, | |
| "step": 705, | |
| "valid_targets_mean": 3050.5, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 0.7457983193277311, | |
| "grad_norm": 0.6606567930836675, | |
| "learning_rate": 3.9995159249394303e-05, | |
| "loss": 0.3167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34521594643592834, | |
| "step": 710, | |
| "valid_targets_mean": 2787.5, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 0.7510504201680672, | |
| "grad_norm": 0.7674014202376552, | |
| "learning_rate": 3.999393814667517e-05, | |
| "loss": 0.3251, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3120829164981842, | |
| "step": 715, | |
| "valid_targets_mean": 2107.4, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 0.7563025210084033, | |
| "grad_norm": 0.778065171637885, | |
| "learning_rate": 3.9992579870603695e-05, | |
| "loss": 0.3154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3578265905380249, | |
| "step": 720, | |
| "valid_targets_mean": 2320.1, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 0.7615546218487395, | |
| "grad_norm": 0.6597642723911712, | |
| "learning_rate": 3.999108443049869e-05, | |
| "loss": 0.3072, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3022955656051636, | |
| "step": 725, | |
| "valid_targets_mean": 2848.9, | |
| "valid_targets_min": 1035 | |
| }, | |
| { | |
| "epoch": 0.7668067226890757, | |
| "grad_norm": 0.7875093632298531, | |
| "learning_rate": 3.998945183661997e-05, | |
| "loss": 0.3373, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2969299256801605, | |
| "step": 730, | |
| "valid_targets_mean": 2284.4, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 0.7720588235294118, | |
| "grad_norm": 0.718806100635945, | |
| "learning_rate": 3.9987682100168376e-05, | |
| "loss": 0.314, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30758941173553467, | |
| "step": 735, | |
| "valid_targets_mean": 3248.9, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 0.7773109243697479, | |
| "grad_norm": 0.6716380195670248, | |
| "learning_rate": 3.99857752332856e-05, | |
| "loss": 0.3062, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29802027344703674, | |
| "step": 740, | |
| "valid_targets_mean": 2760.8, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 0.782563025210084, | |
| "grad_norm": 0.7048509390731609, | |
| "learning_rate": 3.998373124905418e-05, | |
| "loss": 0.3266, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.346552312374115, | |
| "step": 745, | |
| "valid_targets_mean": 2789.3, | |
| "valid_targets_min": 778 | |
| }, | |
| { | |
| "epoch": 0.7878151260504201, | |
| "grad_norm": 0.6889434405961281, | |
| "learning_rate": 3.9981550161497375e-05, | |
| "loss": 0.3285, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30958837270736694, | |
| "step": 750, | |
| "valid_targets_mean": 2730.2, | |
| "valid_targets_min": 815 | |
| }, | |
| { | |
| "epoch": 0.7930672268907563, | |
| "grad_norm": 0.830218274291599, | |
| "learning_rate": 3.9979231985579074e-05, | |
| "loss": 0.3098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29535311460494995, | |
| "step": 755, | |
| "valid_targets_mean": 2651.5, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 0.7983193277310925, | |
| "grad_norm": 0.869991526672995, | |
| "learning_rate": 3.99767767372037e-05, | |
| "loss": 0.3155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2807968854904175, | |
| "step": 760, | |
| "valid_targets_mean": 1711.4, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 0.8035714285714286, | |
| "grad_norm": 0.7081493643105407, | |
| "learning_rate": 3.997418443321609e-05, | |
| "loss": 0.3097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32702767848968506, | |
| "step": 765, | |
| "valid_targets_mean": 2585.1, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 0.8088235294117647, | |
| "grad_norm": 0.8332144275405594, | |
| "learning_rate": 3.997145509140138e-05, | |
| "loss": 0.3095, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31555208563804626, | |
| "step": 770, | |
| "valid_targets_mean": 2041.9, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 0.8140756302521008, | |
| "grad_norm": 0.6009347239627268, | |
| "learning_rate": 3.9968588730484896e-05, | |
| "loss": 0.3284, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34607839584350586, | |
| "step": 775, | |
| "valid_targets_mean": 3614.8, | |
| "valid_targets_min": 1437 | |
| }, | |
| { | |
| "epoch": 0.819327731092437, | |
| "grad_norm": 0.7137919217562437, | |
| "learning_rate": 3.9965585370132013e-05, | |
| "loss": 0.314, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30686646699905396, | |
| "step": 780, | |
| "valid_targets_mean": 2493.3, | |
| "valid_targets_min": 893 | |
| }, | |
| { | |
| "epoch": 0.8245798319327731, | |
| "grad_norm": 0.921127052828424, | |
| "learning_rate": 3.996244503094804e-05, | |
| "loss": 0.3143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29819056391716003, | |
| "step": 785, | |
| "valid_targets_mean": 1954.9, | |
| "valid_targets_min": 880 | |
| }, | |
| { | |
| "epoch": 0.8298319327731093, | |
| "grad_norm": 0.7650761079657684, | |
| "learning_rate": 3.995916773447804e-05, | |
| "loss": 0.3196, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2871814966201782, | |
| "step": 790, | |
| "valid_targets_mean": 2216.4, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 0.8350840336134454, | |
| "grad_norm": 0.7916639791964368, | |
| "learning_rate": 3.995575350320671e-05, | |
| "loss": 0.3237, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3441137373447418, | |
| "step": 795, | |
| "valid_targets_mean": 2616.8, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 0.8403361344537815, | |
| "grad_norm": 0.6287517180487977, | |
| "learning_rate": 3.995220236055824e-05, | |
| "loss": 0.3135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31627726554870605, | |
| "step": 800, | |
| "valid_targets_mean": 3041.9, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 0.8455882352941176, | |
| "grad_norm": 0.6231303563829806, | |
| "learning_rate": 3.99485143308961e-05, | |
| "loss": 0.3099, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3036961853504181, | |
| "step": 805, | |
| "valid_targets_mean": 3131.9, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 0.8508403361344538, | |
| "grad_norm": 0.7183070177297722, | |
| "learning_rate": 3.994468943952296e-05, | |
| "loss": 0.3272, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32938453555107117, | |
| "step": 810, | |
| "valid_targets_mean": 2472.3, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 0.8560924369747899, | |
| "grad_norm": 0.7681753001830797, | |
| "learning_rate": 3.994072771268041e-05, | |
| "loss": 0.3195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3146156072616577, | |
| "step": 815, | |
| "valid_targets_mean": 2485.6, | |
| "valid_targets_min": 903 | |
| }, | |
| { | |
| "epoch": 0.8613445378151261, | |
| "grad_norm": 0.6508689610348719, | |
| "learning_rate": 3.993662917754885e-05, | |
| "loss": 0.3127, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32557421922683716, | |
| "step": 820, | |
| "valid_targets_mean": 2845.4, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 0.8665966386554622, | |
| "grad_norm": 0.7108824995997928, | |
| "learning_rate": 3.993239386224732e-05, | |
| "loss": 0.3082, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27090033888816833, | |
| "step": 825, | |
| "valid_targets_mean": 2473.5, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 0.8718487394957983, | |
| "grad_norm": 0.7733012632993366, | |
| "learning_rate": 3.992802179583322e-05, | |
| "loss": 0.3199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32703638076782227, | |
| "step": 830, | |
| "valid_targets_mean": 2476.1, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 0.8771008403361344, | |
| "grad_norm": 0.725543154459089, | |
| "learning_rate": 3.9923513008302204e-05, | |
| "loss": 0.3169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.34420400857925415, | |
| "step": 835, | |
| "valid_targets_mean": 2456.4, | |
| "valid_targets_min": 1279 | |
| }, | |
| { | |
| "epoch": 0.8823529411764706, | |
| "grad_norm": 0.6577502601354023, | |
| "learning_rate": 3.991886753058792e-05, | |
| "loss": 0.3014, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27701324224472046, | |
| "step": 840, | |
| "valid_targets_mean": 2375.6, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 0.8876050420168067, | |
| "grad_norm": 0.6971018819451299, | |
| "learning_rate": 3.991408539456182e-05, | |
| "loss": 0.3292, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29996004700660706, | |
| "step": 845, | |
| "valid_targets_mean": 2597.3, | |
| "valid_targets_min": 1058 | |
| }, | |
| { | |
| "epoch": 0.8928571428571429, | |
| "grad_norm": 0.6864026260317219, | |
| "learning_rate": 3.990916663303293e-05, | |
| "loss": 0.3211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3099818229675293, | |
| "step": 850, | |
| "valid_targets_mean": 2555.9, | |
| "valid_targets_min": 1232 | |
| }, | |
| { | |
| "epoch": 0.898109243697479, | |
| "grad_norm": 0.6612404353720289, | |
| "learning_rate": 3.990411127974762e-05, | |
| "loss": 0.2985, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32130175828933716, | |
| "step": 855, | |
| "valid_targets_mean": 3308.6, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 0.9033613445378151, | |
| "grad_norm": 0.8109760616029387, | |
| "learning_rate": 3.989891936938939e-05, | |
| "loss": 0.297, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30671462416648865, | |
| "step": 860, | |
| "valid_targets_mean": 1730.1, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 0.9086134453781513, | |
| "grad_norm": 0.7674061973927943, | |
| "learning_rate": 3.9893590937578634e-05, | |
| "loss": 0.3084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.35476604104042053, | |
| "step": 865, | |
| "valid_targets_mean": 1927.0, | |
| "valid_targets_min": 882 | |
| }, | |
| { | |
| "epoch": 0.9138655462184874, | |
| "grad_norm": 0.8917074057700859, | |
| "learning_rate": 3.9888126020872375e-05, | |
| "loss": 0.3193, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3417109251022339, | |
| "step": 870, | |
| "valid_targets_mean": 2367.9, | |
| "valid_targets_min": 956 | |
| }, | |
| { | |
| "epoch": 0.9191176470588235, | |
| "grad_norm": 0.7447344177878483, | |
| "learning_rate": 3.988252465676401e-05, | |
| "loss": 0.332, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3013457655906677, | |
| "step": 875, | |
| "valid_targets_mean": 2151.2, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 0.9243697478991597, | |
| "grad_norm": 0.6389313482178219, | |
| "learning_rate": 3.98767868836831e-05, | |
| "loss": 0.2987, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3207515478134155, | |
| "step": 880, | |
| "valid_targets_mean": 3044.4, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 0.9296218487394958, | |
| "grad_norm": 0.7460485228221652, | |
| "learning_rate": 3.987091274099504e-05, | |
| "loss": 0.3008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.298840194940567, | |
| "step": 885, | |
| "valid_targets_mean": 2523.6, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 0.9348739495798319, | |
| "grad_norm": 0.7123957413614789, | |
| "learning_rate": 3.986490226900084e-05, | |
| "loss": 0.2958, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3151017725467682, | |
| "step": 890, | |
| "valid_targets_mean": 2639.6, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 0.9401260504201681, | |
| "grad_norm": 0.8425889968517829, | |
| "learning_rate": 3.985875550893684e-05, | |
| "loss": 0.3007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3078573942184448, | |
| "step": 895, | |
| "valid_targets_mean": 2052.1, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 0.9453781512605042, | |
| "grad_norm": 0.6910638587235646, | |
| "learning_rate": 3.9852472502974386e-05, | |
| "loss": 0.2981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3093729019165039, | |
| "step": 900, | |
| "valid_targets_mean": 2373.8, | |
| "valid_targets_min": 919 | |
| }, | |
| { | |
| "epoch": 0.9506302521008403, | |
| "grad_norm": 0.67274771832946, | |
| "learning_rate": 3.984605329421961e-05, | |
| "loss": 0.3064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3389376997947693, | |
| "step": 905, | |
| "valid_targets_mean": 2981.9, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 0.9558823529411765, | |
| "grad_norm": 0.7287563442852749, | |
| "learning_rate": 3.983949792671307e-05, | |
| "loss": 0.2871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29178452491760254, | |
| "step": 910, | |
| "valid_targets_mean": 2206.9, | |
| "valid_targets_min": 953 | |
| }, | |
| { | |
| "epoch": 0.9611344537815126, | |
| "grad_norm": 0.6697336132699603, | |
| "learning_rate": 3.9832806445429486e-05, | |
| "loss": 0.3059, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2787833511829376, | |
| "step": 915, | |
| "valid_targets_mean": 2421.2, | |
| "valid_targets_min": 1033 | |
| }, | |
| { | |
| "epoch": 0.9663865546218487, | |
| "grad_norm": 0.7721239770788888, | |
| "learning_rate": 3.982597889627742e-05, | |
| "loss": 0.3084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3243314027786255, | |
| "step": 920, | |
| "valid_targets_mean": 2458.9, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 0.9716386554621849, | |
| "grad_norm": 0.6890378426610851, | |
| "learning_rate": 3.981901532609896e-05, | |
| "loss": 0.313, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3086700439453125, | |
| "step": 925, | |
| "valid_targets_mean": 2401.8, | |
| "valid_targets_min": 926 | |
| }, | |
| { | |
| "epoch": 0.976890756302521, | |
| "grad_norm": 0.7281226712845131, | |
| "learning_rate": 3.9811915782669406e-05, | |
| "loss": 0.3124, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27539515495300293, | |
| "step": 930, | |
| "valid_targets_mean": 2145.1, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 0.9821428571428571, | |
| "grad_norm": 0.7223280692952256, | |
| "learning_rate": 3.980468031469691e-05, | |
| "loss": 0.3144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2952006459236145, | |
| "step": 935, | |
| "valid_targets_mean": 2334.9, | |
| "valid_targets_min": 1055 | |
| }, | |
| { | |
| "epoch": 0.9873949579831933, | |
| "grad_norm": 0.7460709191349191, | |
| "learning_rate": 3.9797308971822206e-05, | |
| "loss": 0.3175, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3261488676071167, | |
| "step": 940, | |
| "valid_targets_mean": 2191.4, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 0.9926470588235294, | |
| "grad_norm": 0.6186791687377395, | |
| "learning_rate": 3.978980180461821e-05, | |
| "loss": 0.3292, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.33339810371398926, | |
| "step": 945, | |
| "valid_targets_mean": 3133.6, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 0.9978991596638656, | |
| "grad_norm": 0.8177731068661213, | |
| "learning_rate": 3.9782158864589696e-05, | |
| "loss": 0.3013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27547362446784973, | |
| "step": 950, | |
| "valid_targets_mean": 1856.8, | |
| "valid_targets_min": 736 | |
| }, | |
| { | |
| "epoch": 1.0031512605042017, | |
| "grad_norm": 0.7256412516546766, | |
| "learning_rate": 3.977438020417293e-05, | |
| "loss": 0.2844, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.305136114358902, | |
| "step": 955, | |
| "valid_targets_mean": 2359.5, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 1.0084033613445378, | |
| "grad_norm": 0.7379697373188653, | |
| "learning_rate": 3.9766465876735354e-05, | |
| "loss": 0.2853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2883484661579132, | |
| "step": 960, | |
| "valid_targets_mean": 2517.8, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 1.013655462184874, | |
| "grad_norm": 0.7450004669339628, | |
| "learning_rate": 3.9758415936575155e-05, | |
| "loss": 0.2694, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2624656856060028, | |
| "step": 965, | |
| "valid_targets_mean": 2335.6, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 1.01890756302521, | |
| "grad_norm": 0.8213378024423946, | |
| "learning_rate": 3.975023043892094e-05, | |
| "loss": 0.2998, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3582037091255188, | |
| "step": 970, | |
| "valid_targets_mean": 2665.7, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 1.0241596638655461, | |
| "grad_norm": 1.1203862991152767, | |
| "learning_rate": 3.974190943993133e-05, | |
| "loss": 0.2855, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2818828821182251, | |
| "step": 975, | |
| "valid_targets_mean": 1555.1, | |
| "valid_targets_min": 668 | |
| }, | |
| { | |
| "epoch": 1.0294117647058822, | |
| "grad_norm": 0.5968672236493667, | |
| "learning_rate": 3.973345299669461e-05, | |
| "loss": 0.2909, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24849054217338562, | |
| "step": 980, | |
| "valid_targets_mean": 3073.6, | |
| "valid_targets_min": 745 | |
| }, | |
| { | |
| "epoch": 1.0346638655462186, | |
| "grad_norm": 0.6412679780590781, | |
| "learning_rate": 3.972486116722828e-05, | |
| "loss": 0.2966, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.270571231842041, | |
| "step": 985, | |
| "valid_targets_mean": 2620.9, | |
| "valid_targets_min": 1058 | |
| }, | |
| { | |
| "epoch": 1.0399159663865547, | |
| "grad_norm": 0.658674789610642, | |
| "learning_rate": 3.971613401047872e-05, | |
| "loss": 0.2928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28596004843711853, | |
| "step": 990, | |
| "valid_targets_mean": 2736.1, | |
| "valid_targets_min": 696 | |
| }, | |
| { | |
| "epoch": 1.0451680672268908, | |
| "grad_norm": 0.708522171289682, | |
| "learning_rate": 3.970727158632075e-05, | |
| "loss": 0.2791, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25710612535476685, | |
| "step": 995, | |
| "valid_targets_mean": 2025.1, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 1.050420168067227, | |
| "grad_norm": 0.8163565196188108, | |
| "learning_rate": 3.969827395555721e-05, | |
| "loss": 0.2842, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3168991506099701, | |
| "step": 1000, | |
| "valid_targets_mean": 2106.7, | |
| "valid_targets_min": 901 | |
| }, | |
| { | |
| "epoch": 1.055672268907563, | |
| "grad_norm": 0.7814582268431446, | |
| "learning_rate": 3.968914117991857e-05, | |
| "loss": 0.2953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.269026517868042, | |
| "step": 1005, | |
| "valid_targets_mean": 2927.3, | |
| "valid_targets_min": 936 | |
| }, | |
| { | |
| "epoch": 1.0609243697478992, | |
| "grad_norm": 0.7483470353521008, | |
| "learning_rate": 3.967987332206249e-05, | |
| "loss": 0.2864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2755717635154724, | |
| "step": 1010, | |
| "valid_targets_mean": 2053.3, | |
| "valid_targets_min": 857 | |
| }, | |
| { | |
| "epoch": 1.0661764705882353, | |
| "grad_norm": 0.6950043296063686, | |
| "learning_rate": 3.967047044557341e-05, | |
| "loss": 0.2927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2624019682407379, | |
| "step": 1015, | |
| "valid_targets_mean": 2560.6, | |
| "valid_targets_min": 918 | |
| }, | |
| { | |
| "epoch": 1.0714285714285714, | |
| "grad_norm": 0.7326135503168667, | |
| "learning_rate": 3.966093261496208e-05, | |
| "loss": 0.2827, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2939094305038452, | |
| "step": 1020, | |
| "valid_targets_mean": 2364.4, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 1.0766806722689075, | |
| "grad_norm": 0.7595424312977608, | |
| "learning_rate": 3.9651259895665146e-05, | |
| "loss": 0.303, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32039862871170044, | |
| "step": 1025, | |
| "valid_targets_mean": 2230.6, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 1.0819327731092436, | |
| "grad_norm": 0.745092373926396, | |
| "learning_rate": 3.9641452354044687e-05, | |
| "loss": 0.2933, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30866172909736633, | |
| "step": 1030, | |
| "valid_targets_mean": 2336.8, | |
| "valid_targets_min": 1140 | |
| }, | |
| { | |
| "epoch": 1.0871848739495797, | |
| "grad_norm": 0.7617543273247012, | |
| "learning_rate": 3.9631510057387765e-05, | |
| "loss": 0.2765, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2596009075641632, | |
| "step": 1035, | |
| "valid_targets_mean": 2062.8, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 1.092436974789916, | |
| "grad_norm": 0.8286121150834354, | |
| "learning_rate": 3.962143307390596e-05, | |
| "loss": 0.2798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2832181453704834, | |
| "step": 1040, | |
| "valid_targets_mean": 2286.1, | |
| "valid_targets_min": 1170 | |
| }, | |
| { | |
| "epoch": 1.0976890756302522, | |
| "grad_norm": 0.8421971274176083, | |
| "learning_rate": 3.961122147273491e-05, | |
| "loss": 0.28, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31647855043411255, | |
| "step": 1045, | |
| "valid_targets_mean": 2367.7, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 1.1029411764705883, | |
| "grad_norm": 0.7098338434234529, | |
| "learning_rate": 3.9600875323933826e-05, | |
| "loss": 0.2787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30434438586235046, | |
| "step": 1050, | |
| "valid_targets_mean": 2628.7, | |
| "valid_targets_min": 1227 | |
| }, | |
| { | |
| "epoch": 1.1081932773109244, | |
| "grad_norm": 0.6866219477953722, | |
| "learning_rate": 3.959039469848502e-05, | |
| "loss": 0.2878, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29877203702926636, | |
| "step": 1055, | |
| "valid_targets_mean": 2690.5, | |
| "valid_targets_min": 666 | |
| }, | |
| { | |
| "epoch": 1.1134453781512605, | |
| "grad_norm": 0.6372603497209481, | |
| "learning_rate": 3.9579779668293416e-05, | |
| "loss": 0.3004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28999483585357666, | |
| "step": 1060, | |
| "valid_targets_mean": 2948.1, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 1.1186974789915967, | |
| "grad_norm": 0.6799822940846726, | |
| "learning_rate": 3.956903030618605e-05, | |
| "loss": 0.2797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27467814087867737, | |
| "step": 1065, | |
| "valid_targets_mean": 2615.4, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 1.1239495798319328, | |
| "grad_norm": 0.7809972116466566, | |
| "learning_rate": 3.955814668591156e-05, | |
| "loss": 0.2837, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27500802278518677, | |
| "step": 1070, | |
| "valid_targets_mean": 2294.4, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 1.129201680672269, | |
| "grad_norm": 0.7524552269972444, | |
| "learning_rate": 3.9547128882139735e-05, | |
| "loss": 0.274, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2802833616733551, | |
| "step": 1075, | |
| "valid_targets_mean": 2288.1, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 1.134453781512605, | |
| "grad_norm": 0.6658052832741743, | |
| "learning_rate": 3.9535976970460925e-05, | |
| "loss": 0.3048, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3079906702041626, | |
| "step": 1080, | |
| "valid_targets_mean": 3495.8, | |
| "valid_targets_min": 1241 | |
| }, | |
| { | |
| "epoch": 1.1397058823529411, | |
| "grad_norm": 0.7347104088170611, | |
| "learning_rate": 3.9524691027385585e-05, | |
| "loss": 0.2912, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31090694665908813, | |
| "step": 1085, | |
| "valid_targets_mean": 2511.8, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 1.1449579831932772, | |
| "grad_norm": 0.6724112220847397, | |
| "learning_rate": 3.95132711303437e-05, | |
| "loss": 0.2997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30762532353401184, | |
| "step": 1090, | |
| "valid_targets_mean": 2850.2, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 1.1502100840336134, | |
| "grad_norm": 0.8334744080367508, | |
| "learning_rate": 3.9501717357684315e-05, | |
| "loss": 0.2775, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28743976354599, | |
| "step": 1095, | |
| "valid_targets_mean": 2099.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 1.1554621848739495, | |
| "grad_norm": 0.7032590588711007, | |
| "learning_rate": 3.9490029788674934e-05, | |
| "loss": 0.2917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2826242446899414, | |
| "step": 1100, | |
| "valid_targets_mean": 2369.9, | |
| "valid_targets_min": 1013 | |
| }, | |
| { | |
| "epoch": 1.1607142857142858, | |
| "grad_norm": 1.221702121249645, | |
| "learning_rate": 3.9478208503501e-05, | |
| "loss": 0.288, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2712562680244446, | |
| "step": 1105, | |
| "valid_targets_mean": 2393.2, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 1.165966386554622, | |
| "grad_norm": 0.7494443041520203, | |
| "learning_rate": 3.946625358326538e-05, | |
| "loss": 0.266, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2519804835319519, | |
| "step": 1110, | |
| "valid_targets_mean": 1945.5, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 1.171218487394958, | |
| "grad_norm": 0.5997164956764384, | |
| "learning_rate": 3.945416510998775e-05, | |
| "loss": 0.2959, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2693081498146057, | |
| "step": 1115, | |
| "valid_targets_mean": 3155.6, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 1.1764705882352942, | |
| "grad_norm": 0.6293715830721176, | |
| "learning_rate": 3.944194316660406e-05, | |
| "loss": 0.2876, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2860058546066284, | |
| "step": 1120, | |
| "valid_targets_mean": 2777.3, | |
| "valid_targets_min": 955 | |
| }, | |
| { | |
| "epoch": 1.1817226890756303, | |
| "grad_norm": 0.7097668355853033, | |
| "learning_rate": 3.942958783696598e-05, | |
| "loss": 0.2818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2959762215614319, | |
| "step": 1125, | |
| "valid_targets_mean": 2465.1, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 1.1869747899159664, | |
| "grad_norm": 0.6825216992106221, | |
| "learning_rate": 3.94170992058403e-05, | |
| "loss": 0.2773, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27206242084503174, | |
| "step": 1130, | |
| "valid_targets_mean": 2581.8, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 1.1922268907563025, | |
| "grad_norm": 0.6192081356673204, | |
| "learning_rate": 3.9404477358908354e-05, | |
| "loss": 0.264, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25278788805007935, | |
| "step": 1135, | |
| "valid_targets_mean": 2664.7, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 1.1974789915966386, | |
| "grad_norm": 0.811292426338912, | |
| "learning_rate": 3.9391722382765445e-05, | |
| "loss": 0.2948, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29716646671295166, | |
| "step": 1140, | |
| "valid_targets_mean": 1980.2, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 1.2027310924369747, | |
| "grad_norm": 0.7668088242832095, | |
| "learning_rate": 3.937883436492025e-05, | |
| "loss": 0.2877, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27448099851608276, | |
| "step": 1145, | |
| "valid_targets_mean": 1986.5, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 1.2079831932773109, | |
| "grad_norm": 0.7413211114263699, | |
| "learning_rate": 3.9365813393794186e-05, | |
| "loss": 0.2983, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2899211049079895, | |
| "step": 1150, | |
| "valid_targets_mean": 2858.4, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 1.213235294117647, | |
| "grad_norm": 0.7546705832538436, | |
| "learning_rate": 3.9352659558720836e-05, | |
| "loss": 0.2826, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2793716490268707, | |
| "step": 1155, | |
| "valid_targets_mean": 2138.5, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 1.2184873949579833, | |
| "grad_norm": 0.782588277733481, | |
| "learning_rate": 3.933937294994535e-05, | |
| "loss": 0.2912, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2977146804332733, | |
| "step": 1160, | |
| "valid_targets_mean": 2275.8, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 1.2237394957983194, | |
| "grad_norm": 0.8153448615714938, | |
| "learning_rate": 3.932595365862379e-05, | |
| "loss": 0.2974, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3228553533554077, | |
| "step": 1165, | |
| "valid_targets_mean": 2634.8, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 1.2289915966386555, | |
| "grad_norm": 0.792990263276749, | |
| "learning_rate": 3.9312401776822504e-05, | |
| "loss": 0.2789, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26295557618141174, | |
| "step": 1170, | |
| "valid_targets_mean": 1794.0, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 1.2342436974789917, | |
| "grad_norm": 0.6797271279062665, | |
| "learning_rate": 3.9298717397517546e-05, | |
| "loss": 0.2678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27940213680267334, | |
| "step": 1175, | |
| "valid_targets_mean": 2708.9, | |
| "valid_targets_min": 1078 | |
| }, | |
| { | |
| "epoch": 1.2394957983193278, | |
| "grad_norm": 0.6849285603022243, | |
| "learning_rate": 3.928490061459396e-05, | |
| "loss": 0.2776, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2747243046760559, | |
| "step": 1180, | |
| "valid_targets_mean": 2251.1, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 1.2447478991596639, | |
| "grad_norm": 0.7245761997254816, | |
| "learning_rate": 3.927095152284521e-05, | |
| "loss": 0.2924, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3488772511482239, | |
| "step": 1185, | |
| "valid_targets_mean": 3074.4, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "grad_norm": 0.7010066220238207, | |
| "learning_rate": 3.925687021797249e-05, | |
| "loss": 0.2834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29797542095184326, | |
| "step": 1190, | |
| "valid_targets_mean": 2721.6, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 1.2552521008403361, | |
| "grad_norm": 1.0179359955088225, | |
| "learning_rate": 3.924265679658407e-05, | |
| "loss": 0.285, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27947139739990234, | |
| "step": 1195, | |
| "valid_targets_mean": 1642.1, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 1.2605042016806722, | |
| "grad_norm": 0.7149247912396931, | |
| "learning_rate": 3.922831135619462e-05, | |
| "loss": 0.2831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2821911573410034, | |
| "step": 1200, | |
| "valid_targets_mean": 2422.2, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 1.2657563025210083, | |
| "grad_norm": 0.7243464387956448, | |
| "learning_rate": 3.9213833995224605e-05, | |
| "loss": 0.2818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2676169276237488, | |
| "step": 1205, | |
| "valid_targets_mean": 2188.3, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 1.2710084033613445, | |
| "grad_norm": 0.7370679575877426, | |
| "learning_rate": 3.919922481299952e-05, | |
| "loss": 0.3017, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32717978954315186, | |
| "step": 1210, | |
| "valid_targets_mean": 2311.9, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 1.2762605042016806, | |
| "grad_norm": 0.6894289625236196, | |
| "learning_rate": 3.918448390974928e-05, | |
| "loss": 0.2759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26075074076652527, | |
| "step": 1215, | |
| "valid_targets_mean": 2090.9, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 1.2815126050420167, | |
| "grad_norm": 0.8026290889025559, | |
| "learning_rate": 3.9169611386607476e-05, | |
| "loss": 0.2928, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30541282892227173, | |
| "step": 1220, | |
| "valid_targets_mean": 2121.8, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 1.2867647058823528, | |
| "grad_norm": 0.7771870350210088, | |
| "learning_rate": 3.9154607345610746e-05, | |
| "loss": 0.273, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2856689691543579, | |
| "step": 1225, | |
| "valid_targets_mean": 2416.0, | |
| "valid_targets_min": 1196 | |
| }, | |
| { | |
| "epoch": 1.2920168067226891, | |
| "grad_norm": 0.673563167861582, | |
| "learning_rate": 3.913947188969801e-05, | |
| "loss": 0.3009, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2647175192832947, | |
| "step": 1230, | |
| "valid_targets_mean": 2471.2, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 1.2972689075630253, | |
| "grad_norm": 0.6531080612213936, | |
| "learning_rate": 3.912420512270981e-05, | |
| "loss": 0.2766, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22836507856845856, | |
| "step": 1235, | |
| "valid_targets_mean": 2253.1, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 1.3025210084033614, | |
| "grad_norm": 0.7035585449641938, | |
| "learning_rate": 3.910880714938757e-05, | |
| "loss": 0.2851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3099633455276489, | |
| "step": 1240, | |
| "valid_targets_mean": 2499.9, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 1.3077731092436975, | |
| "grad_norm": 0.5567277712502462, | |
| "learning_rate": 3.909327807537288e-05, | |
| "loss": 0.2897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28143376111984253, | |
| "step": 1245, | |
| "valid_targets_mean": 3603.4, | |
| "valid_targets_min": 1081 | |
| }, | |
| { | |
| "epoch": 1.3130252100840336, | |
| "grad_norm": 0.7110770433202822, | |
| "learning_rate": 3.90776180072068e-05, | |
| "loss": 0.2744, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2797868251800537, | |
| "step": 1250, | |
| "valid_targets_mean": 2393.1, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 1.3182773109243697, | |
| "grad_norm": 0.5586358952592809, | |
| "learning_rate": 3.906182705232909e-05, | |
| "loss": 0.2778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2782168388366699, | |
| "step": 1255, | |
| "valid_targets_mean": 4218.8, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 1.3235294117647058, | |
| "grad_norm": 0.7634189610982803, | |
| "learning_rate": 3.904590531907751e-05, | |
| "loss": 0.286, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30689549446105957, | |
| "step": 1260, | |
| "valid_targets_mean": 2308.6, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 1.328781512605042, | |
| "grad_norm": 0.660398552740273, | |
| "learning_rate": 3.9029852916687034e-05, | |
| "loss": 0.2895, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29954206943511963, | |
| "step": 1265, | |
| "valid_targets_mean": 2645.9, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 1.334033613445378, | |
| "grad_norm": 0.674497976039509, | |
| "learning_rate": 3.9013669955289136e-05, | |
| "loss": 0.2945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2616213858127594, | |
| "step": 1270, | |
| "valid_targets_mean": 2590.4, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 1.3392857142857144, | |
| "grad_norm": 0.6745461538720665, | |
| "learning_rate": 3.899735654591101e-05, | |
| "loss": 0.2724, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2934377193450928, | |
| "step": 1275, | |
| "valid_targets_mean": 2654.6, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 1.3445378151260505, | |
| "grad_norm": 0.6298474437654481, | |
| "learning_rate": 3.898091280047486e-05, | |
| "loss": 0.2907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26556211709976196, | |
| "step": 1280, | |
| "valid_targets_mean": 3061.3, | |
| "valid_targets_min": 905 | |
| }, | |
| { | |
| "epoch": 1.3497899159663866, | |
| "grad_norm": 0.6789724693265738, | |
| "learning_rate": 3.896433883179703e-05, | |
| "loss": 0.2797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24146893620491028, | |
| "step": 1285, | |
| "valid_targets_mean": 2094.8, | |
| "valid_targets_min": 1109 | |
| }, | |
| { | |
| "epoch": 1.3550420168067228, | |
| "grad_norm": 0.7300952923000248, | |
| "learning_rate": 3.894763475358736e-05, | |
| "loss": 0.2687, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2770267724990845, | |
| "step": 1290, | |
| "valid_targets_mean": 2336.6, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 1.3602941176470589, | |
| "grad_norm": 0.8086297463249519, | |
| "learning_rate": 3.8930800680448275e-05, | |
| "loss": 0.2713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26473838090896606, | |
| "step": 1295, | |
| "valid_targets_mean": 1924.0, | |
| "valid_targets_min": 1099 | |
| }, | |
| { | |
| "epoch": 1.365546218487395, | |
| "grad_norm": 0.6977529571217691, | |
| "learning_rate": 3.891383672787411e-05, | |
| "loss": 0.2761, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2646206021308899, | |
| "step": 1300, | |
| "valid_targets_mean": 2431.1, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 1.370798319327731, | |
| "grad_norm": 0.7033155019588632, | |
| "learning_rate": 3.889674301225025e-05, | |
| "loss": 0.2746, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27550557255744934, | |
| "step": 1305, | |
| "valid_targets_mean": 2446.4, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 1.3760504201680672, | |
| "grad_norm": 0.7011256224002429, | |
| "learning_rate": 3.8879519650852356e-05, | |
| "loss": 0.2798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29224735498428345, | |
| "step": 1310, | |
| "valid_targets_mean": 2337.5, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 1.3813025210084033, | |
| "grad_norm": 0.7551918929706497, | |
| "learning_rate": 3.886216676184555e-05, | |
| "loss": 0.2786, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25635597109794617, | |
| "step": 1315, | |
| "valid_targets_mean": 1767.4, | |
| "valid_targets_min": 903 | |
| }, | |
| { | |
| "epoch": 1.3865546218487395, | |
| "grad_norm": 0.841773127497859, | |
| "learning_rate": 3.8844684464283614e-05, | |
| "loss": 0.2729, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28463175892829895, | |
| "step": 1320, | |
| "valid_targets_mean": 2054.4, | |
| "valid_targets_min": 1058 | |
| }, | |
| { | |
| "epoch": 1.3918067226890756, | |
| "grad_norm": 0.714138459758691, | |
| "learning_rate": 3.882707287810817e-05, | |
| "loss": 0.2829, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3063795566558838, | |
| "step": 1325, | |
| "valid_targets_mean": 2356.1, | |
| "valid_targets_min": 980 | |
| }, | |
| { | |
| "epoch": 1.3970588235294117, | |
| "grad_norm": 0.7951782867715268, | |
| "learning_rate": 3.880933212414786e-05, | |
| "loss": 0.2791, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31758105754852295, | |
| "step": 1330, | |
| "valid_targets_mean": 2118.9, | |
| "valid_targets_min": 685 | |
| }, | |
| { | |
| "epoch": 1.4023109243697478, | |
| "grad_norm": 0.6803907534201502, | |
| "learning_rate": 3.87914623241175e-05, | |
| "loss": 0.2774, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2610444724559784, | |
| "step": 1335, | |
| "valid_targets_mean": 2595.9, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 1.407563025210084, | |
| "grad_norm": 0.7146731724902138, | |
| "learning_rate": 3.877346360061728e-05, | |
| "loss": 0.2903, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2991168200969696, | |
| "step": 1340, | |
| "valid_targets_mean": 2701.4, | |
| "valid_targets_min": 931 | |
| }, | |
| { | |
| "epoch": 1.41281512605042, | |
| "grad_norm": 0.7591356155611764, | |
| "learning_rate": 3.8755336077131894e-05, | |
| "loss": 0.2835, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2531967759132385, | |
| "step": 1345, | |
| "valid_targets_mean": 1822.8, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 1.4180672268907564, | |
| "grad_norm": 0.6930387851247973, | |
| "learning_rate": 3.873707987802967e-05, | |
| "loss": 0.2775, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2844356298446655, | |
| "step": 1350, | |
| "valid_targets_mean": 2694.1, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 1.4233193277310925, | |
| "grad_norm": 0.6030534452348743, | |
| "learning_rate": 3.871869512856179e-05, | |
| "loss": 0.2738, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2641228139400482, | |
| "step": 1355, | |
| "valid_targets_mean": 3138.3, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 1.4285714285714286, | |
| "grad_norm": 0.5730393066215059, | |
| "learning_rate": 3.870018195486138e-05, | |
| "loss": 0.2695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24910803139209747, | |
| "step": 1360, | |
| "valid_targets_mean": 3588.1, | |
| "valid_targets_min": 1099 | |
| }, | |
| { | |
| "epoch": 1.4338235294117647, | |
| "grad_norm": 0.7755395352301468, | |
| "learning_rate": 3.868154048394262e-05, | |
| "loss": 0.276, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2994786202907562, | |
| "step": 1365, | |
| "valid_targets_mean": 2700.9, | |
| "valid_targets_min": 1110 | |
| }, | |
| { | |
| "epoch": 1.4390756302521008, | |
| "grad_norm": 0.7005171055078728, | |
| "learning_rate": 3.8662770843699944e-05, | |
| "loss": 0.2938, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3116563558578491, | |
| "step": 1370, | |
| "valid_targets_mean": 2591.6, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 1.444327731092437, | |
| "grad_norm": 0.7295296113213054, | |
| "learning_rate": 3.8643873162907086e-05, | |
| "loss": 0.2858, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2686575651168823, | |
| "step": 1375, | |
| "valid_targets_mean": 2169.4, | |
| "valid_targets_min": 690 | |
| }, | |
| { | |
| "epoch": 1.449579831932773, | |
| "grad_norm": 0.728303365746196, | |
| "learning_rate": 3.862484757121627e-05, | |
| "loss": 0.3071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.343506395816803, | |
| "step": 1380, | |
| "valid_targets_mean": 2436.4, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 1.4548319327731092, | |
| "grad_norm": 0.7264219557451437, | |
| "learning_rate": 3.860569419915727e-05, | |
| "loss": 0.2952, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27744126319885254, | |
| "step": 1385, | |
| "valid_targets_mean": 2069.1, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 1.4600840336134453, | |
| "grad_norm": 0.7451210015033147, | |
| "learning_rate": 3.858641317813653e-05, | |
| "loss": 0.2592, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.264180064201355, | |
| "step": 1390, | |
| "valid_targets_mean": 2131.1, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 1.4653361344537816, | |
| "grad_norm": 0.7180216886659947, | |
| "learning_rate": 3.8567004640436234e-05, | |
| "loss": 0.2864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28325045108795166, | |
| "step": 1395, | |
| "valid_targets_mean": 2150.3, | |
| "valid_targets_min": 1239 | |
| }, | |
| { | |
| "epoch": 1.4705882352941178, | |
| "grad_norm": 0.6005140811506897, | |
| "learning_rate": 3.8547468719213476e-05, | |
| "loss": 0.291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2805612087249756, | |
| "step": 1400, | |
| "valid_targets_mean": 2986.0, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 1.4758403361344539, | |
| "grad_norm": 0.6701734829120335, | |
| "learning_rate": 3.852780554849925e-05, | |
| "loss": 0.2713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2722683250904083, | |
| "step": 1405, | |
| "valid_targets_mean": 2545.0, | |
| "valid_targets_min": 957 | |
| }, | |
| { | |
| "epoch": 1.48109243697479, | |
| "grad_norm": 0.7156588483612103, | |
| "learning_rate": 3.850801526319761e-05, | |
| "loss": 0.2917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3095453381538391, | |
| "step": 1410, | |
| "valid_targets_mean": 2583.2, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 1.486344537815126, | |
| "grad_norm": 0.6515593357765462, | |
| "learning_rate": 3.84880979990847e-05, | |
| "loss": 0.278, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2939472198486328, | |
| "step": 1415, | |
| "valid_targets_mean": 2874.5, | |
| "valid_targets_min": 1405 | |
| }, | |
| { | |
| "epoch": 1.4915966386554622, | |
| "grad_norm": 0.6808237094478136, | |
| "learning_rate": 3.846805389280782e-05, | |
| "loss": 0.2836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29528388381004333, | |
| "step": 1420, | |
| "valid_targets_mean": 2692.4, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 1.4968487394957983, | |
| "grad_norm": 0.6429218218943833, | |
| "learning_rate": 3.844788308188452e-05, | |
| "loss": 0.2749, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24127134680747986, | |
| "step": 1425, | |
| "valid_targets_mean": 2371.2, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 1.5021008403361344, | |
| "grad_norm": 0.7293161414772861, | |
| "learning_rate": 3.8427585704701634e-05, | |
| "loss": 0.289, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.32644128799438477, | |
| "step": 1430, | |
| "valid_targets_mean": 2395.9, | |
| "valid_targets_min": 885 | |
| }, | |
| { | |
| "epoch": 1.5073529411764706, | |
| "grad_norm": 0.6776284586447868, | |
| "learning_rate": 3.840716190051433e-05, | |
| "loss": 0.2978, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2834451198577881, | |
| "step": 1435, | |
| "valid_targets_mean": 2865.2, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 1.5126050420168067, | |
| "grad_norm": 0.7065935410256674, | |
| "learning_rate": 3.8386611809445155e-05, | |
| "loss": 0.2874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2767036557197571, | |
| "step": 1440, | |
| "valid_targets_mean": 2116.9, | |
| "valid_targets_min": 1347 | |
| }, | |
| { | |
| "epoch": 1.5178571428571428, | |
| "grad_norm": 0.6875344935274045, | |
| "learning_rate": 3.8365935572483095e-05, | |
| "loss": 0.2797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25539255142211914, | |
| "step": 1445, | |
| "valid_targets_mean": 2034.7, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 1.523109243697479, | |
| "grad_norm": 0.6114420882513082, | |
| "learning_rate": 3.834513333148258e-05, | |
| "loss": 0.2925, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3040091097354889, | |
| "step": 1450, | |
| "valid_targets_mean": 3394.8, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 1.528361344537815, | |
| "grad_norm": 0.7351299306831928, | |
| "learning_rate": 3.832420522916252e-05, | |
| "loss": 0.2726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29218441247940063, | |
| "step": 1455, | |
| "valid_targets_mean": 2289.6, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 1.5336134453781511, | |
| "grad_norm": 0.6742529951804918, | |
| "learning_rate": 3.830315140910534e-05, | |
| "loss": 0.2824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27924418449401855, | |
| "step": 1460, | |
| "valid_targets_mean": 2554.1, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 1.5388655462184873, | |
| "grad_norm": 0.642103371520459, | |
| "learning_rate": 3.8281972015755965e-05, | |
| "loss": 0.2912, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28389638662338257, | |
| "step": 1465, | |
| "valid_targets_mean": 2878.4, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 1.5441176470588234, | |
| "grad_norm": 0.6652065712153472, | |
| "learning_rate": 3.826066719442086e-05, | |
| "loss": 0.3033, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28304386138916016, | |
| "step": 1470, | |
| "valid_targets_mean": 3034.9, | |
| "valid_targets_min": 1390 | |
| }, | |
| { | |
| "epoch": 1.5493697478991597, | |
| "grad_norm": 0.6732297614547262, | |
| "learning_rate": 3.823923709126701e-05, | |
| "loss": 0.2739, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27302515506744385, | |
| "step": 1475, | |
| "valid_targets_mean": 2630.1, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 1.5546218487394958, | |
| "grad_norm": 0.7977790582505264, | |
| "learning_rate": 3.821768185332095e-05, | |
| "loss": 0.2752, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23856990039348602, | |
| "step": 1480, | |
| "valid_targets_mean": 1874.7, | |
| "valid_targets_min": 895 | |
| }, | |
| { | |
| "epoch": 1.559873949579832, | |
| "grad_norm": 0.6416684613994287, | |
| "learning_rate": 3.81960016284677e-05, | |
| "loss": 0.2789, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2834779918193817, | |
| "step": 1485, | |
| "valid_targets_mean": 2730.8, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 1.565126050420168, | |
| "grad_norm": 0.7215663679705727, | |
| "learning_rate": 3.817419656544979e-05, | |
| "loss": 0.2713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28470200300216675, | |
| "step": 1490, | |
| "valid_targets_mean": 2297.9, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 1.5703781512605042, | |
| "grad_norm": 0.7210848118654766, | |
| "learning_rate": 3.815226681386626e-05, | |
| "loss": 0.3029, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30118465423583984, | |
| "step": 1495, | |
| "valid_targets_mean": 2422.6, | |
| "valid_targets_min": 672 | |
| }, | |
| { | |
| "epoch": 1.5756302521008403, | |
| "grad_norm": 0.7318400772244765, | |
| "learning_rate": 3.8130212524171576e-05, | |
| "loss": 0.2871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3544023036956787, | |
| "step": 1500, | |
| "valid_targets_mean": 2730.2, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 1.5808823529411766, | |
| "grad_norm": 0.764194530000509, | |
| "learning_rate": 3.810803384767465e-05, | |
| "loss": 0.2668, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22156544029712677, | |
| "step": 1505, | |
| "valid_targets_mean": 1869.6, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 1.5861344537815127, | |
| "grad_norm": 0.7198936509540297, | |
| "learning_rate": 3.808573093653777e-05, | |
| "loss": 0.2803, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29135745763778687, | |
| "step": 1510, | |
| "valid_targets_mean": 2092.8, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 1.5913865546218489, | |
| "grad_norm": 0.664810251317532, | |
| "learning_rate": 3.806330394377556e-05, | |
| "loss": 0.2671, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2656877934932709, | |
| "step": 1515, | |
| "valid_targets_mean": 2300.4, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 1.596638655462185, | |
| "grad_norm": 0.7873381808170635, | |
| "learning_rate": 3.8040753023253956e-05, | |
| "loss": 0.2784, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2537113130092621, | |
| "step": 1520, | |
| "valid_targets_mean": 2188.1, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 1.601890756302521, | |
| "grad_norm": 0.8013762972275075, | |
| "learning_rate": 3.801807832968912e-05, | |
| "loss": 0.2911, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28357455134391785, | |
| "step": 1525, | |
| "valid_targets_mean": 1841.9, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 1.6071428571428572, | |
| "grad_norm": 0.6363572479336133, | |
| "learning_rate": 3.799528001864637e-05, | |
| "loss": 0.2632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28672128915786743, | |
| "step": 1530, | |
| "valid_targets_mean": 2678.1, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 1.6123949579831933, | |
| "grad_norm": 0.626675259486661, | |
| "learning_rate": 3.797235824653918e-05, | |
| "loss": 0.2834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2521238327026367, | |
| "step": 1535, | |
| "valid_targets_mean": 2923.1, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 1.6176470588235294, | |
| "grad_norm": 0.7719429167065398, | |
| "learning_rate": 3.7949313170628006e-05, | |
| "loss": 0.2855, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26524102687835693, | |
| "step": 1540, | |
| "valid_targets_mean": 2188.1, | |
| "valid_targets_min": 889 | |
| }, | |
| { | |
| "epoch": 1.6228991596638656, | |
| "grad_norm": 0.6140414130735539, | |
| "learning_rate": 3.79261449490193e-05, | |
| "loss": 0.2824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28398892283439636, | |
| "step": 1545, | |
| "valid_targets_mean": 2975.3, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 1.6281512605042017, | |
| "grad_norm": 0.6333512043620656, | |
| "learning_rate": 3.7902853740664356e-05, | |
| "loss": 0.2777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25651031732559204, | |
| "step": 1550, | |
| "valid_targets_mean": 2676.1, | |
| "valid_targets_min": 885 | |
| }, | |
| { | |
| "epoch": 1.6334033613445378, | |
| "grad_norm": 0.7945225853691145, | |
| "learning_rate": 3.7879439705358286e-05, | |
| "loss": 0.2886, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28769397735595703, | |
| "step": 1555, | |
| "valid_targets_mean": 2333.7, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 1.638655462184874, | |
| "grad_norm": 0.6613578680004154, | |
| "learning_rate": 3.785590300373884e-05, | |
| "loss": 0.2963, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2788962423801422, | |
| "step": 1560, | |
| "valid_targets_mean": 2716.4, | |
| "valid_targets_min": 1120 | |
| }, | |
| { | |
| "epoch": 1.64390756302521, | |
| "grad_norm": 0.6200210594873011, | |
| "learning_rate": 3.7832243797285385e-05, | |
| "loss": 0.2742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30993425846099854, | |
| "step": 1565, | |
| "valid_targets_mean": 3246.9, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 1.6491596638655461, | |
| "grad_norm": 0.6841959383278441, | |
| "learning_rate": 3.780846224831775e-05, | |
| "loss": 0.2795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29545074701309204, | |
| "step": 1570, | |
| "valid_targets_mean": 2656.0, | |
| "valid_targets_min": 1143 | |
| }, | |
| { | |
| "epoch": 1.6544117647058822, | |
| "grad_norm": 0.646661736839655, | |
| "learning_rate": 3.7784558519995135e-05, | |
| "loss": 0.2807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2950468063354492, | |
| "step": 1575, | |
| "valid_targets_mean": 2778.9, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 1.6596638655462184, | |
| "grad_norm": 0.5963871526060771, | |
| "learning_rate": 3.776053277631496e-05, | |
| "loss": 0.2784, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27574625611305237, | |
| "step": 1580, | |
| "valid_targets_mean": 2941.4, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 1.6649159663865545, | |
| "grad_norm": 0.6789032459708729, | |
| "learning_rate": 3.7736385182111785e-05, | |
| "loss": 0.2766, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28872212767601013, | |
| "step": 1585, | |
| "valid_targets_mean": 2428.8, | |
| "valid_targets_min": 1240 | |
| }, | |
| { | |
| "epoch": 1.6701680672268906, | |
| "grad_norm": 0.8766320854428324, | |
| "learning_rate": 3.771211590305614e-05, | |
| "loss": 0.2937, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27332085371017456, | |
| "step": 1590, | |
| "valid_targets_mean": 2621.5, | |
| "valid_targets_min": 1095 | |
| }, | |
| { | |
| "epoch": 1.675420168067227, | |
| "grad_norm": 0.6841578200409935, | |
| "learning_rate": 3.768772510565342e-05, | |
| "loss": 0.2698, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2623317241668701, | |
| "step": 1595, | |
| "valid_targets_mean": 2145.7, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 1.680672268907563, | |
| "grad_norm": 0.7486173630607642, | |
| "learning_rate": 3.76632129572427e-05, | |
| "loss": 0.2732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.306223064661026, | |
| "step": 1600, | |
| "valid_targets_mean": 2413.9, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 1.6859243697478992, | |
| "grad_norm": 0.6168280825117122, | |
| "learning_rate": 3.763857962599565e-05, | |
| "loss": 0.2747, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26991304755210876, | |
| "step": 1605, | |
| "valid_targets_mean": 2552.2, | |
| "valid_targets_min": 1097 | |
| }, | |
| { | |
| "epoch": 1.6911764705882353, | |
| "grad_norm": 0.6766304351456689, | |
| "learning_rate": 3.761382528091531e-05, | |
| "loss": 0.2759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28340253233909607, | |
| "step": 1610, | |
| "valid_targets_mean": 2665.7, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 1.6964285714285714, | |
| "grad_norm": 0.7062299801799153, | |
| "learning_rate": 3.7588950091834986e-05, | |
| "loss": 0.2863, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27810871601104736, | |
| "step": 1615, | |
| "valid_targets_mean": 2484.9, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 1.7016806722689075, | |
| "grad_norm": 0.6152392266872755, | |
| "learning_rate": 3.756395422941706e-05, | |
| "loss": 0.2823, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3008061647415161, | |
| "step": 1620, | |
| "valid_targets_mean": 3126.2, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 1.7069327731092439, | |
| "grad_norm": 0.5905779852643404, | |
| "learning_rate": 3.753883786515181e-05, | |
| "loss": 0.2714, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26379841566085815, | |
| "step": 1625, | |
| "valid_targets_mean": 3115.7, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 1.71218487394958, | |
| "grad_norm": 0.7318199425494478, | |
| "learning_rate": 3.751360117135628e-05, | |
| "loss": 0.2544, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2550744414329529, | |
| "step": 1630, | |
| "valid_targets_mean": 2223.6, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 1.717436974789916, | |
| "grad_norm": 0.663349889323202, | |
| "learning_rate": 3.7488244321173025e-05, | |
| "loss": 0.2836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2922671437263489, | |
| "step": 1635, | |
| "valid_targets_mean": 2466.1, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 1.7226890756302522, | |
| "grad_norm": 0.7439072694880867, | |
| "learning_rate": 3.746276748856898e-05, | |
| "loss": 0.2728, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2746467590332031, | |
| "step": 1640, | |
| "valid_targets_mean": 2209.7, | |
| "valid_targets_min": 968 | |
| }, | |
| { | |
| "epoch": 1.7279411764705883, | |
| "grad_norm": 0.7242128901414696, | |
| "learning_rate": 3.743717084833425e-05, | |
| "loss": 0.2713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2738949656486511, | |
| "step": 1645, | |
| "valid_targets_mean": 2048.8, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 1.7331932773109244, | |
| "grad_norm": 0.5975483868192742, | |
| "learning_rate": 3.741145457608093e-05, | |
| "loss": 0.2834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2882964611053467, | |
| "step": 1650, | |
| "valid_targets_mean": 3101.3, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 1.7384453781512605, | |
| "grad_norm": 0.7199800936223271, | |
| "learning_rate": 3.738561884824183e-05, | |
| "loss": 0.2895, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.31954213976860046, | |
| "step": 1655, | |
| "valid_targets_mean": 2576.0, | |
| "valid_targets_min": 1213 | |
| }, | |
| { | |
| "epoch": 1.7436974789915967, | |
| "grad_norm": 0.643400369166649, | |
| "learning_rate": 3.735966384206936e-05, | |
| "loss": 0.2691, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2960051894187927, | |
| "step": 1660, | |
| "valid_targets_mean": 3030.7, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 1.7489495798319328, | |
| "grad_norm": 0.5952106042261482, | |
| "learning_rate": 3.733358973563425e-05, | |
| "loss": 0.2851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29441675543785095, | |
| "step": 1665, | |
| "valid_targets_mean": 3160.8, | |
| "valid_targets_min": 1137 | |
| }, | |
| { | |
| "epoch": 1.754201680672269, | |
| "grad_norm": 0.6786980642043301, | |
| "learning_rate": 3.730739670782435e-05, | |
| "loss": 0.2809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2848990559577942, | |
| "step": 1670, | |
| "valid_targets_mean": 2464.0, | |
| "valid_targets_min": 900 | |
| }, | |
| { | |
| "epoch": 1.759453781512605, | |
| "grad_norm": 0.6509095658222819, | |
| "learning_rate": 3.72810849383434e-05, | |
| "loss": 0.2678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26271501183509827, | |
| "step": 1675, | |
| "valid_targets_mean": 2341.4, | |
| "valid_targets_min": 537 | |
| }, | |
| { | |
| "epoch": 1.7647058823529411, | |
| "grad_norm": 0.7898671503621937, | |
| "learning_rate": 3.725465460770978e-05, | |
| "loss": 0.3089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3113194704055786, | |
| "step": 1680, | |
| "valid_targets_mean": 2466.2, | |
| "valid_targets_min": 846 | |
| }, | |
| { | |
| "epoch": 1.7699579831932772, | |
| "grad_norm": 0.6739070075903468, | |
| "learning_rate": 3.7228105897255324e-05, | |
| "loss": 0.2811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2531262934207916, | |
| "step": 1685, | |
| "valid_targets_mean": 2049.8, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 1.7752100840336134, | |
| "grad_norm": 0.6710697879150522, | |
| "learning_rate": 3.7201438989124e-05, | |
| "loss": 0.2761, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28480058908462524, | |
| "step": 1690, | |
| "valid_targets_mean": 2619.4, | |
| "valid_targets_min": 1076 | |
| }, | |
| { | |
| "epoch": 1.7804621848739495, | |
| "grad_norm": 0.662695525206772, | |
| "learning_rate": 3.717465406627074e-05, | |
| "loss": 0.2726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2571936547756195, | |
| "step": 1695, | |
| "valid_targets_mean": 2399.6, | |
| "valid_targets_min": 924 | |
| }, | |
| { | |
| "epoch": 1.7857142857142856, | |
| "grad_norm": 0.6320359884584241, | |
| "learning_rate": 3.714775131246011e-05, | |
| "loss": 0.2818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2644287347793579, | |
| "step": 1700, | |
| "valid_targets_mean": 2833.1, | |
| "valid_targets_min": 1022 | |
| }, | |
| { | |
| "epoch": 1.7909663865546217, | |
| "grad_norm": 0.5608837993978538, | |
| "learning_rate": 3.71207309122651e-05, | |
| "loss": 0.2656, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25060662627220154, | |
| "step": 1705, | |
| "valid_targets_mean": 3011.6, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 1.7962184873949578, | |
| "grad_norm": 0.7749333627981896, | |
| "learning_rate": 3.709359305106585e-05, | |
| "loss": 0.2835, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3079194128513336, | |
| "step": 1710, | |
| "valid_targets_mean": 2177.3, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 1.8014705882352942, | |
| "grad_norm": 0.6704960756054471, | |
| "learning_rate": 3.7066337915048354e-05, | |
| "loss": 0.2633, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24317601323127747, | |
| "step": 1715, | |
| "valid_targets_mean": 2278.3, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 1.8067226890756303, | |
| "grad_norm": 0.726852310035077, | |
| "learning_rate": 3.7038965691203205e-05, | |
| "loss": 0.2867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2838640809059143, | |
| "step": 1720, | |
| "valid_targets_mean": 2385.5, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 1.8119747899159664, | |
| "grad_norm": 0.7307615924485396, | |
| "learning_rate": 3.701147656732431e-05, | |
| "loss": 0.2626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2594029903411865, | |
| "step": 1725, | |
| "valid_targets_mean": 2254.4, | |
| "valid_targets_min": 823 | |
| }, | |
| { | |
| "epoch": 1.8172268907563025, | |
| "grad_norm": 0.6804119390418033, | |
| "learning_rate": 3.6983870732007596e-05, | |
| "loss": 0.2897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30185481905937195, | |
| "step": 1730, | |
| "valid_targets_mean": 2529.1, | |
| "valid_targets_min": 861 | |
| }, | |
| { | |
| "epoch": 1.8224789915966386, | |
| "grad_norm": 0.5971578197210542, | |
| "learning_rate": 3.695614837464972e-05, | |
| "loss": 0.2799, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2714010179042816, | |
| "step": 1735, | |
| "valid_targets_mean": 2919.9, | |
| "valid_targets_min": 934 | |
| }, | |
| { | |
| "epoch": 1.8277310924369747, | |
| "grad_norm": 0.7432330036153248, | |
| "learning_rate": 3.692830968544675e-05, | |
| "loss": 0.2905, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2643352150917053, | |
| "step": 1740, | |
| "valid_targets_mean": 2157.0, | |
| "valid_targets_min": 913 | |
| }, | |
| { | |
| "epoch": 1.832983193277311, | |
| "grad_norm": 0.9451021623758795, | |
| "learning_rate": 3.690035485539291e-05, | |
| "loss": 0.2848, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2754422128200531, | |
| "step": 1745, | |
| "valid_targets_mean": 2272.9, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 1.8382352941176472, | |
| "grad_norm": 0.6448983937098554, | |
| "learning_rate": 3.6872284076279205e-05, | |
| "loss": 0.2735, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27130642533302307, | |
| "step": 1750, | |
| "valid_targets_mean": 2665.2, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 1.8434873949579833, | |
| "grad_norm": 0.6568847222392369, | |
| "learning_rate": 3.684409754069215e-05, | |
| "loss": 0.2887, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28479936718940735, | |
| "step": 1755, | |
| "valid_targets_mean": 2863.5, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 1.8487394957983194, | |
| "grad_norm": 0.9463446748898482, | |
| "learning_rate": 3.681579544201244e-05, | |
| "loss": 0.265, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.267799973487854, | |
| "step": 1760, | |
| "valid_targets_mean": 2591.4, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 1.8539915966386555, | |
| "grad_norm": 0.6626299146737941, | |
| "learning_rate": 3.6787377974413614e-05, | |
| "loss": 0.275, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26858189702033997, | |
| "step": 1765, | |
| "valid_targets_mean": 2298.1, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 1.8592436974789917, | |
| "grad_norm": 0.5511447233080423, | |
| "learning_rate": 3.6758845332860734e-05, | |
| "loss": 0.2641, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25503212213516235, | |
| "step": 1770, | |
| "valid_targets_mean": 3300.6, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 1.8644957983193278, | |
| "grad_norm": 0.6864216219048197, | |
| "learning_rate": 3.673019771310903e-05, | |
| "loss": 0.2724, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29600006341934204, | |
| "step": 1775, | |
| "valid_targets_mean": 2480.2, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 1.8697478991596639, | |
| "grad_norm": 0.6727603790480745, | |
| "learning_rate": 3.670143531170258e-05, | |
| "loss": 0.2847, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2595556974411011, | |
| "step": 1780, | |
| "valid_targets_mean": 2553.4, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 1.875, | |
| "grad_norm": 0.6449743095704825, | |
| "learning_rate": 3.667255832597294e-05, | |
| "loss": 0.2719, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.287334680557251, | |
| "step": 1785, | |
| "valid_targets_mean": 2550.0, | |
| "valid_targets_min": 674 | |
| }, | |
| { | |
| "epoch": 1.8802521008403361, | |
| "grad_norm": 0.7356297744717748, | |
| "learning_rate": 3.664356695403781e-05, | |
| "loss": 0.2709, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2541128993034363, | |
| "step": 1790, | |
| "valid_targets_mean": 1760.9, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 1.8855042016806722, | |
| "grad_norm": 0.6441525341607449, | |
| "learning_rate": 3.661446139479965e-05, | |
| "loss": 0.2722, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2607778310775757, | |
| "step": 1795, | |
| "valid_targets_mean": 2639.6, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 1.8907563025210083, | |
| "grad_norm": 0.6764765078850176, | |
| "learning_rate": 3.658524184794436e-05, | |
| "loss": 0.2808, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2690533399581909, | |
| "step": 1800, | |
| "valid_targets_mean": 2191.6, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 1.8960084033613445, | |
| "grad_norm": 0.6223425261146642, | |
| "learning_rate": 3.655590851393984e-05, | |
| "loss": 0.266, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2553020417690277, | |
| "step": 1805, | |
| "valid_targets_mean": 2840.2, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 1.9012605042016806, | |
| "grad_norm": 0.571683682772899, | |
| "learning_rate": 3.652646159403468e-05, | |
| "loss": 0.2724, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2515188455581665, | |
| "step": 1810, | |
| "valid_targets_mean": 3140.4, | |
| "valid_targets_min": 919 | |
| }, | |
| { | |
| "epoch": 1.9065126050420167, | |
| "grad_norm": 0.6705352110164979, | |
| "learning_rate": 3.649690129025675e-05, | |
| "loss": 0.285, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28106388449668884, | |
| "step": 1815, | |
| "valid_targets_mean": 2872.1, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 1.9117647058823528, | |
| "grad_norm": 0.6674867092826292, | |
| "learning_rate": 3.6467227805411824e-05, | |
| "loss": 0.2805, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2530893385410309, | |
| "step": 1820, | |
| "valid_targets_mean": 2187.0, | |
| "valid_targets_min": 1228 | |
| }, | |
| { | |
| "epoch": 1.917016806722689, | |
| "grad_norm": 0.7791624483418543, | |
| "learning_rate": 3.6437441343082174e-05, | |
| "loss": 0.2576, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25658732652664185, | |
| "step": 1825, | |
| "valid_targets_mean": 1671.2, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 1.9222689075630253, | |
| "grad_norm": 0.7296738608202369, | |
| "learning_rate": 3.6407542107625167e-05, | |
| "loss": 0.2785, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25650477409362793, | |
| "step": 1830, | |
| "valid_targets_mean": 2080.8, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 1.9275210084033614, | |
| "grad_norm": 0.6401628879972454, | |
| "learning_rate": 3.637753030417192e-05, | |
| "loss": 0.2882, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.271770179271698, | |
| "step": 1835, | |
| "valid_targets_mean": 2537.3, | |
| "valid_targets_min": 1114 | |
| }, | |
| { | |
| "epoch": 1.9327731092436975, | |
| "grad_norm": 0.6126369558801827, | |
| "learning_rate": 3.6347406138625805e-05, | |
| "loss": 0.2925, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2938627600669861, | |
| "step": 1840, | |
| "valid_targets_mean": 2981.4, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 1.9380252100840336, | |
| "grad_norm": 0.7218262814733415, | |
| "learning_rate": 3.631716981766111e-05, | |
| "loss": 0.2786, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30355095863342285, | |
| "step": 1845, | |
| "valid_targets_mean": 2301.2, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 1.9432773109243697, | |
| "grad_norm": 0.668031618282193, | |
| "learning_rate": 3.6286821548721594e-05, | |
| "loss": 0.2752, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27426058053970337, | |
| "step": 1850, | |
| "valid_targets_mean": 2243.2, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 1.9485294117647058, | |
| "grad_norm": 0.752574443244231, | |
| "learning_rate": 3.625636154001904e-05, | |
| "loss": 0.2859, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3078942894935608, | |
| "step": 1855, | |
| "valid_targets_mean": 2398.1, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 1.9537815126050422, | |
| "grad_norm": 0.7317868910075894, | |
| "learning_rate": 3.622579000053186e-05, | |
| "loss": 0.2737, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28089436888694763, | |
| "step": 1860, | |
| "valid_targets_mean": 2312.2, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 1.9590336134453783, | |
| "grad_norm": 0.5809494600449272, | |
| "learning_rate": 3.619510714000368e-05, | |
| "loss": 0.2872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28781622648239136, | |
| "step": 1865, | |
| "valid_targets_mean": 3431.6, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 1.9642857142857144, | |
| "grad_norm": 0.6561581539587649, | |
| "learning_rate": 3.616431316894181e-05, | |
| "loss": 0.2752, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2707546651363373, | |
| "step": 1870, | |
| "valid_targets_mean": 2314.8, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 1.9695378151260505, | |
| "grad_norm": 0.6586257109850335, | |
| "learning_rate": 3.613340829861592e-05, | |
| "loss": 0.2674, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28310245275497437, | |
| "step": 1875, | |
| "valid_targets_mean": 2942.6, | |
| "valid_targets_min": 936 | |
| }, | |
| { | |
| "epoch": 1.9747899159663866, | |
| "grad_norm": 0.6401802852139319, | |
| "learning_rate": 3.61023927410565e-05, | |
| "loss": 0.2666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.290630578994751, | |
| "step": 1880, | |
| "valid_targets_mean": 2797.1, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 1.9800420168067228, | |
| "grad_norm": 0.6218936013613373, | |
| "learning_rate": 3.607126670905346e-05, | |
| "loss": 0.2807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27291837334632874, | |
| "step": 1885, | |
| "valid_targets_mean": 2842.4, | |
| "valid_targets_min": 484 | |
| }, | |
| { | |
| "epoch": 1.9852941176470589, | |
| "grad_norm": 0.6591905106890031, | |
| "learning_rate": 3.60400304161546e-05, | |
| "loss": 0.2745, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24094945192337036, | |
| "step": 1890, | |
| "valid_targets_mean": 2038.6, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 1.990546218487395, | |
| "grad_norm": 0.6093929527492841, | |
| "learning_rate": 3.600868407666426e-05, | |
| "loss": 0.2653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.251456618309021, | |
| "step": 1895, | |
| "valid_targets_mean": 2710.1, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 1.995798319327731, | |
| "grad_norm": 0.7825350266466161, | |
| "learning_rate": 3.597722790564172e-05, | |
| "loss": 0.2824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.288199245929718, | |
| "step": 1900, | |
| "valid_targets_mean": 1974.7, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 2.0010504201680672, | |
| "grad_norm": 0.5933761068921561, | |
| "learning_rate": 3.5945662118899824e-05, | |
| "loss": 0.2795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2670100927352905, | |
| "step": 1905, | |
| "valid_targets_mean": 3187.6, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 2.0063025210084033, | |
| "grad_norm": 0.681788431886679, | |
| "learning_rate": 3.5913986933003455e-05, | |
| "loss": 0.2537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23222820460796356, | |
| "step": 1910, | |
| "valid_targets_mean": 2392.1, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 2.0115546218487395, | |
| "grad_norm": 0.6959432299927795, | |
| "learning_rate": 3.5882202565268046e-05, | |
| "loss": 0.2578, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25007233023643494, | |
| "step": 1915, | |
| "valid_targets_mean": 2417.9, | |
| "valid_targets_min": 775 | |
| }, | |
| { | |
| "epoch": 2.0168067226890756, | |
| "grad_norm": 0.7488088088666018, | |
| "learning_rate": 3.58503092337581e-05, | |
| "loss": 0.2415, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26307564973831177, | |
| "step": 1920, | |
| "valid_targets_mean": 2433.2, | |
| "valid_targets_min": 923 | |
| }, | |
| { | |
| "epoch": 2.0220588235294117, | |
| "grad_norm": 0.8762656352125061, | |
| "learning_rate": 3.5818307157285705e-05, | |
| "loss": 0.2404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2252756655216217, | |
| "step": 1925, | |
| "valid_targets_mean": 2069.1, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 2.027310924369748, | |
| "grad_norm": 0.7108878947236019, | |
| "learning_rate": 3.5786196555409006e-05, | |
| "loss": 0.2483, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2223615050315857, | |
| "step": 1930, | |
| "valid_targets_mean": 2248.2, | |
| "valid_targets_min": 1239 | |
| }, | |
| { | |
| "epoch": 2.032563025210084, | |
| "grad_norm": 0.6819496202558551, | |
| "learning_rate": 3.575397764843073e-05, | |
| "loss": 0.2386, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2185153365135193, | |
| "step": 1935, | |
| "valid_targets_mean": 2224.1, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 2.03781512605042, | |
| "grad_norm": 0.6372992890667374, | |
| "learning_rate": 3.572165065739664e-05, | |
| "loss": 0.2442, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2360488474369049, | |
| "step": 1940, | |
| "valid_targets_mean": 2788.4, | |
| "valid_targets_min": 889 | |
| }, | |
| { | |
| "epoch": 2.043067226890756, | |
| "grad_norm": 0.6386451317642459, | |
| "learning_rate": 3.5689215804094053e-05, | |
| "loss": 0.2322, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22324788570404053, | |
| "step": 1945, | |
| "valid_targets_mean": 2618.2, | |
| "valid_targets_min": 994 | |
| }, | |
| { | |
| "epoch": 2.0483193277310923, | |
| "grad_norm": 0.7004427658132236, | |
| "learning_rate": 3.5656673311050285e-05, | |
| "loss": 0.2515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27351114153862, | |
| "step": 1950, | |
| "valid_targets_mean": 2478.5, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 2.0535714285714284, | |
| "grad_norm": 0.7101917295820503, | |
| "learning_rate": 3.562402340153115e-05, | |
| "loss": 0.2297, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2520638406276703, | |
| "step": 1955, | |
| "valid_targets_mean": 2472.2, | |
| "valid_targets_min": 1105 | |
| }, | |
| { | |
| "epoch": 2.0588235294117645, | |
| "grad_norm": 0.6511098248209234, | |
| "learning_rate": 3.559126629953942e-05, | |
| "loss": 0.2551, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2435356080532074, | |
| "step": 1960, | |
| "valid_targets_mean": 3219.1, | |
| "valid_targets_min": 1104 | |
| }, | |
| { | |
| "epoch": 2.064075630252101, | |
| "grad_norm": 0.6358952603701429, | |
| "learning_rate": 3.5558402229813287e-05, | |
| "loss": 0.2449, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2297707498073578, | |
| "step": 1965, | |
| "valid_targets_mean": 2425.1, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 2.069327731092437, | |
| "grad_norm": 0.6868889163392828, | |
| "learning_rate": 3.552543141782483e-05, | |
| "loss": 0.2333, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24342474341392517, | |
| "step": 1970, | |
| "valid_targets_mean": 2674.5, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 2.0745798319327733, | |
| "grad_norm": 0.6503662588808112, | |
| "learning_rate": 3.549235408977843e-05, | |
| "loss": 0.2432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23457615077495575, | |
| "step": 1975, | |
| "valid_targets_mean": 2697.2, | |
| "valid_targets_min": 874 | |
| }, | |
| { | |
| "epoch": 2.0798319327731094, | |
| "grad_norm": 0.822377652612648, | |
| "learning_rate": 3.545917047260928e-05, | |
| "loss": 0.2556, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26807093620300293, | |
| "step": 1980, | |
| "valid_targets_mean": 1928.6, | |
| "valid_targets_min": 930 | |
| }, | |
| { | |
| "epoch": 2.0850840336134455, | |
| "grad_norm": 0.7145196067189729, | |
| "learning_rate": 3.542588079398179e-05, | |
| "loss": 0.2426, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26111432909965515, | |
| "step": 1985, | |
| "valid_targets_mean": 2330.2, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 2.0903361344537816, | |
| "grad_norm": 0.7523096751763567, | |
| "learning_rate": 3.539248528228801e-05, | |
| "loss": 0.2426, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24235697090625763, | |
| "step": 1990, | |
| "valid_targets_mean": 2053.5, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 2.0955882352941178, | |
| "grad_norm": 0.7632438420213485, | |
| "learning_rate": 3.535898416664611e-05, | |
| "loss": 0.2326, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22527560591697693, | |
| "step": 1995, | |
| "valid_targets_mean": 2201.0, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 2.100840336134454, | |
| "grad_norm": 0.7057387201013904, | |
| "learning_rate": 3.5325377676898766e-05, | |
| "loss": 0.233, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25476691126823425, | |
| "step": 2000, | |
| "valid_targets_mean": 2419.0, | |
| "valid_targets_min": 944 | |
| }, | |
| { | |
| "epoch": 2.10609243697479, | |
| "grad_norm": 0.6906581075458723, | |
| "learning_rate": 3.529166604361161e-05, | |
| "loss": 0.2323, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19508974254131317, | |
| "step": 2005, | |
| "valid_targets_mean": 1831.8, | |
| "valid_targets_min": 959 | |
| }, | |
| { | |
| "epoch": 2.111344537815126, | |
| "grad_norm": 0.7527030761728207, | |
| "learning_rate": 3.525784949807162e-05, | |
| "loss": 0.2359, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23376065492630005, | |
| "step": 2010, | |
| "valid_targets_mean": 2055.6, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 2.116596638655462, | |
| "grad_norm": 0.6868567018841076, | |
| "learning_rate": 3.522392827228557e-05, | |
| "loss": 0.2511, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2521514296531677, | |
| "step": 2015, | |
| "valid_targets_mean": 3081.2, | |
| "valid_targets_min": 1079 | |
| }, | |
| { | |
| "epoch": 2.1218487394957983, | |
| "grad_norm": 0.6159147126631715, | |
| "learning_rate": 3.518990259897841e-05, | |
| "loss": 0.2301, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24102434515953064, | |
| "step": 2020, | |
| "valid_targets_mean": 3128.9, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 2.1271008403361344, | |
| "grad_norm": 0.6607408752549182, | |
| "learning_rate": 3.515577271159167e-05, | |
| "loss": 0.2451, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24916204810142517, | |
| "step": 2025, | |
| "valid_targets_mean": 2437.9, | |
| "valid_targets_min": 895 | |
| }, | |
| { | |
| "epoch": 2.1323529411764706, | |
| "grad_norm": 0.6823759669457126, | |
| "learning_rate": 3.512153884428188e-05, | |
| "loss": 0.2427, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24420082569122314, | |
| "step": 2030, | |
| "valid_targets_mean": 2662.5, | |
| "valid_targets_min": 1212 | |
| }, | |
| { | |
| "epoch": 2.1376050420168067, | |
| "grad_norm": 0.8063415806620354, | |
| "learning_rate": 3.5087201231918944e-05, | |
| "loss": 0.2482, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.255155086517334, | |
| "step": 2035, | |
| "valid_targets_mean": 2731.2, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 2.142857142857143, | |
| "grad_norm": 0.7415414698356663, | |
| "learning_rate": 3.505276011008454e-05, | |
| "loss": 0.2435, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23635327816009521, | |
| "step": 2040, | |
| "valid_targets_mean": 2135.6, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 2.148109243697479, | |
| "grad_norm": 0.6907057684258697, | |
| "learning_rate": 3.501821571507048e-05, | |
| "loss": 0.2379, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24406486749649048, | |
| "step": 2045, | |
| "valid_targets_mean": 2447.7, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 2.153361344537815, | |
| "grad_norm": 0.6782470155275483, | |
| "learning_rate": 3.498356828387714e-05, | |
| "loss": 0.2463, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23916223645210266, | |
| "step": 2050, | |
| "valid_targets_mean": 2313.4, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 2.158613445378151, | |
| "grad_norm": 0.5671322606625604, | |
| "learning_rate": 3.494881805421176e-05, | |
| "loss": 0.2438, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19718879461288452, | |
| "step": 2055, | |
| "valid_targets_mean": 2824.4, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 2.1638655462184873, | |
| "grad_norm": 0.9382283657965643, | |
| "learning_rate": 3.4913965264486894e-05, | |
| "loss": 0.2328, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25609293580055237, | |
| "step": 2060, | |
| "valid_targets_mean": 1910.2, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 2.1691176470588234, | |
| "grad_norm": 0.6346830510041134, | |
| "learning_rate": 3.48790101538187e-05, | |
| "loss": 0.2413, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2414173036813736, | |
| "step": 2065, | |
| "valid_targets_mean": 2811.0, | |
| "valid_targets_min": 1046 | |
| }, | |
| { | |
| "epoch": 2.1743697478991595, | |
| "grad_norm": 0.7269704066923304, | |
| "learning_rate": 3.484395296202537e-05, | |
| "loss": 0.2459, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24697639048099518, | |
| "step": 2070, | |
| "valid_targets_mean": 2118.1, | |
| "valid_targets_min": 1008 | |
| }, | |
| { | |
| "epoch": 2.1796218487394956, | |
| "grad_norm": 0.6343101249317197, | |
| "learning_rate": 3.480879392962541e-05, | |
| "loss": 0.2452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23526614904403687, | |
| "step": 2075, | |
| "valid_targets_mean": 2820.8, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 2.184873949579832, | |
| "grad_norm": 0.7287435124218465, | |
| "learning_rate": 3.477353329783606e-05, | |
| "loss": 0.2366, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24225783348083496, | |
| "step": 2080, | |
| "valid_targets_mean": 2625.6, | |
| "valid_targets_min": 1045 | |
| }, | |
| { | |
| "epoch": 2.190126050420168, | |
| "grad_norm": 0.713355174077326, | |
| "learning_rate": 3.4738171308571595e-05, | |
| "loss": 0.2524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.254425972700119, | |
| "step": 2085, | |
| "valid_targets_mean": 2268.8, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 2.1953781512605044, | |
| "grad_norm": 0.6705604877150245, | |
| "learning_rate": 3.470270820444168e-05, | |
| "loss": 0.2453, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2541666328907013, | |
| "step": 2090, | |
| "valid_targets_mean": 2757.6, | |
| "valid_targets_min": 952 | |
| }, | |
| { | |
| "epoch": 2.2006302521008405, | |
| "grad_norm": 0.7233467972990565, | |
| "learning_rate": 3.4667144228749697e-05, | |
| "loss": 0.2656, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25709813833236694, | |
| "step": 2095, | |
| "valid_targets_mean": 2277.6, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 2.2058823529411766, | |
| "grad_norm": 0.7024704278871472, | |
| "learning_rate": 3.463147962549109e-05, | |
| "loss": 0.2302, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.238604336977005, | |
| "step": 2100, | |
| "valid_targets_mean": 2394.7, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 2.2111344537815127, | |
| "grad_norm": 0.7334542386999148, | |
| "learning_rate": 3.459571463935168e-05, | |
| "loss": 0.2423, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25194650888442993, | |
| "step": 2105, | |
| "valid_targets_mean": 2613.2, | |
| "valid_targets_min": 1580 | |
| }, | |
| { | |
| "epoch": 2.216386554621849, | |
| "grad_norm": 0.645760996168686, | |
| "learning_rate": 3.455984951570598e-05, | |
| "loss": 0.2452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22080859541893005, | |
| "step": 2110, | |
| "valid_targets_mean": 2687.3, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 2.221638655462185, | |
| "grad_norm": 0.6519130528193421, | |
| "learning_rate": 3.452388450061556e-05, | |
| "loss": 0.2498, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2937582731246948, | |
| "step": 2115, | |
| "valid_targets_mean": 3205.2, | |
| "valid_targets_min": 1111 | |
| }, | |
| { | |
| "epoch": 2.226890756302521, | |
| "grad_norm": 0.6624235701635455, | |
| "learning_rate": 3.448781984082727e-05, | |
| "loss": 0.2343, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21843945980072021, | |
| "step": 2120, | |
| "valid_targets_mean": 2508.0, | |
| "valid_targets_min": 1057 | |
| }, | |
| { | |
| "epoch": 2.232142857142857, | |
| "grad_norm": 0.7029645390202783, | |
| "learning_rate": 3.4451655783771625e-05, | |
| "loss": 0.2541, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2414093017578125, | |
| "step": 2125, | |
| "valid_targets_mean": 2228.1, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 2.2373949579831933, | |
| "grad_norm": 0.7968386644644923, | |
| "learning_rate": 3.4415392577561064e-05, | |
| "loss": 0.2339, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24058222770690918, | |
| "step": 2130, | |
| "valid_targets_mean": 1892.4, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 2.2426470588235294, | |
| "grad_norm": 0.7056496069912468, | |
| "learning_rate": 3.4379030470988285e-05, | |
| "loss": 0.2477, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24626708030700684, | |
| "step": 2135, | |
| "valid_targets_mean": 2260.1, | |
| "valid_targets_min": 928 | |
| }, | |
| { | |
| "epoch": 2.2478991596638656, | |
| "grad_norm": 0.6124790183288958, | |
| "learning_rate": 3.4342569713524496e-05, | |
| "loss": 0.2476, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22665052115917206, | |
| "step": 2140, | |
| "valid_targets_mean": 3020.3, | |
| "valid_targets_min": 1178 | |
| }, | |
| { | |
| "epoch": 2.2531512605042017, | |
| "grad_norm": 0.7290143698690328, | |
| "learning_rate": 3.4306010555317735e-05, | |
| "loss": 0.2386, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26667648553848267, | |
| "step": 2145, | |
| "valid_targets_mean": 2242.3, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 2.258403361344538, | |
| "grad_norm": 0.6119475782036818, | |
| "learning_rate": 3.4269353247191135e-05, | |
| "loss": 0.2374, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2166629135608673, | |
| "step": 2150, | |
| "valid_targets_mean": 2546.4, | |
| "valid_targets_min": 997 | |
| }, | |
| { | |
| "epoch": 2.263655462184874, | |
| "grad_norm": 0.6765056487027129, | |
| "learning_rate": 3.423259804064122e-05, | |
| "loss": 0.2492, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24621796607971191, | |
| "step": 2155, | |
| "valid_targets_mean": 2518.2, | |
| "valid_targets_min": 1296 | |
| }, | |
| { | |
| "epoch": 2.26890756302521, | |
| "grad_norm": 0.711905698459275, | |
| "learning_rate": 3.419574518783616e-05, | |
| "loss": 0.2287, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24736756086349487, | |
| "step": 2160, | |
| "valid_targets_mean": 3044.6, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 2.274159663865546, | |
| "grad_norm": 0.6672052083995789, | |
| "learning_rate": 3.4158794941614055e-05, | |
| "loss": 0.2546, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24737080931663513, | |
| "step": 2165, | |
| "valid_targets_mean": 3020.8, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 2.2794117647058822, | |
| "grad_norm": 0.7567079909016078, | |
| "learning_rate": 3.4121747555481216e-05, | |
| "loss": 0.2381, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2148715853691101, | |
| "step": 2170, | |
| "valid_targets_mean": 1714.9, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 2.2846638655462184, | |
| "grad_norm": 0.7056947251086683, | |
| "learning_rate": 3.4084603283610365e-05, | |
| "loss": 0.2458, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27144932746887207, | |
| "step": 2175, | |
| "valid_targets_mean": 2400.7, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 2.2899159663865545, | |
| "grad_norm": 0.7646246632019781, | |
| "learning_rate": 3.404736238083897e-05, | |
| "loss": 0.2429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23836296796798706, | |
| "step": 2180, | |
| "valid_targets_mean": 2148.2, | |
| "valid_targets_min": 709 | |
| }, | |
| { | |
| "epoch": 2.2951680672268906, | |
| "grad_norm": 0.6677517363916046, | |
| "learning_rate": 3.401002510266744e-05, | |
| "loss": 0.2433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24747160077095032, | |
| "step": 2185, | |
| "valid_targets_mean": 2523.8, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 2.3004201680672267, | |
| "grad_norm": 0.7787607353169186, | |
| "learning_rate": 3.397259170525739e-05, | |
| "loss": 0.2398, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24741989374160767, | |
| "step": 2190, | |
| "valid_targets_mean": 2161.8, | |
| "valid_targets_min": 1218 | |
| }, | |
| { | |
| "epoch": 2.3056722689075633, | |
| "grad_norm": 0.6898906104395905, | |
| "learning_rate": 3.393506244542991e-05, | |
| "loss": 0.2345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21572214365005493, | |
| "step": 2195, | |
| "valid_targets_mean": 2425.9, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 2.310924369747899, | |
| "grad_norm": 0.7076822593529297, | |
| "learning_rate": 3.389743758066374e-05, | |
| "loss": 0.2481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2650451064109802, | |
| "step": 2200, | |
| "valid_targets_mean": 2744.9, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 2.3161764705882355, | |
| "grad_norm": 0.7536553955769041, | |
| "learning_rate": 3.385971736909357e-05, | |
| "loss": 0.257, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2877483367919922, | |
| "step": 2205, | |
| "valid_targets_mean": 2068.6, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 2.3214285714285716, | |
| "grad_norm": 0.7234207845549456, | |
| "learning_rate": 3.382190206950822e-05, | |
| "loss": 0.2504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.243576318025589, | |
| "step": 2210, | |
| "valid_targets_mean": 2941.5, | |
| "valid_targets_min": 967 | |
| }, | |
| { | |
| "epoch": 2.3266806722689077, | |
| "grad_norm": 0.7515713668566515, | |
| "learning_rate": 3.3783991941348916e-05, | |
| "loss": 0.2464, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.264710009098053, | |
| "step": 2215, | |
| "valid_targets_mean": 2004.6, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 2.331932773109244, | |
| "grad_norm": 0.6251497478316201, | |
| "learning_rate": 3.374598724470745e-05, | |
| "loss": 0.2429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24632088840007782, | |
| "step": 2220, | |
| "valid_targets_mean": 2663.2, | |
| "valid_targets_min": 786 | |
| }, | |
| { | |
| "epoch": 2.33718487394958, | |
| "grad_norm": 0.6823271730426904, | |
| "learning_rate": 3.370788824032441e-05, | |
| "loss": 0.2411, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24375592172145844, | |
| "step": 2225, | |
| "valid_targets_mean": 2528.2, | |
| "valid_targets_min": 1099 | |
| }, | |
| { | |
| "epoch": 2.342436974789916, | |
| "grad_norm": 0.650663602449571, | |
| "learning_rate": 3.366969518958746e-05, | |
| "loss": 0.2552, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2503768801689148, | |
| "step": 2230, | |
| "valid_targets_mean": 2865.1, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 2.347689075630252, | |
| "grad_norm": 0.6601800895017542, | |
| "learning_rate": 3.363140835452945e-05, | |
| "loss": 0.2565, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2578763961791992, | |
| "step": 2235, | |
| "valid_targets_mean": 2738.4, | |
| "valid_targets_min": 1007 | |
| }, | |
| { | |
| "epoch": 2.3529411764705883, | |
| "grad_norm": 0.692317322973956, | |
| "learning_rate": 3.3593027997826654e-05, | |
| "loss": 0.2478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2523569166660309, | |
| "step": 2240, | |
| "valid_targets_mean": 2517.1, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 2.3581932773109244, | |
| "grad_norm": 0.7286731675117749, | |
| "learning_rate": 3.355455438279702e-05, | |
| "loss": 0.228, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20360924303531647, | |
| "step": 2245, | |
| "valid_targets_mean": 1795.9, | |
| "valid_targets_min": 521 | |
| }, | |
| { | |
| "epoch": 2.3634453781512605, | |
| "grad_norm": 0.7006474168718877, | |
| "learning_rate": 3.351598777339827e-05, | |
| "loss": 0.2312, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2347767949104309, | |
| "step": 2250, | |
| "valid_targets_mean": 2666.3, | |
| "valid_targets_min": 1020 | |
| }, | |
| { | |
| "epoch": 2.3686974789915967, | |
| "grad_norm": 0.5939844727687064, | |
| "learning_rate": 3.3477328434226164e-05, | |
| "loss": 0.2413, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23968379199504852, | |
| "step": 2255, | |
| "valid_targets_mean": 3521.1, | |
| "valid_targets_min": 1437 | |
| }, | |
| { | |
| "epoch": 2.3739495798319328, | |
| "grad_norm": 0.7625687687660169, | |
| "learning_rate": 3.343857663051264e-05, | |
| "loss": 0.2371, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20183886587619781, | |
| "step": 2260, | |
| "valid_targets_mean": 1734.0, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 2.379201680672269, | |
| "grad_norm": 0.7219603926351075, | |
| "learning_rate": 3.339973262812403e-05, | |
| "loss": 0.2486, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22381290793418884, | |
| "step": 2265, | |
| "valid_targets_mean": 2065.6, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 2.384453781512605, | |
| "grad_norm": 0.6274958459143466, | |
| "learning_rate": 3.336079669355921e-05, | |
| "loss": 0.2592, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26679128408432007, | |
| "step": 2270, | |
| "valid_targets_mean": 3185.1, | |
| "valid_targets_min": 886 | |
| }, | |
| { | |
| "epoch": 2.389705882352941, | |
| "grad_norm": 0.6780169305716136, | |
| "learning_rate": 3.332176909394777e-05, | |
| "loss": 0.2304, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2543075978755951, | |
| "step": 2275, | |
| "valid_targets_mean": 2658.4, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 2.3949579831932772, | |
| "grad_norm": 0.5895782853276299, | |
| "learning_rate": 3.3282650097048206e-05, | |
| "loss": 0.2562, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23634852468967438, | |
| "step": 2280, | |
| "valid_targets_mean": 3037.2, | |
| "valid_targets_min": 938 | |
| }, | |
| { | |
| "epoch": 2.4002100840336134, | |
| "grad_norm": 0.7590461713759683, | |
| "learning_rate": 3.324343997124606e-05, | |
| "loss": 0.2388, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23115390539169312, | |
| "step": 2285, | |
| "valid_targets_mean": 2907.3, | |
| "valid_targets_min": 1192 | |
| }, | |
| { | |
| "epoch": 2.4054621848739495, | |
| "grad_norm": 0.7830457306094891, | |
| "learning_rate": 3.3204138985552094e-05, | |
| "loss": 0.2334, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27056723833084106, | |
| "step": 2290, | |
| "valid_targets_mean": 1997.1, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 2.4107142857142856, | |
| "grad_norm": 0.6194407127598283, | |
| "learning_rate": 3.316474740960041e-05, | |
| "loss": 0.2445, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19772791862487793, | |
| "step": 2295, | |
| "valid_targets_mean": 2389.4, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 2.4159663865546217, | |
| "grad_norm": 0.6520643182398843, | |
| "learning_rate": 3.312526551364667e-05, | |
| "loss": 0.2378, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22252126038074493, | |
| "step": 2300, | |
| "valid_targets_mean": 2424.2, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 2.421218487394958, | |
| "grad_norm": 0.6888170634914532, | |
| "learning_rate": 3.308569356856616e-05, | |
| "loss": 0.2456, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24550795555114746, | |
| "step": 2305, | |
| "valid_targets_mean": 2487.7, | |
| "valid_targets_min": 701 | |
| }, | |
| { | |
| "epoch": 2.426470588235294, | |
| "grad_norm": 0.5982413076324926, | |
| "learning_rate": 3.3046031845851993e-05, | |
| "loss": 0.2497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25226032733917236, | |
| "step": 2310, | |
| "valid_targets_mean": 3157.9, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 2.43172268907563, | |
| "grad_norm": 0.6571326373833773, | |
| "learning_rate": 3.3006280617613214e-05, | |
| "loss": 0.227, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22074852883815765, | |
| "step": 2315, | |
| "valid_targets_mean": 2177.3, | |
| "valid_targets_min": 651 | |
| }, | |
| { | |
| "epoch": 2.4369747899159666, | |
| "grad_norm": 0.6894533182551907, | |
| "learning_rate": 3.296644015657295e-05, | |
| "loss": 0.2392, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21584966778755188, | |
| "step": 2320, | |
| "valid_targets_mean": 2304.4, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 2.4422268907563023, | |
| "grad_norm": 0.6972919721870979, | |
| "learning_rate": 3.292651073606652e-05, | |
| "loss": 0.2328, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23415690660476685, | |
| "step": 2325, | |
| "valid_targets_mean": 2303.6, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 2.447478991596639, | |
| "grad_norm": 0.6334274393400314, | |
| "learning_rate": 3.288649263003958e-05, | |
| "loss": 0.239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23000667989253998, | |
| "step": 2330, | |
| "valid_targets_mean": 2705.6, | |
| "valid_targets_min": 752 | |
| }, | |
| { | |
| "epoch": 2.452731092436975, | |
| "grad_norm": 0.618380033342548, | |
| "learning_rate": 3.284638611304623e-05, | |
| "loss": 0.237, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2020508050918579, | |
| "step": 2335, | |
| "valid_targets_mean": 2487.5, | |
| "valid_targets_min": 1089 | |
| }, | |
| { | |
| "epoch": 2.457983193277311, | |
| "grad_norm": 0.6426580413501916, | |
| "learning_rate": 3.280619146024714e-05, | |
| "loss": 0.242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23832888901233673, | |
| "step": 2340, | |
| "valid_targets_mean": 2668.8, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 2.463235294117647, | |
| "grad_norm": 0.6496710491874221, | |
| "learning_rate": 3.2765908947407665e-05, | |
| "loss": 0.2465, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24643950164318085, | |
| "step": 2345, | |
| "valid_targets_mean": 3095.7, | |
| "valid_targets_min": 1618 | |
| }, | |
| { | |
| "epoch": 2.4684873949579833, | |
| "grad_norm": 0.6671046199265696, | |
| "learning_rate": 3.2725538850895933e-05, | |
| "loss": 0.2419, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24697118997573853, | |
| "step": 2350, | |
| "valid_targets_mean": 3175.2, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 2.4737394957983194, | |
| "grad_norm": 0.608779379161093, | |
| "learning_rate": 3.268508144768096e-05, | |
| "loss": 0.2383, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2293773591518402, | |
| "step": 2355, | |
| "valid_targets_mean": 2969.8, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 2.4789915966386555, | |
| "grad_norm": 0.7152715550331328, | |
| "learning_rate": 3.264453701533075e-05, | |
| "loss": 0.2418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24297022819519043, | |
| "step": 2360, | |
| "valid_targets_mean": 2333.1, | |
| "valid_targets_min": 717 | |
| }, | |
| { | |
| "epoch": 2.4842436974789917, | |
| "grad_norm": 0.6884766551131825, | |
| "learning_rate": 3.260390583201041e-05, | |
| "loss": 0.2468, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2274142950773239, | |
| "step": 2365, | |
| "valid_targets_mean": 2203.9, | |
| "valid_targets_min": 886 | |
| }, | |
| { | |
| "epoch": 2.4894957983193278, | |
| "grad_norm": 0.6696446459706793, | |
| "learning_rate": 3.2563188176480217e-05, | |
| "loss": 0.2394, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2381383329629898, | |
| "step": 2370, | |
| "valid_targets_mean": 2626.6, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 2.494747899159664, | |
| "grad_norm": 0.6401047773492371, | |
| "learning_rate": 3.252238432809368e-05, | |
| "loss": 0.2466, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2555871307849884, | |
| "step": 2375, | |
| "valid_targets_mean": 2559.3, | |
| "valid_targets_min": 244 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "grad_norm": 0.7122232018012691, | |
| "learning_rate": 3.2481494566795716e-05, | |
| "loss": 0.2501, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2375078648328781, | |
| "step": 2380, | |
| "valid_targets_mean": 2202.1, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 2.505252100840336, | |
| "grad_norm": 0.6808348421830116, | |
| "learning_rate": 3.244051917312061e-05, | |
| "loss": 0.2424, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23621997237205505, | |
| "step": 2385, | |
| "valid_targets_mean": 2670.9, | |
| "valid_targets_min": 769 | |
| }, | |
| { | |
| "epoch": 2.5105042016806722, | |
| "grad_norm": 0.6243398513672342, | |
| "learning_rate": 3.239945842819018e-05, | |
| "loss": 0.2523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23079396784305573, | |
| "step": 2390, | |
| "valid_targets_mean": 2769.6, | |
| "valid_targets_min": 1140 | |
| }, | |
| { | |
| "epoch": 2.5157563025210083, | |
| "grad_norm": 0.6608060434787372, | |
| "learning_rate": 3.2358312613711804e-05, | |
| "loss": 0.2547, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.253095418214798, | |
| "step": 2395, | |
| "valid_targets_mean": 2660.6, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 2.5210084033613445, | |
| "grad_norm": 0.7121224716293856, | |
| "learning_rate": 3.231708201197653e-05, | |
| "loss": 0.2408, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21336033940315247, | |
| "step": 2400, | |
| "valid_targets_mean": 2111.9, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 2.5262605042016806, | |
| "grad_norm": 0.6338382678155583, | |
| "learning_rate": 3.227576690585706e-05, | |
| "loss": 0.2496, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2580583095550537, | |
| "step": 2405, | |
| "valid_targets_mean": 3034.8, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 2.5315126050420167, | |
| "grad_norm": 0.7538479941411893, | |
| "learning_rate": 3.223436757880591e-05, | |
| "loss": 0.2508, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22333016991615295, | |
| "step": 2410, | |
| "valid_targets_mean": 1899.9, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 2.536764705882353, | |
| "grad_norm": 0.6566761719653834, | |
| "learning_rate": 3.219288431485337e-05, | |
| "loss": 0.2396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24630621075630188, | |
| "step": 2415, | |
| "valid_targets_mean": 2439.4, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 2.542016806722689, | |
| "grad_norm": 0.6042493534650838, | |
| "learning_rate": 3.2151317398605646e-05, | |
| "loss": 0.2392, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25575879216194153, | |
| "step": 2420, | |
| "valid_targets_mean": 3347.0, | |
| "valid_targets_min": 1460 | |
| }, | |
| { | |
| "epoch": 2.5472689075630255, | |
| "grad_norm": 0.6284939551656498, | |
| "learning_rate": 3.2109667115242815e-05, | |
| "loss": 0.2523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24401450157165527, | |
| "step": 2425, | |
| "valid_targets_mean": 2538.7, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 2.552521008403361, | |
| "grad_norm": 0.6528228615180638, | |
| "learning_rate": 3.206793375051694e-05, | |
| "loss": 0.2495, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25775042176246643, | |
| "step": 2430, | |
| "valid_targets_mean": 2766.4, | |
| "valid_targets_min": 844 | |
| }, | |
| { | |
| "epoch": 2.5577731092436977, | |
| "grad_norm": 0.6809614766621044, | |
| "learning_rate": 3.2026117590750086e-05, | |
| "loss": 0.2415, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25471678376197815, | |
| "step": 2435, | |
| "valid_targets_mean": 2465.3, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 2.5630252100840334, | |
| "grad_norm": 0.7551843328572223, | |
| "learning_rate": 3.198421892283235e-05, | |
| "loss": 0.2356, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2478971630334854, | |
| "step": 2440, | |
| "valid_targets_mean": 2033.8, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 2.56827731092437, | |
| "grad_norm": 0.732748269558122, | |
| "learning_rate": 3.1942238034219885e-05, | |
| "loss": 0.2358, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25034216046333313, | |
| "step": 2445, | |
| "valid_targets_mean": 2231.6, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 2.5735294117647056, | |
| "grad_norm": 0.6778858165535954, | |
| "learning_rate": 3.190017521293296e-05, | |
| "loss": 0.2494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23393598198890686, | |
| "step": 2450, | |
| "valid_targets_mean": 2473.6, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 2.578781512605042, | |
| "grad_norm": 0.6028467351028476, | |
| "learning_rate": 3.185803074755395e-05, | |
| "loss": 0.2458, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2509484887123108, | |
| "step": 2455, | |
| "valid_targets_mean": 2993.6, | |
| "valid_targets_min": 1079 | |
| }, | |
| { | |
| "epoch": 2.5840336134453783, | |
| "grad_norm": 0.7291609586754345, | |
| "learning_rate": 3.1815804927225363e-05, | |
| "loss": 0.2342, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2330750823020935, | |
| "step": 2460, | |
| "valid_targets_mean": 2394.2, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 2.5892857142857144, | |
| "grad_norm": 0.7117035098015625, | |
| "learning_rate": 3.1773498041647876e-05, | |
| "loss": 0.2477, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2390121966600418, | |
| "step": 2465, | |
| "valid_targets_mean": 2322.9, | |
| "valid_targets_min": 972 | |
| }, | |
| { | |
| "epoch": 2.5945378151260505, | |
| "grad_norm": 0.7267971531615649, | |
| "learning_rate": 3.173111038107834e-05, | |
| "loss": 0.2433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24882632493972778, | |
| "step": 2470, | |
| "valid_targets_mean": 2219.9, | |
| "valid_targets_min": 887 | |
| }, | |
| { | |
| "epoch": 2.5997899159663866, | |
| "grad_norm": 0.6553214599651517, | |
| "learning_rate": 3.168864223632776e-05, | |
| "loss": 0.2386, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23804202675819397, | |
| "step": 2475, | |
| "valid_targets_mean": 2536.0, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 2.6050420168067228, | |
| "grad_norm": 0.6758403503740037, | |
| "learning_rate": 3.1646093898759334e-05, | |
| "loss": 0.235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23084652423858643, | |
| "step": 2480, | |
| "valid_targets_mean": 2361.0, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 2.610294117647059, | |
| "grad_norm": 0.7731853525222528, | |
| "learning_rate": 3.160346566028645e-05, | |
| "loss": 0.2566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27315887808799744, | |
| "step": 2485, | |
| "valid_targets_mean": 1826.8, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 2.615546218487395, | |
| "grad_norm": 0.7568845150659382, | |
| "learning_rate": 3.156075781337068e-05, | |
| "loss": 0.2446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26611223816871643, | |
| "step": 2490, | |
| "valid_targets_mean": 2255.1, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 2.620798319327731, | |
| "grad_norm": 0.7862703099000441, | |
| "learning_rate": 3.1517970651019735e-05, | |
| "loss": 0.2466, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26096102595329285, | |
| "step": 2495, | |
| "valid_targets_mean": 2134.9, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 2.6260504201680672, | |
| "grad_norm": 0.67585915869592, | |
| "learning_rate": 3.1475104466785536e-05, | |
| "loss": 0.2462, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26358675956726074, | |
| "step": 2500, | |
| "valid_targets_mean": 2735.0, | |
| "valid_targets_min": 668 | |
| }, | |
| { | |
| "epoch": 2.6313025210084033, | |
| "grad_norm": 0.6329275844175969, | |
| "learning_rate": 3.143215955476211e-05, | |
| "loss": 0.2491, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24213485419750214, | |
| "step": 2505, | |
| "valid_targets_mean": 2604.1, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 2.6365546218487395, | |
| "grad_norm": 0.6225761182265704, | |
| "learning_rate": 3.138913620958365e-05, | |
| "loss": 0.2581, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2655182480812073, | |
| "step": 2510, | |
| "valid_targets_mean": 3134.6, | |
| "valid_targets_min": 1472 | |
| }, | |
| { | |
| "epoch": 2.6418067226890756, | |
| "grad_norm": 0.6693316147755365, | |
| "learning_rate": 3.134603472642244e-05, | |
| "loss": 0.2612, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2521858513355255, | |
| "step": 2515, | |
| "valid_targets_mean": 2876.2, | |
| "valid_targets_min": 772 | |
| }, | |
| { | |
| "epoch": 2.6470588235294117, | |
| "grad_norm": 0.7037886165836804, | |
| "learning_rate": 3.1302855400986866e-05, | |
| "loss": 0.2518, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26977285742759705, | |
| "step": 2520, | |
| "valid_targets_mean": 2489.3, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 2.652310924369748, | |
| "grad_norm": 0.7880145756624581, | |
| "learning_rate": 3.1259598529519336e-05, | |
| "loss": 0.241, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24999578297138214, | |
| "step": 2525, | |
| "valid_targets_mean": 2124.8, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 2.657563025210084, | |
| "grad_norm": 0.6308106007836234, | |
| "learning_rate": 3.121626440879432e-05, | |
| "loss": 0.2391, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2530635595321655, | |
| "step": 2530, | |
| "valid_targets_mean": 2594.6, | |
| "valid_targets_min": 645 | |
| }, | |
| { | |
| "epoch": 2.66281512605042, | |
| "grad_norm": 0.7372282318557066, | |
| "learning_rate": 3.117285333611625e-05, | |
| "loss": 0.2264, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2395913451910019, | |
| "step": 2535, | |
| "valid_targets_mean": 1904.2, | |
| "valid_targets_min": 965 | |
| }, | |
| { | |
| "epoch": 2.668067226890756, | |
| "grad_norm": 0.667012830669792, | |
| "learning_rate": 3.112936560931752e-05, | |
| "loss": 0.2315, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22841016948223114, | |
| "step": 2540, | |
| "valid_targets_mean": 2184.2, | |
| "valid_targets_min": 942 | |
| }, | |
| { | |
| "epoch": 2.6733193277310923, | |
| "grad_norm": 0.6644445077742529, | |
| "learning_rate": 3.108580152675642e-05, | |
| "loss": 0.2501, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23634028434753418, | |
| "step": 2545, | |
| "valid_targets_mean": 2397.2, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.678571428571429, | |
| "grad_norm": 0.7787179904265334, | |
| "learning_rate": 3.104216138731511e-05, | |
| "loss": 0.2383, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2434334009885788, | |
| "step": 2550, | |
| "valid_targets_mean": 2057.8, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 2.6838235294117645, | |
| "grad_norm": 0.6846974855879137, | |
| "learning_rate": 3.0998445490397546e-05, | |
| "loss": 0.2379, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24789969623088837, | |
| "step": 2555, | |
| "valid_targets_mean": 2343.2, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 2.689075630252101, | |
| "grad_norm": 0.7370020291648643, | |
| "learning_rate": 3.095465413592743e-05, | |
| "loss": 0.2442, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25336140394210815, | |
| "step": 2560, | |
| "valid_targets_mean": 2095.1, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 2.6943277310924367, | |
| "grad_norm": 0.6326410487326921, | |
| "learning_rate": 3.091078762434616e-05, | |
| "loss": 0.241, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2503868043422699, | |
| "step": 2565, | |
| "valid_targets_mean": 2962.9, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 2.6995798319327733, | |
| "grad_norm": 0.6774135014050404, | |
| "learning_rate": 3.086684625661081e-05, | |
| "loss": 0.2402, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2455071359872818, | |
| "step": 2570, | |
| "valid_targets_mean": 2381.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 2.7048319327731094, | |
| "grad_norm": 0.6927859104567512, | |
| "learning_rate": 3.082283033419197e-05, | |
| "loss": 0.2543, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29159125685691833, | |
| "step": 2575, | |
| "valid_targets_mean": 2816.8, | |
| "valid_targets_min": 1047 | |
| }, | |
| { | |
| "epoch": 2.7100840336134455, | |
| "grad_norm": 0.6723232233408306, | |
| "learning_rate": 3.077874015907174e-05, | |
| "loss": 0.2481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2541651725769043, | |
| "step": 2580, | |
| "valid_targets_mean": 2508.1, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 2.7153361344537816, | |
| "grad_norm": 0.6173987363991384, | |
| "learning_rate": 3.073457603374167e-05, | |
| "loss": 0.2295, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2286214530467987, | |
| "step": 2585, | |
| "valid_targets_mean": 3016.7, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 2.7205882352941178, | |
| "grad_norm": 0.7180860813557801, | |
| "learning_rate": 3.069033826120066e-05, | |
| "loss": 0.2476, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23617371916770935, | |
| "step": 2590, | |
| "valid_targets_mean": 2155.6, | |
| "valid_targets_min": 805 | |
| }, | |
| { | |
| "epoch": 2.725840336134454, | |
| "grad_norm": 0.6332083889121407, | |
| "learning_rate": 3.064602714495286e-05, | |
| "loss": 0.2303, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2275339514017105, | |
| "step": 2595, | |
| "valid_targets_mean": 2649.5, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 2.73109243697479, | |
| "grad_norm": 0.5824671695502105, | |
| "learning_rate": 3.060164298900565e-05, | |
| "loss": 0.2474, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21272224187850952, | |
| "step": 2600, | |
| "valid_targets_mean": 2831.7, | |
| "valid_targets_min": 1280 | |
| }, | |
| { | |
| "epoch": 2.736344537815126, | |
| "grad_norm": 0.7999352632938964, | |
| "learning_rate": 3.055718609786749e-05, | |
| "loss": 0.2357, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23128217458724976, | |
| "step": 2605, | |
| "valid_targets_mean": 1753.5, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 2.741596638655462, | |
| "grad_norm": 0.6862220018448458, | |
| "learning_rate": 3.051265677654585e-05, | |
| "loss": 0.2388, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24940389394760132, | |
| "step": 2610, | |
| "valid_targets_mean": 2828.1, | |
| "valid_targets_min": 1251 | |
| }, | |
| { | |
| "epoch": 2.7468487394957983, | |
| "grad_norm": 0.7409612251295014, | |
| "learning_rate": 3.0468055330545162e-05, | |
| "loss": 0.2466, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.226776584982872, | |
| "step": 2615, | |
| "valid_targets_mean": 2611.8, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 2.7521008403361344, | |
| "grad_norm": 0.7045597400364605, | |
| "learning_rate": 3.0423382065864653e-05, | |
| "loss": 0.2308, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22663600742816925, | |
| "step": 2620, | |
| "valid_targets_mean": 1957.2, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 2.7573529411764706, | |
| "grad_norm": 0.7198164206409869, | |
| "learning_rate": 3.0378637288996304e-05, | |
| "loss": 0.2404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23351144790649414, | |
| "step": 2625, | |
| "valid_targets_mean": 2081.1, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 2.7626050420168067, | |
| "grad_norm": 0.5873814818670327, | |
| "learning_rate": 3.033382130692269e-05, | |
| "loss": 0.243, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21516209840774536, | |
| "step": 2630, | |
| "valid_targets_mean": 2769.6, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 2.767857142857143, | |
| "grad_norm": 0.6335582852038116, | |
| "learning_rate": 3.0288934427114955e-05, | |
| "loss": 0.2364, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2048596739768982, | |
| "step": 2635, | |
| "valid_targets_mean": 2169.6, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 2.773109243697479, | |
| "grad_norm": 0.67104459281815, | |
| "learning_rate": 3.0243976957530602e-05, | |
| "loss": 0.2512, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2463095635175705, | |
| "step": 2640, | |
| "valid_targets_mean": 2423.3, | |
| "valid_targets_min": 702 | |
| }, | |
| { | |
| "epoch": 2.778361344537815, | |
| "grad_norm": 0.6291113433855628, | |
| "learning_rate": 3.0198949206611475e-05, | |
| "loss": 0.2302, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.247708261013031, | |
| "step": 2645, | |
| "valid_targets_mean": 2842.7, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 2.783613445378151, | |
| "grad_norm": 0.7183619458186878, | |
| "learning_rate": 3.0153851483281582e-05, | |
| "loss": 0.2478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25846749544143677, | |
| "step": 2650, | |
| "valid_targets_mean": 2008.1, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 2.7888655462184873, | |
| "grad_norm": 0.6296111197357144, | |
| "learning_rate": 3.010868409694499e-05, | |
| "loss": 0.2311, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2236376702785492, | |
| "step": 2655, | |
| "valid_targets_mean": 2598.8, | |
| "valid_targets_min": 1260 | |
| }, | |
| { | |
| "epoch": 2.7941176470588234, | |
| "grad_norm": 0.6152933998116437, | |
| "learning_rate": 3.0063447357483712e-05, | |
| "loss": 0.2306, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22628861665725708, | |
| "step": 2660, | |
| "valid_targets_mean": 2571.6, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 2.79936974789916, | |
| "grad_norm": 0.6612164055765357, | |
| "learning_rate": 3.0018141575255587e-05, | |
| "loss": 0.2464, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22041192650794983, | |
| "step": 2665, | |
| "valid_targets_mean": 2249.2, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 2.8046218487394956, | |
| "grad_norm": 0.7274917143243959, | |
| "learning_rate": 2.9972767061092107e-05, | |
| "loss": 0.2358, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2216116338968277, | |
| "step": 2670, | |
| "valid_targets_mean": 1970.0, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 2.809873949579832, | |
| "grad_norm": 0.6906029702041456, | |
| "learning_rate": 2.9927324126296352e-05, | |
| "loss": 0.2523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23099112510681152, | |
| "step": 2675, | |
| "valid_targets_mean": 2226.4, | |
| "valid_targets_min": 786 | |
| }, | |
| { | |
| "epoch": 2.815126050420168, | |
| "grad_norm": 0.6602585381068012, | |
| "learning_rate": 2.98818130826408e-05, | |
| "loss": 0.2395, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23484909534454346, | |
| "step": 2680, | |
| "valid_targets_mean": 2583.8, | |
| "valid_targets_min": 584 | |
| }, | |
| { | |
| "epoch": 2.8203781512605044, | |
| "grad_norm": 0.700101600296082, | |
| "learning_rate": 2.9836234242365204e-05, | |
| "loss": 0.2446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26351702213287354, | |
| "step": 2685, | |
| "valid_targets_mean": 2604.2, | |
| "valid_targets_min": 1181 | |
| }, | |
| { | |
| "epoch": 2.82563025210084, | |
| "grad_norm": 1.1113414468630831, | |
| "learning_rate": 2.979058791817447e-05, | |
| "loss": 0.2361, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2611852288246155, | |
| "step": 2690, | |
| "valid_targets_mean": 2256.4, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 2.8308823529411766, | |
| "grad_norm": 0.6092742628805521, | |
| "learning_rate": 2.974487442323647e-05, | |
| "loss": 0.2433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22087091207504272, | |
| "step": 2695, | |
| "valid_targets_mean": 2595.6, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 2.8361344537815127, | |
| "grad_norm": 0.6990782326556737, | |
| "learning_rate": 2.9699094071179938e-05, | |
| "loss": 0.2424, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24499669671058655, | |
| "step": 2700, | |
| "valid_targets_mean": 2403.4, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 2.841386554621849, | |
| "grad_norm": 0.6311106363395269, | |
| "learning_rate": 2.9653247176092286e-05, | |
| "loss": 0.2526, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23955629765987396, | |
| "step": 2705, | |
| "valid_targets_mean": 3173.2, | |
| "valid_targets_min": 1337 | |
| }, | |
| { | |
| "epoch": 2.846638655462185, | |
| "grad_norm": 0.6466020830134367, | |
| "learning_rate": 2.9607334052517466e-05, | |
| "loss": 0.2356, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25517964363098145, | |
| "step": 2710, | |
| "valid_targets_mean": 2581.2, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 2.851890756302521, | |
| "grad_norm": 0.6364448114027271, | |
| "learning_rate": 2.9561355015453803e-05, | |
| "loss": 0.2428, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23851974308490753, | |
| "step": 2715, | |
| "valid_targets_mean": 2660.0, | |
| "valid_targets_min": 514 | |
| }, | |
| { | |
| "epoch": 2.857142857142857, | |
| "grad_norm": 0.6924312748250914, | |
| "learning_rate": 2.9515310380351847e-05, | |
| "loss": 0.2427, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25390565395355225, | |
| "step": 2720, | |
| "valid_targets_mean": 2458.5, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 2.8623949579831933, | |
| "grad_norm": 0.6680820681499023, | |
| "learning_rate": 2.9469200463112186e-05, | |
| "loss": 0.2322, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22375771403312683, | |
| "step": 2725, | |
| "valid_targets_mean": 2688.9, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 2.8676470588235294, | |
| "grad_norm": 0.6088973094421416, | |
| "learning_rate": 2.9423025580083298e-05, | |
| "loss": 0.2306, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20682302117347717, | |
| "step": 2730, | |
| "valid_targets_mean": 2755.1, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 2.8728991596638656, | |
| "grad_norm": 0.6859266823062403, | |
| "learning_rate": 2.937678604805938e-05, | |
| "loss": 0.2342, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21406525373458862, | |
| "step": 2735, | |
| "valid_targets_mean": 2067.5, | |
| "valid_targets_min": 1166 | |
| }, | |
| { | |
| "epoch": 2.8781512605042017, | |
| "grad_norm": 0.6634535239376516, | |
| "learning_rate": 2.9330482184278168e-05, | |
| "loss": 0.2381, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24567797780036926, | |
| "step": 2740, | |
| "valid_targets_mean": 2457.8, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 2.883403361344538, | |
| "grad_norm": 0.6340089258781645, | |
| "learning_rate": 2.9284114306418758e-05, | |
| "loss": 0.2389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2264363169670105, | |
| "step": 2745, | |
| "valid_targets_mean": 2580.6, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 2.888655462184874, | |
| "grad_norm": 0.6682756597647231, | |
| "learning_rate": 2.923768273259942e-05, | |
| "loss": 0.2342, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2170996367931366, | |
| "step": 2750, | |
| "valid_targets_mean": 2042.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 2.89390756302521, | |
| "grad_norm": 0.6425815374610044, | |
| "learning_rate": 2.919118778137546e-05, | |
| "loss": 0.2371, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20629505813121796, | |
| "step": 2755, | |
| "valid_targets_mean": 2251.1, | |
| "valid_targets_min": 779 | |
| }, | |
| { | |
| "epoch": 2.899159663865546, | |
| "grad_norm": 0.6998165427866984, | |
| "learning_rate": 2.9144629771736954e-05, | |
| "loss": 0.2626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.27105337381362915, | |
| "step": 2760, | |
| "valid_targets_mean": 2532.9, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 2.9044117647058822, | |
| "grad_norm": 0.6850527077895415, | |
| "learning_rate": 2.9098009023106645e-05, | |
| "loss": 0.2419, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2457822561264038, | |
| "step": 2765, | |
| "valid_targets_mean": 2170.1, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 2.9096638655462184, | |
| "grad_norm": 0.5591265461314032, | |
| "learning_rate": 2.9051325855337684e-05, | |
| "loss": 0.2417, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23795005679130554, | |
| "step": 2770, | |
| "valid_targets_mean": 3337.8, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 2.9149159663865545, | |
| "grad_norm": 0.7253204297962602, | |
| "learning_rate": 2.90045805887115e-05, | |
| "loss": 0.2298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26152512431144714, | |
| "step": 2775, | |
| "valid_targets_mean": 2250.4, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 2.9201680672268906, | |
| "grad_norm": 0.6320606127718346, | |
| "learning_rate": 2.8957773543935518e-05, | |
| "loss": 0.231, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22176562249660492, | |
| "step": 2780, | |
| "valid_targets_mean": 2631.1, | |
| "valid_targets_min": 943 | |
| }, | |
| { | |
| "epoch": 2.9254201680672267, | |
| "grad_norm": 0.6808320182720331, | |
| "learning_rate": 2.8910905042141046e-05, | |
| "loss": 0.2378, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2327442467212677, | |
| "step": 2785, | |
| "valid_targets_mean": 2285.2, | |
| "valid_targets_min": 836 | |
| }, | |
| { | |
| "epoch": 2.9306722689075633, | |
| "grad_norm": 0.7718275961115919, | |
| "learning_rate": 2.8863975404881022e-05, | |
| "loss": 0.2489, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23095104098320007, | |
| "step": 2790, | |
| "valid_targets_mean": 1883.9, | |
| "valid_targets_min": 836 | |
| }, | |
| { | |
| "epoch": 2.935924369747899, | |
| "grad_norm": 0.6901725161531642, | |
| "learning_rate": 2.881698495412781e-05, | |
| "loss": 0.2383, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23618492484092712, | |
| "step": 2795, | |
| "valid_targets_mean": 2379.7, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 2.9411764705882355, | |
| "grad_norm": 0.6510522678878954, | |
| "learning_rate": 2.8769934012271004e-05, | |
| "loss": 0.2418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26842981576919556, | |
| "step": 2800, | |
| "valid_targets_mean": 2957.8, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 2.946428571428571, | |
| "grad_norm": 0.690393878481563, | |
| "learning_rate": 2.8722822902115228e-05, | |
| "loss": 0.2301, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2306583821773529, | |
| "step": 2805, | |
| "valid_targets_mean": 2844.0, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 2.9516806722689077, | |
| "grad_norm": 0.6881209587839068, | |
| "learning_rate": 2.8675651946877884e-05, | |
| "loss": 0.2508, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26074153184890747, | |
| "step": 2810, | |
| "valid_targets_mean": 2694.2, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 2.956932773109244, | |
| "grad_norm": 0.5892580385058059, | |
| "learning_rate": 2.862842147018696e-05, | |
| "loss": 0.2448, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2656477689743042, | |
| "step": 2815, | |
| "valid_targets_mean": 3714.4, | |
| "valid_targets_min": 1073 | |
| }, | |
| { | |
| "epoch": 2.96218487394958, | |
| "grad_norm": 0.6824881929007802, | |
| "learning_rate": 2.858113179607881e-05, | |
| "loss": 0.2421, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23978950083255768, | |
| "step": 2820, | |
| "valid_targets_mean": 2137.4, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 2.967436974789916, | |
| "grad_norm": 0.6773521974686484, | |
| "learning_rate": 2.8533783248995938e-05, | |
| "loss": 0.2435, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2452758550643921, | |
| "step": 2825, | |
| "valid_targets_mean": 2297.1, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 2.972689075630252, | |
| "grad_norm": 0.7308770430795042, | |
| "learning_rate": 2.848637615378475e-05, | |
| "loss": 0.2534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24922780692577362, | |
| "step": 2830, | |
| "valid_targets_mean": 2439.7, | |
| "valid_targets_min": 1055 | |
| }, | |
| { | |
| "epoch": 2.9779411764705883, | |
| "grad_norm": 0.6031497666783858, | |
| "learning_rate": 2.8438910835693314e-05, | |
| "loss": 0.2387, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2477431297302246, | |
| "step": 2835, | |
| "valid_targets_mean": 2889.0, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.9831932773109244, | |
| "grad_norm": 0.7153376315518947, | |
| "learning_rate": 2.8391387620369192e-05, | |
| "loss": 0.2448, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.25468868017196655, | |
| "step": 2840, | |
| "valid_targets_mean": 2295.2, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 2.9884453781512605, | |
| "grad_norm": 0.6969521384499777, | |
| "learning_rate": 2.834380683385714e-05, | |
| "loss": 0.2429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.220721036195755, | |
| "step": 2845, | |
| "valid_targets_mean": 2244.9, | |
| "valid_targets_min": 1073 | |
| }, | |
| { | |
| "epoch": 2.9936974789915967, | |
| "grad_norm": 0.7414684846578835, | |
| "learning_rate": 2.8296168802596892e-05, | |
| "loss": 0.2391, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23130197823047638, | |
| "step": 2850, | |
| "valid_targets_mean": 1981.4, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 2.9989495798319328, | |
| "grad_norm": 0.7478026755802675, | |
| "learning_rate": 2.8248473853420933e-05, | |
| "loss": 0.2411, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2313975691795349, | |
| "step": 2855, | |
| "valid_targets_mean": 1840.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 3.004201680672269, | |
| "grad_norm": 0.6438849416484059, | |
| "learning_rate": 2.8200722313552247e-05, | |
| "loss": 0.2094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20280009508132935, | |
| "step": 2860, | |
| "valid_targets_mean": 2363.8, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 3.009453781512605, | |
| "grad_norm": 0.6700105099862343, | |
| "learning_rate": 2.8152914510602073e-05, | |
| "loss": 0.2048, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21491551399230957, | |
| "step": 2865, | |
| "valid_targets_mean": 2595.0, | |
| "valid_targets_min": 557 | |
| }, | |
| { | |
| "epoch": 3.014705882352941, | |
| "grad_norm": 0.6973369056653054, | |
| "learning_rate": 2.810505077256765e-05, | |
| "loss": 0.2078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19096165895462036, | |
| "step": 2870, | |
| "valid_targets_mean": 2309.4, | |
| "valid_targets_min": 666 | |
| }, | |
| { | |
| "epoch": 3.0199579831932772, | |
| "grad_norm": 0.7391770624716282, | |
| "learning_rate": 2.8057131427829983e-05, | |
| "loss": 0.2055, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22697217762470245, | |
| "step": 2875, | |
| "valid_targets_mean": 2748.4, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 3.0252100840336134, | |
| "grad_norm": 0.7323248663686152, | |
| "learning_rate": 2.800915680515157e-05, | |
| "loss": 0.2135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21774935722351074, | |
| "step": 2880, | |
| "valid_targets_mean": 2176.5, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 3.0304621848739495, | |
| "grad_norm": 0.7293920324073403, | |
| "learning_rate": 2.7961127233674167e-05, | |
| "loss": 0.2025, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19380387663841248, | |
| "step": 2885, | |
| "valid_targets_mean": 2121.6, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 3.0357142857142856, | |
| "grad_norm": 0.7214083612213081, | |
| "learning_rate": 2.7913043042916513e-05, | |
| "loss": 0.2057, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2048916071653366, | |
| "step": 2890, | |
| "valid_targets_mean": 2635.6, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 3.0409663865546217, | |
| "grad_norm": 0.6608114815054316, | |
| "learning_rate": 2.786490456277208e-05, | |
| "loss": 0.2103, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21670818328857422, | |
| "step": 2895, | |
| "valid_targets_mean": 2556.7, | |
| "valid_targets_min": 762 | |
| }, | |
| { | |
| "epoch": 3.046218487394958, | |
| "grad_norm": 0.66860146042256, | |
| "learning_rate": 2.7816712123506796e-05, | |
| "loss": 0.1965, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20460188388824463, | |
| "step": 2900, | |
| "valid_targets_mean": 2566.9, | |
| "valid_targets_min": 420 | |
| }, | |
| { | |
| "epoch": 3.051470588235294, | |
| "grad_norm": 0.7008326884337155, | |
| "learning_rate": 2.776846605575681e-05, | |
| "loss": 0.2132, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.201397106051445, | |
| "step": 2905, | |
| "valid_targets_mean": 2224.4, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 3.05672268907563, | |
| "grad_norm": 0.7720626933651689, | |
| "learning_rate": 2.772016669052618e-05, | |
| "loss": 0.2133, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2363039255142212, | |
| "step": 2910, | |
| "valid_targets_mean": 2557.2, | |
| "valid_targets_min": 1106 | |
| }, | |
| { | |
| "epoch": 3.0619747899159666, | |
| "grad_norm": 0.7021404124625371, | |
| "learning_rate": 2.767181435918464e-05, | |
| "loss": 0.1966, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20766621828079224, | |
| "step": 2915, | |
| "valid_targets_mean": 2482.2, | |
| "valid_targets_min": 855 | |
| }, | |
| { | |
| "epoch": 3.0672268907563027, | |
| "grad_norm": 0.6343762530485658, | |
| "learning_rate": 2.7623409393465298e-05, | |
| "loss": 0.2071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19036218523979187, | |
| "step": 2920, | |
| "valid_targets_mean": 3076.4, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 3.072478991596639, | |
| "grad_norm": 0.7035625376410721, | |
| "learning_rate": 2.7574952125462386e-05, | |
| "loss": 0.1964, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18567273020744324, | |
| "step": 2925, | |
| "valid_targets_mean": 2296.6, | |
| "valid_targets_min": 467 | |
| }, | |
| { | |
| "epoch": 3.077731092436975, | |
| "grad_norm": 0.6382148687024268, | |
| "learning_rate": 2.7526442887628962e-05, | |
| "loss": 0.2128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19346240162849426, | |
| "step": 2930, | |
| "valid_targets_mean": 2762.5, | |
| "valid_targets_min": 937 | |
| }, | |
| { | |
| "epoch": 3.082983193277311, | |
| "grad_norm": 0.7008906642109496, | |
| "learning_rate": 2.7477882012774633e-05, | |
| "loss": 0.2187, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21867555379867554, | |
| "step": 2935, | |
| "valid_targets_mean": 2581.9, | |
| "valid_targets_min": 1101 | |
| }, | |
| { | |
| "epoch": 3.088235294117647, | |
| "grad_norm": 0.708331450124728, | |
| "learning_rate": 2.742926983406328e-05, | |
| "loss": 0.2003, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22323349118232727, | |
| "step": 2940, | |
| "valid_targets_mean": 2543.1, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 3.0934873949579833, | |
| "grad_norm": 0.7455067327316252, | |
| "learning_rate": 2.7380606685010767e-05, | |
| "loss": 0.2035, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21157895028591156, | |
| "step": 2945, | |
| "valid_targets_mean": 2247.4, | |
| "valid_targets_min": 701 | |
| }, | |
| { | |
| "epoch": 3.0987394957983194, | |
| "grad_norm": 0.6753873272175467, | |
| "learning_rate": 2.7331892899482654e-05, | |
| "loss": 0.2167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2144676148891449, | |
| "step": 2950, | |
| "valid_targets_mean": 2689.9, | |
| "valid_targets_min": 767 | |
| }, | |
| { | |
| "epoch": 3.1039915966386555, | |
| "grad_norm": 0.6865230940428809, | |
| "learning_rate": 2.7283128811691893e-05, | |
| "loss": 0.2062, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2236787974834442, | |
| "step": 2955, | |
| "valid_targets_mean": 2606.0, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 3.1092436974789917, | |
| "grad_norm": 0.6404580815887296, | |
| "learning_rate": 2.7234314756196564e-05, | |
| "loss": 0.1962, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.197324737906456, | |
| "step": 2960, | |
| "valid_targets_mean": 2785.6, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 3.1144957983193278, | |
| "grad_norm": 0.7140771160670712, | |
| "learning_rate": 2.718545106789755e-05, | |
| "loss": 0.2037, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19954288005828857, | |
| "step": 2965, | |
| "valid_targets_mean": 2234.5, | |
| "valid_targets_min": 872 | |
| }, | |
| { | |
| "epoch": 3.119747899159664, | |
| "grad_norm": 0.6975932021440397, | |
| "learning_rate": 2.713653808203626e-05, | |
| "loss": 0.1975, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2170845866203308, | |
| "step": 2970, | |
| "valid_targets_mean": 2639.3, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 3.125, | |
| "grad_norm": 0.7113689274916534, | |
| "learning_rate": 2.7087576134192305e-05, | |
| "loss": 0.2064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21494010090827942, | |
| "step": 2975, | |
| "valid_targets_mean": 2403.4, | |
| "valid_targets_min": 1159 | |
| }, | |
| { | |
| "epoch": 3.130252100840336, | |
| "grad_norm": 0.6838797403752815, | |
| "learning_rate": 2.7038565560281252e-05, | |
| "loss": 0.2072, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22361846268177032, | |
| "step": 2980, | |
| "valid_targets_mean": 2727.6, | |
| "valid_targets_min": 1111 | |
| }, | |
| { | |
| "epoch": 3.1355042016806722, | |
| "grad_norm": 0.7301967352422043, | |
| "learning_rate": 2.6989506696552224e-05, | |
| "loss": 0.2226, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24528783559799194, | |
| "step": 2985, | |
| "valid_targets_mean": 2384.6, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 3.1407563025210083, | |
| "grad_norm": 0.7300962730703254, | |
| "learning_rate": 2.6940399879585685e-05, | |
| "loss": 0.2226, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24462322890758514, | |
| "step": 2990, | |
| "valid_targets_mean": 2379.9, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 3.1460084033613445, | |
| "grad_norm": 0.6314374147339737, | |
| "learning_rate": 2.6891245446291093e-05, | |
| "loss": 0.1947, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20750781893730164, | |
| "step": 2995, | |
| "valid_targets_mean": 3385.5, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 3.1512605042016806, | |
| "grad_norm": 0.6709712599868535, | |
| "learning_rate": 2.6842043733904578e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19627419114112854, | |
| "step": 3000, | |
| "valid_targets_mean": 2626.9, | |
| "valid_targets_min": 1022 | |
| }, | |
| { | |
| "epoch": 3.1565126050420167, | |
| "grad_norm": 0.7079636790584424, | |
| "learning_rate": 2.6792795079986634e-05, | |
| "loss": 0.2089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21414107084274292, | |
| "step": 3005, | |
| "valid_targets_mean": 2522.1, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 3.161764705882353, | |
| "grad_norm": 0.8013316017451616, | |
| "learning_rate": 2.6743499822419833e-05, | |
| "loss": 0.2065, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21229714155197144, | |
| "step": 3010, | |
| "valid_targets_mean": 1945.2, | |
| "valid_targets_min": 1078 | |
| }, | |
| { | |
| "epoch": 3.167016806722689, | |
| "grad_norm": 0.7573033558532629, | |
| "learning_rate": 2.669415829940646e-05, | |
| "loss": 0.2004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22666743397712708, | |
| "step": 3015, | |
| "valid_targets_mean": 2352.1, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 3.172268907563025, | |
| "grad_norm": 0.6967416593034284, | |
| "learning_rate": 2.6644770849466226e-05, | |
| "loss": 0.2107, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19799959659576416, | |
| "step": 3020, | |
| "valid_targets_mean": 2473.9, | |
| "valid_targets_min": 926 | |
| }, | |
| { | |
| "epoch": 3.177521008403361, | |
| "grad_norm": 0.6958940312721166, | |
| "learning_rate": 2.6595337811433925e-05, | |
| "loss": 0.2213, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21963286399841309, | |
| "step": 3025, | |
| "valid_targets_mean": 2855.2, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 3.1827731092436973, | |
| "grad_norm": 0.6644738321935247, | |
| "learning_rate": 2.654585952445714e-05, | |
| "loss": 0.2058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19399645924568176, | |
| "step": 3030, | |
| "valid_targets_mean": 2638.8, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 3.1880252100840334, | |
| "grad_norm": 0.7306469351103843, | |
| "learning_rate": 2.6496336327993882e-05, | |
| "loss": 0.2144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20161199569702148, | |
| "step": 3035, | |
| "valid_targets_mean": 2136.2, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 3.19327731092437, | |
| "grad_norm": 0.7466263828328548, | |
| "learning_rate": 2.6446768561810264e-05, | |
| "loss": 0.209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20817327499389648, | |
| "step": 3040, | |
| "valid_targets_mean": 2153.1, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 3.198529411764706, | |
| "grad_norm": 0.6197811979812847, | |
| "learning_rate": 2.6397156565978198e-05, | |
| "loss": 0.2048, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18749669194221497, | |
| "step": 3045, | |
| "valid_targets_mean": 2755.1, | |
| "valid_targets_min": 823 | |
| }, | |
| { | |
| "epoch": 3.203781512605042, | |
| "grad_norm": 0.6980373354580599, | |
| "learning_rate": 2.6347500680873043e-05, | |
| "loss": 0.2155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20377440750598907, | |
| "step": 3050, | |
| "valid_targets_mean": 2491.8, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 3.2090336134453783, | |
| "grad_norm": 0.7085865520639477, | |
| "learning_rate": 2.6297801247171258e-05, | |
| "loss": 0.2129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2216644287109375, | |
| "step": 3055, | |
| "valid_targets_mean": 2806.9, | |
| "valid_targets_min": 1071 | |
| }, | |
| { | |
| "epoch": 3.2142857142857144, | |
| "grad_norm": 0.7205089790035474, | |
| "learning_rate": 2.6248058605848097e-05, | |
| "loss": 0.2212, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22407683730125427, | |
| "step": 3060, | |
| "valid_targets_mean": 2472.9, | |
| "valid_targets_min": 769 | |
| }, | |
| { | |
| "epoch": 3.2195378151260505, | |
| "grad_norm": 0.737404594127863, | |
| "learning_rate": 2.6198273098175234e-05, | |
| "loss": 0.2204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20326296985149384, | |
| "step": 3065, | |
| "valid_targets_mean": 2019.1, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 3.2247899159663866, | |
| "grad_norm": 0.6073130718222292, | |
| "learning_rate": 2.614844506571844e-05, | |
| "loss": 0.2108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20230062305927277, | |
| "step": 3070, | |
| "valid_targets_mean": 3304.4, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.2300420168067228, | |
| "grad_norm": 0.7085944292327786, | |
| "learning_rate": 2.609857485033525e-05, | |
| "loss": 0.215, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20458903908729553, | |
| "step": 3075, | |
| "valid_targets_mean": 2500.9, | |
| "valid_targets_min": 902 | |
| }, | |
| { | |
| "epoch": 3.235294117647059, | |
| "grad_norm": 0.7171287135570944, | |
| "learning_rate": 2.6048662794172595e-05, | |
| "loss": 0.2095, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20819462835788727, | |
| "step": 3080, | |
| "valid_targets_mean": 2273.6, | |
| "valid_targets_min": 787 | |
| }, | |
| { | |
| "epoch": 3.240546218487395, | |
| "grad_norm": 0.7241451356443119, | |
| "learning_rate": 2.599870923966447e-05, | |
| "loss": 0.2064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2197907567024231, | |
| "step": 3085, | |
| "valid_targets_mean": 2637.4, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 3.245798319327731, | |
| "grad_norm": 0.6305118521703866, | |
| "learning_rate": 2.5948714529529565e-05, | |
| "loss": 0.191, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18660598993301392, | |
| "step": 3090, | |
| "valid_targets_mean": 3157.4, | |
| "valid_targets_min": 1567 | |
| }, | |
| { | |
| "epoch": 3.2510504201680672, | |
| "grad_norm": 0.6893413514689206, | |
| "learning_rate": 2.5898679006768958e-05, | |
| "loss": 0.2081, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19289371371269226, | |
| "step": 3095, | |
| "valid_targets_mean": 2322.9, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 3.2563025210084033, | |
| "grad_norm": 0.7101789326050002, | |
| "learning_rate": 2.5848603014663702e-05, | |
| "loss": 0.2077, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2187972515821457, | |
| "step": 3100, | |
| "valid_targets_mean": 2543.6, | |
| "valid_targets_min": 1024 | |
| }, | |
| { | |
| "epoch": 3.2615546218487395, | |
| "grad_norm": 0.9990787057867775, | |
| "learning_rate": 2.579848689677252e-05, | |
| "loss": 0.2064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21138152480125427, | |
| "step": 3105, | |
| "valid_targets_mean": 2269.6, | |
| "valid_targets_min": 982 | |
| }, | |
| { | |
| "epoch": 3.2668067226890756, | |
| "grad_norm": 0.8018203500723405, | |
| "learning_rate": 2.5748330996929414e-05, | |
| "loss": 0.2023, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20356349647045135, | |
| "step": 3110, | |
| "valid_targets_mean": 1958.8, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 3.2720588235294117, | |
| "grad_norm": 0.6603494412466163, | |
| "learning_rate": 2.5698135659241334e-05, | |
| "loss": 0.2031, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18327882885932922, | |
| "step": 3115, | |
| "valid_targets_mean": 2493.4, | |
| "valid_targets_min": 626 | |
| }, | |
| { | |
| "epoch": 3.277310924369748, | |
| "grad_norm": 0.762235597580697, | |
| "learning_rate": 2.564790122808579e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20926137268543243, | |
| "step": 3120, | |
| "valid_targets_mean": 2595.1, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 3.282563025210084, | |
| "grad_norm": 0.6535295283093114, | |
| "learning_rate": 2.5597628048108504e-05, | |
| "loss": 0.2168, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19361212849617004, | |
| "step": 3125, | |
| "valid_targets_mean": 2594.5, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 3.28781512605042, | |
| "grad_norm": 0.7359578809709093, | |
| "learning_rate": 2.5547316464221058e-05, | |
| "loss": 0.2171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24306191504001617, | |
| "step": 3130, | |
| "valid_targets_mean": 2595.2, | |
| "valid_targets_min": 762 | |
| }, | |
| { | |
| "epoch": 3.293067226890756, | |
| "grad_norm": 0.7555698425148304, | |
| "learning_rate": 2.5496966821598493e-05, | |
| "loss": 0.2028, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18649283051490784, | |
| "step": 3135, | |
| "valid_targets_mean": 2025.6, | |
| "valid_targets_min": 1099 | |
| }, | |
| { | |
| "epoch": 3.2983193277310923, | |
| "grad_norm": 0.6867271312637763, | |
| "learning_rate": 2.5446579465676977e-05, | |
| "loss": 0.1993, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2280460000038147, | |
| "step": 3140, | |
| "valid_targets_mean": 2837.4, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 3.3035714285714284, | |
| "grad_norm": 0.7890879070663028, | |
| "learning_rate": 2.5396154742151407e-05, | |
| "loss": 0.2058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24599486589431763, | |
| "step": 3145, | |
| "valid_targets_mean": 2421.9, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 3.3088235294117645, | |
| "grad_norm": 0.7043690415607949, | |
| "learning_rate": 2.5345692996973075e-05, | |
| "loss": 0.2029, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20645210146903992, | |
| "step": 3150, | |
| "valid_targets_mean": 2250.8, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 3.314075630252101, | |
| "grad_norm": 0.6920998160391696, | |
| "learning_rate": 2.5295194576347233e-05, | |
| "loss": 0.2121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22358179092407227, | |
| "step": 3155, | |
| "valid_targets_mean": 2739.9, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 3.3193277310924367, | |
| "grad_norm": 0.8209408504141388, | |
| "learning_rate": 2.5244659826730774e-05, | |
| "loss": 0.2082, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22088845074176788, | |
| "step": 3160, | |
| "valid_targets_mean": 2667.9, | |
| "valid_targets_min": 696 | |
| }, | |
| { | |
| "epoch": 3.3245798319327733, | |
| "grad_norm": 0.73587278467102, | |
| "learning_rate": 2.519408909482984e-05, | |
| "loss": 0.2101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20288720726966858, | |
| "step": 3165, | |
| "valid_targets_mean": 2124.1, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 3.3298319327731094, | |
| "grad_norm": 0.6501152833208236, | |
| "learning_rate": 2.5143482727597427e-05, | |
| "loss": 0.2008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20123985409736633, | |
| "step": 3170, | |
| "valid_targets_mean": 3260.6, | |
| "valid_targets_min": 1042 | |
| }, | |
| { | |
| "epoch": 3.3350840336134455, | |
| "grad_norm": 0.6791329217706075, | |
| "learning_rate": 2.509284107223102e-05, | |
| "loss": 0.1986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21437102556228638, | |
| "step": 3175, | |
| "valid_targets_mean": 2677.3, | |
| "valid_targets_min": 1019 | |
| }, | |
| { | |
| "epoch": 3.3403361344537816, | |
| "grad_norm": 0.8312153231004512, | |
| "learning_rate": 2.50421644761702e-05, | |
| "loss": 0.2154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2082928568124771, | |
| "step": 3180, | |
| "valid_targets_mean": 1766.3, | |
| "valid_targets_min": 947 | |
| }, | |
| { | |
| "epoch": 3.3455882352941178, | |
| "grad_norm": 0.6346713038488362, | |
| "learning_rate": 2.499145328709429e-05, | |
| "loss": 0.2047, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2058856189250946, | |
| "step": 3185, | |
| "valid_targets_mean": 2967.8, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 3.350840336134454, | |
| "grad_norm": 0.7841149043918871, | |
| "learning_rate": 2.4940707852919924e-05, | |
| "loss": 0.2103, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21573182940483093, | |
| "step": 3190, | |
| "valid_targets_mean": 1853.8, | |
| "valid_targets_min": 1039 | |
| }, | |
| { | |
| "epoch": 3.35609243697479, | |
| "grad_norm": 0.7289604889723915, | |
| "learning_rate": 2.488992852179868e-05, | |
| "loss": 0.2001, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21095168590545654, | |
| "step": 3195, | |
| "valid_targets_mean": 2192.4, | |
| "valid_targets_min": 1030 | |
| }, | |
| { | |
| "epoch": 3.361344537815126, | |
| "grad_norm": 0.7426897994171837, | |
| "learning_rate": 2.483911564211472e-05, | |
| "loss": 0.2013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2094847857952118, | |
| "step": 3200, | |
| "valid_targets_mean": 2182.3, | |
| "valid_targets_min": 1090 | |
| }, | |
| { | |
| "epoch": 3.366596638655462, | |
| "grad_norm": 0.6905790936704768, | |
| "learning_rate": 2.4788269562482355e-05, | |
| "loss": 0.2038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18250685930252075, | |
| "step": 3205, | |
| "valid_targets_mean": 2181.2, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 3.3718487394957983, | |
| "grad_norm": 0.6711044771695471, | |
| "learning_rate": 2.4737390631743675e-05, | |
| "loss": 0.217, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21213862299919128, | |
| "step": 3210, | |
| "valid_targets_mean": 2841.9, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 3.3771008403361344, | |
| "grad_norm": 0.7172706019446755, | |
| "learning_rate": 2.4686479198966146e-05, | |
| "loss": 0.2113, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21013203263282776, | |
| "step": 3215, | |
| "valid_targets_mean": 2435.9, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 3.3823529411764706, | |
| "grad_norm": 0.746681881677102, | |
| "learning_rate": 2.463553561344024e-05, | |
| "loss": 0.2193, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2209860384464264, | |
| "step": 3220, | |
| "valid_targets_mean": 2233.9, | |
| "valid_targets_min": 929 | |
| }, | |
| { | |
| "epoch": 3.3876050420168067, | |
| "grad_norm": 0.7545727322641577, | |
| "learning_rate": 2.4584560224677013e-05, | |
| "loss": 0.2176, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2379944920539856, | |
| "step": 3225, | |
| "valid_targets_mean": 2649.4, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 3.392857142857143, | |
| "grad_norm": 0.6782835854882808, | |
| "learning_rate": 2.4533553382405702e-05, | |
| "loss": 0.211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20617589354515076, | |
| "step": 3230, | |
| "valid_targets_mean": 2802.6, | |
| "valid_targets_min": 882 | |
| }, | |
| { | |
| "epoch": 3.398109243697479, | |
| "grad_norm": 0.7578767328316001, | |
| "learning_rate": 2.4482515436571365e-05, | |
| "loss": 0.2172, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22788771986961365, | |
| "step": 3235, | |
| "valid_targets_mean": 2326.2, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 3.403361344537815, | |
| "grad_norm": 0.6541735808021184, | |
| "learning_rate": 2.4431446737332437e-05, | |
| "loss": 0.2121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20903077721595764, | |
| "step": 3240, | |
| "valid_targets_mean": 3203.4, | |
| "valid_targets_min": 1003 | |
| }, | |
| { | |
| "epoch": 3.408613445378151, | |
| "grad_norm": 0.7811038952636883, | |
| "learning_rate": 2.438034763505833e-05, | |
| "loss": 0.2166, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21477511525154114, | |
| "step": 3245, | |
| "valid_targets_mean": 1979.8, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 3.4138655462184873, | |
| "grad_norm": 0.6625538840028757, | |
| "learning_rate": 2.432921848032707e-05, | |
| "loss": 0.2044, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2320900559425354, | |
| "step": 3250, | |
| "valid_targets_mean": 2910.2, | |
| "valid_targets_min": 1315 | |
| }, | |
| { | |
| "epoch": 3.4191176470588234, | |
| "grad_norm": 0.7485969796152663, | |
| "learning_rate": 2.427805962392286e-05, | |
| "loss": 0.2025, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20698726177215576, | |
| "step": 3255, | |
| "valid_targets_mean": 1875.1, | |
| "valid_targets_min": 863 | |
| }, | |
| { | |
| "epoch": 3.4243697478991595, | |
| "grad_norm": 0.7348995050217967, | |
| "learning_rate": 2.4226871416833683e-05, | |
| "loss": 0.1988, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20574046671390533, | |
| "step": 3260, | |
| "valid_targets_mean": 2167.6, | |
| "valid_targets_min": 1058 | |
| }, | |
| { | |
| "epoch": 3.4296218487394956, | |
| "grad_norm": 0.6743471501486973, | |
| "learning_rate": 2.4175654210248848e-05, | |
| "loss": 0.2121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19565929472446442, | |
| "step": 3265, | |
| "valid_targets_mean": 2611.9, | |
| "valid_targets_min": 806 | |
| }, | |
| { | |
| "epoch": 3.434873949579832, | |
| "grad_norm": 0.6301404143473273, | |
| "learning_rate": 2.4124408355556677e-05, | |
| "loss": 0.2177, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22094877064228058, | |
| "step": 3270, | |
| "valid_targets_mean": 3081.8, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 3.440126050420168, | |
| "grad_norm": 0.6526940719618316, | |
| "learning_rate": 2.4073134204342014e-05, | |
| "loss": 0.2089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20152367651462555, | |
| "step": 3275, | |
| "valid_targets_mean": 2381.3, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 3.4453781512605044, | |
| "grad_norm": 0.6659235538473005, | |
| "learning_rate": 2.4021832108383837e-05, | |
| "loss": 0.2207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20738565921783447, | |
| "step": 3280, | |
| "valid_targets_mean": 2625.6, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 3.4506302521008405, | |
| "grad_norm": 0.7166624142705825, | |
| "learning_rate": 2.3970502419652854e-05, | |
| "loss": 0.2073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19094321131706238, | |
| "step": 3285, | |
| "valid_targets_mean": 2041.8, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 3.4558823529411766, | |
| "grad_norm": 0.7984671258808161, | |
| "learning_rate": 2.3919145490309076e-05, | |
| "loss": 0.2116, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21888110041618347, | |
| "step": 3290, | |
| "valid_targets_mean": 1798.3, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 3.4611344537815127, | |
| "grad_norm": 0.6963556467492686, | |
| "learning_rate": 2.38677616726994e-05, | |
| "loss": 0.1958, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17585915327072144, | |
| "step": 3295, | |
| "valid_targets_mean": 2040.5, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 3.466386554621849, | |
| "grad_norm": 0.5743174603598479, | |
| "learning_rate": 2.381635131935521e-05, | |
| "loss": 0.205, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1667209416627884, | |
| "step": 3300, | |
| "valid_targets_mean": 3134.2, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 3.471638655462185, | |
| "grad_norm": 0.7243626640826193, | |
| "learning_rate": 2.3764914782989926e-05, | |
| "loss": 0.2135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22822029888629913, | |
| "step": 3305, | |
| "valid_targets_mean": 2619.9, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 3.476890756302521, | |
| "grad_norm": 0.669561945697959, | |
| "learning_rate": 2.3713452416496625e-05, | |
| "loss": 0.2097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20612934231758118, | |
| "step": 3310, | |
| "valid_targets_mean": 2525.0, | |
| "valid_targets_min": 717 | |
| }, | |
| { | |
| "epoch": 3.482142857142857, | |
| "grad_norm": 0.7414140882682606, | |
| "learning_rate": 2.366196457294558e-05, | |
| "loss": 0.2137, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21093863248825073, | |
| "step": 3315, | |
| "valid_targets_mean": 2090.9, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 3.4873949579831933, | |
| "grad_norm": 0.6133977770277803, | |
| "learning_rate": 2.3610451605581858e-05, | |
| "loss": 0.2061, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1882474273443222, | |
| "step": 3320, | |
| "valid_targets_mean": 3069.8, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 3.4926470588235294, | |
| "grad_norm": 0.6648952110036311, | |
| "learning_rate": 2.3558913867822905e-05, | |
| "loss": 0.2057, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20279830694198608, | |
| "step": 3325, | |
| "valid_targets_mean": 2729.6, | |
| "valid_targets_min": 484 | |
| }, | |
| { | |
| "epoch": 3.4978991596638656, | |
| "grad_norm": 0.7839692483507994, | |
| "learning_rate": 2.3507351713256105e-05, | |
| "loss": 0.2052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21825388073921204, | |
| "step": 3330, | |
| "valid_targets_mean": 1961.4, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 3.5031512605042017, | |
| "grad_norm": 0.6420384291551596, | |
| "learning_rate": 2.3455765495636353e-05, | |
| "loss": 0.2015, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20120246708393097, | |
| "step": 3335, | |
| "valid_targets_mean": 2840.8, | |
| "valid_targets_min": 815 | |
| }, | |
| { | |
| "epoch": 3.508403361344538, | |
| "grad_norm": 0.7332433509193245, | |
| "learning_rate": 2.3404155568883643e-05, | |
| "loss": 0.214, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23752212524414062, | |
| "step": 3340, | |
| "valid_targets_mean": 2502.9, | |
| "valid_targets_min": 1052 | |
| }, | |
| { | |
| "epoch": 3.513655462184874, | |
| "grad_norm": 0.7036254623505526, | |
| "learning_rate": 2.3352522287080625e-05, | |
| "loss": 0.2107, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22437232732772827, | |
| "step": 3345, | |
| "valid_targets_mean": 2470.9, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 3.51890756302521, | |
| "grad_norm": 0.674241478442284, | |
| "learning_rate": 2.3300866004470182e-05, | |
| "loss": 0.2121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23106199502944946, | |
| "step": 3350, | |
| "valid_targets_mean": 2678.7, | |
| "valid_targets_min": 1190 | |
| }, | |
| { | |
| "epoch": 3.524159663865546, | |
| "grad_norm": 0.6913901681322876, | |
| "learning_rate": 2.324918707545302e-05, | |
| "loss": 0.2085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20624200999736786, | |
| "step": 3355, | |
| "valid_targets_mean": 2525.9, | |
| "valid_targets_min": 886 | |
| }, | |
| { | |
| "epoch": 3.5294117647058822, | |
| "grad_norm": 0.7215505447514452, | |
| "learning_rate": 2.3197485854585183e-05, | |
| "loss": 0.2074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18115992844104767, | |
| "step": 3360, | |
| "valid_targets_mean": 1991.8, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 3.5346638655462184, | |
| "grad_norm": 0.6339782709302462, | |
| "learning_rate": 2.3145762696575675e-05, | |
| "loss": 0.2014, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20273572206497192, | |
| "step": 3365, | |
| "valid_targets_mean": 3340.9, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 3.5399159663865545, | |
| "grad_norm": 0.6976739706673805, | |
| "learning_rate": 2.3094017956283995e-05, | |
| "loss": 0.1995, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2007465362548828, | |
| "step": 3370, | |
| "valid_targets_mean": 2454.3, | |
| "valid_targets_min": 943 | |
| }, | |
| { | |
| "epoch": 3.5451680672268906, | |
| "grad_norm": 0.6726999837744964, | |
| "learning_rate": 2.3042251988717728e-05, | |
| "loss": 0.2006, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16553646326065063, | |
| "step": 3375, | |
| "valid_targets_mean": 2192.6, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 3.5504201680672267, | |
| "grad_norm": 0.7301827373994265, | |
| "learning_rate": 2.2990465149030077e-05, | |
| "loss": 0.2013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19199733436107635, | |
| "step": 3380, | |
| "valid_targets_mean": 2214.3, | |
| "valid_targets_min": 897 | |
| }, | |
| { | |
| "epoch": 3.5556722689075633, | |
| "grad_norm": 0.6670894246533184, | |
| "learning_rate": 2.2938657792517446e-05, | |
| "loss": 0.221, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2090798318386078, | |
| "step": 3385, | |
| "valid_targets_mean": 2772.3, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 3.560924369747899, | |
| "grad_norm": 0.7737810251679192, | |
| "learning_rate": 2.2886830274617017e-05, | |
| "loss": 0.2142, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24221691489219666, | |
| "step": 3390, | |
| "valid_targets_mean": 2221.1, | |
| "valid_targets_min": 861 | |
| }, | |
| { | |
| "epoch": 3.5661764705882355, | |
| "grad_norm": 0.6681259679390096, | |
| "learning_rate": 2.2834982950904267e-05, | |
| "loss": 0.2107, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19659334421157837, | |
| "step": 3395, | |
| "valid_targets_mean": 2447.6, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 3.571428571428571, | |
| "grad_norm": 0.8318412772228536, | |
| "learning_rate": 2.2783116177090573e-05, | |
| "loss": 0.2149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22690819203853607, | |
| "step": 3400, | |
| "valid_targets_mean": 2100.6, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 3.5766806722689077, | |
| "grad_norm": 0.6541716254669749, | |
| "learning_rate": 2.2731230309020762e-05, | |
| "loss": 0.2164, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19614872336387634, | |
| "step": 3405, | |
| "valid_targets_mean": 2690.9, | |
| "valid_targets_min": 896 | |
| }, | |
| { | |
| "epoch": 3.581932773109244, | |
| "grad_norm": 0.735048933666179, | |
| "learning_rate": 2.2679325702670642e-05, | |
| "loss": 0.2098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22025522589683533, | |
| "step": 3410, | |
| "valid_targets_mean": 2054.1, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 3.58718487394958, | |
| "grad_norm": 0.7432848251883631, | |
| "learning_rate": 2.2627402714144586e-05, | |
| "loss": 0.2058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21356913447380066, | |
| "step": 3415, | |
| "valid_targets_mean": 2400.3, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 3.592436974789916, | |
| "grad_norm": 0.6273611826621119, | |
| "learning_rate": 2.2575461699673085e-05, | |
| "loss": 0.2052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21526393294334412, | |
| "step": 3420, | |
| "valid_targets_mean": 3208.0, | |
| "valid_targets_min": 1402 | |
| }, | |
| { | |
| "epoch": 3.597689075630252, | |
| "grad_norm": 0.6901843944101652, | |
| "learning_rate": 2.2523503015610303e-05, | |
| "loss": 0.2088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21407434344291687, | |
| "step": 3425, | |
| "valid_targets_mean": 2472.9, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 3.6029411764705883, | |
| "grad_norm": 0.6750339364655951, | |
| "learning_rate": 2.2471527018431637e-05, | |
| "loss": 0.2146, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2302270233631134, | |
| "step": 3430, | |
| "valid_targets_mean": 3292.7, | |
| "valid_targets_min": 943 | |
| }, | |
| { | |
| "epoch": 3.6081932773109244, | |
| "grad_norm": 0.6870012851485073, | |
| "learning_rate": 2.2419534064731238e-05, | |
| "loss": 0.2054, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2101989984512329, | |
| "step": 3435, | |
| "valid_targets_mean": 2430.9, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 3.6134453781512605, | |
| "grad_norm": 0.7652411173135991, | |
| "learning_rate": 2.236752451121963e-05, | |
| "loss": 0.2181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20987604558467865, | |
| "step": 3440, | |
| "valid_targets_mean": 2773.8, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 3.6186974789915967, | |
| "grad_norm": 0.96702355486614, | |
| "learning_rate": 2.2315498714721198e-05, | |
| "loss": 0.2094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2208380401134491, | |
| "step": 3445, | |
| "valid_targets_mean": 2227.6, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 3.6239495798319328, | |
| "grad_norm": 0.7568581037232849, | |
| "learning_rate": 2.2263457032171762e-05, | |
| "loss": 0.2144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20502294600009918, | |
| "step": 3450, | |
| "valid_targets_mean": 1974.8, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 3.629201680672269, | |
| "grad_norm": 0.6324142287116146, | |
| "learning_rate": 2.2211399820616154e-05, | |
| "loss": 0.2017, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1848364919424057, | |
| "step": 3455, | |
| "valid_targets_mean": 2530.7, | |
| "valid_targets_min": 1356 | |
| }, | |
| { | |
| "epoch": 3.634453781512605, | |
| "grad_norm": 0.6787789057247168, | |
| "learning_rate": 2.2159327437205733e-05, | |
| "loss": 0.2202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20259186625480652, | |
| "step": 3460, | |
| "valid_targets_mean": 2469.3, | |
| "valid_targets_min": 843 | |
| }, | |
| { | |
| "epoch": 3.639705882352941, | |
| "grad_norm": 0.6108458629326181, | |
| "learning_rate": 2.210724023919595e-05, | |
| "loss": 0.21, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21205033361911774, | |
| "step": 3465, | |
| "valid_targets_mean": 3304.5, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 3.6449579831932772, | |
| "grad_norm": 0.7630702002273175, | |
| "learning_rate": 2.205513858394389e-05, | |
| "loss": 0.2093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22506915032863617, | |
| "step": 3470, | |
| "valid_targets_mean": 2128.4, | |
| "valid_targets_min": 982 | |
| }, | |
| { | |
| "epoch": 3.6502100840336134, | |
| "grad_norm": 0.6853988147626505, | |
| "learning_rate": 2.2003022828905832e-05, | |
| "loss": 0.2047, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2102399468421936, | |
| "step": 3475, | |
| "valid_targets_mean": 2369.8, | |
| "valid_targets_min": 942 | |
| }, | |
| { | |
| "epoch": 3.6554621848739495, | |
| "grad_norm": 0.8205924816380262, | |
| "learning_rate": 2.19508933316348e-05, | |
| "loss": 0.2094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2225526124238968, | |
| "step": 3480, | |
| "valid_targets_mean": 2051.9, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 3.6607142857142856, | |
| "grad_norm": 0.7855685679107322, | |
| "learning_rate": 2.189875044977808e-05, | |
| "loss": 0.2142, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21352198719978333, | |
| "step": 3485, | |
| "valid_targets_mean": 2024.9, | |
| "valid_targets_min": 868 | |
| }, | |
| { | |
| "epoch": 3.6659663865546217, | |
| "grad_norm": 0.6891565684644114, | |
| "learning_rate": 2.1846594541074796e-05, | |
| "loss": 0.2067, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19370394945144653, | |
| "step": 3490, | |
| "valid_targets_mean": 2303.7, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 3.671218487394958, | |
| "grad_norm": 0.66281123426459, | |
| "learning_rate": 2.179442596335345e-05, | |
| "loss": 0.2158, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21961744129657745, | |
| "step": 3495, | |
| "valid_targets_mean": 2835.9, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 3.6764705882352944, | |
| "grad_norm": 0.6917997631532197, | |
| "learning_rate": 2.174224507452945e-05, | |
| "loss": 0.2227, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20729771256446838, | |
| "step": 3500, | |
| "valid_targets_mean": 2409.9, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 3.68172268907563, | |
| "grad_norm": 0.7215177976056998, | |
| "learning_rate": 2.1690052232602677e-05, | |
| "loss": 0.2213, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24402323365211487, | |
| "step": 3505, | |
| "valid_targets_mean": 2753.1, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 3.6869747899159666, | |
| "grad_norm": 0.6763541211099533, | |
| "learning_rate": 2.1637847795655022e-05, | |
| "loss": 0.2118, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21992330253124237, | |
| "step": 3510, | |
| "valid_targets_mean": 2720.9, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 3.6922268907563023, | |
| "grad_norm": 0.6816841950766274, | |
| "learning_rate": 2.1585632121847925e-05, | |
| "loss": 0.2169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19951927661895752, | |
| "step": 3515, | |
| "valid_targets_mean": 2837.1, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 3.697478991596639, | |
| "grad_norm": 0.695061597525997, | |
| "learning_rate": 2.153340556941991e-05, | |
| "loss": 0.2147, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23091833293437958, | |
| "step": 3520, | |
| "valid_targets_mean": 2830.2, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 3.7027310924369745, | |
| "grad_norm": 0.7145487481152861, | |
| "learning_rate": 2.1481168496684145e-05, | |
| "loss": 0.2152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22669626772403717, | |
| "step": 3525, | |
| "valid_targets_mean": 2385.2, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 3.707983193277311, | |
| "grad_norm": 0.773517958966538, | |
| "learning_rate": 2.142892126202597e-05, | |
| "loss": 0.2083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20061250030994415, | |
| "step": 3530, | |
| "valid_targets_mean": 1901.8, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 3.713235294117647, | |
| "grad_norm": 0.6514006030844219, | |
| "learning_rate": 2.137666422390045e-05, | |
| "loss": 0.2088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21307383477687836, | |
| "step": 3535, | |
| "valid_targets_mean": 2973.2, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 3.7184873949579833, | |
| "grad_norm": 0.6294462036244911, | |
| "learning_rate": 2.132439774082991e-05, | |
| "loss": 0.2197, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2212301641702652, | |
| "step": 3540, | |
| "valid_targets_mean": 3056.4, | |
| "valid_targets_min": 1194 | |
| }, | |
| { | |
| "epoch": 3.7237394957983194, | |
| "grad_norm": 0.654887046400427, | |
| "learning_rate": 2.1272122171401467e-05, | |
| "loss": 0.1945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18347257375717163, | |
| "step": 3545, | |
| "valid_targets_mean": 2574.4, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 3.7289915966386555, | |
| "grad_norm": 0.6809633747210387, | |
| "learning_rate": 2.1219837874264573e-05, | |
| "loss": 0.2024, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18207016587257385, | |
| "step": 3550, | |
| "valid_targets_mean": 2467.8, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 3.7342436974789917, | |
| "grad_norm": 0.8490088878626119, | |
| "learning_rate": 2.116754520812857e-05, | |
| "loss": 0.2161, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21279117465019226, | |
| "step": 3555, | |
| "valid_targets_mean": 1938.0, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 3.7394957983193278, | |
| "grad_norm": 0.7362797471450374, | |
| "learning_rate": 2.111524453176022e-05, | |
| "loss": 0.2187, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2274627536535263, | |
| "step": 3560, | |
| "valid_targets_mean": 2261.9, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 3.744747899159664, | |
| "grad_norm": 0.7547351585834484, | |
| "learning_rate": 2.1062936203981227e-05, | |
| "loss": 0.2208, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21145398914813995, | |
| "step": 3565, | |
| "valid_targets_mean": 2079.0, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "grad_norm": 0.6040282948873019, | |
| "learning_rate": 2.10106205836658e-05, | |
| "loss": 0.2067, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17291516065597534, | |
| "step": 3570, | |
| "valid_targets_mean": 2807.0, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 3.755252100840336, | |
| "grad_norm": 0.7692191031763723, | |
| "learning_rate": 2.095829802973817e-05, | |
| "loss": 0.2149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21016576886177063, | |
| "step": 3575, | |
| "valid_targets_mean": 2692.1, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 3.7605042016806722, | |
| "grad_norm": 0.6937553318754113, | |
| "learning_rate": 2.090596890117016e-05, | |
| "loss": 0.211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20397567749023438, | |
| "step": 3580, | |
| "valid_targets_mean": 2325.8, | |
| "valid_targets_min": 915 | |
| }, | |
| { | |
| "epoch": 3.7657563025210083, | |
| "grad_norm": 0.7298575781771081, | |
| "learning_rate": 2.0853633556978674e-05, | |
| "loss": 0.2089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2305256426334381, | |
| "step": 3585, | |
| "valid_targets_mean": 2357.3, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 3.7710084033613445, | |
| "grad_norm": 0.8229634474031364, | |
| "learning_rate": 2.0801292356223265e-05, | |
| "loss": 0.2235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23045924305915833, | |
| "step": 3590, | |
| "valid_targets_mean": 1879.1, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 3.7762605042016806, | |
| "grad_norm": 0.6135522966100759, | |
| "learning_rate": 2.0748945658003695e-05, | |
| "loss": 0.2116, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20716413855552673, | |
| "step": 3595, | |
| "valid_targets_mean": 3463.8, | |
| "valid_targets_min": 244 | |
| }, | |
| { | |
| "epoch": 3.7815126050420167, | |
| "grad_norm": 0.8109456213693356, | |
| "learning_rate": 2.0696593821457392e-05, | |
| "loss": 0.2199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21201831102371216, | |
| "step": 3600, | |
| "valid_targets_mean": 2110.6, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 3.786764705882353, | |
| "grad_norm": 0.7309833863162327, | |
| "learning_rate": 2.064423720575709e-05, | |
| "loss": 0.2172, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22593368589878082, | |
| "step": 3605, | |
| "valid_targets_mean": 2323.6, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 3.792016806722689, | |
| "grad_norm": 0.6453442893544274, | |
| "learning_rate": 2.0591876170108257e-05, | |
| "loss": 0.2124, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22857289016246796, | |
| "step": 3610, | |
| "valid_targets_mean": 2964.2, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 3.7972689075630255, | |
| "grad_norm": 1.045593114030177, | |
| "learning_rate": 2.0539511073746743e-05, | |
| "loss": 0.1914, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2170926183462143, | |
| "step": 3615, | |
| "valid_targets_mean": 2493.9, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 3.802521008403361, | |
| "grad_norm": 0.685780352452399, | |
| "learning_rate": 2.0487142275936213e-05, | |
| "loss": 0.2002, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1872934252023697, | |
| "step": 3620, | |
| "valid_targets_mean": 2356.4, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 3.8077731092436977, | |
| "grad_norm": 0.7579039345943492, | |
| "learning_rate": 2.0434770135965743e-05, | |
| "loss": 0.2121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.200333833694458, | |
| "step": 3625, | |
| "valid_targets_mean": 2157.2, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 3.8130252100840334, | |
| "grad_norm": 0.6470297697973766, | |
| "learning_rate": 2.0382395013147347e-05, | |
| "loss": 0.2022, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19545529782772064, | |
| "step": 3630, | |
| "valid_targets_mean": 2711.9, | |
| "valid_targets_min": 1144 | |
| }, | |
| { | |
| "epoch": 3.81827731092437, | |
| "grad_norm": 0.653362214398331, | |
| "learning_rate": 2.033001726681349e-05, | |
| "loss": 0.2199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21826599538326263, | |
| "step": 3635, | |
| "valid_targets_mean": 2853.1, | |
| "valid_targets_min": 1078 | |
| }, | |
| { | |
| "epoch": 3.8235294117647056, | |
| "grad_norm": 0.6867053178122183, | |
| "learning_rate": 2.027763725631463e-05, | |
| "loss": 0.2023, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19404883682727814, | |
| "step": 3640, | |
| "valid_targets_mean": 2482.8, | |
| "valid_targets_min": 1067 | |
| }, | |
| { | |
| "epoch": 3.828781512605042, | |
| "grad_norm": 0.7218696655921097, | |
| "learning_rate": 2.0225255341016794e-05, | |
| "loss": 0.2105, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1934603750705719, | |
| "step": 3645, | |
| "valid_targets_mean": 2152.1, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 3.8340336134453783, | |
| "grad_norm": 0.7091556745123918, | |
| "learning_rate": 2.017287188029904e-05, | |
| "loss": 0.2038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21000255644321442, | |
| "step": 3650, | |
| "valid_targets_mean": 2308.8, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 3.8392857142857144, | |
| "grad_norm": 0.6815200778696167, | |
| "learning_rate": 2.0120487233551035e-05, | |
| "loss": 0.2187, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23103326559066772, | |
| "step": 3655, | |
| "valid_targets_mean": 2734.1, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 3.8445378151260505, | |
| "grad_norm": 0.6280352447022034, | |
| "learning_rate": 2.006810176017059e-05, | |
| "loss": 0.2085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21302129328250885, | |
| "step": 3660, | |
| "valid_targets_mean": 3324.1, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 3.8497899159663866, | |
| "grad_norm": 0.7066023629427282, | |
| "learning_rate": 2.0015715819561205e-05, | |
| "loss": 0.2138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21427252888679504, | |
| "step": 3665, | |
| "valid_targets_mean": 2452.4, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 3.8550420168067228, | |
| "grad_norm": 0.647316148786966, | |
| "learning_rate": 1.9963329771129558e-05, | |
| "loss": 0.1952, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1968604177236557, | |
| "step": 3670, | |
| "valid_targets_mean": 2530.9, | |
| "valid_targets_min": 1254 | |
| }, | |
| { | |
| "epoch": 3.860294117647059, | |
| "grad_norm": 0.6784605187933433, | |
| "learning_rate": 1.9910943974283066e-05, | |
| "loss": 0.2143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2061229795217514, | |
| "step": 3675, | |
| "valid_targets_mean": 2622.5, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 3.865546218487395, | |
| "grad_norm": 0.6920269949000379, | |
| "learning_rate": 1.9858558788427447e-05, | |
| "loss": 0.1973, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18967095017433167, | |
| "step": 3680, | |
| "valid_targets_mean": 2135.2, | |
| "valid_targets_min": 513 | |
| }, | |
| { | |
| "epoch": 3.870798319327731, | |
| "grad_norm": 0.7052959882420897, | |
| "learning_rate": 1.9806174572964205e-05, | |
| "loss": 0.2198, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20895016193389893, | |
| "step": 3685, | |
| "valid_targets_mean": 2395.5, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 3.8760504201680672, | |
| "grad_norm": 0.6385724766539387, | |
| "learning_rate": 1.975379168728819e-05, | |
| "loss": 0.2136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2081984281539917, | |
| "step": 3690, | |
| "valid_targets_mean": 2853.7, | |
| "valid_targets_min": 936 | |
| }, | |
| { | |
| "epoch": 3.8813025210084033, | |
| "grad_norm": 0.72510115733088, | |
| "learning_rate": 1.9701410490785128e-05, | |
| "loss": 0.2249, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22268277406692505, | |
| "step": 3695, | |
| "valid_targets_mean": 2524.2, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 3.8865546218487395, | |
| "grad_norm": 0.7123147276422185, | |
| "learning_rate": 1.9649031342829178e-05, | |
| "loss": 0.2055, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22262287139892578, | |
| "step": 3700, | |
| "valid_targets_mean": 2540.8, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.8918067226890756, | |
| "grad_norm": 0.6390925839310951, | |
| "learning_rate": 1.9596654602780398e-05, | |
| "loss": 0.2161, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19909490644931793, | |
| "step": 3705, | |
| "valid_targets_mean": 2773.2, | |
| "valid_targets_min": 853 | |
| }, | |
| { | |
| "epoch": 3.8970588235294117, | |
| "grad_norm": 0.6412477203773937, | |
| "learning_rate": 1.9544280629982364e-05, | |
| "loss": 0.2038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1946112960577011, | |
| "step": 3710, | |
| "valid_targets_mean": 2642.8, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 3.902310924369748, | |
| "grad_norm": 0.716484891233997, | |
| "learning_rate": 1.949190978375966e-05, | |
| "loss": 0.2122, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23176735639572144, | |
| "step": 3715, | |
| "valid_targets_mean": 2558.9, | |
| "valid_targets_min": 1150 | |
| }, | |
| { | |
| "epoch": 3.907563025210084, | |
| "grad_norm": 0.6768189969402229, | |
| "learning_rate": 1.9439542423415413e-05, | |
| "loss": 0.2018, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19778092205524445, | |
| "step": 3720, | |
| "valid_targets_mean": 2772.1, | |
| "valid_targets_min": 1107 | |
| }, | |
| { | |
| "epoch": 3.91281512605042, | |
| "grad_norm": 0.6207653385694745, | |
| "learning_rate": 1.9387178908228838e-05, | |
| "loss": 0.2181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20920640230178833, | |
| "step": 3725, | |
| "valid_targets_mean": 3297.2, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 3.918067226890756, | |
| "grad_norm": 0.7356716818302198, | |
| "learning_rate": 1.933481959745276e-05, | |
| "loss": 0.2218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24886155128479004, | |
| "step": 3730, | |
| "valid_targets_mean": 2577.4, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 3.9233193277310923, | |
| "grad_norm": 0.6771424238422388, | |
| "learning_rate": 1.9282464850311184e-05, | |
| "loss": 0.2115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22557204961776733, | |
| "step": 3735, | |
| "valid_targets_mean": 2883.5, | |
| "valid_targets_min": 474 | |
| }, | |
| { | |
| "epoch": 3.928571428571429, | |
| "grad_norm": 0.7641048688976471, | |
| "learning_rate": 1.9230115025996764e-05, | |
| "loss": 0.208, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2032201588153839, | |
| "step": 3740, | |
| "valid_targets_mean": 1858.1, | |
| "valid_targets_min": 1143 | |
| }, | |
| { | |
| "epoch": 3.9338235294117645, | |
| "grad_norm": 0.5906821588122515, | |
| "learning_rate": 1.9177770483668416e-05, | |
| "loss": 0.2203, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1867651641368866, | |
| "step": 3745, | |
| "valid_targets_mean": 2936.9, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 3.939075630252101, | |
| "grad_norm": 0.6203552861446046, | |
| "learning_rate": 1.912543158244881e-05, | |
| "loss": 0.2128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19798970222473145, | |
| "step": 3750, | |
| "valid_targets_mean": 3003.4, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 3.9443277310924367, | |
| "grad_norm": 0.7646340475457648, | |
| "learning_rate": 1.9073098681421895e-05, | |
| "loss": 0.2129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19918084144592285, | |
| "step": 3755, | |
| "valid_targets_mean": 1886.4, | |
| "valid_targets_min": 545 | |
| }, | |
| { | |
| "epoch": 3.9495798319327733, | |
| "grad_norm": 0.6926521998600635, | |
| "learning_rate": 1.902077213963048e-05, | |
| "loss": 0.1995, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2141200602054596, | |
| "step": 3760, | |
| "valid_targets_mean": 2899.2, | |
| "valid_targets_min": 896 | |
| }, | |
| { | |
| "epoch": 3.9548319327731094, | |
| "grad_norm": 0.7157783415052118, | |
| "learning_rate": 1.896845231607372e-05, | |
| "loss": 0.1966, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19222521781921387, | |
| "step": 3765, | |
| "valid_targets_mean": 2013.7, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 3.9600840336134455, | |
| "grad_norm": 0.8266386898112025, | |
| "learning_rate": 1.8916139569704704e-05, | |
| "loss": 0.2116, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22533836960792542, | |
| "step": 3770, | |
| "valid_targets_mean": 2034.5, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 3.9653361344537816, | |
| "grad_norm": 0.8096944452658449, | |
| "learning_rate": 1.886383425942795e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18177619576454163, | |
| "step": 3775, | |
| "valid_targets_mean": 1980.1, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 3.9705882352941178, | |
| "grad_norm": 0.6708795853596977, | |
| "learning_rate": 1.8811536744096956e-05, | |
| "loss": 0.2115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2508021593093872, | |
| "step": 3780, | |
| "valid_targets_mean": 2865.2, | |
| "valid_targets_min": 712 | |
| }, | |
| { | |
| "epoch": 3.975840336134454, | |
| "grad_norm": 0.6124622098856916, | |
| "learning_rate": 1.8759247382511748e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21529731154441833, | |
| "step": 3785, | |
| "valid_targets_mean": 3260.9, | |
| "valid_targets_min": 772 | |
| }, | |
| { | |
| "epoch": 3.98109243697479, | |
| "grad_norm": 0.7245013981014534, | |
| "learning_rate": 1.8706966533416413e-05, | |
| "loss": 0.2103, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2099657505750656, | |
| "step": 3790, | |
| "valid_targets_mean": 2156.7, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 3.986344537815126, | |
| "grad_norm": 0.7199624429982473, | |
| "learning_rate": 1.8654694555496624e-05, | |
| "loss": 0.2138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2163430154323578, | |
| "step": 3795, | |
| "valid_targets_mean": 2900.4, | |
| "valid_targets_min": 1205 | |
| }, | |
| { | |
| "epoch": 3.991596638655462, | |
| "grad_norm": 0.7313432594209842, | |
| "learning_rate": 1.860243180737721e-05, | |
| "loss": 0.2097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24018210172653198, | |
| "step": 3800, | |
| "valid_targets_mean": 2467.2, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 3.9968487394957983, | |
| "grad_norm": 0.686344735243992, | |
| "learning_rate": 1.8550178647619664e-05, | |
| "loss": 0.2064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2029697597026825, | |
| "step": 3805, | |
| "valid_targets_mean": 2462.6, | |
| "valid_targets_min": 853 | |
| }, | |
| { | |
| "epoch": 4.0021008403361344, | |
| "grad_norm": 0.5964545073234324, | |
| "learning_rate": 1.84979354347197e-05, | |
| "loss": 0.1972, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15941348671913147, | |
| "step": 3810, | |
| "valid_targets_mean": 2904.0, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 4.007352941176471, | |
| "grad_norm": 0.6310111711662433, | |
| "learning_rate": 1.8445702527104782e-05, | |
| "loss": 0.1758, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17425957322120667, | |
| "step": 3815, | |
| "valid_targets_mean": 2903.4, | |
| "valid_targets_min": 1572 | |
| }, | |
| { | |
| "epoch": 4.012605042016807, | |
| "grad_norm": 0.8382816261281872, | |
| "learning_rate": 1.8393480283131677e-05, | |
| "loss": 0.1876, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1987411081790924, | |
| "step": 3820, | |
| "valid_targets_mean": 2029.4, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 4.017857142857143, | |
| "grad_norm": 0.6828047242423465, | |
| "learning_rate": 1.8341269061083996e-05, | |
| "loss": 0.1826, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16214141249656677, | |
| "step": 3825, | |
| "valid_targets_mean": 2665.1, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 4.023109243697479, | |
| "grad_norm": 0.6544969992637274, | |
| "learning_rate": 1.8289069219169717e-05, | |
| "loss": 0.1698, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15575842559337616, | |
| "step": 3830, | |
| "valid_targets_mean": 2708.9, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 4.0283613445378155, | |
| "grad_norm": 0.665567823015263, | |
| "learning_rate": 1.8236881115518766e-05, | |
| "loss": 0.1769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1710759848356247, | |
| "step": 3835, | |
| "valid_targets_mean": 2501.2, | |
| "valid_targets_min": 671 | |
| }, | |
| { | |
| "epoch": 4.033613445378151, | |
| "grad_norm": 0.8048649946622554, | |
| "learning_rate": 1.8184705108180516e-05, | |
| "loss": 0.1964, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21742048859596252, | |
| "step": 3840, | |
| "valid_targets_mean": 2245.1, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 4.038865546218488, | |
| "grad_norm": 0.6730348979572065, | |
| "learning_rate": 1.8132541555121353e-05, | |
| "loss": 0.1745, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17370149493217468, | |
| "step": 3845, | |
| "valid_targets_mean": 3004.6, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 4.044117647058823, | |
| "grad_norm": 0.6773064122209941, | |
| "learning_rate": 1.808039081422223e-05, | |
| "loss": 0.1759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.177533358335495, | |
| "step": 3850, | |
| "valid_targets_mean": 2959.3, | |
| "valid_targets_min": 923 | |
| }, | |
| { | |
| "epoch": 4.04936974789916, | |
| "grad_norm": 0.8972828803367779, | |
| "learning_rate": 1.802825324327618e-05, | |
| "loss": 0.1866, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19859957695007324, | |
| "step": 3855, | |
| "valid_targets_mean": 1707.2, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 4.054621848739496, | |
| "grad_norm": 0.6816271881225173, | |
| "learning_rate": 1.7976129199985886e-05, | |
| "loss": 0.1863, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19027729332447052, | |
| "step": 3860, | |
| "valid_targets_mean": 2623.9, | |
| "valid_targets_min": 1048 | |
| }, | |
| { | |
| "epoch": 4.059873949579832, | |
| "grad_norm": 0.7082107239302485, | |
| "learning_rate": 1.7924019041961228e-05, | |
| "loss": 0.179, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17302575707435608, | |
| "step": 3865, | |
| "valid_targets_mean": 2540.1, | |
| "valid_targets_min": 792 | |
| }, | |
| { | |
| "epoch": 4.065126050420168, | |
| "grad_norm": 0.781860364268105, | |
| "learning_rate": 1.7871923126716827e-05, | |
| "loss": 0.1827, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17314311861991882, | |
| "step": 3870, | |
| "valid_targets_mean": 1902.4, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 4.070378151260504, | |
| "grad_norm": 0.770240331697946, | |
| "learning_rate": 1.7819841811669573e-05, | |
| "loss": 0.1726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17767006158828735, | |
| "step": 3875, | |
| "valid_targets_mean": 2162.2, | |
| "valid_targets_min": 990 | |
| }, | |
| { | |
| "epoch": 4.07563025210084, | |
| "grad_norm": 0.7494050596565602, | |
| "learning_rate": 1.7767775454136194e-05, | |
| "loss": 0.1684, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16591396927833557, | |
| "step": 3880, | |
| "valid_targets_mean": 2518.5, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 4.080882352941177, | |
| "grad_norm": 0.6697129284162905, | |
| "learning_rate": 1.7715724411330806e-05, | |
| "loss": 0.1757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1829383820295334, | |
| "step": 3885, | |
| "valid_targets_mean": 3106.3, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 4.086134453781512, | |
| "grad_norm": 0.7339734064772027, | |
| "learning_rate": 1.7663689040362446e-05, | |
| "loss": 0.1872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18020042777061462, | |
| "step": 3890, | |
| "valid_targets_mean": 2812.9, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 4.091386554621849, | |
| "grad_norm": 0.7376435988375549, | |
| "learning_rate": 1.761166969823262e-05, | |
| "loss": 0.1798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15077045559883118, | |
| "step": 3895, | |
| "valid_targets_mean": 2092.4, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 4.0966386554621845, | |
| "grad_norm": 0.6730357795335743, | |
| "learning_rate": 1.7559666741832896e-05, | |
| "loss": 0.187, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16643144190311432, | |
| "step": 3900, | |
| "valid_targets_mean": 2449.2, | |
| "valid_targets_min": 982 | |
| }, | |
| { | |
| "epoch": 4.101890756302521, | |
| "grad_norm": 0.8297442949389751, | |
| "learning_rate": 1.750768052794239e-05, | |
| "loss": 0.1927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18952782452106476, | |
| "step": 3905, | |
| "valid_targets_mean": 1901.9, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 4.107142857142857, | |
| "grad_norm": 0.794021570909151, | |
| "learning_rate": 1.7455711413225372e-05, | |
| "loss": 0.1846, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19502568244934082, | |
| "step": 3910, | |
| "valid_targets_mean": 2269.1, | |
| "valid_targets_min": 930 | |
| }, | |
| { | |
| "epoch": 4.112394957983193, | |
| "grad_norm": 0.8070144150239097, | |
| "learning_rate": 1.7403759754228794e-05, | |
| "loss": 0.1795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17317187786102295, | |
| "step": 3915, | |
| "valid_targets_mean": 1938.3, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 4.117647058823529, | |
| "grad_norm": 0.6999096413433571, | |
| "learning_rate": 1.7351825907379837e-05, | |
| "loss": 0.1803, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1811293363571167, | |
| "step": 3920, | |
| "valid_targets_mean": 2730.1, | |
| "valid_targets_min": 912 | |
| }, | |
| { | |
| "epoch": 4.1228991596638656, | |
| "grad_norm": 0.7320665392327763, | |
| "learning_rate": 1.7299910228983513e-05, | |
| "loss": 0.1786, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1840408742427826, | |
| "step": 3925, | |
| "valid_targets_mean": 2366.4, | |
| "valid_targets_min": 1159 | |
| }, | |
| { | |
| "epoch": 4.128151260504202, | |
| "grad_norm": 0.7286921165522647, | |
| "learning_rate": 1.7248013075220128e-05, | |
| "loss": 0.1764, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17183949053287506, | |
| "step": 3930, | |
| "valid_targets_mean": 2257.6, | |
| "valid_targets_min": 886 | |
| }, | |
| { | |
| "epoch": 4.133403361344538, | |
| "grad_norm": 0.7113378690439488, | |
| "learning_rate": 1.7196134802142945e-05, | |
| "loss": 0.1846, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1910371482372284, | |
| "step": 3935, | |
| "valid_targets_mean": 2610.0, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 4.138655462184874, | |
| "grad_norm": 0.9815704878476788, | |
| "learning_rate": 1.7144275765675673e-05, | |
| "loss": 0.1737, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1488598883152008, | |
| "step": 3940, | |
| "valid_targets_mean": 2683.5, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 4.14390756302521, | |
| "grad_norm": 0.6552007544862622, | |
| "learning_rate": 1.7092436321610042e-05, | |
| "loss": 0.1755, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17244458198547363, | |
| "step": 3945, | |
| "valid_targets_mean": 3017.2, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 4.149159663865547, | |
| "grad_norm": 0.7250853653334564, | |
| "learning_rate": 1.704061682560336e-05, | |
| "loss": 0.171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17147307097911835, | |
| "step": 3950, | |
| "valid_targets_mean": 2484.5, | |
| "valid_targets_min": 1216 | |
| }, | |
| { | |
| "epoch": 4.154411764705882, | |
| "grad_norm": 0.7570116394721385, | |
| "learning_rate": 1.698881763317609e-05, | |
| "loss": 0.1856, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19652274250984192, | |
| "step": 3955, | |
| "valid_targets_mean": 2825.2, | |
| "valid_targets_min": 1066 | |
| }, | |
| { | |
| "epoch": 4.159663865546219, | |
| "grad_norm": 0.7477767810091138, | |
| "learning_rate": 1.69370390997094e-05, | |
| "loss": 0.1868, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1591658890247345, | |
| "step": 3960, | |
| "valid_targets_mean": 2386.2, | |
| "valid_targets_min": 551 | |
| }, | |
| { | |
| "epoch": 4.1649159663865545, | |
| "grad_norm": 0.7313664660917486, | |
| "learning_rate": 1.6885281580442677e-05, | |
| "loss": 0.1798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18173372745513916, | |
| "step": 3965, | |
| "valid_targets_mean": 2462.9, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 4.170168067226891, | |
| "grad_norm": 0.6878567291997353, | |
| "learning_rate": 1.6833545430471194e-05, | |
| "loss": 0.1817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16087912023067474, | |
| "step": 3970, | |
| "valid_targets_mean": 2534.8, | |
| "valid_targets_min": 920 | |
| }, | |
| { | |
| "epoch": 4.175420168067227, | |
| "grad_norm": 0.7273228101827134, | |
| "learning_rate": 1.678183100474359e-05, | |
| "loss": 0.1814, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1902090162038803, | |
| "step": 3975, | |
| "valid_targets_mean": 2782.7, | |
| "valid_targets_min": 1105 | |
| }, | |
| { | |
| "epoch": 4.180672268907563, | |
| "grad_norm": 0.7924963064350935, | |
| "learning_rate": 1.6730138658059448e-05, | |
| "loss": 0.1784, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18615125119686127, | |
| "step": 3980, | |
| "valid_targets_mean": 2283.1, | |
| "valid_targets_min": 936 | |
| }, | |
| { | |
| "epoch": 4.185924369747899, | |
| "grad_norm": 0.7668376984218435, | |
| "learning_rate": 1.667846874506689e-05, | |
| "loss": 0.1724, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18692851066589355, | |
| "step": 3985, | |
| "valid_targets_mean": 2412.2, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 4.1911764705882355, | |
| "grad_norm": 0.6979922592014087, | |
| "learning_rate": 1.662682162026012e-05, | |
| "loss": 0.183, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16892990469932556, | |
| "step": 3990, | |
| "valid_targets_mean": 2284.2, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 4.196428571428571, | |
| "grad_norm": 0.7050681864554844, | |
| "learning_rate": 1.6575197637976995e-05, | |
| "loss": 0.179, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19187793135643005, | |
| "step": 3995, | |
| "valid_targets_mean": 3008.6, | |
| "valid_targets_min": 775 | |
| }, | |
| { | |
| "epoch": 4.201680672268908, | |
| "grad_norm": 0.7383233513848826, | |
| "learning_rate": 1.652359715239661e-05, | |
| "loss": 0.1767, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17365893721580505, | |
| "step": 4000, | |
| "valid_targets_mean": 2399.9, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 4.206932773109243, | |
| "grad_norm": 0.7797983416724591, | |
| "learning_rate": 1.6472020517536828e-05, | |
| "loss": 0.1804, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17422497272491455, | |
| "step": 4005, | |
| "valid_targets_mean": 3114.1, | |
| "valid_targets_min": 839 | |
| }, | |
| { | |
| "epoch": 4.21218487394958, | |
| "grad_norm": 0.7864125547244055, | |
| "learning_rate": 1.6420468087251907e-05, | |
| "loss": 0.1807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16865600645542145, | |
| "step": 4010, | |
| "valid_targets_mean": 1723.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 4.217436974789916, | |
| "grad_norm": 0.7327097751903167, | |
| "learning_rate": 1.6368940215230026e-05, | |
| "loss": 0.1779, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16839268803596497, | |
| "step": 4015, | |
| "valid_targets_mean": 2459.2, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 4.222689075630252, | |
| "grad_norm": 0.8238296453072825, | |
| "learning_rate": 1.6317437254990875e-05, | |
| "loss": 0.1839, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1884918212890625, | |
| "step": 4020, | |
| "valid_targets_mean": 2194.0, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 4.227941176470588, | |
| "grad_norm": 0.6460591633506324, | |
| "learning_rate": 1.626595955988325e-05, | |
| "loss": 0.1829, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20153161883354187, | |
| "step": 4025, | |
| "valid_targets_mean": 3779.9, | |
| "valid_targets_min": 1333 | |
| }, | |
| { | |
| "epoch": 4.233193277310924, | |
| "grad_norm": 0.7457345391565217, | |
| "learning_rate": 1.6214507483082593e-05, | |
| "loss": 0.1892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19580884277820587, | |
| "step": 4030, | |
| "valid_targets_mean": 2798.9, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 4.23844537815126, | |
| "grad_norm": 0.9377584690672369, | |
| "learning_rate": 1.6163081377588588e-05, | |
| "loss": 0.1822, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19530422985553741, | |
| "step": 4035, | |
| "valid_targets_mean": 2320.8, | |
| "valid_targets_min": 1065 | |
| }, | |
| { | |
| "epoch": 4.243697478991597, | |
| "grad_norm": 0.7042255194986914, | |
| "learning_rate": 1.6111681596222732e-05, | |
| "loss": 0.1858, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1799120008945465, | |
| "step": 4040, | |
| "valid_targets_mean": 2469.2, | |
| "valid_targets_min": 1188 | |
| }, | |
| { | |
| "epoch": 4.248949579831933, | |
| "grad_norm": 0.8657610459161533, | |
| "learning_rate": 1.606030849162593e-05, | |
| "loss": 0.1878, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17773385345935822, | |
| "step": 4045, | |
| "valid_targets_mean": 2067.2, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 4.254201680672269, | |
| "grad_norm": 0.7170358815089117, | |
| "learning_rate": 1.600896241625605e-05, | |
| "loss": 0.1819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17556703090667725, | |
| "step": 4050, | |
| "valid_targets_mean": 2699.3, | |
| "valid_targets_min": 1230 | |
| }, | |
| { | |
| "epoch": 4.2594537815126055, | |
| "grad_norm": 0.6997369617036321, | |
| "learning_rate": 1.5957643722385532e-05, | |
| "loss": 0.1853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18936964869499207, | |
| "step": 4055, | |
| "valid_targets_mean": 2953.2, | |
| "valid_targets_min": 1086 | |
| }, | |
| { | |
| "epoch": 4.264705882352941, | |
| "grad_norm": 0.6298850135767894, | |
| "learning_rate": 1.590635276209896e-05, | |
| "loss": 0.1793, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15947186946868896, | |
| "step": 4060, | |
| "valid_targets_mean": 3146.7, | |
| "valid_targets_min": 902 | |
| }, | |
| { | |
| "epoch": 4.269957983193278, | |
| "grad_norm": 0.7757297711293648, | |
| "learning_rate": 1.5855089887290634e-05, | |
| "loss": 0.1883, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1859544813632965, | |
| "step": 4065, | |
| "valid_targets_mean": 2376.6, | |
| "valid_targets_min": 889 | |
| }, | |
| { | |
| "epoch": 4.275210084033613, | |
| "grad_norm": 0.707196240386275, | |
| "learning_rate": 1.5803855449662175e-05, | |
| "loss": 0.1838, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17105159163475037, | |
| "step": 4070, | |
| "valid_targets_mean": 3088.3, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 4.28046218487395, | |
| "grad_norm": 0.6412532070399728, | |
| "learning_rate": 1.5752649800720096e-05, | |
| "loss": 0.1827, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17575950920581818, | |
| "step": 4075, | |
| "valid_targets_mean": 3077.6, | |
| "valid_targets_min": 929 | |
| }, | |
| { | |
| "epoch": 4.285714285714286, | |
| "grad_norm": 0.7333503639293483, | |
| "learning_rate": 1.5701473291773395e-05, | |
| "loss": 0.1833, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16692572832107544, | |
| "step": 4080, | |
| "valid_targets_mean": 2213.4, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 4.290966386554622, | |
| "grad_norm": 0.7789248084874639, | |
| "learning_rate": 1.5650326273931166e-05, | |
| "loss": 0.1898, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18778863549232483, | |
| "step": 4085, | |
| "valid_targets_mean": 2436.8, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 4.296218487394958, | |
| "grad_norm": 0.7608523839784216, | |
| "learning_rate": 1.559920909810016e-05, | |
| "loss": 0.1905, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18575254082679749, | |
| "step": 4090, | |
| "valid_targets_mean": 2774.9, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 4.301470588235294, | |
| "grad_norm": 0.7894327029271562, | |
| "learning_rate": 1.5548122114982393e-05, | |
| "loss": 0.1838, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20460262894630432, | |
| "step": 4095, | |
| "valid_targets_mean": 2156.1, | |
| "valid_targets_min": 628 | |
| }, | |
| { | |
| "epoch": 4.30672268907563, | |
| "grad_norm": 0.7108371500394565, | |
| "learning_rate": 1.5497065675072728e-05, | |
| "loss": 0.174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1754794716835022, | |
| "step": 4100, | |
| "valid_targets_mean": 2560.9, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 4.311974789915967, | |
| "grad_norm": 0.7197950949390775, | |
| "learning_rate": 1.5446040128656485e-05, | |
| "loss": 0.182, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1896936446428299, | |
| "step": 4105, | |
| "valid_targets_mean": 2634.4, | |
| "valid_targets_min": 1256 | |
| }, | |
| { | |
| "epoch": 4.317226890756302, | |
| "grad_norm": 0.815627208196532, | |
| "learning_rate": 1.539504582580704e-05, | |
| "loss": 0.1888, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1940707117319107, | |
| "step": 4110, | |
| "valid_targets_mean": 2417.6, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 4.322478991596639, | |
| "grad_norm": 0.6715019204570937, | |
| "learning_rate": 1.5344083116383395e-05, | |
| "loss": 0.1814, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16882187128067017, | |
| "step": 4115, | |
| "valid_targets_mean": 3112.1, | |
| "valid_targets_min": 910 | |
| }, | |
| { | |
| "epoch": 4.3277310924369745, | |
| "grad_norm": 0.8201097153549575, | |
| "learning_rate": 1.529315235002781e-05, | |
| "loss": 0.1853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18270252645015717, | |
| "step": 4120, | |
| "valid_targets_mean": 2153.9, | |
| "valid_targets_min": 973 | |
| }, | |
| { | |
| "epoch": 4.332983193277311, | |
| "grad_norm": 0.6915688934573623, | |
| "learning_rate": 1.5242253876163399e-05, | |
| "loss": 0.1911, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18757754564285278, | |
| "step": 4125, | |
| "valid_targets_mean": 3134.6, | |
| "valid_targets_min": 1432 | |
| }, | |
| { | |
| "epoch": 4.338235294117647, | |
| "grad_norm": 0.7700752339493685, | |
| "learning_rate": 1.5191388043991712e-05, | |
| "loss": 0.1789, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20721203088760376, | |
| "step": 4130, | |
| "valid_targets_mean": 2372.6, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 4.343487394957983, | |
| "grad_norm": 0.7313076777239362, | |
| "learning_rate": 1.5140555202490359e-05, | |
| "loss": 0.1817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16860508918762207, | |
| "step": 4135, | |
| "valid_targets_mean": 2125.1, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 4.348739495798319, | |
| "grad_norm": 0.7070894615662773, | |
| "learning_rate": 1.5089755700410602e-05, | |
| "loss": 0.1916, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1806333363056183, | |
| "step": 4140, | |
| "valid_targets_mean": 2950.4, | |
| "valid_targets_min": 1042 | |
| }, | |
| { | |
| "epoch": 4.3539915966386555, | |
| "grad_norm": 0.7047900654504996, | |
| "learning_rate": 1.5038989886274992e-05, | |
| "loss": 0.183, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17181357741355896, | |
| "step": 4145, | |
| "valid_targets_mean": 2456.4, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 4.359243697478991, | |
| "grad_norm": 0.7947773694315373, | |
| "learning_rate": 1.498825810837492e-05, | |
| "loss": 0.1824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1729154884815216, | |
| "step": 4150, | |
| "valid_targets_mean": 2260.0, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 4.364495798319328, | |
| "grad_norm": 0.7438599144833491, | |
| "learning_rate": 1.4937560714768283e-05, | |
| "loss": 0.1797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17960473895072937, | |
| "step": 4155, | |
| "valid_targets_mean": 2432.4, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 4.369747899159664, | |
| "grad_norm": 0.771452228573944, | |
| "learning_rate": 1.4886898053277086e-05, | |
| "loss": 0.179, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21052835881710052, | |
| "step": 4160, | |
| "valid_targets_mean": 2578.2, | |
| "valid_targets_min": 769 | |
| }, | |
| { | |
| "epoch": 4.375, | |
| "grad_norm": 0.6592959235985212, | |
| "learning_rate": 1.4836270471485029e-05, | |
| "loss": 0.1706, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18025822937488556, | |
| "step": 4165, | |
| "valid_targets_mean": 3174.4, | |
| "valid_targets_min": 850 | |
| }, | |
| { | |
| "epoch": 4.380252100840336, | |
| "grad_norm": 0.818888534502337, | |
| "learning_rate": 1.4785678316735144e-05, | |
| "loss": 0.1788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1800573468208313, | |
| "step": 4170, | |
| "valid_targets_mean": 2206.2, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 4.385504201680672, | |
| "grad_norm": 0.6949510487488129, | |
| "learning_rate": 1.4735121936127406e-05, | |
| "loss": 0.1753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15901532769203186, | |
| "step": 4175, | |
| "valid_targets_mean": 2460.3, | |
| "valid_targets_min": 309 | |
| }, | |
| { | |
| "epoch": 4.390756302521009, | |
| "grad_norm": 0.7369699549502026, | |
| "learning_rate": 1.4684601676516366e-05, | |
| "loss": 0.1779, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1716357171535492, | |
| "step": 4180, | |
| "valid_targets_mean": 2640.8, | |
| "valid_targets_min": 626 | |
| }, | |
| { | |
| "epoch": 4.3960084033613445, | |
| "grad_norm": 0.8430612932856147, | |
| "learning_rate": 1.4634117884508738e-05, | |
| "loss": 0.1757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17793697118759155, | |
| "step": 4185, | |
| "valid_targets_mean": 1770.9, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 4.401260504201681, | |
| "grad_norm": 0.7931502126995679, | |
| "learning_rate": 1.458367090646105e-05, | |
| "loss": 0.1797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20347827672958374, | |
| "step": 4190, | |
| "valid_targets_mean": 2522.5, | |
| "valid_targets_min": 985 | |
| }, | |
| { | |
| "epoch": 4.406512605042017, | |
| "grad_norm": 0.7167848190563175, | |
| "learning_rate": 1.4533261088477257e-05, | |
| "loss": 0.1841, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18484370410442352, | |
| "step": 4195, | |
| "valid_targets_mean": 2692.8, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 4.411764705882353, | |
| "grad_norm": 0.7138856506553442, | |
| "learning_rate": 1.448288877640637e-05, | |
| "loss": 0.1824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1668294370174408, | |
| "step": 4200, | |
| "valid_targets_mean": 2518.8, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 4.417016806722689, | |
| "grad_norm": 0.6504444444519508, | |
| "learning_rate": 1.4432554315840082e-05, | |
| "loss": 0.1783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18564704060554504, | |
| "step": 4205, | |
| "valid_targets_mean": 3140.6, | |
| "valid_targets_min": 1135 | |
| }, | |
| { | |
| "epoch": 4.4222689075630255, | |
| "grad_norm": 0.8499270398136098, | |
| "learning_rate": 1.438225805211039e-05, | |
| "loss": 0.1818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18404464423656464, | |
| "step": 4210, | |
| "valid_targets_mean": 1649.0, | |
| "valid_targets_min": 644 | |
| }, | |
| { | |
| "epoch": 4.427521008403361, | |
| "grad_norm": 0.715092529952324, | |
| "learning_rate": 1.4332000330287248e-05, | |
| "loss": 0.1821, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1709214150905609, | |
| "step": 4215, | |
| "valid_targets_mean": 2421.9, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 4.432773109243698, | |
| "grad_norm": 0.7288642734691909, | |
| "learning_rate": 1.428178149517617e-05, | |
| "loss": 0.1939, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20450027287006378, | |
| "step": 4220, | |
| "valid_targets_mean": 2933.5, | |
| "valid_targets_min": 1024 | |
| }, | |
| { | |
| "epoch": 4.438025210084033, | |
| "grad_norm": 0.795039692170866, | |
| "learning_rate": 1.4231601891315876e-05, | |
| "loss": 0.192, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18002474308013916, | |
| "step": 4225, | |
| "valid_targets_mean": 2059.5, | |
| "valid_targets_min": 1001 | |
| }, | |
| { | |
| "epoch": 4.44327731092437, | |
| "grad_norm": 0.988082973084519, | |
| "learning_rate": 1.418146186297594e-05, | |
| "loss": 0.1796, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18671777844429016, | |
| "step": 4230, | |
| "valid_targets_mean": 1872.4, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 4.448529411764706, | |
| "grad_norm": 0.7931578128546755, | |
| "learning_rate": 1.413136175415441e-05, | |
| "loss": 0.1714, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19192029535770416, | |
| "step": 4235, | |
| "valid_targets_mean": 2010.4, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 4.453781512605042, | |
| "grad_norm": 0.8177205538850968, | |
| "learning_rate": 1.4081301908575457e-05, | |
| "loss": 0.1848, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17905864119529724, | |
| "step": 4240, | |
| "valid_targets_mean": 1845.2, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 4.459033613445378, | |
| "grad_norm": 0.7284329325277255, | |
| "learning_rate": 1.4031282669687035e-05, | |
| "loss": 0.1836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1842656135559082, | |
| "step": 4245, | |
| "valid_targets_mean": 2472.1, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 4.464285714285714, | |
| "grad_norm": 0.7240658652686646, | |
| "learning_rate": 1.3981304380658474e-05, | |
| "loss": 0.1841, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19007527828216553, | |
| "step": 4250, | |
| "valid_targets_mean": 2486.7, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 4.46953781512605, | |
| "grad_norm": 0.7916628114284047, | |
| "learning_rate": 1.3931367384378183e-05, | |
| "loss": 0.1787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2028101086616516, | |
| "step": 4255, | |
| "valid_targets_mean": 2322.4, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 4.474789915966387, | |
| "grad_norm": 0.7379642382110667, | |
| "learning_rate": 1.3881472023451262e-05, | |
| "loss": 0.1867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18040382862091064, | |
| "step": 4260, | |
| "valid_targets_mean": 2429.3, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 4.480042016806722, | |
| "grad_norm": 0.7027762915252015, | |
| "learning_rate": 1.3831618640197153e-05, | |
| "loss": 0.1781, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1830647736787796, | |
| "step": 4265, | |
| "valid_targets_mean": 2796.7, | |
| "valid_targets_min": 887 | |
| }, | |
| { | |
| "epoch": 4.485294117647059, | |
| "grad_norm": 0.716947336725011, | |
| "learning_rate": 1.3781807576647315e-05, | |
| "loss": 0.1826, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1749706119298935, | |
| "step": 4270, | |
| "valid_targets_mean": 2428.2, | |
| "valid_targets_min": 805 | |
| }, | |
| { | |
| "epoch": 4.4905462184873945, | |
| "grad_norm": 0.7292281203695309, | |
| "learning_rate": 1.373203917454285e-05, | |
| "loss": 0.1799, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1867130696773529, | |
| "step": 4275, | |
| "valid_targets_mean": 2838.0, | |
| "valid_targets_min": 1197 | |
| }, | |
| { | |
| "epoch": 4.495798319327731, | |
| "grad_norm": 0.7290851774469034, | |
| "learning_rate": 1.3682313775332186e-05, | |
| "loss": 0.1899, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16877129673957825, | |
| "step": 4280, | |
| "valid_targets_mean": 2704.8, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 4.501050420168067, | |
| "grad_norm": 0.9037910643162406, | |
| "learning_rate": 1.3632631720168705e-05, | |
| "loss": 0.1801, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1778022050857544, | |
| "step": 4285, | |
| "valid_targets_mean": 1608.9, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 4.506302521008403, | |
| "grad_norm": 0.7874366555714657, | |
| "learning_rate": 1.358299334990842e-05, | |
| "loss": 0.1771, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1755077838897705, | |
| "step": 4290, | |
| "valid_targets_mean": 2098.4, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 4.51155462184874, | |
| "grad_norm": 0.8040088928448513, | |
| "learning_rate": 1.3533399005107635e-05, | |
| "loss": 0.198, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1978076994419098, | |
| "step": 4295, | |
| "valid_targets_mean": 2012.0, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 4.516806722689076, | |
| "grad_norm": 0.6756346527241319, | |
| "learning_rate": 1.3483849026020603e-05, | |
| "loss": 0.1768, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17955875396728516, | |
| "step": 4300, | |
| "valid_targets_mean": 2901.0, | |
| "valid_targets_min": 857 | |
| }, | |
| { | |
| "epoch": 4.522058823529412, | |
| "grad_norm": 0.7127125041012732, | |
| "learning_rate": 1.3434343752597199e-05, | |
| "loss": 0.1741, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18654534220695496, | |
| "step": 4305, | |
| "valid_targets_mean": 2637.3, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 4.527310924369748, | |
| "grad_norm": 0.6197339001895822, | |
| "learning_rate": 1.3384883524480576e-05, | |
| "loss": 0.1813, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15983977913856506, | |
| "step": 4310, | |
| "valid_targets_mean": 3360.6, | |
| "valid_targets_min": 991 | |
| }, | |
| { | |
| "epoch": 4.532563025210084, | |
| "grad_norm": 0.7186207642971952, | |
| "learning_rate": 1.333546868100486e-05, | |
| "loss": 0.1745, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18633624911308289, | |
| "step": 4315, | |
| "valid_targets_mean": 2568.3, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 4.53781512605042, | |
| "grad_norm": 0.6506719666430806, | |
| "learning_rate": 1.3286099561192784e-05, | |
| "loss": 0.1806, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20066654682159424, | |
| "step": 4320, | |
| "valid_targets_mean": 3334.8, | |
| "valid_targets_min": 1078 | |
| }, | |
| { | |
| "epoch": 4.543067226890757, | |
| "grad_norm": 0.7139868060577005, | |
| "learning_rate": 1.3236776503753397e-05, | |
| "loss": 0.1825, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1832045167684555, | |
| "step": 4325, | |
| "valid_targets_mean": 2449.9, | |
| "valid_targets_min": 889 | |
| }, | |
| { | |
| "epoch": 4.548319327731092, | |
| "grad_norm": 0.8346149773596157, | |
| "learning_rate": 1.3187499847079713e-05, | |
| "loss": 0.1764, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17674115300178528, | |
| "step": 4330, | |
| "valid_targets_mean": 1668.5, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 4.553571428571429, | |
| "grad_norm": 0.8581192326236325, | |
| "learning_rate": 1.3138269929246427e-05, | |
| "loss": 0.1752, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.169841468334198, | |
| "step": 4335, | |
| "valid_targets_mean": 2480.5, | |
| "valid_targets_min": 758 | |
| }, | |
| { | |
| "epoch": 4.5588235294117645, | |
| "grad_norm": 0.6712603189184432, | |
| "learning_rate": 1.308908708800753e-05, | |
| "loss": 0.1947, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17790794372558594, | |
| "step": 4340, | |
| "valid_targets_mean": 3218.3, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 4.564075630252101, | |
| "grad_norm": 0.8382631585908171, | |
| "learning_rate": 1.3039951660794063e-05, | |
| "loss": 0.188, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20454202592372894, | |
| "step": 4345, | |
| "valid_targets_mean": 2153.2, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 4.569327731092437, | |
| "grad_norm": 0.7788139909395001, | |
| "learning_rate": 1.2990863984711773e-05, | |
| "loss": 0.181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.187415212392807, | |
| "step": 4350, | |
| "valid_targets_mean": 2221.0, | |
| "valid_targets_min": 784 | |
| }, | |
| { | |
| "epoch": 4.574579831932773, | |
| "grad_norm": 0.8228381865776829, | |
| "learning_rate": 1.294182439653878e-05, | |
| "loss": 0.1882, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2052648365497589, | |
| "step": 4355, | |
| "valid_targets_mean": 2024.1, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 4.579831932773109, | |
| "grad_norm": 0.7969097110247945, | |
| "learning_rate": 1.2892833232723297e-05, | |
| "loss": 0.1863, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19444799423217773, | |
| "step": 4360, | |
| "valid_targets_mean": 2430.7, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 4.5850840336134455, | |
| "grad_norm": 0.8950927024396919, | |
| "learning_rate": 1.2843890829381303e-05, | |
| "loss": 0.1819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1735416203737259, | |
| "step": 4365, | |
| "valid_targets_mean": 1948.1, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 4.590336134453781, | |
| "grad_norm": 0.7524319791833006, | |
| "learning_rate": 1.2794997522294263e-05, | |
| "loss": 0.1787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17769590020179749, | |
| "step": 4370, | |
| "valid_targets_mean": 2317.9, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 4.595588235294118, | |
| "grad_norm": 0.8185328703075273, | |
| "learning_rate": 1.2746153646906766e-05, | |
| "loss": 0.1722, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17284256219863892, | |
| "step": 4375, | |
| "valid_targets_mean": 1852.3, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 4.600840336134453, | |
| "grad_norm": 0.8366151911421962, | |
| "learning_rate": 1.2697359538324303e-05, | |
| "loss": 0.1839, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17999669909477234, | |
| "step": 4380, | |
| "valid_targets_mean": 2349.4, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 4.60609243697479, | |
| "grad_norm": 0.74572952017287, | |
| "learning_rate": 1.2648615531310907e-05, | |
| "loss": 0.1805, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18219012022018433, | |
| "step": 4385, | |
| "valid_targets_mean": 2272.4, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 4.6113445378151265, | |
| "grad_norm": 0.7591238610620517, | |
| "learning_rate": 1.259992196028688e-05, | |
| "loss": 0.1772, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1870465874671936, | |
| "step": 4390, | |
| "valid_targets_mean": 2377.6, | |
| "valid_targets_min": 884 | |
| }, | |
| { | |
| "epoch": 4.616596638655462, | |
| "grad_norm": 0.7984813374731973, | |
| "learning_rate": 1.2551279159326495e-05, | |
| "loss": 0.1853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1903630644083023, | |
| "step": 4395, | |
| "valid_targets_mean": 2266.4, | |
| "valid_targets_min": 868 | |
| }, | |
| { | |
| "epoch": 4.621848739495798, | |
| "grad_norm": 0.7169284378785025, | |
| "learning_rate": 1.2502687462155709e-05, | |
| "loss": 0.1767, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17516914010047913, | |
| "step": 4400, | |
| "valid_targets_mean": 2642.4, | |
| "valid_targets_min": 1050 | |
| }, | |
| { | |
| "epoch": 4.6271008403361344, | |
| "grad_norm": 0.7778626068632182, | |
| "learning_rate": 1.2454147202149865e-05, | |
| "loss": 0.1873, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18522366881370544, | |
| "step": 4405, | |
| "valid_targets_mean": 2276.4, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 4.632352941176471, | |
| "grad_norm": 0.7163788723425591, | |
| "learning_rate": 1.2405658712331409e-05, | |
| "loss": 0.1718, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17094354331493378, | |
| "step": 4410, | |
| "valid_targets_mean": 2692.1, | |
| "valid_targets_min": 1078 | |
| }, | |
| { | |
| "epoch": 4.637605042016807, | |
| "grad_norm": 0.7021250334619412, | |
| "learning_rate": 1.2357222325367604e-05, | |
| "loss": 0.1798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2073124796152115, | |
| "step": 4415, | |
| "valid_targets_mean": 3277.5, | |
| "valid_targets_min": 998 | |
| }, | |
| { | |
| "epoch": 4.642857142857143, | |
| "grad_norm": 0.7042673790509252, | |
| "learning_rate": 1.2308838373568249e-05, | |
| "loss": 0.1788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18093574047088623, | |
| "step": 4420, | |
| "valid_targets_mean": 3193.1, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 4.648109243697479, | |
| "grad_norm": 0.7080962295239951, | |
| "learning_rate": 1.2260507188883395e-05, | |
| "loss": 0.1802, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.179480642080307, | |
| "step": 4425, | |
| "valid_targets_mean": 2543.3, | |
| "valid_targets_min": 967 | |
| }, | |
| { | |
| "epoch": 4.6533613445378155, | |
| "grad_norm": 0.7498653248193045, | |
| "learning_rate": 1.2212229102901077e-05, | |
| "loss": 0.1869, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1921127885580063, | |
| "step": 4430, | |
| "valid_targets_mean": 2489.6, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 4.658613445378151, | |
| "grad_norm": 0.7571679788787566, | |
| "learning_rate": 1.2164004446845037e-05, | |
| "loss": 0.1803, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19102567434310913, | |
| "step": 4435, | |
| "valid_targets_mean": 2422.0, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 4.663865546218488, | |
| "grad_norm": 0.7458838101603535, | |
| "learning_rate": 1.2115833551572438e-05, | |
| "loss": 0.1937, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18319690227508545, | |
| "step": 4440, | |
| "valid_targets_mean": 2450.1, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 4.669117647058823, | |
| "grad_norm": 0.7046359584542661, | |
| "learning_rate": 1.2067716747571616e-05, | |
| "loss": 0.1881, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1766650229692459, | |
| "step": 4445, | |
| "valid_targets_mean": 2313.4, | |
| "valid_targets_min": 843 | |
| }, | |
| { | |
| "epoch": 4.67436974789916, | |
| "grad_norm": 0.8345834489540149, | |
| "learning_rate": 1.201965436495978e-05, | |
| "loss": 0.179, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19209042191505432, | |
| "step": 4450, | |
| "valid_targets_mean": 2150.6, | |
| "valid_targets_min": 962 | |
| }, | |
| { | |
| "epoch": 4.679621848739496, | |
| "grad_norm": 0.763774531100772, | |
| "learning_rate": 1.1971646733480784e-05, | |
| "loss": 0.1822, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20099681615829468, | |
| "step": 4455, | |
| "valid_targets_mean": 2721.1, | |
| "valid_targets_min": 1033 | |
| }, | |
| { | |
| "epoch": 4.684873949579832, | |
| "grad_norm": 0.6918104106739509, | |
| "learning_rate": 1.1923694182502848e-05, | |
| "loss": 0.1841, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17281074821949005, | |
| "step": 4460, | |
| "valid_targets_mean": 2406.6, | |
| "valid_targets_min": 978 | |
| }, | |
| { | |
| "epoch": 4.690126050420168, | |
| "grad_norm": 0.7551738929861327, | |
| "learning_rate": 1.1875797041016286e-05, | |
| "loss": 0.1825, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1859922856092453, | |
| "step": 4465, | |
| "valid_targets_mean": 2412.7, | |
| "valid_targets_min": 911 | |
| }, | |
| { | |
| "epoch": 4.695378151260504, | |
| "grad_norm": 0.7758417255586617, | |
| "learning_rate": 1.1827955637631283e-05, | |
| "loss": 0.1811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16596679389476776, | |
| "step": 4470, | |
| "valid_targets_mean": 2247.4, | |
| "valid_targets_min": 833 | |
| }, | |
| { | |
| "epoch": 4.70063025210084, | |
| "grad_norm": 0.6806231945697778, | |
| "learning_rate": 1.1780170300575602e-05, | |
| "loss": 0.1812, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1946016550064087, | |
| "step": 4475, | |
| "valid_targets_mean": 3193.5, | |
| "valid_targets_min": 1143 | |
| }, | |
| { | |
| "epoch": 4.705882352941177, | |
| "grad_norm": 0.7375240636905129, | |
| "learning_rate": 1.1732441357692353e-05, | |
| "loss": 0.1871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17887581884860992, | |
| "step": 4480, | |
| "valid_targets_mean": 2464.6, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 4.711134453781512, | |
| "grad_norm": 0.7620340981837231, | |
| "learning_rate": 1.1684769136437742e-05, | |
| "loss": 0.1778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19154343008995056, | |
| "step": 4485, | |
| "valid_targets_mean": 2281.2, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 4.716386554621849, | |
| "grad_norm": 0.82056252467419, | |
| "learning_rate": 1.1637153963878815e-05, | |
| "loss": 0.1817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19052766263484955, | |
| "step": 4490, | |
| "valid_targets_mean": 2145.5, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 4.7216386554621845, | |
| "grad_norm": 0.7301286521046133, | |
| "learning_rate": 1.1589596166691231e-05, | |
| "loss": 0.1766, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1829635053873062, | |
| "step": 4495, | |
| "valid_targets_mean": 2482.4, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 4.726890756302521, | |
| "grad_norm": 0.7903970658437431, | |
| "learning_rate": 1.1542096071157012e-05, | |
| "loss": 0.1753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1784076690673828, | |
| "step": 4500, | |
| "valid_targets_mean": 2605.2, | |
| "valid_targets_min": 950 | |
| }, | |
| { | |
| "epoch": 4.732142857142857, | |
| "grad_norm": 0.7637658355202966, | |
| "learning_rate": 1.1494654003162285e-05, | |
| "loss": 0.1745, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18634071946144104, | |
| "step": 4505, | |
| "valid_targets_mean": 2553.0, | |
| "valid_targets_min": 950 | |
| }, | |
| { | |
| "epoch": 4.737394957983193, | |
| "grad_norm": 0.7221781006332623, | |
| "learning_rate": 1.1447270288195089e-05, | |
| "loss": 0.1778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17665636539459229, | |
| "step": 4510, | |
| "valid_targets_mean": 2349.6, | |
| "valid_targets_min": 847 | |
| }, | |
| { | |
| "epoch": 4.742647058823529, | |
| "grad_norm": 0.720008738864535, | |
| "learning_rate": 1.1399945251343114e-05, | |
| "loss": 0.1815, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.183954656124115, | |
| "step": 4515, | |
| "valid_targets_mean": 2651.1, | |
| "valid_targets_min": 963 | |
| }, | |
| { | |
| "epoch": 4.7478991596638656, | |
| "grad_norm": 0.7439764441972336, | |
| "learning_rate": 1.1352679217291457e-05, | |
| "loss": 0.1842, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18053632974624634, | |
| "step": 4520, | |
| "valid_targets_mean": 2257.0, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 4.753151260504202, | |
| "grad_norm": 0.6873632051406735, | |
| "learning_rate": 1.1305472510320419e-05, | |
| "loss": 0.1783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16327188909053802, | |
| "step": 4525, | |
| "valid_targets_mean": 2640.4, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 4.758403361344538, | |
| "grad_norm": 0.7442947567531205, | |
| "learning_rate": 1.1258325454303286e-05, | |
| "loss": 0.195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1675795316696167, | |
| "step": 4530, | |
| "valid_targets_mean": 2390.1, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 4.7636554621848735, | |
| "grad_norm": 0.7354254137750549, | |
| "learning_rate": 1.1211238372704073e-05, | |
| "loss": 0.1887, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17392656207084656, | |
| "step": 4535, | |
| "valid_targets_mean": 2248.0, | |
| "valid_targets_min": 1064 | |
| }, | |
| { | |
| "epoch": 4.76890756302521, | |
| "grad_norm": 0.7224478673738945, | |
| "learning_rate": 1.1164211588575339e-05, | |
| "loss": 0.1845, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16614243388175964, | |
| "step": 4540, | |
| "valid_targets_mean": 2204.6, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 4.774159663865547, | |
| "grad_norm": 0.6780397730456732, | |
| "learning_rate": 1.1117245424555967e-05, | |
| "loss": 0.1836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17192646861076355, | |
| "step": 4545, | |
| "valid_targets_mean": 2732.8, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 4.779411764705882, | |
| "grad_norm": 0.8114966919737321, | |
| "learning_rate": 1.1070340202868915e-05, | |
| "loss": 0.1715, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16627684235572815, | |
| "step": 4550, | |
| "valid_targets_mean": 2037.4, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 4.784663865546219, | |
| "grad_norm": 0.7089064475714499, | |
| "learning_rate": 1.1023496245319056e-05, | |
| "loss": 0.1743, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20096446573734283, | |
| "step": 4555, | |
| "valid_targets_mean": 3190.9, | |
| "valid_targets_min": 1018 | |
| }, | |
| { | |
| "epoch": 4.7899159663865545, | |
| "grad_norm": 0.7629594666078792, | |
| "learning_rate": 1.0976713873290925e-05, | |
| "loss": 0.1789, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18434454500675201, | |
| "step": 4560, | |
| "valid_targets_mean": 2231.0, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 4.795168067226891, | |
| "grad_norm": 0.7206103511159119, | |
| "learning_rate": 1.0929993407746543e-05, | |
| "loss": 0.1787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1781184822320938, | |
| "step": 4565, | |
| "valid_targets_mean": 2413.1, | |
| "valid_targets_min": 924 | |
| }, | |
| { | |
| "epoch": 4.800420168067227, | |
| "grad_norm": 0.7319103361120156, | |
| "learning_rate": 1.0883335169223212e-05, | |
| "loss": 0.181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19576598703861237, | |
| "step": 4570, | |
| "valid_targets_mean": 2644.1, | |
| "valid_targets_min": 1339 | |
| }, | |
| { | |
| "epoch": 4.805672268907563, | |
| "grad_norm": 0.7224561194750161, | |
| "learning_rate": 1.0836739477831297e-05, | |
| "loss": 0.1769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16441801190376282, | |
| "step": 4575, | |
| "valid_targets_mean": 2237.8, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 4.810924369747899, | |
| "grad_norm": 0.8825086906465849, | |
| "learning_rate": 1.0790206653252055e-05, | |
| "loss": 0.169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18621914088726044, | |
| "step": 4580, | |
| "valid_targets_mean": 2011.2, | |
| "valid_targets_min": 631 | |
| }, | |
| { | |
| "epoch": 4.8161764705882355, | |
| "grad_norm": 0.6887944713418872, | |
| "learning_rate": 1.0743737014735434e-05, | |
| "loss": 0.1796, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18349093198776245, | |
| "step": 4585, | |
| "valid_targets_mean": 2805.9, | |
| "valid_targets_min": 595 | |
| }, | |
| { | |
| "epoch": 4.821428571428571, | |
| "grad_norm": 0.8266516711323478, | |
| "learning_rate": 1.0697330881097857e-05, | |
| "loss": 0.1889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18345323204994202, | |
| "step": 4590, | |
| "valid_targets_mean": 2207.9, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 4.826680672268908, | |
| "grad_norm": 0.7454608401910962, | |
| "learning_rate": 1.0650988570720077e-05, | |
| "loss": 0.1695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16886943578720093, | |
| "step": 4595, | |
| "valid_targets_mean": 2130.4, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 4.831932773109243, | |
| "grad_norm": 0.7525541827091601, | |
| "learning_rate": 1.0604710401544963e-05, | |
| "loss": 0.1889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2049073427915573, | |
| "step": 4600, | |
| "valid_targets_mean": 2671.1, | |
| "valid_targets_min": 938 | |
| }, | |
| { | |
| "epoch": 4.83718487394958, | |
| "grad_norm": 0.7221773649936342, | |
| "learning_rate": 1.0558496691075318e-05, | |
| "loss": 0.1874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18932510912418365, | |
| "step": 4605, | |
| "valid_targets_mean": 2382.8, | |
| "valid_targets_min": 533 | |
| }, | |
| { | |
| "epoch": 4.842436974789916, | |
| "grad_norm": 0.6282371291333917, | |
| "learning_rate": 1.0512347756371726e-05, | |
| "loss": 0.1732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18067455291748047, | |
| "step": 4610, | |
| "valid_targets_mean": 3488.6, | |
| "valid_targets_min": 1483 | |
| }, | |
| { | |
| "epoch": 4.847689075630252, | |
| "grad_norm": 0.7736804746763539, | |
| "learning_rate": 1.0466263914050362e-05, | |
| "loss": 0.1866, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.187209352850914, | |
| "step": 4615, | |
| "valid_targets_mean": 2130.1, | |
| "valid_targets_min": 1190 | |
| }, | |
| { | |
| "epoch": 4.852941176470588, | |
| "grad_norm": 0.8194923770615844, | |
| "learning_rate": 1.0420245480280804e-05, | |
| "loss": 0.1711, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19902583956718445, | |
| "step": 4620, | |
| "valid_targets_mean": 2030.2, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 4.858193277310924, | |
| "grad_norm": 0.7237311731603411, | |
| "learning_rate": 1.0374292770783891e-05, | |
| "loss": 0.1885, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18148672580718994, | |
| "step": 4625, | |
| "valid_targets_mean": 2961.3, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 4.86344537815126, | |
| "grad_norm": 0.6669660493298567, | |
| "learning_rate": 1.0328406100829542e-05, | |
| "loss": 0.1876, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17993488907814026, | |
| "step": 4630, | |
| "valid_targets_mean": 3014.6, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 4.868697478991597, | |
| "grad_norm": 0.8700432106817276, | |
| "learning_rate": 1.0282585785234578e-05, | |
| "loss": 0.1833, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18658801913261414, | |
| "step": 4635, | |
| "valid_targets_mean": 1932.2, | |
| "valid_targets_min": 930 | |
| }, | |
| { | |
| "epoch": 4.873949579831933, | |
| "grad_norm": 0.6217143473547485, | |
| "learning_rate": 1.023683213836061e-05, | |
| "loss": 0.1788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1737050712108612, | |
| "step": 4640, | |
| "valid_targets_mean": 3531.2, | |
| "valid_targets_min": 844 | |
| }, | |
| { | |
| "epoch": 4.879201680672269, | |
| "grad_norm": 0.9539626771397886, | |
| "learning_rate": 1.0191145474111823e-05, | |
| "loss": 0.1846, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19349545240402222, | |
| "step": 4645, | |
| "valid_targets_mean": 1414.2, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 4.884453781512605, | |
| "grad_norm": 0.6923182425799058, | |
| "learning_rate": 1.0145526105932872e-05, | |
| "loss": 0.1712, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16115731000900269, | |
| "step": 4650, | |
| "valid_targets_mean": 2663.4, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 4.889705882352941, | |
| "grad_norm": 0.6914625755370188, | |
| "learning_rate": 1.0099974346806714e-05, | |
| "loss": 0.1916, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1792561411857605, | |
| "step": 4655, | |
| "valid_targets_mean": 2781.5, | |
| "valid_targets_min": 1252 | |
| }, | |
| { | |
| "epoch": 4.894957983193278, | |
| "grad_norm": 0.7734782088863728, | |
| "learning_rate": 1.0054490509252423e-05, | |
| "loss": 0.191, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20624291896820068, | |
| "step": 4660, | |
| "valid_targets_mean": 2440.4, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 4.900210084033613, | |
| "grad_norm": 0.7941952415974891, | |
| "learning_rate": 1.0009074905323118e-05, | |
| "loss": 0.1686, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17345239222049713, | |
| "step": 4665, | |
| "valid_targets_mean": 2916.6, | |
| "valid_targets_min": 1297 | |
| }, | |
| { | |
| "epoch": 4.90546218487395, | |
| "grad_norm": 0.8030163177897169, | |
| "learning_rate": 9.963727846603784e-06, | |
| "loss": 0.1794, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18559759855270386, | |
| "step": 4670, | |
| "valid_targets_mean": 2144.8, | |
| "valid_targets_min": 987 | |
| }, | |
| { | |
| "epoch": 4.910714285714286, | |
| "grad_norm": 0.7030553547126519, | |
| "learning_rate": 9.918449644209087e-06, | |
| "loss": 0.1852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18912020325660706, | |
| "step": 4675, | |
| "valid_targets_mean": 2720.7, | |
| "valid_targets_min": 1084 | |
| }, | |
| { | |
| "epoch": 4.915966386554622, | |
| "grad_norm": 0.6845418285443712, | |
| "learning_rate": 9.873240608781341e-06, | |
| "loss": 0.182, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.202968031167984, | |
| "step": 4680, | |
| "valid_targets_mean": 3153.9, | |
| "valid_targets_min": 1043 | |
| }, | |
| { | |
| "epoch": 4.921218487394958, | |
| "grad_norm": 0.7655672446200298, | |
| "learning_rate": 9.828101050488308e-06, | |
| "loss": 0.1873, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18019559979438782, | |
| "step": 4685, | |
| "valid_targets_mean": 1993.6, | |
| "valid_targets_min": 1195 | |
| }, | |
| { | |
| "epoch": 4.926470588235294, | |
| "grad_norm": 0.7622908992963848, | |
| "learning_rate": 9.783031279021063e-06, | |
| "loss": 0.1795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19854578375816345, | |
| "step": 4690, | |
| "valid_targets_mean": 2434.5, | |
| "valid_targets_min": 1013 | |
| }, | |
| { | |
| "epoch": 4.93172268907563, | |
| "grad_norm": 0.7356813022171075, | |
| "learning_rate": 9.738031603591926e-06, | |
| "loss": 0.1864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1860363483428955, | |
| "step": 4695, | |
| "valid_targets_mean": 2400.0, | |
| "valid_targets_min": 934 | |
| }, | |
| { | |
| "epoch": 4.936974789915967, | |
| "grad_norm": 0.720810615748148, | |
| "learning_rate": 9.69310233293227e-06, | |
| "loss": 0.1853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1856434941291809, | |
| "step": 4700, | |
| "valid_targets_mean": 2621.8, | |
| "valid_targets_min": 1276 | |
| }, | |
| { | |
| "epoch": 4.942226890756302, | |
| "grad_norm": 0.7063545010234002, | |
| "learning_rate": 9.648243775290476e-06, | |
| "loss": 0.1826, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16022886335849762, | |
| "step": 4705, | |
| "valid_targets_mean": 2576.8, | |
| "valid_targets_min": 1221 | |
| }, | |
| { | |
| "epoch": 4.947478991596639, | |
| "grad_norm": 0.7386993073314914, | |
| "learning_rate": 9.60345623842974e-06, | |
| "loss": 0.1793, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17438361048698425, | |
| "step": 4710, | |
| "valid_targets_mean": 2572.9, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 4.9527310924369745, | |
| "grad_norm": 0.7347227314045449, | |
| "learning_rate": 9.558740029626046e-06, | |
| "loss": 0.1749, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1749304234981537, | |
| "step": 4715, | |
| "valid_targets_mean": 2584.6, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 4.957983193277311, | |
| "grad_norm": 0.728403778635231, | |
| "learning_rate": 9.51409545566597e-06, | |
| "loss": 0.1797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18584081530570984, | |
| "step": 4720, | |
| "valid_targets_mean": 2500.6, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 4.963235294117647, | |
| "grad_norm": 0.7578264772501891, | |
| "learning_rate": 9.469522822844663e-06, | |
| "loss": 0.1826, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1682163029909134, | |
| "step": 4725, | |
| "valid_targets_mean": 2061.4, | |
| "valid_targets_min": 691 | |
| }, | |
| { | |
| "epoch": 4.968487394957983, | |
| "grad_norm": 0.7750388844604879, | |
| "learning_rate": 9.425022436963664e-06, | |
| "loss": 0.1932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19586464762687683, | |
| "step": 4730, | |
| "valid_targets_mean": 2407.8, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 4.973739495798319, | |
| "grad_norm": 0.7766444281597565, | |
| "learning_rate": 9.380594603328875e-06, | |
| "loss": 0.1984, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2102508395910263, | |
| "step": 4735, | |
| "valid_targets_mean": 2843.3, | |
| "valid_targets_min": 639 | |
| }, | |
| { | |
| "epoch": 4.9789915966386555, | |
| "grad_norm": 0.760942443434771, | |
| "learning_rate": 9.336239626748432e-06, | |
| "loss": 0.1806, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17757880687713623, | |
| "step": 4740, | |
| "valid_targets_mean": 2666.6, | |
| "valid_targets_min": 825 | |
| }, | |
| { | |
| "epoch": 4.984243697478991, | |
| "grad_norm": 0.7731058692289628, | |
| "learning_rate": 9.291957811530602e-06, | |
| "loss": 0.1837, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18547788262367249, | |
| "step": 4745, | |
| "valid_targets_mean": 2214.3, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 4.989495798319328, | |
| "grad_norm": 0.700031528165842, | |
| "learning_rate": 9.247749461481712e-06, | |
| "loss": 0.1696, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18496280908584595, | |
| "step": 4750, | |
| "valid_targets_mean": 2794.8, | |
| "valid_targets_min": 945 | |
| }, | |
| { | |
| "epoch": 4.994747899159664, | |
| "grad_norm": 0.6273143784305558, | |
| "learning_rate": 9.203614879904083e-06, | |
| "loss": 0.1798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15987840294837952, | |
| "step": 4755, | |
| "valid_targets_mean": 3185.1, | |
| "valid_targets_min": 1250 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.6788365442881951, | |
| "learning_rate": 9.159554369593901e-06, | |
| "loss": 0.1866, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17320957779884338, | |
| "step": 4760, | |
| "valid_targets_mean": 2633.3, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 5.005252100840337, | |
| "grad_norm": 0.7483605516753753, | |
| "learning_rate": 9.115568232839193e-06, | |
| "loss": 0.1509, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12966476380825043, | |
| "step": 4765, | |
| "valid_targets_mean": 1753.9, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 5.010504201680672, | |
| "grad_norm": 0.7211521050334696, | |
| "learning_rate": 9.071656771417721e-06, | |
| "loss": 0.1588, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16123877465724945, | |
| "step": 4770, | |
| "valid_targets_mean": 2470.6, | |
| "valid_targets_min": 1058 | |
| }, | |
| { | |
| "epoch": 5.015756302521009, | |
| "grad_norm": 0.7296273456769096, | |
| "learning_rate": 9.027820286594902e-06, | |
| "loss": 0.1619, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14505019783973694, | |
| "step": 4775, | |
| "valid_targets_mean": 2658.1, | |
| "valid_targets_min": 1128 | |
| }, | |
| { | |
| "epoch": 5.0210084033613445, | |
| "grad_norm": 0.7817862758275839, | |
| "learning_rate": 8.984059079121785e-06, | |
| "loss": 0.1634, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15351414680480957, | |
| "step": 4780, | |
| "valid_targets_mean": 2114.3, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 5.026260504201681, | |
| "grad_norm": 0.7668268893222531, | |
| "learning_rate": 8.940373449232935e-06, | |
| "loss": 0.1666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16739562153816223, | |
| "step": 4785, | |
| "valid_targets_mean": 2364.6, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 5.031512605042017, | |
| "grad_norm": 0.820070132627084, | |
| "learning_rate": 8.896763696644405e-06, | |
| "loss": 0.1568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15973633527755737, | |
| "step": 4790, | |
| "valid_targets_mean": 1928.4, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 5.036764705882353, | |
| "grad_norm": 0.7252481874686856, | |
| "learning_rate": 8.853230120551693e-06, | |
| "loss": 0.1583, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14669166505336761, | |
| "step": 4795, | |
| "valid_targets_mean": 2317.4, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 5.042016806722689, | |
| "grad_norm": 0.8044647525929012, | |
| "learning_rate": 8.809773019627635e-06, | |
| "loss": 0.1566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17486470937728882, | |
| "step": 4800, | |
| "valid_targets_mean": 2453.2, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 5.0472689075630255, | |
| "grad_norm": 0.7168891105498033, | |
| "learning_rate": 8.766392692020413e-06, | |
| "loss": 0.1741, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16313612461090088, | |
| "step": 4805, | |
| "valid_targets_mean": 3045.1, | |
| "valid_targets_min": 1015 | |
| }, | |
| { | |
| "epoch": 5.052521008403361, | |
| "grad_norm": 0.8391271783579383, | |
| "learning_rate": 8.723089435351497e-06, | |
| "loss": 0.1626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17179730534553528, | |
| "step": 4810, | |
| "valid_targets_mean": 2010.1, | |
| "valid_targets_min": 836 | |
| }, | |
| { | |
| "epoch": 5.057773109243698, | |
| "grad_norm": 0.7301522003326225, | |
| "learning_rate": 8.679863546713559e-06, | |
| "loss": 0.1576, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17533841729164124, | |
| "step": 4815, | |
| "valid_targets_mean": 3249.5, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 5.063025210084033, | |
| "grad_norm": 0.886339846842999, | |
| "learning_rate": 8.636715322668498e-06, | |
| "loss": 0.1679, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16565924882888794, | |
| "step": 4820, | |
| "valid_targets_mean": 2003.6, | |
| "valid_targets_min": 908 | |
| }, | |
| { | |
| "epoch": 5.06827731092437, | |
| "grad_norm": 0.7726449898609571, | |
| "learning_rate": 8.593645059245365e-06, | |
| "loss": 0.1668, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15337593853473663, | |
| "step": 4825, | |
| "valid_targets_mean": 2167.6, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 5.073529411764706, | |
| "grad_norm": 0.7695203697189681, | |
| "learning_rate": 8.550653051938333e-06, | |
| "loss": 0.1626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1670958697795868, | |
| "step": 4830, | |
| "valid_targets_mean": 2386.7, | |
| "valid_targets_min": 755 | |
| }, | |
| { | |
| "epoch": 5.078781512605042, | |
| "grad_norm": 0.8074962356870891, | |
| "learning_rate": 8.507739595704695e-06, | |
| "loss": 0.1523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1489427387714386, | |
| "step": 4835, | |
| "valid_targets_mean": 2025.5, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 5.084033613445378, | |
| "grad_norm": 0.8716776759288114, | |
| "learning_rate": 8.464904984962832e-06, | |
| "loss": 0.1634, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16994109749794006, | |
| "step": 4840, | |
| "valid_targets_mean": 2000.9, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 5.089285714285714, | |
| "grad_norm": 0.8076509459810718, | |
| "learning_rate": 8.422149513590151e-06, | |
| "loss": 0.1732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17325329780578613, | |
| "step": 4845, | |
| "valid_targets_mean": 2370.3, | |
| "valid_targets_min": 537 | |
| }, | |
| { | |
| "epoch": 5.09453781512605, | |
| "grad_norm": 0.8070190471856774, | |
| "learning_rate": 8.37947347492115e-06, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15762285888195038, | |
| "step": 4850, | |
| "valid_targets_mean": 2040.4, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 5.099789915966387, | |
| "grad_norm": 0.74092533277044, | |
| "learning_rate": 8.33687716174532e-06, | |
| "loss": 0.1603, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16209125518798828, | |
| "step": 4855, | |
| "valid_targets_mean": 2598.8, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 5.105042016806722, | |
| "grad_norm": 0.860251504616201, | |
| "learning_rate": 8.294360866305192e-06, | |
| "loss": 0.1628, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15497906506061554, | |
| "step": 4860, | |
| "valid_targets_mean": 2211.9, | |
| "valid_targets_min": 1015 | |
| }, | |
| { | |
| "epoch": 5.110294117647059, | |
| "grad_norm": 0.7907729718003471, | |
| "learning_rate": 8.251924880294317e-06, | |
| "loss": 0.1666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16414563357830048, | |
| "step": 4865, | |
| "valid_targets_mean": 2080.1, | |
| "valid_targets_min": 919 | |
| }, | |
| { | |
| "epoch": 5.1155462184873945, | |
| "grad_norm": 0.8047837776866085, | |
| "learning_rate": 8.20956949485527e-06, | |
| "loss": 0.1599, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15285125374794006, | |
| "step": 4870, | |
| "valid_targets_mean": 1991.6, | |
| "valid_targets_min": 571 | |
| }, | |
| { | |
| "epoch": 5.120798319327731, | |
| "grad_norm": 0.7304681966773338, | |
| "learning_rate": 8.167295000577622e-06, | |
| "loss": 0.1693, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15932945907115936, | |
| "step": 4875, | |
| "valid_targets_mean": 2356.6, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 5.126050420168067, | |
| "grad_norm": 0.7843064737707304, | |
| "learning_rate": 8.125101687496e-06, | |
| "loss": 0.1662, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1509605348110199, | |
| "step": 4880, | |
| "valid_targets_mean": 2156.1, | |
| "valid_targets_min": 945 | |
| }, | |
| { | |
| "epoch": 5.131302521008403, | |
| "grad_norm": 0.6507131542748491, | |
| "learning_rate": 8.082989845088038e-06, | |
| "loss": 0.1609, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.168035089969635, | |
| "step": 4885, | |
| "valid_targets_mean": 3697.9, | |
| "valid_targets_min": 1139 | |
| }, | |
| { | |
| "epoch": 5.13655462184874, | |
| "grad_norm": 0.7983692941356186, | |
| "learning_rate": 8.040959762272441e-06, | |
| "loss": 0.1711, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17778906226158142, | |
| "step": 4890, | |
| "valid_targets_mean": 2277.1, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 5.141806722689076, | |
| "grad_norm": 0.717526944817668, | |
| "learning_rate": 7.999011727406995e-06, | |
| "loss": 0.1605, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1474049985408783, | |
| "step": 4895, | |
| "valid_targets_mean": 3174.6, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 5.147058823529412, | |
| "grad_norm": 0.6501295927520986, | |
| "learning_rate": 7.957146028286524e-06, | |
| "loss": 0.159, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15974193811416626, | |
| "step": 4900, | |
| "valid_targets_mean": 3719.8, | |
| "valid_targets_min": 786 | |
| }, | |
| { | |
| "epoch": 5.152310924369748, | |
| "grad_norm": 0.7634163772813823, | |
| "learning_rate": 7.915362952141017e-06, | |
| "loss": 0.1627, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13993379473686218, | |
| "step": 4905, | |
| "valid_targets_mean": 2612.3, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 5.157563025210084, | |
| "grad_norm": 0.8229169359259055, | |
| "learning_rate": 7.873662785633594e-06, | |
| "loss": 0.1627, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17392058670520782, | |
| "step": 4910, | |
| "valid_targets_mean": 2420.1, | |
| "valid_targets_min": 685 | |
| }, | |
| { | |
| "epoch": 5.16281512605042, | |
| "grad_norm": 0.8391334241227872, | |
| "learning_rate": 7.832045814858538e-06, | |
| "loss": 0.1662, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15847104787826538, | |
| "step": 4915, | |
| "valid_targets_mean": 2340.0, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 5.168067226890757, | |
| "grad_norm": 0.7113772435226458, | |
| "learning_rate": 7.790512325339367e-06, | |
| "loss": 0.1679, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17543965578079224, | |
| "step": 4920, | |
| "valid_targets_mean": 2999.2, | |
| "valid_targets_min": 1254 | |
| }, | |
| { | |
| "epoch": 5.173319327731092, | |
| "grad_norm": 0.8373320616460568, | |
| "learning_rate": 7.74906260202685e-06, | |
| "loss": 0.1666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16569611430168152, | |
| "step": 4925, | |
| "valid_targets_mean": 2456.8, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 5.178571428571429, | |
| "grad_norm": 0.756975477682416, | |
| "learning_rate": 7.707696929297053e-06, | |
| "loss": 0.162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15648490190505981, | |
| "step": 4930, | |
| "valid_targets_mean": 2480.5, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 5.1838235294117645, | |
| "grad_norm": 0.841074398063682, | |
| "learning_rate": 7.666415590949382e-06, | |
| "loss": 0.1585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16663537919521332, | |
| "step": 4935, | |
| "valid_targets_mean": 2205.2, | |
| "valid_targets_min": 817 | |
| }, | |
| { | |
| "epoch": 5.189075630252101, | |
| "grad_norm": 0.8588889722655582, | |
| "learning_rate": 7.625218870204676e-06, | |
| "loss": 0.1608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16082826256752014, | |
| "step": 4940, | |
| "valid_targets_mean": 2507.4, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.194327731092437, | |
| "grad_norm": 0.7390042440173223, | |
| "learning_rate": 7.584107049703195e-06, | |
| "loss": 0.1645, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15865691006183624, | |
| "step": 4945, | |
| "valid_targets_mean": 2532.7, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 5.199579831932773, | |
| "grad_norm": 0.7806217716341791, | |
| "learning_rate": 7.5430804115027615e-06, | |
| "loss": 0.1606, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15922147035598755, | |
| "step": 4950, | |
| "valid_targets_mean": 2130.4, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 5.204831932773109, | |
| "grad_norm": 0.6872497366416, | |
| "learning_rate": 7.502139237076744e-06, | |
| "loss": 0.1603, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16006314754486084, | |
| "step": 4955, | |
| "valid_targets_mean": 3239.2, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 5.2100840336134455, | |
| "grad_norm": 0.8265925780625734, | |
| "learning_rate": 7.461283807312194e-06, | |
| "loss": 0.159, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15462356805801392, | |
| "step": 4960, | |
| "valid_targets_mean": 2074.3, | |
| "valid_targets_min": 736 | |
| }, | |
| { | |
| "epoch": 5.215336134453781, | |
| "grad_norm": 0.8400103721164567, | |
| "learning_rate": 7.420514402507886e-06, | |
| "loss": 0.1716, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16285106539726257, | |
| "step": 4965, | |
| "valid_targets_mean": 2134.1, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 5.220588235294118, | |
| "grad_norm": 0.7819781182832531, | |
| "learning_rate": 7.379831302372389e-06, | |
| "loss": 0.1633, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15541312098503113, | |
| "step": 4970, | |
| "valid_targets_mean": 2262.2, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 5.225840336134453, | |
| "grad_norm": 0.8105251128609028, | |
| "learning_rate": 7.3392347860221556e-06, | |
| "loss": 0.1707, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16982173919677734, | |
| "step": 4975, | |
| "valid_targets_mean": 2223.4, | |
| "valid_targets_min": 999 | |
| }, | |
| { | |
| "epoch": 5.23109243697479, | |
| "grad_norm": 0.6863440387762159, | |
| "learning_rate": 7.298725131979629e-06, | |
| "loss": 0.1632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15579064190387726, | |
| "step": 4980, | |
| "valid_targets_mean": 3175.9, | |
| "valid_targets_min": 1177 | |
| }, | |
| { | |
| "epoch": 5.236344537815126, | |
| "grad_norm": 0.7441930568222063, | |
| "learning_rate": 7.258302618171287e-06, | |
| "loss": 0.1701, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1511012315750122, | |
| "step": 4985, | |
| "valid_targets_mean": 2616.9, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 5.241596638655462, | |
| "grad_norm": 0.7374029049491165, | |
| "learning_rate": 7.217967521925775e-06, | |
| "loss": 0.1621, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16436655819416046, | |
| "step": 4990, | |
| "valid_targets_mean": 2690.6, | |
| "valid_targets_min": 874 | |
| }, | |
| { | |
| "epoch": 5.246848739495798, | |
| "grad_norm": 0.7497430314524601, | |
| "learning_rate": 7.177720119971998e-06, | |
| "loss": 0.1527, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17379173636436462, | |
| "step": 4995, | |
| "valid_targets_mean": 2911.5, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 5.2521008403361344, | |
| "grad_norm": 0.7942479630297596, | |
| "learning_rate": 7.137560688437184e-06, | |
| "loss": 0.1658, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1384781152009964, | |
| "step": 5000, | |
| "valid_targets_mean": 1960.1, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.257352941176471, | |
| "grad_norm": 0.6952857832526612, | |
| "learning_rate": 7.097489502845047e-06, | |
| "loss": 0.1552, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12435892224311829, | |
| "step": 5005, | |
| "valid_targets_mean": 2232.8, | |
| "valid_targets_min": 420 | |
| }, | |
| { | |
| "epoch": 5.262605042016807, | |
| "grad_norm": 0.7259264839559267, | |
| "learning_rate": 7.0575068381138525e-06, | |
| "loss": 0.1608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15408116579055786, | |
| "step": 5010, | |
| "valid_targets_mean": 2326.8, | |
| "valid_targets_min": 1105 | |
| }, | |
| { | |
| "epoch": 5.267857142857143, | |
| "grad_norm": 0.6907832100331577, | |
| "learning_rate": 7.0176129685545414e-06, | |
| "loss": 0.1601, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1581760197877884, | |
| "step": 5015, | |
| "valid_targets_mean": 2910.8, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 5.273109243697479, | |
| "grad_norm": 0.7359189102249872, | |
| "learning_rate": 6.977808167868867e-06, | |
| "loss": 0.1655, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1630137860774994, | |
| "step": 5020, | |
| "valid_targets_mean": 3007.1, | |
| "valid_targets_min": 1082 | |
| }, | |
| { | |
| "epoch": 5.2783613445378155, | |
| "grad_norm": 0.7813513338182236, | |
| "learning_rate": 6.9380927091475085e-06, | |
| "loss": 0.1568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16749534010887146, | |
| "step": 5025, | |
| "valid_targets_mean": 2539.6, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 5.283613445378151, | |
| "grad_norm": 0.7597593894886342, | |
| "learning_rate": 6.898466864868165e-06, | |
| "loss": 0.1606, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16621197760105133, | |
| "step": 5030, | |
| "valid_targets_mean": 2721.4, | |
| "valid_targets_min": 1009 | |
| }, | |
| { | |
| "epoch": 5.288865546218488, | |
| "grad_norm": 0.7182742303119539, | |
| "learning_rate": 6.858930906893751e-06, | |
| "loss": 0.1631, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15185761451721191, | |
| "step": 5035, | |
| "valid_targets_mean": 2491.1, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 5.294117647058823, | |
| "grad_norm": 0.8593841083426136, | |
| "learning_rate": 6.819485106470454e-06, | |
| "loss": 0.1597, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15788263082504272, | |
| "step": 5040, | |
| "valid_targets_mean": 1728.0, | |
| "valid_targets_min": 846 | |
| }, | |
| { | |
| "epoch": 5.29936974789916, | |
| "grad_norm": 0.8035500930290126, | |
| "learning_rate": 6.780129734225949e-06, | |
| "loss": 0.1486, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15254664421081543, | |
| "step": 5045, | |
| "valid_targets_mean": 2170.8, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 5.304621848739496, | |
| "grad_norm": 0.8459834255980975, | |
| "learning_rate": 6.740865060167483e-06, | |
| "loss": 0.1595, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16684821248054504, | |
| "step": 5050, | |
| "valid_targets_mean": 1869.9, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 5.309873949579832, | |
| "grad_norm": 0.7852698744711136, | |
| "learning_rate": 6.701691353680049e-06, | |
| "loss": 0.1664, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1623380482196808, | |
| "step": 5055, | |
| "valid_targets_mean": 2586.2, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 5.315126050420168, | |
| "grad_norm": 0.7811887298191118, | |
| "learning_rate": 6.662608883524542e-06, | |
| "loss": 0.155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13154323399066925, | |
| "step": 5060, | |
| "valid_targets_mean": 1648.8, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 5.320378151260504, | |
| "grad_norm": 0.660895377255489, | |
| "learning_rate": 6.6236179178359115e-06, | |
| "loss": 0.1642, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15457244217395782, | |
| "step": 5065, | |
| "valid_targets_mean": 3175.9, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 5.32563025210084, | |
| "grad_norm": 0.7245039671860787, | |
| "learning_rate": 6.5847187241213e-06, | |
| "loss": 0.1534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14748461544513702, | |
| "step": 5070, | |
| "valid_targets_mean": 2657.9, | |
| "valid_targets_min": 878 | |
| }, | |
| { | |
| "epoch": 5.330882352941177, | |
| "grad_norm": 0.7753382710745726, | |
| "learning_rate": 6.5459115692582386e-06, | |
| "loss": 0.1565, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16546081006526947, | |
| "step": 5075, | |
| "valid_targets_mean": 2395.2, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 5.336134453781512, | |
| "grad_norm": 0.7698920766313514, | |
| "learning_rate": 6.507196719492819e-06, | |
| "loss": 0.1547, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1609654426574707, | |
| "step": 5080, | |
| "valid_targets_mean": 2615.0, | |
| "valid_targets_min": 881 | |
| }, | |
| { | |
| "epoch": 5.341386554621849, | |
| "grad_norm": 0.8553168767751825, | |
| "learning_rate": 6.468574440437801e-06, | |
| "loss": 0.1612, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15425896644592285, | |
| "step": 5085, | |
| "valid_targets_mean": 1917.0, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 5.3466386554621845, | |
| "grad_norm": 0.7379979219387762, | |
| "learning_rate": 6.43004499707089e-06, | |
| "loss": 0.1704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18386310338974, | |
| "step": 5090, | |
| "valid_targets_mean": 2954.1, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 5.351890756302521, | |
| "grad_norm": 0.800532265924272, | |
| "learning_rate": 6.3916086537328545e-06, | |
| "loss": 0.1569, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1512497365474701, | |
| "step": 5095, | |
| "valid_targets_mean": 2645.6, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 5.357142857142857, | |
| "grad_norm": 0.7258456117778306, | |
| "learning_rate": 6.3532656741257145e-06, | |
| "loss": 0.1597, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15273120999336243, | |
| "step": 5100, | |
| "valid_targets_mean": 2574.7, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 5.362394957983193, | |
| "grad_norm": 0.732365954356259, | |
| "learning_rate": 6.315016321310965e-06, | |
| "loss": 0.1657, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16224431991577148, | |
| "step": 5105, | |
| "valid_targets_mean": 2581.6, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 5.367647058823529, | |
| "grad_norm": 0.7789962344565333, | |
| "learning_rate": 6.276860857707727e-06, | |
| "loss": 0.1719, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17684108018875122, | |
| "step": 5110, | |
| "valid_targets_mean": 2496.4, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 5.3728991596638656, | |
| "grad_norm": 0.8532814228814727, | |
| "learning_rate": 6.238799545090994e-06, | |
| "loss": 0.1649, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1521385908126831, | |
| "step": 5115, | |
| "valid_targets_mean": 2414.0, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 5.378151260504202, | |
| "grad_norm": 0.7899952744078172, | |
| "learning_rate": 6.20083264458981e-06, | |
| "loss": 0.1474, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1502818763256073, | |
| "step": 5120, | |
| "valid_targets_mean": 2204.8, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 5.383403361344538, | |
| "grad_norm": 0.6387635570636864, | |
| "learning_rate": 6.162960416685469e-06, | |
| "loss": 0.1602, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1471373438835144, | |
| "step": 5125, | |
| "valid_targets_mean": 3084.1, | |
| "valid_targets_min": 1468 | |
| }, | |
| { | |
| "epoch": 5.388655462184874, | |
| "grad_norm": 0.8485693482893446, | |
| "learning_rate": 6.125183121209741e-06, | |
| "loss": 0.1578, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1780264526605606, | |
| "step": 5130, | |
| "valid_targets_mean": 2274.9, | |
| "valid_targets_min": 1094 | |
| }, | |
| { | |
| "epoch": 5.39390756302521, | |
| "grad_norm": 0.7495677095087009, | |
| "learning_rate": 6.087501017343107e-06, | |
| "loss": 0.1644, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15651904046535492, | |
| "step": 5135, | |
| "valid_targets_mean": 2542.9, | |
| "valid_targets_min": 763 | |
| }, | |
| { | |
| "epoch": 5.399159663865547, | |
| "grad_norm": 0.7435758493446045, | |
| "learning_rate": 6.0499143636129384e-06, | |
| "loss": 0.163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1455836445093155, | |
| "step": 5140, | |
| "valid_targets_mean": 2512.5, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 5.404411764705882, | |
| "grad_norm": 0.8993390562265813, | |
| "learning_rate": 6.012423417891766e-06, | |
| "loss": 0.1672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17336681485176086, | |
| "step": 5145, | |
| "valid_targets_mean": 2680.2, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 5.409663865546219, | |
| "grad_norm": 0.8004525963625007, | |
| "learning_rate": 5.975028437395493e-06, | |
| "loss": 0.1648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15126681327819824, | |
| "step": 5150, | |
| "valid_targets_mean": 2013.7, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 5.4149159663865545, | |
| "grad_norm": 0.7799255786777086, | |
| "learning_rate": 5.937729678681612e-06, | |
| "loss": 0.1576, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15724611282348633, | |
| "step": 5155, | |
| "valid_targets_mean": 2525.8, | |
| "valid_targets_min": 1280 | |
| }, | |
| { | |
| "epoch": 5.420168067226891, | |
| "grad_norm": 0.7841845158497852, | |
| "learning_rate": 5.9005273976474684e-06, | |
| "loss": 0.171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15903203189373016, | |
| "step": 5160, | |
| "valid_targets_mean": 2379.1, | |
| "valid_targets_min": 1163 | |
| }, | |
| { | |
| "epoch": 5.425420168067227, | |
| "grad_norm": 0.6717665505168996, | |
| "learning_rate": 5.8634218495285126e-06, | |
| "loss": 0.1577, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14495673775672913, | |
| "step": 5165, | |
| "valid_targets_mean": 2827.6, | |
| "valid_targets_min": 1100 | |
| }, | |
| { | |
| "epoch": 5.430672268907563, | |
| "grad_norm": 0.8298875239928819, | |
| "learning_rate": 5.8264132888965155e-06, | |
| "loss": 0.1552, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14547504484653473, | |
| "step": 5170, | |
| "valid_targets_mean": 2200.4, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 5.435924369747899, | |
| "grad_norm": 0.7078895304712665, | |
| "learning_rate": 5.789501969657853e-06, | |
| "loss": 0.1562, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14443817734718323, | |
| "step": 5175, | |
| "valid_targets_mean": 2713.6, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 5.4411764705882355, | |
| "grad_norm": 0.7468395911409705, | |
| "learning_rate": 5.752688145051761e-06, | |
| "loss": 0.1645, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15606167912483215, | |
| "step": 5180, | |
| "valid_targets_mean": 2896.7, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 5.446428571428571, | |
| "grad_norm": 0.7339528998149712, | |
| "learning_rate": 5.715972067648561e-06, | |
| "loss": 0.1598, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16431309282779694, | |
| "step": 5185, | |
| "valid_targets_mean": 2671.4, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 5.451680672268908, | |
| "grad_norm": 0.7898995773556446, | |
| "learning_rate": 5.679353989347989e-06, | |
| "loss": 0.1562, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14948245882987976, | |
| "step": 5190, | |
| "valid_targets_mean": 2137.9, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 5.456932773109243, | |
| "grad_norm": 0.8149040646058742, | |
| "learning_rate": 5.642834161377408e-06, | |
| "loss": 0.1599, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16367627680301666, | |
| "step": 5195, | |
| "valid_targets_mean": 2087.3, | |
| "valid_targets_min": 1040 | |
| }, | |
| { | |
| "epoch": 5.46218487394958, | |
| "grad_norm": 0.6719154088237885, | |
| "learning_rate": 5.606412834290121e-06, | |
| "loss": 0.1574, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14343780279159546, | |
| "step": 5200, | |
| "valid_targets_mean": 3016.2, | |
| "valid_targets_min": 1041 | |
| }, | |
| { | |
| "epoch": 5.467436974789916, | |
| "grad_norm": 0.8508194897949711, | |
| "learning_rate": 5.570090257963645e-06, | |
| "loss": 0.1638, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16275382041931152, | |
| "step": 5205, | |
| "valid_targets_mean": 1842.1, | |
| "valid_targets_min": 1076 | |
| }, | |
| { | |
| "epoch": 5.472689075630252, | |
| "grad_norm": 0.7598842769868193, | |
| "learning_rate": 5.533866681597981e-06, | |
| "loss": 0.155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1676841378211975, | |
| "step": 5210, | |
| "valid_targets_mean": 2668.4, | |
| "valid_targets_min": 900 | |
| }, | |
| { | |
| "epoch": 5.477941176470588, | |
| "grad_norm": 0.8473153688144975, | |
| "learning_rate": 5.497742353713928e-06, | |
| "loss": 0.1689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17238308489322662, | |
| "step": 5215, | |
| "valid_targets_mean": 2029.6, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 5.483193277310924, | |
| "grad_norm": 0.737492773817561, | |
| "learning_rate": 5.461717522151371e-06, | |
| "loss": 0.1524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15577836334705353, | |
| "step": 5220, | |
| "valid_targets_mean": 2799.9, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 5.48844537815126, | |
| "grad_norm": 0.7013802768068832, | |
| "learning_rate": 5.4257924340675535e-06, | |
| "loss": 0.1649, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16768768429756165, | |
| "step": 5225, | |
| "valid_targets_mean": 3200.6, | |
| "valid_targets_min": 1303 | |
| }, | |
| { | |
| "epoch": 5.493697478991597, | |
| "grad_norm": 0.6898526895385486, | |
| "learning_rate": 5.389967335935429e-06, | |
| "loss": 0.1524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15428847074508667, | |
| "step": 5230, | |
| "valid_targets_mean": 2979.0, | |
| "valid_targets_min": 735 | |
| }, | |
| { | |
| "epoch": 5.498949579831933, | |
| "grad_norm": 0.8269332396210477, | |
| "learning_rate": 5.3542424735419306e-06, | |
| "loss": 0.1662, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1691667139530182, | |
| "step": 5235, | |
| "valid_targets_mean": 2126.2, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 5.504201680672269, | |
| "grad_norm": 0.7582781505938827, | |
| "learning_rate": 5.318618091986294e-06, | |
| "loss": 0.1595, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16203898191452026, | |
| "step": 5240, | |
| "valid_targets_mean": 2414.2, | |
| "valid_targets_min": 758 | |
| }, | |
| { | |
| "epoch": 5.509453781512605, | |
| "grad_norm": 0.964837540275934, | |
| "learning_rate": 5.283094435678398e-06, | |
| "loss": 0.1643, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18285033106803894, | |
| "step": 5245, | |
| "valid_targets_mean": 2520.3, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 5.514705882352941, | |
| "grad_norm": 0.7407984840086744, | |
| "learning_rate": 5.2476717483370685e-06, | |
| "loss": 0.1691, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18044058978557587, | |
| "step": 5250, | |
| "valid_targets_mean": 2980.8, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 5.519957983193278, | |
| "grad_norm": 0.670814241203273, | |
| "learning_rate": 5.212350272988394e-06, | |
| "loss": 0.1597, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13862532377243042, | |
| "step": 5255, | |
| "valid_targets_mean": 2928.0, | |
| "valid_targets_min": 752 | |
| }, | |
| { | |
| "epoch": 5.525210084033613, | |
| "grad_norm": 0.7638249491401231, | |
| "learning_rate": 5.177130251964088e-06, | |
| "loss": 0.1669, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15651783347129822, | |
| "step": 5260, | |
| "valid_targets_mean": 2388.1, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 5.53046218487395, | |
| "grad_norm": 0.7464606207804972, | |
| "learning_rate": 5.1420119268998085e-06, | |
| "loss": 0.1541, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15144014358520508, | |
| "step": 5265, | |
| "valid_targets_mean": 2414.2, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 5.535714285714286, | |
| "grad_norm": 0.6723804956667677, | |
| "learning_rate": 5.106995538733488e-06, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12311984598636627, | |
| "step": 5270, | |
| "valid_targets_mean": 2505.5, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 5.540966386554622, | |
| "grad_norm": 0.7514104012472849, | |
| "learning_rate": 5.07208132770372e-06, | |
| "loss": 0.1568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16300344467163086, | |
| "step": 5275, | |
| "valid_targets_mean": 2621.9, | |
| "valid_targets_min": 1151 | |
| }, | |
| { | |
| "epoch": 5.546218487394958, | |
| "grad_norm": 0.8709704943028632, | |
| "learning_rate": 5.037269533348075e-06, | |
| "loss": 0.1651, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15614300966262817, | |
| "step": 5280, | |
| "valid_targets_mean": 2086.8, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 5.551470588235294, | |
| "grad_norm": 0.8145404233606028, | |
| "learning_rate": 5.002560394501463e-06, | |
| "loss": 0.1657, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1704893410205841, | |
| "step": 5285, | |
| "valid_targets_mean": 2252.9, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 5.55672268907563, | |
| "grad_norm": 0.7858067778473078, | |
| "learning_rate": 4.9679541492945135e-06, | |
| "loss": 0.1585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17581012845039368, | |
| "step": 5290, | |
| "valid_targets_mean": 2497.4, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 5.561974789915967, | |
| "grad_norm": 0.7444934360144145, | |
| "learning_rate": 4.933451035151913e-06, | |
| "loss": 0.1596, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16977301239967346, | |
| "step": 5295, | |
| "valid_targets_mean": 2621.8, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 5.567226890756302, | |
| "grad_norm": 0.7002188781201552, | |
| "learning_rate": 4.899051288790806e-06, | |
| "loss": 0.1643, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15018996596336365, | |
| "step": 5300, | |
| "valid_targets_mean": 2713.4, | |
| "valid_targets_min": 839 | |
| }, | |
| { | |
| "epoch": 5.572478991596639, | |
| "grad_norm": 0.7598676053646428, | |
| "learning_rate": 4.86475514621916e-06, | |
| "loss": 0.165, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16991931200027466, | |
| "step": 5305, | |
| "valid_targets_mean": 2667.0, | |
| "valid_targets_min": 900 | |
| }, | |
| { | |
| "epoch": 5.5777310924369745, | |
| "grad_norm": 0.8820515115776142, | |
| "learning_rate": 4.830562842734112e-06, | |
| "loss": 0.1591, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1688452661037445, | |
| "step": 5310, | |
| "valid_targets_mean": 2134.6, | |
| "valid_targets_min": 903 | |
| }, | |
| { | |
| "epoch": 5.582983193277311, | |
| "grad_norm": 0.7149756144412622, | |
| "learning_rate": 4.796474612920421e-06, | |
| "loss": 0.1604, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15258319675922394, | |
| "step": 5315, | |
| "valid_targets_mean": 2937.4, | |
| "valid_targets_min": 513 | |
| }, | |
| { | |
| "epoch": 5.588235294117647, | |
| "grad_norm": 0.6941030908245607, | |
| "learning_rate": 4.762490690648813e-06, | |
| "loss": 0.1652, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17828604578971863, | |
| "step": 5320, | |
| "valid_targets_mean": 3366.6, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 5.593487394957983, | |
| "grad_norm": 0.7755309453396569, | |
| "learning_rate": 4.728611309074374e-06, | |
| "loss": 0.1583, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16606014966964722, | |
| "step": 5325, | |
| "valid_targets_mean": 2534.5, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 5.598739495798319, | |
| "grad_norm": 0.7161574736206265, | |
| "learning_rate": 4.694836700634975e-06, | |
| "loss": 0.1648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1472625732421875, | |
| "step": 5330, | |
| "valid_targets_mean": 2363.2, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 5.6039915966386555, | |
| "grad_norm": 0.9658673174845026, | |
| "learning_rate": 4.661167097049668e-06, | |
| "loss": 0.1695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15956471860408783, | |
| "step": 5335, | |
| "valid_targets_mean": 2047.2, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 5.609243697478991, | |
| "grad_norm": 0.8292596385955369, | |
| "learning_rate": 4.62760272931708e-06, | |
| "loss": 0.1604, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16373419761657715, | |
| "step": 5340, | |
| "valid_targets_mean": 2055.4, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 5.614495798319328, | |
| "grad_norm": 0.7872816299655447, | |
| "learning_rate": 4.594143827713844e-06, | |
| "loss": 0.1601, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16614991426467896, | |
| "step": 5345, | |
| "valid_targets_mean": 2267.2, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 5.619747899159664, | |
| "grad_norm": 0.6885239670075206, | |
| "learning_rate": 4.56079062179303e-06, | |
| "loss": 0.1614, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1568869948387146, | |
| "step": 5350, | |
| "valid_targets_mean": 3092.1, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 5.625, | |
| "grad_norm": 0.7063945748122799, | |
| "learning_rate": 4.527543340382538e-06, | |
| "loss": 0.1653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15817028284072876, | |
| "step": 5355, | |
| "valid_targets_mean": 3193.8, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 5.630252100840336, | |
| "grad_norm": 0.7483586106813811, | |
| "learning_rate": 4.494402211583569e-06, | |
| "loss": 0.1674, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1763601005077362, | |
| "step": 5360, | |
| "valid_targets_mean": 2585.2, | |
| "valid_targets_min": 702 | |
| }, | |
| { | |
| "epoch": 5.635504201680672, | |
| "grad_norm": 0.7266651690365762, | |
| "learning_rate": 4.461367462769009e-06, | |
| "loss": 0.1649, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15703195333480835, | |
| "step": 5365, | |
| "valid_targets_mean": 2579.4, | |
| "valid_targets_min": 943 | |
| }, | |
| { | |
| "epoch": 5.640756302521009, | |
| "grad_norm": 0.7723418200616802, | |
| "learning_rate": 4.428439320581923e-06, | |
| "loss": 0.1672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16179069876670837, | |
| "step": 5370, | |
| "valid_targets_mean": 2583.6, | |
| "valid_targets_min": 1207 | |
| }, | |
| { | |
| "epoch": 5.6460084033613445, | |
| "grad_norm": 0.6951598907006209, | |
| "learning_rate": 4.3956180109339705e-06, | |
| "loss": 0.1631, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15667061507701874, | |
| "step": 5375, | |
| "valid_targets_mean": 2790.9, | |
| "valid_targets_min": 985 | |
| }, | |
| { | |
| "epoch": 5.651260504201681, | |
| "grad_norm": 0.7579446080195453, | |
| "learning_rate": 4.362903759003849e-06, | |
| "loss": 0.1601, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15197469294071198, | |
| "step": 5380, | |
| "valid_targets_mean": 2506.1, | |
| "valid_targets_min": 853 | |
| }, | |
| { | |
| "epoch": 5.656512605042017, | |
| "grad_norm": 0.7299857797114069, | |
| "learning_rate": 4.33029678923576e-06, | |
| "loss": 0.1683, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1476059854030609, | |
| "step": 5385, | |
| "valid_targets_mean": 2806.4, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 5.661764705882353, | |
| "grad_norm": 0.7781274130758863, | |
| "learning_rate": 4.297797325337889e-06, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1483781933784485, | |
| "step": 5390, | |
| "valid_targets_mean": 2407.4, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 5.667016806722689, | |
| "grad_norm": 0.7313088256357814, | |
| "learning_rate": 4.265405590280822e-06, | |
| "loss": 0.1619, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15789197385311127, | |
| "step": 5395, | |
| "valid_targets_mean": 3041.6, | |
| "valid_targets_min": 1206 | |
| }, | |
| { | |
| "epoch": 5.6722689075630255, | |
| "grad_norm": 0.8103212447323822, | |
| "learning_rate": 4.233121806296072e-06, | |
| "loss": 0.1651, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17392036318778992, | |
| "step": 5400, | |
| "valid_targets_mean": 2288.4, | |
| "valid_targets_min": 533 | |
| }, | |
| { | |
| "epoch": 5.677521008403361, | |
| "grad_norm": 0.7595385849740667, | |
| "learning_rate": 4.200946194874514e-06, | |
| "loss": 0.1584, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15691381692886353, | |
| "step": 5405, | |
| "valid_targets_mean": 2395.4, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 5.682773109243698, | |
| "grad_norm": 0.7278254832783921, | |
| "learning_rate": 4.168878976764872e-06, | |
| "loss": 0.164, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1525121033191681, | |
| "step": 5410, | |
| "valid_targets_mean": 2631.4, | |
| "valid_targets_min": 1157 | |
| }, | |
| { | |
| "epoch": 5.688025210084033, | |
| "grad_norm": 0.8021682749292429, | |
| "learning_rate": 4.136920371972228e-06, | |
| "loss": 0.1625, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14895716309547424, | |
| "step": 5415, | |
| "valid_targets_mean": 1936.1, | |
| "valid_targets_min": 912 | |
| }, | |
| { | |
| "epoch": 5.69327731092437, | |
| "grad_norm": 0.6953948453787546, | |
| "learning_rate": 4.105070599756482e-06, | |
| "loss": 0.1585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15940669178962708, | |
| "step": 5420, | |
| "valid_targets_mean": 2777.6, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 5.698529411764706, | |
| "grad_norm": 0.770187756365488, | |
| "learning_rate": 4.073329878630856e-06, | |
| "loss": 0.1667, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16655011475086212, | |
| "step": 5425, | |
| "valid_targets_mean": 2557.6, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 5.703781512605042, | |
| "grad_norm": 0.7698518711299432, | |
| "learning_rate": 4.041698426360425e-06, | |
| "loss": 0.1695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1650160849094391, | |
| "step": 5430, | |
| "valid_targets_mean": 2342.6, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 5.709033613445378, | |
| "grad_norm": 0.7840326454397952, | |
| "learning_rate": 4.010176459960569e-06, | |
| "loss": 0.1603, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.161130890250206, | |
| "step": 5435, | |
| "valid_targets_mean": 2237.3, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 5.714285714285714, | |
| "grad_norm": 0.6973943107045614, | |
| "learning_rate": 3.978764195695528e-06, | |
| "loss": 0.1572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14485879242420197, | |
| "step": 5440, | |
| "valid_targets_mean": 2737.0, | |
| "valid_targets_min": 977 | |
| }, | |
| { | |
| "epoch": 5.71953781512605, | |
| "grad_norm": 0.8240293274129915, | |
| "learning_rate": 3.947461849076912e-06, | |
| "loss": 0.1688, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19433113932609558, | |
| "step": 5445, | |
| "valid_targets_mean": 2314.3, | |
| "valid_targets_min": 977 | |
| }, | |
| { | |
| "epoch": 5.724789915966387, | |
| "grad_norm": 0.8114689626823407, | |
| "learning_rate": 3.916269634862193e-06, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15822897851467133, | |
| "step": 5450, | |
| "valid_targets_mean": 2607.1, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 5.730042016806722, | |
| "grad_norm": 0.8364269359269342, | |
| "learning_rate": 3.885187767053269e-06, | |
| "loss": 0.1544, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1558850109577179, | |
| "step": 5455, | |
| "valid_targets_mean": 1694.3, | |
| "valid_targets_min": 838 | |
| }, | |
| { | |
| "epoch": 5.735294117647059, | |
| "grad_norm": 0.8692400632474553, | |
| "learning_rate": 3.854216458894973e-06, | |
| "loss": 0.1636, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16702617704868317, | |
| "step": 5460, | |
| "valid_targets_mean": 2202.1, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 5.740546218487395, | |
| "grad_norm": 0.7101368299314894, | |
| "learning_rate": 3.823355922873606e-06, | |
| "loss": 0.1624, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14841680228710175, | |
| "step": 5465, | |
| "valid_targets_mean": 2778.5, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 5.745798319327731, | |
| "grad_norm": 0.7291139806638217, | |
| "learning_rate": 3.792606370715508e-06, | |
| "loss": 0.1557, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1669558882713318, | |
| "step": 5470, | |
| "valid_targets_mean": 2786.6, | |
| "valid_targets_min": 1024 | |
| }, | |
| { | |
| "epoch": 5.751050420168067, | |
| "grad_norm": 0.8199168939791637, | |
| "learning_rate": 3.761968013385584e-06, | |
| "loss": 0.1648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15248140692710876, | |
| "step": 5475, | |
| "valid_targets_mean": 1965.8, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 5.756302521008403, | |
| "grad_norm": 0.7220120926134376, | |
| "learning_rate": 3.731441061085843e-06, | |
| "loss": 0.1679, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15842077136039734, | |
| "step": 5480, | |
| "valid_targets_mean": 2855.1, | |
| "valid_targets_min": 1118 | |
| }, | |
| { | |
| "epoch": 5.76155462184874, | |
| "grad_norm": 0.7691870517732631, | |
| "learning_rate": 3.701025723253988e-06, | |
| "loss": 0.1504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14176911115646362, | |
| "step": 5485, | |
| "valid_targets_mean": 2196.8, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 5.766806722689076, | |
| "grad_norm": 0.6895030971240736, | |
| "learning_rate": 3.67072220856197e-06, | |
| "loss": 0.1692, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15497374534606934, | |
| "step": 5490, | |
| "valid_targets_mean": 2655.5, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 5.772058823529412, | |
| "grad_norm": 0.7338262928726779, | |
| "learning_rate": 3.640530724914517e-06, | |
| "loss": 0.1609, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16165027022361755, | |
| "step": 5495, | |
| "valid_targets_mean": 2627.9, | |
| "valid_targets_min": 949 | |
| }, | |
| { | |
| "epoch": 5.777310924369748, | |
| "grad_norm": 0.7406660843906033, | |
| "learning_rate": 3.61045147944777e-06, | |
| "loss": 0.1612, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14221808314323425, | |
| "step": 5500, | |
| "valid_targets_mean": 2517.7, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 5.782563025210084, | |
| "grad_norm": 0.7608572974428726, | |
| "learning_rate": 3.5804846785278315e-06, | |
| "loss": 0.159, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15930047631263733, | |
| "step": 5505, | |
| "valid_targets_mean": 2888.9, | |
| "valid_targets_min": 1200 | |
| }, | |
| { | |
| "epoch": 5.78781512605042, | |
| "grad_norm": 0.7935824693797842, | |
| "learning_rate": 3.5506305277493326e-06, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1513899266719818, | |
| "step": 5510, | |
| "valid_targets_mean": 2279.6, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 5.793067226890757, | |
| "grad_norm": 0.790428932578931, | |
| "learning_rate": 3.520889231934057e-06, | |
| "loss": 0.1624, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1564992070198059, | |
| "step": 5515, | |
| "valid_targets_mean": 2152.1, | |
| "valid_targets_min": 1278 | |
| }, | |
| { | |
| "epoch": 5.798319327731092, | |
| "grad_norm": 0.8386478323895583, | |
| "learning_rate": 3.4912609951295084e-06, | |
| "loss": 0.1692, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1771148443222046, | |
| "step": 5520, | |
| "valid_targets_mean": 2142.2, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 5.803571428571429, | |
| "grad_norm": 0.8169074903755393, | |
| "learning_rate": 3.461746020607526e-06, | |
| "loss": 0.1659, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16074536740779877, | |
| "step": 5525, | |
| "valid_targets_mean": 2207.6, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 5.8088235294117645, | |
| "grad_norm": 0.7542209212555967, | |
| "learning_rate": 3.432344510862895e-06, | |
| "loss": 0.1612, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17460983991622925, | |
| "step": 5530, | |
| "valid_targets_mean": 2285.8, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 5.814075630252101, | |
| "grad_norm": 0.9980332994549753, | |
| "learning_rate": 3.403056667611917e-06, | |
| "loss": 0.1583, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17128345370292664, | |
| "step": 5535, | |
| "valid_targets_mean": 2691.8, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 5.819327731092437, | |
| "grad_norm": 0.7583521616551634, | |
| "learning_rate": 3.373882691791088e-06, | |
| "loss": 0.1609, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15085886418819427, | |
| "step": 5540, | |
| "valid_targets_mean": 2659.0, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 5.824579831932773, | |
| "grad_norm": 0.8477434832925683, | |
| "learning_rate": 3.344822783555679e-06, | |
| "loss": 0.1623, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16110014915466309, | |
| "step": 5545, | |
| "valid_targets_mean": 1985.5, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.829831932773109, | |
| "grad_norm": 0.7643511912140655, | |
| "learning_rate": 3.315877142278363e-06, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16038095951080322, | |
| "step": 5550, | |
| "valid_targets_mean": 2380.6, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 5.8350840336134455, | |
| "grad_norm": 0.7518728544106923, | |
| "learning_rate": 3.287045966547866e-06, | |
| "loss": 0.1632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17358054220676422, | |
| "step": 5555, | |
| "valid_targets_mean": 2760.2, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 5.840336134453781, | |
| "grad_norm": 0.7124048514016621, | |
| "learning_rate": 3.258329454167599e-06, | |
| "loss": 0.159, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1632152795791626, | |
| "step": 5560, | |
| "valid_targets_mean": 2886.0, | |
| "valid_targets_min": 710 | |
| }, | |
| { | |
| "epoch": 5.845588235294118, | |
| "grad_norm": 0.8638758876946544, | |
| "learning_rate": 3.2297278021542833e-06, | |
| "loss": 0.1732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18239599466323853, | |
| "step": 5565, | |
| "valid_targets_mean": 2045.9, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 5.850840336134453, | |
| "grad_norm": 0.8074069942626705, | |
| "learning_rate": 3.201241206736614e-06, | |
| "loss": 0.1683, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19121696054935455, | |
| "step": 5570, | |
| "valid_targets_mean": 2512.3, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 5.85609243697479, | |
| "grad_norm": 0.8094730562236274, | |
| "learning_rate": 3.1728698633539266e-06, | |
| "loss": 0.1632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1697961837053299, | |
| "step": 5575, | |
| "valid_targets_mean": 2204.6, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 5.8613445378151265, | |
| "grad_norm": 0.7593441846784588, | |
| "learning_rate": 3.1446139666548212e-06, | |
| "loss": 0.1631, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1595013290643692, | |
| "step": 5580, | |
| "valid_targets_mean": 2633.0, | |
| "valid_targets_min": 707 | |
| }, | |
| { | |
| "epoch": 5.866596638655462, | |
| "grad_norm": 0.7957651638135713, | |
| "learning_rate": 3.1164737104958707e-06, | |
| "loss": 0.1599, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15644320845603943, | |
| "step": 5585, | |
| "valid_targets_mean": 1890.6, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 5.871848739495798, | |
| "grad_norm": 0.8090053281592035, | |
| "learning_rate": 3.0884492879402452e-06, | |
| "loss": 0.1633, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17713585495948792, | |
| "step": 5590, | |
| "valid_targets_mean": 2193.5, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 5.8771008403361344, | |
| "grad_norm": 0.7500689718315519, | |
| "learning_rate": 3.0605408912564184e-06, | |
| "loss": 0.1647, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16628164052963257, | |
| "step": 5595, | |
| "valid_targets_mean": 2580.8, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 5.882352941176471, | |
| "grad_norm": 0.8192677926893119, | |
| "learning_rate": 3.032748711916851e-06, | |
| "loss": 0.1657, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16836728155612946, | |
| "step": 5600, | |
| "valid_targets_mean": 2461.7, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 5.887605042016807, | |
| "grad_norm": 0.7874830104284019, | |
| "learning_rate": 3.0050729405966495e-06, | |
| "loss": 0.1632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16090510785579681, | |
| "step": 5605, | |
| "valid_targets_mean": 2267.6, | |
| "valid_targets_min": 244 | |
| }, | |
| { | |
| "epoch": 5.892857142857143, | |
| "grad_norm": 0.7833656235582289, | |
| "learning_rate": 2.9775137671722754e-06, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1637769341468811, | |
| "step": 5610, | |
| "valid_targets_mean": 2443.2, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 5.898109243697479, | |
| "grad_norm": 0.7228128266155345, | |
| "learning_rate": 2.9500713807202565e-06, | |
| "loss": 0.154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15642684698104858, | |
| "step": 5615, | |
| "valid_targets_mean": 2638.7, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 5.9033613445378155, | |
| "grad_norm": 0.6752375605078036, | |
| "learning_rate": 2.9227459695158543e-06, | |
| "loss": 0.1565, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15065337717533112, | |
| "step": 5620, | |
| "valid_targets_mean": 3194.4, | |
| "valid_targets_min": 1263 | |
| }, | |
| { | |
| "epoch": 5.908613445378151, | |
| "grad_norm": 0.8638709102103872, | |
| "learning_rate": 2.895537721031809e-06, | |
| "loss": 0.1601, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16943012177944183, | |
| "step": 5625, | |
| "valid_targets_mean": 1833.1, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 5.913865546218488, | |
| "grad_norm": 0.7646968532660349, | |
| "learning_rate": 2.8684468219370364e-06, | |
| "loss": 0.1532, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14894887804985046, | |
| "step": 5630, | |
| "valid_targets_mean": 2673.2, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.919117647058823, | |
| "grad_norm": 0.7612881911075914, | |
| "learning_rate": 2.841473458095334e-06, | |
| "loss": 0.1668, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14882472157478333, | |
| "step": 5635, | |
| "valid_targets_mean": 2316.6, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 5.92436974789916, | |
| "grad_norm": 0.7425817529470462, | |
| "learning_rate": 2.8146178145641355e-06, | |
| "loss": 0.1608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1679328829050064, | |
| "step": 5640, | |
| "valid_targets_mean": 2502.5, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 5.929621848739496, | |
| "grad_norm": 0.9027720359549637, | |
| "learning_rate": 2.787880075593217e-06, | |
| "loss": 0.1669, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15756511688232422, | |
| "step": 5645, | |
| "valid_targets_mean": 2311.6, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 5.934873949579832, | |
| "grad_norm": 0.6980735937613616, | |
| "learning_rate": 2.7612604246234355e-06, | |
| "loss": 0.1608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16025885939598083, | |
| "step": 5650, | |
| "valid_targets_mean": 2912.9, | |
| "valid_targets_min": 854 | |
| }, | |
| { | |
| "epoch": 5.940126050420168, | |
| "grad_norm": 0.7529592237919616, | |
| "learning_rate": 2.7347590442854867e-06, | |
| "loss": 0.16, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17594562470912933, | |
| "step": 5655, | |
| "valid_targets_mean": 2683.8, | |
| "valid_targets_min": 1005 | |
| }, | |
| { | |
| "epoch": 5.945378151260504, | |
| "grad_norm": 0.7696301031774723, | |
| "learning_rate": 2.7083761163986433e-06, | |
| "loss": 0.1683, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17107313871383667, | |
| "step": 5660, | |
| "valid_targets_mean": 2896.2, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 5.95063025210084, | |
| "grad_norm": 0.7798340627342268, | |
| "learning_rate": 2.682111821969493e-06, | |
| "loss": 0.1654, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16536115109920502, | |
| "step": 5665, | |
| "valid_targets_mean": 2587.5, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 5.955882352941177, | |
| "grad_norm": 0.7149524883372318, | |
| "learning_rate": 2.6559663411907233e-06, | |
| "loss": 0.1604, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1709938943386078, | |
| "step": 5670, | |
| "valid_targets_mean": 2945.9, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 5.961134453781512, | |
| "grad_norm": 0.7165019494853891, | |
| "learning_rate": 2.629939853439856e-06, | |
| "loss": 0.1664, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16053658723831177, | |
| "step": 5675, | |
| "valid_targets_mean": 2915.6, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 5.966386554621849, | |
| "grad_norm": 0.9589101754966618, | |
| "learning_rate": 2.604032537278052e-06, | |
| "loss": 0.157, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15422767400741577, | |
| "step": 5680, | |
| "valid_targets_mean": 1772.9, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 5.9716386554621845, | |
| "grad_norm": 0.7874586260627973, | |
| "learning_rate": 2.5782445704488447e-06, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14686915278434753, | |
| "step": 5685, | |
| "valid_targets_mean": 2175.2, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 5.976890756302521, | |
| "grad_norm": 0.8291299717722125, | |
| "learning_rate": 2.5525761298769515e-06, | |
| "loss": 0.1672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17729215323925018, | |
| "step": 5690, | |
| "valid_targets_mean": 2213.9, | |
| "valid_targets_min": 1434 | |
| }, | |
| { | |
| "epoch": 5.982142857142857, | |
| "grad_norm": 0.8550297198565612, | |
| "learning_rate": 2.527027391667054e-06, | |
| "loss": 0.1688, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18570925295352936, | |
| "step": 5695, | |
| "valid_targets_mean": 2631.9, | |
| "valid_targets_min": 840 | |
| }, | |
| { | |
| "epoch": 5.987394957983193, | |
| "grad_norm": 0.8225215202521778, | |
| "learning_rate": 2.501598531102587e-06, | |
| "loss": 0.1619, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1654691994190216, | |
| "step": 5700, | |
| "valid_targets_mean": 2116.9, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 5.992647058823529, | |
| "grad_norm": 0.7818605889350404, | |
| "learning_rate": 2.47628972264452e-06, | |
| "loss": 0.1597, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15820591151714325, | |
| "step": 5705, | |
| "valid_targets_mean": 2376.9, | |
| "valid_targets_min": 775 | |
| }, | |
| { | |
| "epoch": 5.9978991596638656, | |
| "grad_norm": 0.7205251991716778, | |
| "learning_rate": 2.4511011399301944e-06, | |
| "loss": 0.1681, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16490915417671204, | |
| "step": 5710, | |
| "valid_targets_mean": 3045.2, | |
| "valid_targets_min": 921 | |
| }, | |
| { | |
| "epoch": 6.003151260504202, | |
| "grad_norm": 0.79777985742783, | |
| "learning_rate": 2.4260329557721107e-06, | |
| "loss": 0.1556, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16908371448516846, | |
| "step": 5715, | |
| "valid_targets_mean": 2234.9, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 6.008403361344538, | |
| "grad_norm": 0.7056977699167971, | |
| "learning_rate": 2.4010853421567237e-06, | |
| "loss": 0.1435, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1342272162437439, | |
| "step": 5720, | |
| "valid_targets_mean": 2620.8, | |
| "valid_targets_min": 1276 | |
| }, | |
| { | |
| "epoch": 6.013655462184874, | |
| "grad_norm": 0.7206882042436742, | |
| "learning_rate": 2.376258470243309e-06, | |
| "loss": 0.1564, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1475621461868286, | |
| "step": 5725, | |
| "valid_targets_mean": 2910.7, | |
| "valid_targets_min": 998 | |
| }, | |
| { | |
| "epoch": 6.01890756302521, | |
| "grad_norm": 0.6958104794523656, | |
| "learning_rate": 2.351552510362758e-06, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14622151851654053, | |
| "step": 5730, | |
| "valid_targets_mean": 3744.4, | |
| "valid_targets_min": 1153 | |
| }, | |
| { | |
| "epoch": 6.024159663865547, | |
| "grad_norm": 0.74834161850093, | |
| "learning_rate": 2.326967632016406e-06, | |
| "loss": 0.1535, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1483282744884491, | |
| "step": 5735, | |
| "valid_targets_mean": 2573.9, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.029411764705882, | |
| "grad_norm": 0.7566577246672258, | |
| "learning_rate": 2.302504003874888e-06, | |
| "loss": 0.1518, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14986348152160645, | |
| "step": 5740, | |
| "valid_targets_mean": 2467.8, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 6.034663865546219, | |
| "grad_norm": 0.7394059807325174, | |
| "learning_rate": 2.2781617937769675e-06, | |
| "loss": 0.1563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15996532142162323, | |
| "step": 5745, | |
| "valid_targets_mean": 3097.9, | |
| "valid_targets_min": 914 | |
| }, | |
| { | |
| "epoch": 6.0399159663865545, | |
| "grad_norm": 0.7291797979840863, | |
| "learning_rate": 2.2539411687283907e-06, | |
| "loss": 0.1588, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15175995230674744, | |
| "step": 5750, | |
| "valid_targets_mean": 2689.8, | |
| "valid_targets_min": 1214 | |
| }, | |
| { | |
| "epoch": 6.045168067226891, | |
| "grad_norm": 0.8021207951201383, | |
| "learning_rate": 2.229842294900744e-06, | |
| "loss": 0.1484, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15454252064228058, | |
| "step": 5755, | |
| "valid_targets_mean": 2140.5, | |
| "valid_targets_min": 829 | |
| }, | |
| { | |
| "epoch": 6.050420168067227, | |
| "grad_norm": 0.7912228519081135, | |
| "learning_rate": 2.205865337630302e-06, | |
| "loss": 0.1542, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15131813287734985, | |
| "step": 5760, | |
| "valid_targets_mean": 2908.2, | |
| "valid_targets_min": 1220 | |
| }, | |
| { | |
| "epoch": 6.055672268907563, | |
| "grad_norm": 0.7907789459812723, | |
| "learning_rate": 2.182010461416899e-06, | |
| "loss": 0.1562, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1566312164068222, | |
| "step": 5765, | |
| "valid_targets_mean": 2503.7, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 6.060924369747899, | |
| "grad_norm": 0.8839642676042035, | |
| "learning_rate": 2.158277829922817e-06, | |
| "loss": 0.15, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1570267528295517, | |
| "step": 5770, | |
| "valid_targets_mean": 1742.3, | |
| "valid_targets_min": 544 | |
| }, | |
| { | |
| "epoch": 6.0661764705882355, | |
| "grad_norm": 0.8625467723504546, | |
| "learning_rate": 2.1346676059716253e-06, | |
| "loss": 0.1558, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14902624487876892, | |
| "step": 5775, | |
| "valid_targets_mean": 2078.1, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 6.071428571428571, | |
| "grad_norm": 0.7087088554579924, | |
| "learning_rate": 2.1111799515471066e-06, | |
| "loss": 0.144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13471050560474396, | |
| "step": 5780, | |
| "valid_targets_mean": 2473.5, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 6.076680672268908, | |
| "grad_norm": 0.7327436066837598, | |
| "learning_rate": 2.0878150277921195e-06, | |
| "loss": 0.1461, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13805945217609406, | |
| "step": 5785, | |
| "valid_targets_mean": 2741.4, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 6.081932773109243, | |
| "grad_norm": 0.8315227818730387, | |
| "learning_rate": 2.064572995007492e-06, | |
| "loss": 0.1534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14630141854286194, | |
| "step": 5790, | |
| "valid_targets_mean": 2002.4, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 6.08718487394958, | |
| "grad_norm": 0.8401948839242035, | |
| "learning_rate": 2.0414540126509274e-06, | |
| "loss": 0.1529, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16226598620414734, | |
| "step": 5795, | |
| "valid_targets_mean": 2131.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 6.092436974789916, | |
| "grad_norm": 0.8776927085638933, | |
| "learning_rate": 2.0184582393359253e-06, | |
| "loss": 0.1492, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15649349987506866, | |
| "step": 5800, | |
| "valid_targets_mean": 1861.4, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 6.097689075630252, | |
| "grad_norm": 0.8211711286578806, | |
| "learning_rate": 1.99558583283066e-06, | |
| "loss": 0.1557, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14332887530326843, | |
| "step": 5805, | |
| "valid_targets_mean": 2022.8, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.102941176470588, | |
| "grad_norm": 0.732033944801811, | |
| "learning_rate": 1.9728369500569333e-06, | |
| "loss": 0.1415, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13911320269107819, | |
| "step": 5810, | |
| "valid_targets_mean": 2577.7, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 6.108193277310924, | |
| "grad_norm": 0.8099862293022083, | |
| "learning_rate": 1.950211747089077e-06, | |
| "loss": 0.1567, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14890515804290771, | |
| "step": 5815, | |
| "valid_targets_mean": 2239.7, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 6.11344537815126, | |
| "grad_norm": 0.8280954650156288, | |
| "learning_rate": 1.927710379152881e-06, | |
| "loss": 0.1514, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15409252047538757, | |
| "step": 5820, | |
| "valid_targets_mean": 2037.0, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 6.118697478991597, | |
| "grad_norm": 0.7720826623857413, | |
| "learning_rate": 1.9053330006245452e-06, | |
| "loss": 0.1487, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1435692012310028, | |
| "step": 5825, | |
| "valid_targets_mean": 2505.6, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 6.123949579831933, | |
| "grad_norm": 0.8602830964405195, | |
| "learning_rate": 1.8830797650296006e-06, | |
| "loss": 0.1536, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15629276633262634, | |
| "step": 5830, | |
| "valid_targets_mean": 1843.5, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 6.129201680672269, | |
| "grad_norm": 0.7886878487521178, | |
| "learning_rate": 1.8609508250418628e-06, | |
| "loss": 0.163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17373371124267578, | |
| "step": 5835, | |
| "valid_targets_mean": 2652.3, | |
| "valid_targets_min": 1395 | |
| }, | |
| { | |
| "epoch": 6.1344537815126055, | |
| "grad_norm": 0.8219543982430842, | |
| "learning_rate": 1.8389463324823986e-06, | |
| "loss": 0.1471, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14603424072265625, | |
| "step": 5840, | |
| "valid_targets_mean": 2208.8, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 6.139705882352941, | |
| "grad_norm": 0.8240205796466581, | |
| "learning_rate": 1.81706643831846e-06, | |
| "loss": 0.151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1447804570198059, | |
| "step": 5845, | |
| "valid_targets_mean": 2197.1, | |
| "valid_targets_min": 787 | |
| }, | |
| { | |
| "epoch": 6.144957983193278, | |
| "grad_norm": 0.7863349767944662, | |
| "learning_rate": 1.7953112926624715e-06, | |
| "loss": 0.1574, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14369264245033264, | |
| "step": 5850, | |
| "valid_targets_mean": 2488.8, | |
| "valid_targets_min": 875 | |
| }, | |
| { | |
| "epoch": 6.150210084033613, | |
| "grad_norm": 0.7481918069687116, | |
| "learning_rate": 1.7736810447709852e-06, | |
| "loss": 0.1404, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14041663706302643, | |
| "step": 5855, | |
| "valid_targets_mean": 2419.5, | |
| "valid_targets_min": 631 | |
| }, | |
| { | |
| "epoch": 6.15546218487395, | |
| "grad_norm": 0.7548879647168044, | |
| "learning_rate": 1.7521758430436553e-06, | |
| "loss": 0.1568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15782803297042847, | |
| "step": 5860, | |
| "valid_targets_mean": 2704.7, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 6.160714285714286, | |
| "grad_norm": 0.8818736318364171, | |
| "learning_rate": 1.7307958350222366e-06, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1596248298883438, | |
| "step": 5865, | |
| "valid_targets_mean": 2219.8, | |
| "valid_targets_min": 836 | |
| }, | |
| { | |
| "epoch": 6.165966386554622, | |
| "grad_norm": 0.8306669309571554, | |
| "learning_rate": 1.7095411673895545e-06, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16560135781764984, | |
| "step": 5870, | |
| "valid_targets_mean": 2339.6, | |
| "valid_targets_min": 672 | |
| }, | |
| { | |
| "epoch": 6.171218487394958, | |
| "grad_norm": 0.7839020315929305, | |
| "learning_rate": 1.6884119859685034e-06, | |
| "loss": 0.1571, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16316699981689453, | |
| "step": 5875, | |
| "valid_targets_mean": 2607.8, | |
| "valid_targets_min": 644 | |
| }, | |
| { | |
| "epoch": 6.176470588235294, | |
| "grad_norm": 0.7562882938551625, | |
| "learning_rate": 1.6674084357210562e-06, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1440119594335556, | |
| "step": 5880, | |
| "valid_targets_mean": 2218.6, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 6.18172268907563, | |
| "grad_norm": 0.7626842826181003, | |
| "learning_rate": 1.6465306607472608e-06, | |
| "loss": 0.1489, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13738971948623657, | |
| "step": 5885, | |
| "valid_targets_mean": 2323.1, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 6.186974789915967, | |
| "grad_norm": 0.7942359627663949, | |
| "learning_rate": 1.625778804284246e-06, | |
| "loss": 0.146, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14948052167892456, | |
| "step": 5890, | |
| "valid_targets_mean": 2511.6, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 6.192226890756302, | |
| "grad_norm": 0.8431028429315225, | |
| "learning_rate": 1.6051530087052558e-06, | |
| "loss": 0.1459, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15867841243743896, | |
| "step": 5895, | |
| "valid_targets_mean": 2291.1, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 6.197478991596639, | |
| "grad_norm": 0.8666438002772852, | |
| "learning_rate": 1.5846534155186511e-06, | |
| "loss": 0.148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14912620186805725, | |
| "step": 5900, | |
| "valid_targets_mean": 1881.3, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 6.2027310924369745, | |
| "grad_norm": 0.8110997063724646, | |
| "learning_rate": 1.564280165366956e-06, | |
| "loss": 0.1573, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1503579318523407, | |
| "step": 5905, | |
| "valid_targets_mean": 2042.6, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 6.207983193277311, | |
| "grad_norm": 0.7289206797878542, | |
| "learning_rate": 1.544033398025886e-06, | |
| "loss": 0.153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15290409326553345, | |
| "step": 5910, | |
| "valid_targets_mean": 2841.8, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 6.213235294117647, | |
| "grad_norm": 0.8826762189160628, | |
| "learning_rate": 1.5239132524033951e-06, | |
| "loss": 0.1446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15168313682079315, | |
| "step": 5915, | |
| "valid_targets_mean": 2273.7, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 6.218487394957983, | |
| "grad_norm": 0.6729261276809203, | |
| "learning_rate": 1.50391986653871e-06, | |
| "loss": 0.1476, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1368652582168579, | |
| "step": 5920, | |
| "valid_targets_mean": 3169.6, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 6.223739495798319, | |
| "grad_norm": 0.7965190893498632, | |
| "learning_rate": 1.4840533776013954e-06, | |
| "loss": 0.1552, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1640886664390564, | |
| "step": 5925, | |
| "valid_targets_mean": 2498.8, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 6.2289915966386555, | |
| "grad_norm": 0.7757965453046244, | |
| "learning_rate": 1.4643139218904035e-06, | |
| "loss": 0.152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14865760505199432, | |
| "step": 5930, | |
| "valid_targets_mean": 2731.4, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 6.234243697478991, | |
| "grad_norm": 0.7962903371122022, | |
| "learning_rate": 1.4447016348331477e-06, | |
| "loss": 0.1482, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13591375946998596, | |
| "step": 5935, | |
| "valid_targets_mean": 2047.8, | |
| "valid_targets_min": 781 | |
| }, | |
| { | |
| "epoch": 6.239495798319328, | |
| "grad_norm": 0.7557635968086313, | |
| "learning_rate": 1.4252166509845777e-06, | |
| "loss": 0.1585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15759332478046417, | |
| "step": 5940, | |
| "valid_targets_mean": 2775.9, | |
| "valid_targets_min": 875 | |
| }, | |
| { | |
| "epoch": 6.244747899159663, | |
| "grad_norm": 0.7204075261247243, | |
| "learning_rate": 1.4058591040262259e-06, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14412742853164673, | |
| "step": 5945, | |
| "valid_targets_mean": 2885.9, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "grad_norm": 0.7821960625396808, | |
| "learning_rate": 1.3866291267653309e-06, | |
| "loss": 0.1514, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15519963204860687, | |
| "step": 5950, | |
| "valid_targets_mean": 2386.9, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 6.255252100840336, | |
| "grad_norm": 0.7211652670433405, | |
| "learning_rate": 1.3675268511339067e-06, | |
| "loss": 0.1501, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14521880447864532, | |
| "step": 5955, | |
| "valid_targets_mean": 2996.6, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 6.260504201680672, | |
| "grad_norm": 0.8118712246185013, | |
| "learning_rate": 1.3485524081878288e-06, | |
| "loss": 0.1498, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16166505217552185, | |
| "step": 5960, | |
| "valid_targets_mean": 2691.8, | |
| "valid_targets_min": 937 | |
| }, | |
| { | |
| "epoch": 6.265756302521009, | |
| "grad_norm": 0.859100591202131, | |
| "learning_rate": 1.3297059281059533e-06, | |
| "loss": 0.1561, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1694069802761078, | |
| "step": 5965, | |
| "valid_targets_mean": 2484.4, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 6.2710084033613445, | |
| "grad_norm": 0.6880675870848753, | |
| "learning_rate": 1.3109875401892146e-06, | |
| "loss": 0.1491, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13731975853443146, | |
| "step": 5970, | |
| "valid_targets_mean": 3048.2, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 6.276260504201681, | |
| "grad_norm": 0.78207546197356, | |
| "learning_rate": 1.2923973728597373e-06, | |
| "loss": 0.1425, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13927319645881653, | |
| "step": 5975, | |
| "valid_targets_mean": 2147.8, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 6.281512605042017, | |
| "grad_norm": 0.9014126476888846, | |
| "learning_rate": 1.2739355536599508e-06, | |
| "loss": 0.1429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1446988582611084, | |
| "step": 5980, | |
| "valid_targets_mean": 1929.1, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 6.286764705882353, | |
| "grad_norm": 0.7420292923016639, | |
| "learning_rate": 1.2556022092517362e-06, | |
| "loss": 0.1521, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14270836114883423, | |
| "step": 5985, | |
| "valid_targets_mean": 2397.7, | |
| "valid_targets_min": 966 | |
| }, | |
| { | |
| "epoch": 6.292016806722689, | |
| "grad_norm": 0.7955077467970862, | |
| "learning_rate": 1.237397465415524e-06, | |
| "loss": 0.1453, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1445448398590088, | |
| "step": 5990, | |
| "valid_targets_mean": 2025.7, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 6.2972689075630255, | |
| "grad_norm": 0.8287126470993438, | |
| "learning_rate": 1.2193214470494664e-06, | |
| "loss": 0.1517, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1324065625667572, | |
| "step": 5995, | |
| "valid_targets_mean": 1985.8, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 6.302521008403361, | |
| "grad_norm": 0.7561550704661556, | |
| "learning_rate": 1.201374278168541e-06, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15818336606025696, | |
| "step": 6000, | |
| "valid_targets_mean": 2369.2, | |
| "valid_targets_min": 806 | |
| }, | |
| { | |
| "epoch": 6.307773109243698, | |
| "grad_norm": 0.7629710579022473, | |
| "learning_rate": 1.1835560819037429e-06, | |
| "loss": 0.1427, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12634852528572083, | |
| "step": 6005, | |
| "valid_targets_mean": 2487.6, | |
| "valid_targets_min": 870 | |
| }, | |
| { | |
| "epoch": 6.313025210084033, | |
| "grad_norm": 0.8264383119430665, | |
| "learning_rate": 1.1658669805012112e-06, | |
| "loss": 0.144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14036674797534943, | |
| "step": 6010, | |
| "valid_targets_mean": 1953.5, | |
| "valid_targets_min": 639 | |
| }, | |
| { | |
| "epoch": 6.31827731092437, | |
| "grad_norm": 0.8768123841384081, | |
| "learning_rate": 1.1483070953213904e-06, | |
| "loss": 0.148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1594979465007782, | |
| "step": 6015, | |
| "valid_targets_mean": 2033.9, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 6.323529411764706, | |
| "grad_norm": 0.8008028472783635, | |
| "learning_rate": 1.130876546838211e-06, | |
| "loss": 0.1592, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16434642672538757, | |
| "step": 6020, | |
| "valid_targets_mean": 2647.7, | |
| "valid_targets_min": 1279 | |
| }, | |
| { | |
| "epoch": 6.328781512605042, | |
| "grad_norm": 0.778716540961378, | |
| "learning_rate": 1.1135754546382627e-06, | |
| "loss": 0.1439, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15541379153728485, | |
| "step": 6025, | |
| "valid_targets_mean": 2563.2, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.334033613445378, | |
| "grad_norm": 0.7658767727670047, | |
| "learning_rate": 1.0964039374199564e-06, | |
| "loss": 0.1476, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14019645750522614, | |
| "step": 6030, | |
| "valid_targets_mean": 2198.1, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 6.339285714285714, | |
| "grad_norm": 0.7848881321896162, | |
| "learning_rate": 1.07936211299273e-06, | |
| "loss": 0.1504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15517699718475342, | |
| "step": 6035, | |
| "valid_targets_mean": 2547.2, | |
| "valid_targets_min": 825 | |
| }, | |
| { | |
| "epoch": 6.34453781512605, | |
| "grad_norm": 1.1562955954823186, | |
| "learning_rate": 1.0624500982762376e-06, | |
| "loss": 0.1611, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1641463041305542, | |
| "step": 6040, | |
| "valid_targets_mean": 2856.1, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 6.349789915966387, | |
| "grad_norm": 1.243869663850212, | |
| "learning_rate": 1.045668009299523e-06, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1688794195652008, | |
| "step": 6045, | |
| "valid_targets_mean": 1660.8, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 6.355042016806722, | |
| "grad_norm": 0.9029029473166272, | |
| "learning_rate": 1.029015961200266e-06, | |
| "loss": 0.1539, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16218698024749756, | |
| "step": 6050, | |
| "valid_targets_mean": 1781.9, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 6.360294117647059, | |
| "grad_norm": 0.839913681660097, | |
| "learning_rate": 1.012494068223946e-06, | |
| "loss": 0.1545, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14709152281284332, | |
| "step": 6055, | |
| "valid_targets_mean": 1895.1, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 6.3655462184873945, | |
| "grad_norm": 0.7974623565020086, | |
| "learning_rate": 9.961024437230948e-07, | |
| "loss": 0.1493, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13561251759529114, | |
| "step": 6060, | |
| "valid_targets_mean": 2064.5, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 6.370798319327731, | |
| "grad_norm": 0.709502267033278, | |
| "learning_rate": 9.798412001565038e-07, | |
| "loss": 0.1591, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16175958514213562, | |
| "step": 6065, | |
| "valid_targets_mean": 3175.9, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 6.376050420168067, | |
| "grad_norm": 0.9037744133249849, | |
| "learning_rate": 9.63710449088453e-07, | |
| "loss": 0.1499, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15407252311706543, | |
| "step": 6070, | |
| "valid_targets_mean": 1908.1, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 6.381302521008403, | |
| "grad_norm": 0.797641245566909, | |
| "learning_rate": 9.477103011879408e-07, | |
| "loss": 0.1475, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15560175478458405, | |
| "step": 6075, | |
| "valid_targets_mean": 2465.2, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 6.38655462184874, | |
| "grad_norm": 0.7237125765315087, | |
| "learning_rate": 9.318408662279399e-07, | |
| "loss": 0.1527, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14476794004440308, | |
| "step": 6080, | |
| "valid_targets_mean": 2732.1, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 6.391806722689076, | |
| "grad_norm": 0.7575017424625796, | |
| "learning_rate": 9.161022530846253e-07, | |
| "loss": 0.1525, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13442182540893555, | |
| "step": 6085, | |
| "valid_targets_mean": 2915.2, | |
| "valid_targets_min": 1127 | |
| }, | |
| { | |
| "epoch": 6.397058823529412, | |
| "grad_norm": 0.8267114749782785, | |
| "learning_rate": 9.004945697366452e-07, | |
| "loss": 0.1447, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14276978373527527, | |
| "step": 6090, | |
| "valid_targets_mean": 2019.4, | |
| "valid_targets_min": 909 | |
| }, | |
| { | |
| "epoch": 6.402310924369748, | |
| "grad_norm": 0.7237677872215658, | |
| "learning_rate": 8.850179232643641e-07, | |
| "loss": 0.1501, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14779332280158997, | |
| "step": 6095, | |
| "valid_targets_mean": 2826.4, | |
| "valid_targets_min": 898 | |
| }, | |
| { | |
| "epoch": 6.407563025210084, | |
| "grad_norm": 0.8400966461486415, | |
| "learning_rate": 8.696724198491369e-07, | |
| "loss": 0.1537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15353678166866302, | |
| "step": 6100, | |
| "valid_targets_mean": 2215.9, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 6.41281512605042, | |
| "grad_norm": 0.7477003864167849, | |
| "learning_rate": 8.544581647725825e-07, | |
| "loss": 0.1461, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1523742377758026, | |
| "step": 6105, | |
| "valid_targets_mean": 2844.5, | |
| "valid_targets_min": 921 | |
| }, | |
| { | |
| "epoch": 6.418067226890757, | |
| "grad_norm": 0.8178307661739669, | |
| "learning_rate": 8.393752624158603e-07, | |
| "loss": 0.1498, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14179328083992004, | |
| "step": 6110, | |
| "valid_targets_mean": 1932.0, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 6.423319327731092, | |
| "grad_norm": 0.6634779755671151, | |
| "learning_rate": 8.244238162589435e-07, | |
| "loss": 0.148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14001640677452087, | |
| "step": 6115, | |
| "valid_targets_mean": 3367.4, | |
| "valid_targets_min": 1289 | |
| }, | |
| { | |
| "epoch": 6.428571428571429, | |
| "grad_norm": 0.7620021165612041, | |
| "learning_rate": 8.096039288799251e-07, | |
| "loss": 0.1563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15005478262901306, | |
| "step": 6120, | |
| "valid_targets_mean": 2588.8, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 6.4338235294117645, | |
| "grad_norm": 0.7934644918251871, | |
| "learning_rate": 7.949157019543064e-07, | |
| "loss": 0.1559, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17349065840244293, | |
| "step": 6125, | |
| "valid_targets_mean": 2498.0, | |
| "valid_targets_min": 909 | |
| }, | |
| { | |
| "epoch": 6.439075630252101, | |
| "grad_norm": 0.7690205077229771, | |
| "learning_rate": 7.803592362542911e-07, | |
| "loss": 0.1503, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16106484830379486, | |
| "step": 6130, | |
| "valid_targets_mean": 2574.9, | |
| "valid_targets_min": 1357 | |
| }, | |
| { | |
| "epoch": 6.444327731092437, | |
| "grad_norm": 0.8681115870866193, | |
| "learning_rate": 7.659346316481086e-07, | |
| "loss": 0.1505, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14451289176940918, | |
| "step": 6135, | |
| "valid_targets_mean": 1931.4, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 6.449579831932773, | |
| "grad_norm": 0.7601643918944068, | |
| "learning_rate": 7.516419870993296e-07, | |
| "loss": 0.1493, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15384207665920258, | |
| "step": 6140, | |
| "valid_targets_mean": 2423.4, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 6.454831932773109, | |
| "grad_norm": 0.7370958836376663, | |
| "learning_rate": 7.374814006661667e-07, | |
| "loss": 0.152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1404162049293518, | |
| "step": 6145, | |
| "valid_targets_mean": 2909.4, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.4600840336134455, | |
| "grad_norm": 0.8042467319801916, | |
| "learning_rate": 7.234529695008241e-07, | |
| "loss": 0.1459, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15728093683719635, | |
| "step": 6150, | |
| "valid_targets_mean": 2223.1, | |
| "valid_targets_min": 979 | |
| }, | |
| { | |
| "epoch": 6.465336134453781, | |
| "grad_norm": 0.7816915471785587, | |
| "learning_rate": 7.095567898488175e-07, | |
| "loss": 0.156, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16815775632858276, | |
| "step": 6155, | |
| "valid_targets_mean": 2402.7, | |
| "valid_targets_min": 1133 | |
| }, | |
| { | |
| "epoch": 6.470588235294118, | |
| "grad_norm": 0.9016085189910406, | |
| "learning_rate": 6.957929570483224e-07, | |
| "loss": 0.1494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16057243943214417, | |
| "step": 6160, | |
| "valid_targets_mean": 2306.2, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 6.475840336134453, | |
| "grad_norm": 0.7983769298179959, | |
| "learning_rate": 6.821615655295155e-07, | |
| "loss": 0.1605, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16977722942829132, | |
| "step": 6165, | |
| "valid_targets_mean": 2523.9, | |
| "valid_targets_min": 604 | |
| }, | |
| { | |
| "epoch": 6.48109243697479, | |
| "grad_norm": 0.8711158720727042, | |
| "learning_rate": 6.686627088139253e-07, | |
| "loss": 0.1579, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17021444439888, | |
| "step": 6170, | |
| "valid_targets_mean": 2276.9, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 6.486344537815126, | |
| "grad_norm": 0.7653732434373907, | |
| "learning_rate": 6.552964795137895e-07, | |
| "loss": 0.1529, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16731613874435425, | |
| "step": 6175, | |
| "valid_targets_mean": 2977.2, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 6.491596638655462, | |
| "grad_norm": 0.7898921304427059, | |
| "learning_rate": 6.420629693314317e-07, | |
| "loss": 0.148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14303874969482422, | |
| "step": 6180, | |
| "valid_targets_mean": 2086.4, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 6.496848739495798, | |
| "grad_norm": 0.8022544168817034, | |
| "learning_rate": 6.289622690586151e-07, | |
| "loss": 0.1472, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15002703666687012, | |
| "step": 6185, | |
| "valid_targets_mean": 2577.0, | |
| "valid_targets_min": 967 | |
| }, | |
| { | |
| "epoch": 6.5021008403361344, | |
| "grad_norm": 0.7765235327781467, | |
| "learning_rate": 6.159944685759289e-07, | |
| "loss": 0.1598, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1665143221616745, | |
| "step": 6190, | |
| "valid_targets_mean": 2581.1, | |
| "valid_targets_min": 1206 | |
| }, | |
| { | |
| "epoch": 6.507352941176471, | |
| "grad_norm": 0.6652766905185039, | |
| "learning_rate": 6.031596568521792e-07, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15390489995479584, | |
| "step": 6195, | |
| "valid_targets_mean": 3575.8, | |
| "valid_targets_min": 936 | |
| }, | |
| { | |
| "epoch": 6.512605042016807, | |
| "grad_norm": 0.823897053974955, | |
| "learning_rate": 5.904579219437567e-07, | |
| "loss": 0.1504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14515843987464905, | |
| "step": 6200, | |
| "valid_targets_mean": 2476.0, | |
| "valid_targets_min": 897 | |
| }, | |
| { | |
| "epoch": 6.517857142857143, | |
| "grad_norm": 0.8693729158649441, | |
| "learning_rate": 5.778893509940497e-07, | |
| "loss": 0.1456, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14359763264656067, | |
| "step": 6205, | |
| "valid_targets_mean": 1601.5, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 6.523109243697479, | |
| "grad_norm": 0.7360524356141547, | |
| "learning_rate": 5.654540302328526e-07, | |
| "loss": 0.1464, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1544869840145111, | |
| "step": 6210, | |
| "valid_targets_mean": 3031.3, | |
| "valid_targets_min": 868 | |
| }, | |
| { | |
| "epoch": 6.5283613445378155, | |
| "grad_norm": 0.8214329921549752, | |
| "learning_rate": 5.531520449757465e-07, | |
| "loss": 0.1569, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15624156594276428, | |
| "step": 6215, | |
| "valid_targets_mean": 2231.3, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 6.533613445378151, | |
| "grad_norm": 0.7526069627643952, | |
| "learning_rate": 5.409834796235447e-07, | |
| "loss": 0.1503, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14001137018203735, | |
| "step": 6220, | |
| "valid_targets_mean": 2640.0, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.538865546218488, | |
| "grad_norm": 0.8221429966093985, | |
| "learning_rate": 5.28948417661701e-07, | |
| "loss": 0.153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1630651354789734, | |
| "step": 6225, | |
| "valid_targets_mean": 2143.7, | |
| "valid_targets_min": 668 | |
| }, | |
| { | |
| "epoch": 6.544117647058823, | |
| "grad_norm": 0.8173958277347929, | |
| "learning_rate": 5.170469416597223e-07, | |
| "loss": 0.1617, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14818143844604492, | |
| "step": 6230, | |
| "valid_targets_mean": 2997.8, | |
| "valid_targets_min": 1061 | |
| }, | |
| { | |
| "epoch": 6.54936974789916, | |
| "grad_norm": 0.7574217024984009, | |
| "learning_rate": 5.05279133270633e-07, | |
| "loss": 0.1438, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1506175696849823, | |
| "step": 6235, | |
| "valid_targets_mean": 2813.8, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 6.554621848739496, | |
| "grad_norm": 0.8789551921829356, | |
| "learning_rate": 4.936450732303866e-07, | |
| "loss": 0.1648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17505599558353424, | |
| "step": 6240, | |
| "valid_targets_mean": 2185.9, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 6.559873949579832, | |
| "grad_norm": 0.7800209416470806, | |
| "learning_rate": 4.821448413573237e-07, | |
| "loss": 0.1542, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1431932896375656, | |
| "step": 6245, | |
| "valid_targets_mean": 2216.9, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 6.565126050420168, | |
| "grad_norm": 0.7715463308797006, | |
| "learning_rate": 4.70778516551631e-07, | |
| "loss": 0.162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15793776512145996, | |
| "step": 6250, | |
| "valid_targets_mean": 2604.6, | |
| "valid_targets_min": 873 | |
| }, | |
| { | |
| "epoch": 6.570378151260504, | |
| "grad_norm": 0.7929402427060284, | |
| "learning_rate": 4.595461767947829e-07, | |
| "loss": 0.1543, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17367711663246155, | |
| "step": 6255, | |
| "valid_targets_mean": 2793.5, | |
| "valid_targets_min": 890 | |
| }, | |
| { | |
| "epoch": 6.57563025210084, | |
| "grad_norm": 0.7558379818924563, | |
| "learning_rate": 4.484478991490249e-07, | |
| "loss": 0.1504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1597421020269394, | |
| "step": 6260, | |
| "valid_targets_mean": 2706.1, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 6.580882352941177, | |
| "grad_norm": 0.8203283121654529, | |
| "learning_rate": 4.374837597568338e-07, | |
| "loss": 0.1526, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13600069284439087, | |
| "step": 6265, | |
| "valid_targets_mean": 2013.9, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 6.586134453781512, | |
| "grad_norm": 0.7685017317264076, | |
| "learning_rate": 4.266538338403936e-07, | |
| "loss": 0.147, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1418771594762802, | |
| "step": 6270, | |
| "valid_targets_mean": 2087.0, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 6.591386554621849, | |
| "grad_norm": 0.7689705022584826, | |
| "learning_rate": 4.159581957010894e-07, | |
| "loss": 0.149, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15910068154335022, | |
| "step": 6275, | |
| "valid_targets_mean": 2749.8, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 6.5966386554621845, | |
| "grad_norm": 0.8350274594286426, | |
| "learning_rate": 4.0539691871898545e-07, | |
| "loss": 0.1635, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14839161932468414, | |
| "step": 6280, | |
| "valid_targets_mean": 2080.6, | |
| "valid_targets_min": 941 | |
| }, | |
| { | |
| "epoch": 6.601890756302521, | |
| "grad_norm": 0.7123586297209112, | |
| "learning_rate": 3.949700753523344e-07, | |
| "loss": 0.1468, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14631910622119904, | |
| "step": 6285, | |
| "valid_targets_mean": 2960.0, | |
| "valid_targets_min": 763 | |
| }, | |
| { | |
| "epoch": 6.607142857142857, | |
| "grad_norm": 0.7608218792468966, | |
| "learning_rate": 3.8467773713707134e-07, | |
| "loss": 0.1494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14922720193862915, | |
| "step": 6290, | |
| "valid_targets_mean": 2525.2, | |
| "valid_targets_min": 1231 | |
| }, | |
| { | |
| "epoch": 6.612394957983193, | |
| "grad_norm": 0.781295609634859, | |
| "learning_rate": 3.7451997468632486e-07, | |
| "loss": 0.142, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1524946689605713, | |
| "step": 6295, | |
| "valid_targets_mean": 2446.9, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.617647058823529, | |
| "grad_norm": 1.3621692596047696, | |
| "learning_rate": 3.6449685768993327e-07, | |
| "loss": 0.1563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15170569717884064, | |
| "step": 6300, | |
| "valid_targets_mean": 2007.3, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 6.6228991596638656, | |
| "grad_norm": 0.9081109225494257, | |
| "learning_rate": 3.5460845491396944e-07, | |
| "loss": 0.1547, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.154941126704216, | |
| "step": 6305, | |
| "valid_targets_mean": 2046.6, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 6.628151260504202, | |
| "grad_norm": 0.7545604154002317, | |
| "learning_rate": 3.448548342002589e-07, | |
| "loss": 0.1534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15161868929862976, | |
| "step": 6310, | |
| "valid_targets_mean": 2837.8, | |
| "valid_targets_min": 973 | |
| }, | |
| { | |
| "epoch": 6.633403361344538, | |
| "grad_norm": 0.8680496469961246, | |
| "learning_rate": 3.352360624659312e-07, | |
| "loss": 0.16, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1834527552127838, | |
| "step": 6315, | |
| "valid_targets_mean": 2255.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 6.6386554621848735, | |
| "grad_norm": 0.7171097775844976, | |
| "learning_rate": 3.2575220570294276e-07, | |
| "loss": 0.1529, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15503129363059998, | |
| "step": 6320, | |
| "valid_targets_mean": 2962.4, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 6.64390756302521, | |
| "grad_norm": 0.6658600775544034, | |
| "learning_rate": 3.164033289776369e-07, | |
| "loss": 0.1423, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12674590945243835, | |
| "step": 6325, | |
| "valid_targets_mean": 2999.2, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 6.649159663865547, | |
| "grad_norm": 0.7748035457490327, | |
| "learning_rate": 3.071894964302935e-07, | |
| "loss": 0.1477, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16250720620155334, | |
| "step": 6330, | |
| "valid_targets_mean": 2583.9, | |
| "valid_targets_min": 850 | |
| }, | |
| { | |
| "epoch": 6.654411764705882, | |
| "grad_norm": 0.7894017447041788, | |
| "learning_rate": 2.981107712746867e-07, | |
| "loss": 0.1448, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14615963399410248, | |
| "step": 6335, | |
| "valid_targets_mean": 2359.9, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 6.659663865546219, | |
| "grad_norm": 0.7275796769976358, | |
| "learning_rate": 2.891672157976522e-07, | |
| "loss": 0.1389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1456836611032486, | |
| "step": 6340, | |
| "valid_targets_mean": 2687.1, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 6.6649159663865545, | |
| "grad_norm": 0.7464610444598423, | |
| "learning_rate": 2.803588913586608e-07, | |
| "loss": 0.154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15692494809627533, | |
| "step": 6345, | |
| "valid_targets_mean": 2862.6, | |
| "valid_targets_min": 1185 | |
| }, | |
| { | |
| "epoch": 6.670168067226891, | |
| "grad_norm": 0.7409587755798085, | |
| "learning_rate": 2.716858583894033e-07, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14683926105499268, | |
| "step": 6350, | |
| "valid_targets_mean": 2565.6, | |
| "valid_targets_min": 1187 | |
| }, | |
| { | |
| "epoch": 6.675420168067227, | |
| "grad_norm": 0.8489660086658728, | |
| "learning_rate": 2.6314817639335964e-07, | |
| "loss": 0.1486, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14783814549446106, | |
| "step": 6355, | |
| "valid_targets_mean": 2178.1, | |
| "valid_targets_min": 1163 | |
| }, | |
| { | |
| "epoch": 6.680672268907563, | |
| "grad_norm": 0.7446724408854097, | |
| "learning_rate": 2.547459039454103e-07, | |
| "loss": 0.1542, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13613933324813843, | |
| "step": 6360, | |
| "valid_targets_mean": 2619.4, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 6.685924369747899, | |
| "grad_norm": 0.7468301149926823, | |
| "learning_rate": 2.4647909869142117e-07, | |
| "loss": 0.15, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15531302988529205, | |
| "step": 6365, | |
| "valid_targets_mean": 2749.7, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 6.6911764705882355, | |
| "grad_norm": 0.8038823609900482, | |
| "learning_rate": 2.3834781734784817e-07, | |
| "loss": 0.1534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14846262335777283, | |
| "step": 6370, | |
| "valid_targets_mean": 2125.0, | |
| "valid_targets_min": 852 | |
| }, | |
| { | |
| "epoch": 6.696428571428571, | |
| "grad_norm": 0.7769587079233549, | |
| "learning_rate": 2.3035211570135995e-07, | |
| "loss": 0.1528, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1341462880373001, | |
| "step": 6375, | |
| "valid_targets_mean": 1821.1, | |
| "valid_targets_min": 958 | |
| }, | |
| { | |
| "epoch": 6.701680672268908, | |
| "grad_norm": 0.7298168762864212, | |
| "learning_rate": 2.224920486084403e-07, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15344950556755066, | |
| "step": 6380, | |
| "valid_targets_mean": 2839.2, | |
| "valid_targets_min": 1147 | |
| }, | |
| { | |
| "epoch": 6.706932773109243, | |
| "grad_norm": 0.778338115342936, | |
| "learning_rate": 2.1476766999502408e-07, | |
| "loss": 0.1428, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13469842076301575, | |
| "step": 6385, | |
| "valid_targets_mean": 2178.6, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 6.71218487394958, | |
| "grad_norm": 0.816451492189734, | |
| "learning_rate": 2.071790328561152e-07, | |
| "loss": 0.1497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15425826609134674, | |
| "step": 6390, | |
| "valid_targets_mean": 2324.4, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 6.717436974789916, | |
| "grad_norm": 0.8201419970370695, | |
| "learning_rate": 1.997261892554403e-07, | |
| "loss": 0.1563, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15658767521381378, | |
| "step": 6395, | |
| "valid_targets_mean": 2310.4, | |
| "valid_targets_min": 912 | |
| }, | |
| { | |
| "epoch": 6.722689075630252, | |
| "grad_norm": 0.7500376618247141, | |
| "learning_rate": 1.9240919032506688e-07, | |
| "loss": 0.147, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14455173909664154, | |
| "step": 6400, | |
| "valid_targets_mean": 2628.5, | |
| "valid_targets_min": 474 | |
| }, | |
| { | |
| "epoch": 6.727941176470588, | |
| "grad_norm": 0.7584572682030627, | |
| "learning_rate": 1.8522808626507683e-07, | |
| "loss": 0.151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15599125623703003, | |
| "step": 6405, | |
| "valid_targets_mean": 2658.8, | |
| "valid_targets_min": 420 | |
| }, | |
| { | |
| "epoch": 6.733193277310924, | |
| "grad_norm": 0.7357895023796207, | |
| "learning_rate": 1.781829263432111e-07, | |
| "loss": 0.1594, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14581039547920227, | |
| "step": 6410, | |
| "valid_targets_mean": 2959.5, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 6.73844537815126, | |
| "grad_norm": 0.6891830564600709, | |
| "learning_rate": 1.7127375889452569e-07, | |
| "loss": 0.1528, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13923324644565582, | |
| "step": 6415, | |
| "valid_targets_mean": 2834.0, | |
| "valid_targets_min": 1194 | |
| }, | |
| { | |
| "epoch": 6.743697478991597, | |
| "grad_norm": 0.7199729752533398, | |
| "learning_rate": 1.6450063132107396e-07, | |
| "loss": 0.1479, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1373451054096222, | |
| "step": 6420, | |
| "valid_targets_mean": 2572.8, | |
| "valid_targets_min": 626 | |
| }, | |
| { | |
| "epoch": 6.748949579831933, | |
| "grad_norm": 0.7647019920185888, | |
| "learning_rate": 1.5786359009156928e-07, | |
| "loss": 0.1433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15924891829490662, | |
| "step": 6425, | |
| "valid_targets_mean": 2990.7, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 6.754201680672269, | |
| "grad_norm": 0.8310727483362449, | |
| "learning_rate": 1.5136268074107398e-07, | |
| "loss": 0.157, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1688929796218872, | |
| "step": 6430, | |
| "valid_targets_mean": 2370.5, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 6.759453781512605, | |
| "grad_norm": 0.825061605955483, | |
| "learning_rate": 1.44997947870682e-07, | |
| "loss": 0.1639, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16266357898712158, | |
| "step": 6435, | |
| "valid_targets_mean": 2281.9, | |
| "valid_targets_min": 1074 | |
| }, | |
| { | |
| "epoch": 6.764705882352941, | |
| "grad_norm": 0.8533595418248784, | |
| "learning_rate": 1.3876943514721465e-07, | |
| "loss": 0.1525, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16743487119674683, | |
| "step": 6440, | |
| "valid_targets_mean": 2262.5, | |
| "valid_targets_min": 970 | |
| }, | |
| { | |
| "epoch": 6.769957983193278, | |
| "grad_norm": 0.7748683785377023, | |
| "learning_rate": 1.3267718530292296e-07, | |
| "loss": 0.1554, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14080220460891724, | |
| "step": 6445, | |
| "valid_targets_mean": 2343.4, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 6.775210084033613, | |
| "grad_norm": 0.8320606022762239, | |
| "learning_rate": 1.267212401351925e-07, | |
| "loss": 0.16, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1550496518611908, | |
| "step": 6450, | |
| "valid_targets_mean": 2402.6, | |
| "valid_targets_min": 857 | |
| }, | |
| { | |
| "epoch": 6.78046218487395, | |
| "grad_norm": 0.7474317684737926, | |
| "learning_rate": 1.2090164050625907e-07, | |
| "loss": 0.1736, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16620126366615295, | |
| "step": 6455, | |
| "valid_targets_mean": 3082.7, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 6.785714285714286, | |
| "grad_norm": 0.9565980511087515, | |
| "learning_rate": 1.1521842634292013e-07, | |
| "loss": 0.1559, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.160188689827919, | |
| "step": 6460, | |
| "valid_targets_mean": 1479.5, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 6.790966386554622, | |
| "grad_norm": 0.8532947647691757, | |
| "learning_rate": 1.0967163663627044e-07, | |
| "loss": 0.15, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15419529378414154, | |
| "step": 6465, | |
| "valid_targets_mean": 2052.6, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 6.796218487394958, | |
| "grad_norm": 0.8255781206219238, | |
| "learning_rate": 1.0426130944143353e-07, | |
| "loss": 0.151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14306730031967163, | |
| "step": 6470, | |
| "valid_targets_mean": 1963.2, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 6.801470588235294, | |
| "grad_norm": 0.8837049669576724, | |
| "learning_rate": 9.898748187729513e-08, | |
| "loss": 0.1549, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15094450116157532, | |
| "step": 6475, | |
| "valid_targets_mean": 2181.6, | |
| "valid_targets_min": 919 | |
| }, | |
| { | |
| "epoch": 6.80672268907563, | |
| "grad_norm": 0.7790666833598698, | |
| "learning_rate": 9.385019012625007e-08, | |
| "loss": 0.1469, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14933770895004272, | |
| "step": 6480, | |
| "valid_targets_mean": 2469.9, | |
| "valid_targets_min": 1015 | |
| }, | |
| { | |
| "epoch": 6.811974789915967, | |
| "grad_norm": 0.9087222050552412, | |
| "learning_rate": 8.884946943395811e-08, | |
| "loss": 0.1573, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16928185522556305, | |
| "step": 6485, | |
| "valid_targets_mean": 1823.7, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 6.817226890756302, | |
| "grad_norm": 0.7419966743487549, | |
| "learning_rate": 8.398535410910402e-08, | |
| "loss": 0.161, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16752246022224426, | |
| "step": 6490, | |
| "valid_targets_mean": 3067.3, | |
| "valid_targets_min": 796 | |
| }, | |
| { | |
| "epoch": 6.822478991596639, | |
| "grad_norm": 0.7768443736140654, | |
| "learning_rate": 7.925787752314674e-08, | |
| "loss": 0.1552, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14006970822811127, | |
| "step": 6495, | |
| "valid_targets_mean": 2286.3, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 6.8277310924369745, | |
| "grad_norm": 0.7839422460590042, | |
| "learning_rate": 7.466707211010838e-08, | |
| "loss": 0.1526, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15227356553077698, | |
| "step": 6500, | |
| "valid_targets_mean": 2435.8, | |
| "valid_targets_min": 794 | |
| }, | |
| { | |
| "epoch": 6.832983193277311, | |
| "grad_norm": 0.8085913813487312, | |
| "learning_rate": 7.02129693663478e-08, | |
| "loss": 0.1468, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14806930720806122, | |
| "step": 6505, | |
| "valid_targets_mean": 2763.4, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 6.838235294117647, | |
| "grad_norm": 0.830479023779582, | |
| "learning_rate": 6.589559985033189e-08, | |
| "loss": 0.1577, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15857365727424622, | |
| "step": 6510, | |
| "valid_targets_mean": 2199.1, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 6.843487394957983, | |
| "grad_norm": 0.7805037615428879, | |
| "learning_rate": 6.171499318244234e-08, | |
| "loss": 0.1557, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1644633412361145, | |
| "step": 6515, | |
| "valid_targets_mean": 2826.9, | |
| "valid_targets_min": 1091 | |
| }, | |
| { | |
| "epoch": 6.848739495798319, | |
| "grad_norm": 0.8243868421023237, | |
| "learning_rate": 5.767117804476696e-08, | |
| "loss": 0.1468, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1619555652141571, | |
| "step": 6520, | |
| "valid_targets_mean": 2291.1, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 6.8539915966386555, | |
| "grad_norm": 0.6718196143534856, | |
| "learning_rate": 5.376418218089541e-08, | |
| "loss": 0.1468, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1486797034740448, | |
| "step": 6525, | |
| "valid_targets_mean": 3316.1, | |
| "valid_targets_min": 645 | |
| }, | |
| { | |
| "epoch": 6.859243697478991, | |
| "grad_norm": 0.7639600711267872, | |
| "learning_rate": 4.99940323957393e-08, | |
| "loss": 0.1506, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15114453434944153, | |
| "step": 6530, | |
| "valid_targets_mean": 2355.1, | |
| "valid_targets_min": 930 | |
| }, | |
| { | |
| "epoch": 6.864495798319328, | |
| "grad_norm": 0.7434040074285687, | |
| "learning_rate": 4.63607545553435e-08, | |
| "loss": 0.1491, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16190582513809204, | |
| "step": 6535, | |
| "valid_targets_mean": 2631.9, | |
| "valid_targets_min": 997 | |
| }, | |
| { | |
| "epoch": 6.869747899159664, | |
| "grad_norm": 0.7576726635038536, | |
| "learning_rate": 4.2864373586706254e-08, | |
| "loss": 0.1539, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1667475700378418, | |
| "step": 6540, | |
| "valid_targets_mean": 3037.7, | |
| "valid_targets_min": 1256 | |
| }, | |
| { | |
| "epoch": 6.875, | |
| "grad_norm": 0.7949042840267881, | |
| "learning_rate": 3.950491347761487e-08, | |
| "loss": 0.1534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15799729526042938, | |
| "step": 6545, | |
| "valid_targets_mean": 2813.3, | |
| "valid_targets_min": 981 | |
| }, | |
| { | |
| "epoch": 6.880252100840336, | |
| "grad_norm": 0.7599470557620672, | |
| "learning_rate": 3.628239727647254e-08, | |
| "loss": 0.1517, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14834260940551758, | |
| "step": 6550, | |
| "valid_targets_mean": 2786.1, | |
| "valid_targets_min": 551 | |
| }, | |
| { | |
| "epoch": 6.885504201680672, | |
| "grad_norm": 0.7438096344765326, | |
| "learning_rate": 3.319684709215176e-08, | |
| "loss": 0.1488, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1484309434890747, | |
| "step": 6555, | |
| "valid_targets_mean": 2842.8, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 6.890756302521009, | |
| "grad_norm": 0.8034592781374419, | |
| "learning_rate": 3.024828409383007e-08, | |
| "loss": 0.1601, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1651192605495453, | |
| "step": 6560, | |
| "valid_targets_mean": 2401.8, | |
| "valid_targets_min": 1118 | |
| }, | |
| { | |
| "epoch": 6.8960084033613445, | |
| "grad_norm": 0.7481979263488755, | |
| "learning_rate": 2.743672851085233e-08, | |
| "loss": 0.1447, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13949503004550934, | |
| "step": 6565, | |
| "valid_targets_mean": 2675.6, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 6.901260504201681, | |
| "grad_norm": 0.8092072610442079, | |
| "learning_rate": 2.4762199632588634e-08, | |
| "loss": 0.1466, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17036761343479156, | |
| "step": 6570, | |
| "valid_targets_mean": 2485.1, | |
| "valid_targets_min": 909 | |
| }, | |
| { | |
| "epoch": 6.906512605042017, | |
| "grad_norm": 0.8049873065525076, | |
| "learning_rate": 2.2224715808309983e-08, | |
| "loss": 0.156, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15256905555725098, | |
| "step": 6575, | |
| "valid_targets_mean": 2644.2, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 6.911764705882353, | |
| "grad_norm": 0.8617962635473019, | |
| "learning_rate": 1.9824294447043923e-08, | |
| "loss": 0.1571, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15142729878425598, | |
| "step": 6580, | |
| "valid_targets_mean": 2083.2, | |
| "valid_targets_min": 512 | |
| }, | |
| { | |
| "epoch": 6.917016806722689, | |
| "grad_norm": 0.8008776691217604, | |
| "learning_rate": 1.7560952017481313e-08, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1459037810564041, | |
| "step": 6585, | |
| "valid_targets_mean": 2554.2, | |
| "valid_targets_min": 952 | |
| }, | |
| { | |
| "epoch": 6.9222689075630255, | |
| "grad_norm": 0.7191951703364022, | |
| "learning_rate": 1.5434704047836424e-08, | |
| "loss": 0.1475, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.145742267370224, | |
| "step": 6590, | |
| "valid_targets_mean": 2792.4, | |
| "valid_targets_min": 651 | |
| }, | |
| { | |
| "epoch": 6.927521008403361, | |
| "grad_norm": 0.7871859947545846, | |
| "learning_rate": 1.344556512576256e-08, | |
| "loss": 0.1575, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1482214331626892, | |
| "step": 6595, | |
| "valid_targets_mean": 2457.0, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.932773109243698, | |
| "grad_norm": 0.7643202727978451, | |
| "learning_rate": 1.1593548898236606e-08, | |
| "loss": 0.162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15590517222881317, | |
| "step": 6600, | |
| "valid_targets_mean": 2423.8, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.938025210084033, | |
| "grad_norm": 0.8242802199893152, | |
| "learning_rate": 9.878668071474639e-09, | |
| "loss": 0.1556, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1575925648212433, | |
| "step": 6605, | |
| "valid_targets_mean": 2134.1, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 6.94327731092437, | |
| "grad_norm": 0.7939290019589629, | |
| "learning_rate": 8.3009344108409e-09, | |
| "loss": 0.1626, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1599077582359314, | |
| "step": 6610, | |
| "valid_targets_mean": 2462.1, | |
| "valid_targets_min": 1122 | |
| }, | |
| { | |
| "epoch": 6.948529411764706, | |
| "grad_norm": 0.8359337734301073, | |
| "learning_rate": 6.860358740763406e-09, | |
| "loss": 0.1607, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15786179900169373, | |
| "step": 6615, | |
| "valid_targets_mean": 2346.8, | |
| "valid_targets_min": 628 | |
| }, | |
| { | |
| "epoch": 6.953781512605042, | |
| "grad_norm": 0.8230966514103626, | |
| "learning_rate": 5.55695094467179e-09, | |
| "loss": 0.1569, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15767702460289001, | |
| "step": 6620, | |
| "valid_targets_mean": 2301.7, | |
| "valid_targets_min": 667 | |
| }, | |
| { | |
| "epoch": 6.959033613445378, | |
| "grad_norm": 0.8839955512653498, | |
| "learning_rate": 4.3907199649151355e-09, | |
| "loss": 0.1555, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14466795325279236, | |
| "step": 6625, | |
| "valid_targets_mean": 1904.7, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 6.964285714285714, | |
| "grad_norm": 0.7315947556803873, | |
| "learning_rate": 3.361673802708687e-09, | |
| "loss": 0.1543, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1519755721092224, | |
| "step": 6630, | |
| "valid_targets_mean": 2944.4, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 6.96953781512605, | |
| "grad_norm": 0.9022304224894817, | |
| "learning_rate": 2.469819518080563e-09, | |
| "loss": 0.1613, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1760965883731842, | |
| "step": 6635, | |
| "valid_targets_mean": 2122.0, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 6.974789915966387, | |
| "grad_norm": 0.9761476002877811, | |
| "learning_rate": 1.7151632298140209e-09, | |
| "loss": 0.1531, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14784833788871765, | |
| "step": 6640, | |
| "valid_targets_mean": 2816.7, | |
| "valid_targets_min": 1127 | |
| }, | |
| { | |
| "epoch": 6.980042016806722, | |
| "grad_norm": 0.7716322233500594, | |
| "learning_rate": 1.0977101154163727e-09, | |
| "loss": 0.1524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.155085951089859, | |
| "step": 6645, | |
| "valid_targets_mean": 2472.3, | |
| "valid_targets_min": 1348 | |
| }, | |
| { | |
| "epoch": 6.985294117647059, | |
| "grad_norm": 0.8463421638101505, | |
| "learning_rate": 6.174644110767958e-10, | |
| "loss": 0.1534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16529741883277893, | |
| "step": 6650, | |
| "valid_targets_mean": 2150.7, | |
| "valid_targets_min": 842 | |
| }, | |
| { | |
| "epoch": 6.990546218487395, | |
| "grad_norm": 0.7405547503921337, | |
| "learning_rate": 2.744294116419077e-10, | |
| "loss": 0.146, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13631612062454224, | |
| "step": 6655, | |
| "valid_targets_mean": 2381.4, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.995798319327731, | |
| "grad_norm": 0.8184309324867913, | |
| "learning_rate": 6.860747058468064e-11, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15874159336090088, | |
| "step": 6660, | |
| "valid_targets_mean": 2120.4, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1511024534702301, | |
| "step": 6664, | |
| "total_flos": 1189037529694208.0, | |
| "train_loss": 0.22791378844041926, | |
| "train_runtime": 28576.9639, | |
| "train_samples_per_second": 3.731, | |
| "train_steps_per_second": 0.233, | |
| "valid_targets_mean": 3122.5, | |
| "valid_targets_min": 647 | |
| } | |
| ], | |
| "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": 1189037529694208.0, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |