Text Generation
Transformers
Safetensors
qwen3
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Instructions to use DCAgent2/nl2bash-bugs-over5050 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use DCAgent2/nl2bash-bugs-over5050 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="DCAgent2/nl2bash-bugs-over5050") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("DCAgent2/nl2bash-bugs-over5050") model = AutoModelForCausalLM.from_pretrained("DCAgent2/nl2bash-bugs-over5050") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use DCAgent2/nl2bash-bugs-over5050 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "DCAgent2/nl2bash-bugs-over5050" # 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-bugs-over5050", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/DCAgent2/nl2bash-bugs-over5050
- SGLang
How to use DCAgent2/nl2bash-bugs-over5050 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-bugs-over5050" \ --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-bugs-over5050", "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-bugs-over5050" \ --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-bugs-over5050", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use DCAgent2/nl2bash-bugs-over5050 with Docker Model Runner:
docker model run hf.co/DCAgent2/nl2bash-bugs-over5050
| { | |
| "best_global_step": null, | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 7.0, | |
| "eval_steps": 500, | |
| "global_step": 7140, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.004901960784313725, | |
| "grad_norm": 10.20928865918908, | |
| "learning_rate": 2.2408963585434175e-07, | |
| "loss": 0.6185, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5775138139724731, | |
| "step": 5, | |
| "valid_targets_mean": 2510.1, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 0.00980392156862745, | |
| "grad_norm": 9.23497201065913, | |
| "learning_rate": 5.042016806722689e-07, | |
| "loss": 0.5949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.573584794998169, | |
| "step": 10, | |
| "valid_targets_mean": 2837.1, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 0.014705882352941176, | |
| "grad_norm": 11.158774276653146, | |
| "learning_rate": 7.843137254901962e-07, | |
| "loss": 0.644, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6744149327278137, | |
| "step": 15, | |
| "valid_targets_mean": 1654.6, | |
| "valid_targets_min": 516 | |
| }, | |
| { | |
| "epoch": 0.0196078431372549, | |
| "grad_norm": 9.567542813890764, | |
| "learning_rate": 1.0644257703081233e-06, | |
| "loss": 0.6143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.6221225261688232, | |
| "step": 20, | |
| "valid_targets_mean": 2297.6, | |
| "valid_targets_min": 998 | |
| }, | |
| { | |
| "epoch": 0.024509803921568627, | |
| "grad_norm": 10.35194264022352, | |
| "learning_rate": 1.3445378151260504e-06, | |
| "loss": 0.5931, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.7201113700866699, | |
| "step": 25, | |
| "valid_targets_mean": 1942.3, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 0.029411764705882353, | |
| "grad_norm": 6.019435455219015, | |
| "learning_rate": 1.6246498599439778e-06, | |
| "loss": 0.5995, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5141686201095581, | |
| "step": 30, | |
| "valid_targets_mean": 2379.7, | |
| "valid_targets_min": 608 | |
| }, | |
| { | |
| "epoch": 0.03431372549019608, | |
| "grad_norm": 5.347753558046135, | |
| "learning_rate": 1.904761904761905e-06, | |
| "loss": 0.5534, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5356568694114685, | |
| "step": 35, | |
| "valid_targets_mean": 2166.6, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 0.0392156862745098, | |
| "grad_norm": 4.620742455700474, | |
| "learning_rate": 2.184873949579832e-06, | |
| "loss": 0.4874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.5212767124176025, | |
| "step": 40, | |
| "valid_targets_mean": 1865.9, | |
| "valid_targets_min": 524 | |
| }, | |
| { | |
| "epoch": 0.04411764705882353, | |
| "grad_norm": 3.4604614828952194, | |
| "learning_rate": 2.4649859943977594e-06, | |
| "loss": 0.4418, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4115924835205078, | |
| "step": 45, | |
| "valid_targets_mean": 2360.3, | |
| "valid_targets_min": 836 | |
| }, | |
| { | |
| "epoch": 0.049019607843137254, | |
| "grad_norm": 2.926821476599717, | |
| "learning_rate": 2.7450980392156867e-06, | |
| "loss": 0.426, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4429447650909424, | |
| "step": 50, | |
| "valid_targets_mean": 1996.1, | |
| "valid_targets_min": 968 | |
| }, | |
| { | |
| "epoch": 0.05392156862745098, | |
| "grad_norm": 1.9890421518892802, | |
| "learning_rate": 3.0252100840336137e-06, | |
| "loss": 0.4483, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4572969973087311, | |
| "step": 55, | |
| "valid_targets_mean": 1957.2, | |
| "valid_targets_min": 464 | |
| }, | |
| { | |
| "epoch": 0.058823529411764705, | |
| "grad_norm": 1.6865611538561944, | |
| "learning_rate": 3.305322128851541e-06, | |
| "loss": 0.3748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.39461690187454224, | |
| "step": 60, | |
| "valid_targets_mean": 1830.6, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 0.06372549019607843, | |
| "grad_norm": 1.3527111817446995, | |
| "learning_rate": 3.585434173669468e-06, | |
| "loss": 0.3612, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.4330475330352783, | |
| "step": 65, | |
| "valid_targets_mean": 1900.5, | |
| "valid_targets_min": 664 | |
| }, | |
| { | |
| "epoch": 0.06862745098039216, | |
| "grad_norm": 1.19163364826996, | |
| "learning_rate": 3.865546218487396e-06, | |
| "loss": 0.372, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3685574531555176, | |
| "step": 70, | |
| "valid_targets_mean": 1673.1, | |
| "valid_targets_min": 497 | |
| }, | |
| { | |
| "epoch": 0.07352941176470588, | |
| "grad_norm": 0.9397906346537187, | |
| "learning_rate": 4.145658263305323e-06, | |
| "loss": 0.3291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3561197519302368, | |
| "step": 75, | |
| "valid_targets_mean": 2279.2, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 0.0784313725490196, | |
| "grad_norm": 0.727700887008284, | |
| "learning_rate": 4.4257703081232496e-06, | |
| "loss": 0.3366, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2575701177120209, | |
| "step": 80, | |
| "valid_targets_mean": 2609.6, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 0.08333333333333333, | |
| "grad_norm": 0.8886691451113053, | |
| "learning_rate": 4.705882352941177e-06, | |
| "loss": 0.2968, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29151153564453125, | |
| "step": 85, | |
| "valid_targets_mean": 1739.9, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 0.08823529411764706, | |
| "grad_norm": 0.7084752420049867, | |
| "learning_rate": 4.985994397759104e-06, | |
| "loss": 0.3002, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2468690723180771, | |
| "step": 90, | |
| "valid_targets_mean": 2406.2, | |
| "valid_targets_min": 795 | |
| }, | |
| { | |
| "epoch": 0.09313725490196079, | |
| "grad_norm": 0.8183678064858451, | |
| "learning_rate": 5.266106442577032e-06, | |
| "loss": 0.3088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2936941981315613, | |
| "step": 95, | |
| "valid_targets_mean": 2030.6, | |
| "valid_targets_min": 609 | |
| }, | |
| { | |
| "epoch": 0.09803921568627451, | |
| "grad_norm": 0.8376415918043708, | |
| "learning_rate": 5.546218487394959e-06, | |
| "loss": 0.2971, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3431953489780426, | |
| "step": 100, | |
| "valid_targets_mean": 2092.2, | |
| "valid_targets_min": 794 | |
| }, | |
| { | |
| "epoch": 0.10294117647058823, | |
| "grad_norm": 0.6813076714555519, | |
| "learning_rate": 5.826330532212886e-06, | |
| "loss": 0.2635, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2565889358520508, | |
| "step": 105, | |
| "valid_targets_mean": 2320.1, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 0.10784313725490197, | |
| "grad_norm": 0.7381189363611336, | |
| "learning_rate": 6.106442577030814e-06, | |
| "loss": 0.2656, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2865998148918152, | |
| "step": 110, | |
| "valid_targets_mean": 2249.0, | |
| "valid_targets_min": 1219 | |
| }, | |
| { | |
| "epoch": 0.11274509803921569, | |
| "grad_norm": 0.6544891537578502, | |
| "learning_rate": 6.386554621848739e-06, | |
| "loss": 0.3004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23120339214801788, | |
| "step": 115, | |
| "valid_targets_mean": 2653.5, | |
| "valid_targets_min": 337 | |
| }, | |
| { | |
| "epoch": 0.11764705882352941, | |
| "grad_norm": 0.7137286431407412, | |
| "learning_rate": 6.666666666666667e-06, | |
| "loss": 0.2751, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.28938379883766174, | |
| "step": 120, | |
| "valid_targets_mean": 2329.7, | |
| "valid_targets_min": 540 | |
| }, | |
| { | |
| "epoch": 0.12254901960784313, | |
| "grad_norm": 0.7257208667414642, | |
| "learning_rate": 6.946778711484594e-06, | |
| "loss": 0.2771, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2607773244380951, | |
| "step": 125, | |
| "valid_targets_mean": 2316.9, | |
| "valid_targets_min": 619 | |
| }, | |
| { | |
| "epoch": 0.12745098039215685, | |
| "grad_norm": 0.7230258011962319, | |
| "learning_rate": 7.226890756302521e-06, | |
| "loss": 0.2603, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21410706639289856, | |
| "step": 130, | |
| "valid_targets_mean": 2270.0, | |
| "valid_targets_min": 665 | |
| }, | |
| { | |
| "epoch": 0.1323529411764706, | |
| "grad_norm": 0.5759410981582812, | |
| "learning_rate": 7.507002801120449e-06, | |
| "loss": 0.2497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20663133263587952, | |
| "step": 135, | |
| "valid_targets_mean": 3114.1, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 0.13725490196078433, | |
| "grad_norm": 0.7572665669740622, | |
| "learning_rate": 7.787114845938376e-06, | |
| "loss": 0.26, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2611826956272125, | |
| "step": 140, | |
| "valid_targets_mean": 2193.1, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 0.14215686274509803, | |
| "grad_norm": 0.7714401933508285, | |
| "learning_rate": 8.067226890756303e-06, | |
| "loss": 0.2457, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2228604406118393, | |
| "step": 145, | |
| "valid_targets_mean": 2312.3, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 0.14705882352941177, | |
| "grad_norm": 0.7473849126557415, | |
| "learning_rate": 8.34733893557423e-06, | |
| "loss": 0.264, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.26429224014282227, | |
| "step": 150, | |
| "valid_targets_mean": 2596.1, | |
| "valid_targets_min": 582 | |
| }, | |
| { | |
| "epoch": 0.15196078431372548, | |
| "grad_norm": 0.8790059261164762, | |
| "learning_rate": 8.627450980392157e-06, | |
| "loss": 0.2516, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2524535357952118, | |
| "step": 155, | |
| "valid_targets_mean": 1923.6, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 0.1568627450980392, | |
| "grad_norm": 0.6881021835316934, | |
| "learning_rate": 8.907563025210085e-06, | |
| "loss": 0.2303, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22635899484157562, | |
| "step": 160, | |
| "valid_targets_mean": 1896.2, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 0.16176470588235295, | |
| "grad_norm": 0.690388606612483, | |
| "learning_rate": 9.187675070028012e-06, | |
| "loss": 0.2504, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21718770265579224, | |
| "step": 165, | |
| "valid_targets_mean": 1991.8, | |
| "valid_targets_min": 633 | |
| }, | |
| { | |
| "epoch": 0.16666666666666666, | |
| "grad_norm": 0.8425889137005795, | |
| "learning_rate": 9.467787114845938e-06, | |
| "loss": 0.2414, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23774409294128418, | |
| "step": 170, | |
| "valid_targets_mean": 1627.6, | |
| "valid_targets_min": 468 | |
| }, | |
| { | |
| "epoch": 0.1715686274509804, | |
| "grad_norm": 0.7091511823027294, | |
| "learning_rate": 9.747899159663867e-06, | |
| "loss": 0.2156, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1945105493068695, | |
| "step": 175, | |
| "valid_targets_mean": 2712.4, | |
| "valid_targets_min": 815 | |
| }, | |
| { | |
| "epoch": 0.17647058823529413, | |
| "grad_norm": 0.7962789255731376, | |
| "learning_rate": 1.0028011204481793e-05, | |
| "loss": 0.237, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.30094513297080994, | |
| "step": 180, | |
| "valid_targets_mean": 1933.2, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 0.18137254901960784, | |
| "grad_norm": 0.8709581990029245, | |
| "learning_rate": 1.030812324929972e-05, | |
| "loss": 0.2629, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24878257513046265, | |
| "step": 185, | |
| "valid_targets_mean": 2173.3, | |
| "valid_targets_min": 524 | |
| }, | |
| { | |
| "epoch": 0.18627450980392157, | |
| "grad_norm": 0.672591990213361, | |
| "learning_rate": 1.0588235294117648e-05, | |
| "loss": 0.2283, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21655958890914917, | |
| "step": 190, | |
| "valid_targets_mean": 2481.3, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 0.19117647058823528, | |
| "grad_norm": 0.5077039092070841, | |
| "learning_rate": 1.0868347338935574e-05, | |
| "loss": 0.213, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18501999974250793, | |
| "step": 195, | |
| "valid_targets_mean": 3169.5, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 0.19607843137254902, | |
| "grad_norm": 0.6393343276686302, | |
| "learning_rate": 1.1148459383753503e-05, | |
| "loss": 0.2507, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20450522005558014, | |
| "step": 200, | |
| "valid_targets_mean": 2330.5, | |
| "valid_targets_min": 752 | |
| }, | |
| { | |
| "epoch": 0.20098039215686275, | |
| "grad_norm": 0.7377664172401819, | |
| "learning_rate": 1.1428571428571429e-05, | |
| "loss": 0.2432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.24033306539058685, | |
| "step": 205, | |
| "valid_targets_mean": 2058.1, | |
| "valid_targets_min": 821 | |
| }, | |
| { | |
| "epoch": 0.20588235294117646, | |
| "grad_norm": 0.7084884926960634, | |
| "learning_rate": 1.1708683473389357e-05, | |
| "loss": 0.2375, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23557499051094055, | |
| "step": 210, | |
| "valid_targets_mean": 2050.4, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 0.2107843137254902, | |
| "grad_norm": 1.0188002373707512, | |
| "learning_rate": 1.1988795518207284e-05, | |
| "loss": 0.2432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29192501306533813, | |
| "step": 215, | |
| "valid_targets_mean": 1503.8, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 0.21568627450980393, | |
| "grad_norm": 0.6626325047822198, | |
| "learning_rate": 1.2268907563025212e-05, | |
| "loss": 0.2158, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20884019136428833, | |
| "step": 220, | |
| "valid_targets_mean": 2474.2, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 0.22058823529411764, | |
| "grad_norm": 0.6367170883319628, | |
| "learning_rate": 1.2549019607843138e-05, | |
| "loss": 0.2521, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2049025297164917, | |
| "step": 225, | |
| "valid_targets_mean": 2471.3, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 0.22549019607843138, | |
| "grad_norm": 0.7420048828864545, | |
| "learning_rate": 1.2829131652661067e-05, | |
| "loss": 0.2128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2391299456357956, | |
| "step": 230, | |
| "valid_targets_mean": 2442.0, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 0.23039215686274508, | |
| "grad_norm": 0.9380517123817401, | |
| "learning_rate": 1.3109243697478993e-05, | |
| "loss": 0.2315, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2618740200996399, | |
| "step": 235, | |
| "valid_targets_mean": 1533.4, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 0.23529411764705882, | |
| "grad_norm": 0.6743370799947248, | |
| "learning_rate": 1.338935574229692e-05, | |
| "loss": 0.2154, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2558346390724182, | |
| "step": 240, | |
| "valid_targets_mean": 2312.7, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 0.24019607843137256, | |
| "grad_norm": 0.6221079049005317, | |
| "learning_rate": 1.3669467787114848e-05, | |
| "loss": 0.2151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18583598732948303, | |
| "step": 245, | |
| "valid_targets_mean": 2520.4, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 0.24509803921568626, | |
| "grad_norm": 0.8012148275172146, | |
| "learning_rate": 1.3949579831932774e-05, | |
| "loss": 0.2112, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22044017910957336, | |
| "step": 250, | |
| "valid_targets_mean": 2009.4, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "grad_norm": 0.6180778632070758, | |
| "learning_rate": 1.4229691876750703e-05, | |
| "loss": 0.1938, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18992581963539124, | |
| "step": 255, | |
| "valid_targets_mean": 2631.8, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 0.2549019607843137, | |
| "grad_norm": 0.8993994771670876, | |
| "learning_rate": 1.4509803921568629e-05, | |
| "loss": 0.2093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21689380705356598, | |
| "step": 260, | |
| "valid_targets_mean": 1633.8, | |
| "valid_targets_min": 324 | |
| }, | |
| { | |
| "epoch": 0.25980392156862747, | |
| "grad_norm": 0.8889748801968507, | |
| "learning_rate": 1.4789915966386557e-05, | |
| "loss": 0.2005, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2001081109046936, | |
| "step": 265, | |
| "valid_targets_mean": 2276.2, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 0.2647058823529412, | |
| "grad_norm": 0.8677882826227222, | |
| "learning_rate": 1.5070028011204482e-05, | |
| "loss": 0.1971, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19296003878116608, | |
| "step": 270, | |
| "valid_targets_mean": 2239.6, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 0.2696078431372549, | |
| "grad_norm": 0.786172042943967, | |
| "learning_rate": 1.535014005602241e-05, | |
| "loss": 0.2033, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20809459686279297, | |
| "step": 275, | |
| "valid_targets_mean": 1825.3, | |
| "valid_targets_min": 701 | |
| }, | |
| { | |
| "epoch": 0.27450980392156865, | |
| "grad_norm": 0.6695572664647559, | |
| "learning_rate": 1.5630252100840337e-05, | |
| "loss": 0.2001, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18365135788917542, | |
| "step": 280, | |
| "valid_targets_mean": 2497.1, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 0.27941176470588236, | |
| "grad_norm": 0.8627662588551201, | |
| "learning_rate": 1.5910364145658263e-05, | |
| "loss": 0.2118, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22431831061840057, | |
| "step": 285, | |
| "valid_targets_mean": 2399.4, | |
| "valid_targets_min": 1180 | |
| }, | |
| { | |
| "epoch": 0.28431372549019607, | |
| "grad_norm": 0.8410412603652976, | |
| "learning_rate": 1.6190476190476193e-05, | |
| "loss": 0.2094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20205965638160706, | |
| "step": 290, | |
| "valid_targets_mean": 1758.8, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 0.28921568627450983, | |
| "grad_norm": 0.749243931310914, | |
| "learning_rate": 1.647058823529412e-05, | |
| "loss": 0.2169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20068684220314026, | |
| "step": 295, | |
| "valid_targets_mean": 1806.8, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 0.29411764705882354, | |
| "grad_norm": 1.1731365321469984, | |
| "learning_rate": 1.6750700280112046e-05, | |
| "loss": 0.2401, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.3024446964263916, | |
| "step": 300, | |
| "valid_targets_mean": 1756.7, | |
| "valid_targets_min": 458 | |
| }, | |
| { | |
| "epoch": 0.29901960784313725, | |
| "grad_norm": 0.5547867627065093, | |
| "learning_rate": 1.7030812324929973e-05, | |
| "loss": 0.2019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1652289479970932, | |
| "step": 305, | |
| "valid_targets_mean": 2662.3, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 0.30392156862745096, | |
| "grad_norm": 0.6045886405973347, | |
| "learning_rate": 1.7310924369747902e-05, | |
| "loss": 0.1927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14137142896652222, | |
| "step": 310, | |
| "valid_targets_mean": 3014.4, | |
| "valid_targets_min": 952 | |
| }, | |
| { | |
| "epoch": 0.3088235294117647, | |
| "grad_norm": 0.6337281237898196, | |
| "learning_rate": 1.759103641456583e-05, | |
| "loss": 0.1976, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1614706665277481, | |
| "step": 315, | |
| "valid_targets_mean": 2457.1, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 0.3137254901960784, | |
| "grad_norm": 0.7994521931285774, | |
| "learning_rate": 1.7871148459383755e-05, | |
| "loss": 0.2204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.29927748441696167, | |
| "step": 320, | |
| "valid_targets_mean": 2256.4, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 0.31862745098039214, | |
| "grad_norm": 0.6877920338669534, | |
| "learning_rate": 1.8151260504201682e-05, | |
| "loss": 0.2215, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18461309373378754, | |
| "step": 325, | |
| "valid_targets_mean": 2309.9, | |
| "valid_targets_min": 357 | |
| }, | |
| { | |
| "epoch": 0.3235294117647059, | |
| "grad_norm": 0.8324024546914552, | |
| "learning_rate": 1.843137254901961e-05, | |
| "loss": 0.1982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19263777136802673, | |
| "step": 330, | |
| "valid_targets_mean": 1769.6, | |
| "valid_targets_min": 569 | |
| }, | |
| { | |
| "epoch": 0.3284313725490196, | |
| "grad_norm": 0.9991213846200081, | |
| "learning_rate": 1.8711484593837535e-05, | |
| "loss": 0.2142, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2420213520526886, | |
| "step": 335, | |
| "valid_targets_mean": 1401.4, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 0.3333333333333333, | |
| "grad_norm": 0.6238168794305411, | |
| "learning_rate": 1.899159663865546e-05, | |
| "loss": 0.2151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17140337824821472, | |
| "step": 340, | |
| "valid_targets_mean": 2009.0, | |
| "valid_targets_min": 580 | |
| }, | |
| { | |
| "epoch": 0.3382352941176471, | |
| "grad_norm": 0.6728967307166305, | |
| "learning_rate": 1.927170868347339e-05, | |
| "loss": 0.1932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19629782438278198, | |
| "step": 345, | |
| "valid_targets_mean": 2363.2, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 0.3431372549019608, | |
| "grad_norm": 0.6374538764985028, | |
| "learning_rate": 1.9551820728291318e-05, | |
| "loss": 0.1994, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1695806384086609, | |
| "step": 350, | |
| "valid_targets_mean": 2499.8, | |
| "valid_targets_min": 781 | |
| }, | |
| { | |
| "epoch": 0.3480392156862745, | |
| "grad_norm": 0.7086989901290757, | |
| "learning_rate": 1.9831932773109244e-05, | |
| "loss": 0.1895, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22146835923194885, | |
| "step": 355, | |
| "valid_targets_mean": 2031.7, | |
| "valid_targets_min": 779 | |
| }, | |
| { | |
| "epoch": 0.35294117647058826, | |
| "grad_norm": 0.8759913313729994, | |
| "learning_rate": 2.011204481792717e-05, | |
| "loss": 0.2152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2633894681930542, | |
| "step": 360, | |
| "valid_targets_mean": 2092.2, | |
| "valid_targets_min": 815 | |
| }, | |
| { | |
| "epoch": 0.35784313725490197, | |
| "grad_norm": 0.5654576745582779, | |
| "learning_rate": 2.0392156862745097e-05, | |
| "loss": 0.1925, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15165485441684723, | |
| "step": 365, | |
| "valid_targets_mean": 2529.6, | |
| "valid_targets_min": 551 | |
| }, | |
| { | |
| "epoch": 0.3627450980392157, | |
| "grad_norm": 0.6916207113885678, | |
| "learning_rate": 2.0672268907563027e-05, | |
| "loss": 0.2389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19615793228149414, | |
| "step": 370, | |
| "valid_targets_mean": 2093.1, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 0.36764705882352944, | |
| "grad_norm": 0.8691395575013329, | |
| "learning_rate": 2.0952380952380954e-05, | |
| "loss": 0.1845, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2099282443523407, | |
| "step": 375, | |
| "valid_targets_mean": 2353.9, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 0.37254901960784315, | |
| "grad_norm": 0.5336835398911135, | |
| "learning_rate": 2.123249299719888e-05, | |
| "loss": 0.1775, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15454989671707153, | |
| "step": 380, | |
| "valid_targets_mean": 2968.2, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 0.37745098039215685, | |
| "grad_norm": 0.6264965338350452, | |
| "learning_rate": 2.1512605042016807e-05, | |
| "loss": 0.196, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19402390718460083, | |
| "step": 385, | |
| "valid_targets_mean": 2708.5, | |
| "valid_targets_min": 1008 | |
| }, | |
| { | |
| "epoch": 0.38235294117647056, | |
| "grad_norm": 0.6054769755806432, | |
| "learning_rate": 2.1792717086834733e-05, | |
| "loss": 0.1782, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16077178716659546, | |
| "step": 390, | |
| "valid_targets_mean": 2332.9, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 0.3872549019607843, | |
| "grad_norm": 0.6871014526711019, | |
| "learning_rate": 2.2072829131652663e-05, | |
| "loss": 0.2083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19740639626979828, | |
| "step": 395, | |
| "valid_targets_mean": 2283.4, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 0.39215686274509803, | |
| "grad_norm": 0.6946444755001282, | |
| "learning_rate": 2.235294117647059e-05, | |
| "loss": 0.1962, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.177214115858078, | |
| "step": 400, | |
| "valid_targets_mean": 2111.4, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 0.39705882352941174, | |
| "grad_norm": 0.7039019894945975, | |
| "learning_rate": 2.2633053221288516e-05, | |
| "loss": 0.1985, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19564135372638702, | |
| "step": 405, | |
| "valid_targets_mean": 2236.6, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 0.4019607843137255, | |
| "grad_norm": 0.7862930643002248, | |
| "learning_rate": 2.2913165266106443e-05, | |
| "loss": 0.1785, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17840075492858887, | |
| "step": 410, | |
| "valid_targets_mean": 1734.5, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 0.4068627450980392, | |
| "grad_norm": 0.6757610019636033, | |
| "learning_rate": 2.3193277310924373e-05, | |
| "loss": 0.1725, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.145016148686409, | |
| "step": 415, | |
| "valid_targets_mean": 1852.9, | |
| "valid_targets_min": 814 | |
| }, | |
| { | |
| "epoch": 0.4117647058823529, | |
| "grad_norm": 0.5408656516456666, | |
| "learning_rate": 2.34733893557423e-05, | |
| "loss": 0.1746, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1631423830986023, | |
| "step": 420, | |
| "valid_targets_mean": 3106.1, | |
| "valid_targets_min": 1037 | |
| }, | |
| { | |
| "epoch": 0.4166666666666667, | |
| "grad_norm": 0.5558016162889798, | |
| "learning_rate": 2.3753501400560226e-05, | |
| "loss": 0.1949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1663081794977188, | |
| "step": 425, | |
| "valid_targets_mean": 2749.5, | |
| "valid_targets_min": 457 | |
| }, | |
| { | |
| "epoch": 0.4215686274509804, | |
| "grad_norm": 0.5880655917189165, | |
| "learning_rate": 2.4033613445378152e-05, | |
| "loss": 0.1873, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16312886774539948, | |
| "step": 430, | |
| "valid_targets_mean": 2623.3, | |
| "valid_targets_min": 745 | |
| }, | |
| { | |
| "epoch": 0.4264705882352941, | |
| "grad_norm": 0.7344574592722877, | |
| "learning_rate": 2.431372549019608e-05, | |
| "loss": 0.2279, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2649971544742584, | |
| "step": 435, | |
| "valid_targets_mean": 2264.2, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 0.43137254901960786, | |
| "grad_norm": 0.6302873204915442, | |
| "learning_rate": 2.459383753501401e-05, | |
| "loss": 0.1704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1764337718486786, | |
| "step": 440, | |
| "valid_targets_mean": 2357.9, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 0.4362745098039216, | |
| "grad_norm": 0.7622364081121474, | |
| "learning_rate": 2.4873949579831935e-05, | |
| "loss": 0.198, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1960497498512268, | |
| "step": 445, | |
| "valid_targets_mean": 1879.4, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 0.4411764705882353, | |
| "grad_norm": 0.7533017366222997, | |
| "learning_rate": 2.515406162464986e-05, | |
| "loss": 0.1812, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2160189002752304, | |
| "step": 450, | |
| "valid_targets_mean": 1974.1, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 0.44607843137254904, | |
| "grad_norm": 0.634266879787467, | |
| "learning_rate": 2.5434173669467788e-05, | |
| "loss": 0.18, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2114112675189972, | |
| "step": 455, | |
| "valid_targets_mean": 2598.1, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 0.45098039215686275, | |
| "grad_norm": 0.6398501184472496, | |
| "learning_rate": 2.5714285714285718e-05, | |
| "loss": 0.1651, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14820167422294617, | |
| "step": 460, | |
| "valid_targets_mean": 2284.9, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 0.45588235294117646, | |
| "grad_norm": 0.6685908833242333, | |
| "learning_rate": 2.5994397759103644e-05, | |
| "loss": 0.1851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17728768289089203, | |
| "step": 465, | |
| "valid_targets_mean": 2262.2, | |
| "valid_targets_min": 563 | |
| }, | |
| { | |
| "epoch": 0.46078431372549017, | |
| "grad_norm": 0.5843431974090029, | |
| "learning_rate": 2.627450980392157e-05, | |
| "loss": 0.1904, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.181320920586586, | |
| "step": 470, | |
| "valid_targets_mean": 2510.9, | |
| "valid_targets_min": 775 | |
| }, | |
| { | |
| "epoch": 0.46568627450980393, | |
| "grad_norm": 0.4774970152103178, | |
| "learning_rate": 2.6554621848739497e-05, | |
| "loss": 0.1706, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12849435210227966, | |
| "step": 475, | |
| "valid_targets_mean": 3068.2, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 0.47058823529411764, | |
| "grad_norm": 0.5934817185344419, | |
| "learning_rate": 2.6834733893557424e-05, | |
| "loss": 0.1807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1902085244655609, | |
| "step": 480, | |
| "valid_targets_mean": 2842.9, | |
| "valid_targets_min": 961 | |
| }, | |
| { | |
| "epoch": 0.47549019607843135, | |
| "grad_norm": 0.8021530951461361, | |
| "learning_rate": 2.7114845938375354e-05, | |
| "loss": 0.1932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22121350467205048, | |
| "step": 485, | |
| "valid_targets_mean": 1811.6, | |
| "valid_targets_min": 568 | |
| }, | |
| { | |
| "epoch": 0.4803921568627451, | |
| "grad_norm": 0.7090056473413902, | |
| "learning_rate": 2.739495798319328e-05, | |
| "loss": 0.1885, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20266397297382355, | |
| "step": 490, | |
| "valid_targets_mean": 2183.2, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 0.4852941176470588, | |
| "grad_norm": 0.5565085839690194, | |
| "learning_rate": 2.7675070028011207e-05, | |
| "loss": 0.1893, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16087017953395844, | |
| "step": 495, | |
| "valid_targets_mean": 2603.2, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 0.49019607843137253, | |
| "grad_norm": 0.716565785096359, | |
| "learning_rate": 2.7955182072829133e-05, | |
| "loss": 0.1912, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2089020311832428, | |
| "step": 500, | |
| "valid_targets_mean": 2193.8, | |
| "valid_targets_min": 502 | |
| }, | |
| { | |
| "epoch": 0.4950980392156863, | |
| "grad_norm": 0.6524679837642853, | |
| "learning_rate": 2.8235294117647063e-05, | |
| "loss": 0.1897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19976428151130676, | |
| "step": 505, | |
| "valid_targets_mean": 2135.9, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "grad_norm": 0.5937060048137175, | |
| "learning_rate": 2.851540616246499e-05, | |
| "loss": 0.1961, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1664942055940628, | |
| "step": 510, | |
| "valid_targets_mean": 2461.0, | |
| "valid_targets_min": 559 | |
| }, | |
| { | |
| "epoch": 0.5049019607843137, | |
| "grad_norm": 0.7057958054146954, | |
| "learning_rate": 2.8795518207282916e-05, | |
| "loss": 0.2083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19686059653759003, | |
| "step": 515, | |
| "valid_targets_mean": 2053.7, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 0.5098039215686274, | |
| "grad_norm": 0.8351575004625197, | |
| "learning_rate": 2.9075630252100843e-05, | |
| "loss": 0.1949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2226218283176422, | |
| "step": 520, | |
| "valid_targets_mean": 1817.9, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 0.5147058823529411, | |
| "grad_norm": 0.9431458056923483, | |
| "learning_rate": 2.935574229691877e-05, | |
| "loss": 0.1727, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17681443691253662, | |
| "step": 525, | |
| "valid_targets_mean": 2532.9, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 0.5196078431372549, | |
| "grad_norm": 0.6557987732231455, | |
| "learning_rate": 2.96358543417367e-05, | |
| "loss": 0.1787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1878250688314438, | |
| "step": 530, | |
| "valid_targets_mean": 2121.4, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 0.5245098039215687, | |
| "grad_norm": 0.6775473688447424, | |
| "learning_rate": 2.9915966386554626e-05, | |
| "loss": 0.1879, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19594591856002808, | |
| "step": 535, | |
| "valid_targets_mean": 2717.2, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 0.5294117647058824, | |
| "grad_norm": 0.7493448799041711, | |
| "learning_rate": 3.0196078431372552e-05, | |
| "loss": 0.1726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17323686182498932, | |
| "step": 540, | |
| "valid_targets_mean": 2848.9, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 0.5343137254901961, | |
| "grad_norm": 0.6055245402355067, | |
| "learning_rate": 3.047619047619048e-05, | |
| "loss": 0.1914, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18262027204036713, | |
| "step": 545, | |
| "valid_targets_mean": 2384.7, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 0.5392156862745098, | |
| "grad_norm": 0.9951931981576938, | |
| "learning_rate": 3.075630252100841e-05, | |
| "loss": 0.2105, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19094133377075195, | |
| "step": 550, | |
| "valid_targets_mean": 1880.9, | |
| "valid_targets_min": 536 | |
| }, | |
| { | |
| "epoch": 0.5441176470588235, | |
| "grad_norm": 0.5361891619372162, | |
| "learning_rate": 3.1036414565826335e-05, | |
| "loss": 0.1902, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18571898341178894, | |
| "step": 555, | |
| "valid_targets_mean": 2702.5, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 0.5490196078431373, | |
| "grad_norm": 0.7382597364245482, | |
| "learning_rate": 3.131652661064426e-05, | |
| "loss": 0.1831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20009863376617432, | |
| "step": 560, | |
| "valid_targets_mean": 1987.9, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 0.553921568627451, | |
| "grad_norm": 0.49271312001793305, | |
| "learning_rate": 3.159663865546219e-05, | |
| "loss": 0.1723, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13765659928321838, | |
| "step": 565, | |
| "valid_targets_mean": 2502.9, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 0.5588235294117647, | |
| "grad_norm": 0.6602153309510624, | |
| "learning_rate": 3.1876750700280114e-05, | |
| "loss": 0.1843, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17571693658828735, | |
| "step": 570, | |
| "valid_targets_mean": 1902.5, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 0.5637254901960784, | |
| "grad_norm": 0.5846536461117092, | |
| "learning_rate": 3.215686274509804e-05, | |
| "loss": 0.1798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15587720274925232, | |
| "step": 575, | |
| "valid_targets_mean": 2710.6, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 0.5686274509803921, | |
| "grad_norm": 0.614800842745419, | |
| "learning_rate": 3.243697478991597e-05, | |
| "loss": 0.1917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20170342922210693, | |
| "step": 580, | |
| "valid_targets_mean": 2353.9, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 0.5735294117647058, | |
| "grad_norm": 0.7805665554727105, | |
| "learning_rate": 3.2717086834733894e-05, | |
| "loss": 0.189, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21049509942531586, | |
| "step": 585, | |
| "valid_targets_mean": 1816.6, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 0.5784313725490197, | |
| "grad_norm": 0.7649845650415136, | |
| "learning_rate": 3.299719887955182e-05, | |
| "loss": 0.1875, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2008473128080368, | |
| "step": 590, | |
| "valid_targets_mean": 1650.5, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 0.5833333333333334, | |
| "grad_norm": 0.6457466336174561, | |
| "learning_rate": 3.3277310924369754e-05, | |
| "loss": 0.1805, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17701813578605652, | |
| "step": 595, | |
| "valid_targets_mean": 1995.9, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 0.5882352941176471, | |
| "grad_norm": 0.5562439326577053, | |
| "learning_rate": 3.355742296918768e-05, | |
| "loss": 0.1843, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1580512523651123, | |
| "step": 600, | |
| "valid_targets_mean": 2415.7, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 0.5931372549019608, | |
| "grad_norm": 0.56877429322566, | |
| "learning_rate": 3.383753501400561e-05, | |
| "loss": 0.1753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1642981469631195, | |
| "step": 605, | |
| "valid_targets_mean": 2748.0, | |
| "valid_targets_min": 1005 | |
| }, | |
| { | |
| "epoch": 0.5980392156862745, | |
| "grad_norm": 0.548679259946226, | |
| "learning_rate": 3.411764705882353e-05, | |
| "loss": 0.1778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17865321040153503, | |
| "step": 610, | |
| "valid_targets_mean": 2435.0, | |
| "valid_targets_min": 546 | |
| }, | |
| { | |
| "epoch": 0.6029411764705882, | |
| "grad_norm": 0.631089481529247, | |
| "learning_rate": 3.439775910364146e-05, | |
| "loss": 0.1835, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19993621110916138, | |
| "step": 615, | |
| "valid_targets_mean": 2594.4, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 0.6078431372549019, | |
| "grad_norm": 0.5810814829277858, | |
| "learning_rate": 3.4677871148459386e-05, | |
| "loss": 0.1795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1496606171131134, | |
| "step": 620, | |
| "valid_targets_mean": 2420.2, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 0.6127450980392157, | |
| "grad_norm": 0.6657496107127471, | |
| "learning_rate": 3.495798319327731e-05, | |
| "loss": 0.1959, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1836290955543518, | |
| "step": 625, | |
| "valid_targets_mean": 2130.9, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 0.6176470588235294, | |
| "grad_norm": 0.5600295328636472, | |
| "learning_rate": 3.523809523809524e-05, | |
| "loss": 0.1579, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14458300173282623, | |
| "step": 630, | |
| "valid_targets_mean": 2431.5, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 0.6225490196078431, | |
| "grad_norm": 0.49976024934437707, | |
| "learning_rate": 3.5518207282913166e-05, | |
| "loss": 0.1814, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17662914097309113, | |
| "step": 635, | |
| "valid_targets_mean": 3142.9, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 0.6274509803921569, | |
| "grad_norm": 0.6240957673215978, | |
| "learning_rate": 3.57983193277311e-05, | |
| "loss": 0.1766, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20896753668785095, | |
| "step": 640, | |
| "valid_targets_mean": 2600.9, | |
| "valid_targets_min": 1251 | |
| }, | |
| { | |
| "epoch": 0.6323529411764706, | |
| "grad_norm": 0.7125271464926058, | |
| "learning_rate": 3.6078431372549025e-05, | |
| "loss": 0.1722, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17587466537952423, | |
| "step": 645, | |
| "valid_targets_mean": 2187.2, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 0.6372549019607843, | |
| "grad_norm": 0.681219377242786, | |
| "learning_rate": 3.635854341736695e-05, | |
| "loss": 0.1839, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1565675586462021, | |
| "step": 650, | |
| "valid_targets_mean": 1880.1, | |
| "valid_targets_min": 578 | |
| }, | |
| { | |
| "epoch": 0.6421568627450981, | |
| "grad_norm": 0.540781252427908, | |
| "learning_rate": 3.663865546218488e-05, | |
| "loss": 0.1924, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16999506950378418, | |
| "step": 655, | |
| "valid_targets_mean": 2553.5, | |
| "valid_targets_min": 631 | |
| }, | |
| { | |
| "epoch": 0.6470588235294118, | |
| "grad_norm": 0.5599693819997887, | |
| "learning_rate": 3.6918767507002805e-05, | |
| "loss": 0.1795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16493026912212372, | |
| "step": 660, | |
| "valid_targets_mean": 2483.2, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 0.6519607843137255, | |
| "grad_norm": 0.6013502015737507, | |
| "learning_rate": 3.719887955182073e-05, | |
| "loss": 0.1566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13749179244041443, | |
| "step": 665, | |
| "valid_targets_mean": 2236.2, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 0.6568627450980392, | |
| "grad_norm": 0.8440572077136491, | |
| "learning_rate": 3.747899159663866e-05, | |
| "loss": 0.1867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20680469274520874, | |
| "step": 670, | |
| "valid_targets_mean": 1549.4, | |
| "valid_targets_min": 645 | |
| }, | |
| { | |
| "epoch": 0.6617647058823529, | |
| "grad_norm": 0.49979150890191865, | |
| "learning_rate": 3.7759103641456584e-05, | |
| "loss": 0.1651, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1284269243478775, | |
| "step": 675, | |
| "valid_targets_mean": 2789.7, | |
| "valid_targets_min": 920 | |
| }, | |
| { | |
| "epoch": 0.6666666666666666, | |
| "grad_norm": 0.5527405271398915, | |
| "learning_rate": 3.803921568627451e-05, | |
| "loss": 0.1717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16107848286628723, | |
| "step": 680, | |
| "valid_targets_mean": 2213.6, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 0.6715686274509803, | |
| "grad_norm": 0.45853328959246625, | |
| "learning_rate": 3.8319327731092444e-05, | |
| "loss": 0.1474, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14330020546913147, | |
| "step": 685, | |
| "valid_targets_mean": 2835.4, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 0.6764705882352942, | |
| "grad_norm": 0.6033883535487085, | |
| "learning_rate": 3.859943977591037e-05, | |
| "loss": 0.1871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.170333594083786, | |
| "step": 690, | |
| "valid_targets_mean": 2644.1, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 0.6813725490196079, | |
| "grad_norm": 0.48758852132532277, | |
| "learning_rate": 3.88795518207283e-05, | |
| "loss": 0.1645, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12590698897838593, | |
| "step": 695, | |
| "valid_targets_mean": 3071.6, | |
| "valid_targets_min": 545 | |
| }, | |
| { | |
| "epoch": 0.6862745098039216, | |
| "grad_norm": 0.605661287906157, | |
| "learning_rate": 3.915966386554622e-05, | |
| "loss": 0.1915, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17962878942489624, | |
| "step": 700, | |
| "valid_targets_mean": 1955.1, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 0.6911764705882353, | |
| "grad_norm": 0.5020242290872657, | |
| "learning_rate": 3.943977591036415e-05, | |
| "loss": 0.1733, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1520138829946518, | |
| "step": 705, | |
| "valid_targets_mean": 2484.9, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 0.696078431372549, | |
| "grad_norm": 0.6111840324281157, | |
| "learning_rate": 3.971988795518208e-05, | |
| "loss": 0.1591, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15317299962043762, | |
| "step": 710, | |
| "valid_targets_mean": 2317.1, | |
| "valid_targets_min": 834 | |
| }, | |
| { | |
| "epoch": 0.7009803921568627, | |
| "grad_norm": 0.5706697841930912, | |
| "learning_rate": 4e-05, | |
| "loss": 0.1783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1541549265384674, | |
| "step": 715, | |
| "valid_targets_mean": 2519.8, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 0.7058823529411765, | |
| "grad_norm": 0.6242837873790525, | |
| "learning_rate": 3.9999940247222794e-05, | |
| "loss": 0.1756, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17146050930023193, | |
| "step": 720, | |
| "valid_targets_mean": 2097.7, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 0.7107843137254902, | |
| "grad_norm": 0.6320865522891047, | |
| "learning_rate": 3.99997609892482e-05, | |
| "loss": 0.1757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16714760661125183, | |
| "step": 725, | |
| "valid_targets_mean": 2100.1, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 0.7156862745098039, | |
| "grad_norm": 0.6259492059085117, | |
| "learning_rate": 3.9999462227147346e-05, | |
| "loss": 0.1609, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1846911460161209, | |
| "step": 730, | |
| "valid_targets_mean": 2340.9, | |
| "valid_targets_min": 717 | |
| }, | |
| { | |
| "epoch": 0.7205882352941176, | |
| "grad_norm": 0.5649591929759424, | |
| "learning_rate": 3.999904396270541e-05, | |
| "loss": 0.1444, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14328107237815857, | |
| "step": 735, | |
| "valid_targets_mean": 1948.3, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 0.7254901960784313, | |
| "grad_norm": 0.5989473820452854, | |
| "learning_rate": 3.999850619842165e-05, | |
| "loss": 0.1785, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19438403844833374, | |
| "step": 740, | |
| "valid_targets_mean": 2225.5, | |
| "valid_targets_min": 898 | |
| }, | |
| { | |
| "epoch": 0.7303921568627451, | |
| "grad_norm": 0.5877131427477353, | |
| "learning_rate": 3.9997848937509344e-05, | |
| "loss": 0.1702, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17539003491401672, | |
| "step": 745, | |
| "valid_targets_mean": 2403.7, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 0.7352941176470589, | |
| "grad_norm": 0.6133368132701967, | |
| "learning_rate": 3.9997072183895813e-05, | |
| "loss": 0.1873, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17177073657512665, | |
| "step": 750, | |
| "valid_targets_mean": 2129.9, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 0.7401960784313726, | |
| "grad_norm": 0.6243605903426748, | |
| "learning_rate": 3.999617594222238e-05, | |
| "loss": 0.1639, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20242561399936676, | |
| "step": 755, | |
| "valid_targets_mean": 1891.8, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 0.7450980392156863, | |
| "grad_norm": 0.5781945029504527, | |
| "learning_rate": 3.999516021784433e-05, | |
| "loss": 0.2009, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2293742150068283, | |
| "step": 760, | |
| "valid_targets_mean": 2518.5, | |
| "valid_targets_min": 849 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "grad_norm": 0.5246759562973626, | |
| "learning_rate": 3.999402501683091e-05, | |
| "loss": 0.1678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1456177830696106, | |
| "step": 765, | |
| "valid_targets_mean": 2379.1, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 0.7549019607843137, | |
| "grad_norm": 1.3105893523432963, | |
| "learning_rate": 3.9992770345965246e-05, | |
| "loss": 0.1574, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16315758228302002, | |
| "step": 770, | |
| "valid_targets_mean": 2101.8, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 0.7598039215686274, | |
| "grad_norm": 0.48875268576869824, | |
| "learning_rate": 3.999139621274435e-05, | |
| "loss": 0.1653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14056408405303955, | |
| "step": 775, | |
| "valid_targets_mean": 2588.0, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 0.7647058823529411, | |
| "grad_norm": 0.5815354015567381, | |
| "learning_rate": 3.998990262537906e-05, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1769942045211792, | |
| "step": 780, | |
| "valid_targets_mean": 2086.6, | |
| "valid_targets_min": 674 | |
| }, | |
| { | |
| "epoch": 0.7696078431372549, | |
| "grad_norm": 0.5580499155747178, | |
| "learning_rate": 3.998828959279397e-05, | |
| "loss": 0.1846, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16807660460472107, | |
| "step": 785, | |
| "valid_targets_mean": 2295.8, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 0.7745098039215687, | |
| "grad_norm": 0.6543312744336789, | |
| "learning_rate": 3.998655712462739e-05, | |
| "loss": 0.1942, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23026910424232483, | |
| "step": 790, | |
| "valid_targets_mean": 2343.8, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 0.7794117647058824, | |
| "grad_norm": 0.5133434881456423, | |
| "learning_rate": 3.9984705231231305e-05, | |
| "loss": 0.1665, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1757199764251709, | |
| "step": 795, | |
| "valid_targets_mean": 2698.2, | |
| "valid_targets_min": 947 | |
| }, | |
| { | |
| "epoch": 0.7843137254901961, | |
| "grad_norm": 0.4829365196833064, | |
| "learning_rate": 3.9982733923671285e-05, | |
| "loss": 0.1634, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1341038942337036, | |
| "step": 800, | |
| "valid_targets_mean": 2727.9, | |
| "valid_targets_min": 934 | |
| }, | |
| { | |
| "epoch": 0.7892156862745098, | |
| "grad_norm": 0.5478275817278946, | |
| "learning_rate": 3.998064321372645e-05, | |
| "loss": 0.1589, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17312784492969513, | |
| "step": 805, | |
| "valid_targets_mean": 2547.9, | |
| "valid_targets_min": 1030 | |
| }, | |
| { | |
| "epoch": 0.7941176470588235, | |
| "grad_norm": 0.5569452846395607, | |
| "learning_rate": 3.9978433113889374e-05, | |
| "loss": 0.1723, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19361044466495514, | |
| "step": 810, | |
| "valid_targets_mean": 2531.6, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 0.7990196078431373, | |
| "grad_norm": 0.5416020837107107, | |
| "learning_rate": 3.997610363736601e-05, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.156791090965271, | |
| "step": 815, | |
| "valid_targets_mean": 2386.6, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 0.803921568627451, | |
| "grad_norm": 0.5762887091632245, | |
| "learning_rate": 3.997365479807562e-05, | |
| "loss": 0.1728, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19870856404304504, | |
| "step": 820, | |
| "valid_targets_mean": 2393.2, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 0.8088235294117647, | |
| "grad_norm": 0.6286459807846932, | |
| "learning_rate": 3.997108661065072e-05, | |
| "loss": 0.1834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2093677967786789, | |
| "step": 825, | |
| "valid_targets_mean": 2101.6, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 0.8137254901960784, | |
| "grad_norm": 0.4993822078208516, | |
| "learning_rate": 3.996839909043693e-05, | |
| "loss": 0.178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13117878139019012, | |
| "step": 830, | |
| "valid_targets_mean": 2134.5, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 0.8186274509803921, | |
| "grad_norm": 0.6887032555539322, | |
| "learning_rate": 3.996559225349292e-05, | |
| "loss": 0.1667, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18981419503688812, | |
| "step": 835, | |
| "valid_targets_mean": 1885.4, | |
| "valid_targets_min": 767 | |
| }, | |
| { | |
| "epoch": 0.8235294117647058, | |
| "grad_norm": 0.4824669004092395, | |
| "learning_rate": 3.996266611659034e-05, | |
| "loss": 0.204, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1455826759338379, | |
| "step": 840, | |
| "valid_targets_mean": 2453.9, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 0.8284313725490197, | |
| "grad_norm": 0.6045939644368779, | |
| "learning_rate": 3.995962069721367e-05, | |
| "loss": 0.1713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1643049120903015, | |
| "step": 845, | |
| "valid_targets_mean": 1984.4, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 0.8333333333333334, | |
| "grad_norm": 0.6432148110358881, | |
| "learning_rate": 3.995645601356011e-05, | |
| "loss": 0.1677, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1854993999004364, | |
| "step": 850, | |
| "valid_targets_mean": 1761.9, | |
| "valid_targets_min": 526 | |
| }, | |
| { | |
| "epoch": 0.8382352941176471, | |
| "grad_norm": 0.5344791129446362, | |
| "learning_rate": 3.995317208453955e-05, | |
| "loss": 0.1658, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14876899123191833, | |
| "step": 855, | |
| "valid_targets_mean": 2241.1, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 0.8431372549019608, | |
| "grad_norm": 0.6242488703641825, | |
| "learning_rate": 3.994976892977437e-05, | |
| "loss": 0.191, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17894625663757324, | |
| "step": 860, | |
| "valid_targets_mean": 2074.5, | |
| "valid_targets_min": 735 | |
| }, | |
| { | |
| "epoch": 0.8480392156862745, | |
| "grad_norm": 0.7029980773261847, | |
| "learning_rate": 3.994624656959937e-05, | |
| "loss": 0.1824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1853296309709549, | |
| "step": 865, | |
| "valid_targets_mean": 1878.5, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 0.8529411764705882, | |
| "grad_norm": 0.5157657178693743, | |
| "learning_rate": 3.994260502506162e-05, | |
| "loss": 0.1537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1443462073802948, | |
| "step": 870, | |
| "valid_targets_mean": 2428.6, | |
| "valid_targets_min": 471 | |
| }, | |
| { | |
| "epoch": 0.8578431372549019, | |
| "grad_norm": 0.614121702051816, | |
| "learning_rate": 3.9938844317920364e-05, | |
| "loss": 0.1602, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19413134455680847, | |
| "step": 875, | |
| "valid_targets_mean": 1949.9, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 0.8627450980392157, | |
| "grad_norm": 0.584377991874969, | |
| "learning_rate": 3.9934964470646875e-05, | |
| "loss": 0.1795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2134077399969101, | |
| "step": 880, | |
| "valid_targets_mean": 2511.8, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 0.8676470588235294, | |
| "grad_norm": 0.5150088309759572, | |
| "learning_rate": 3.993096550642431e-05, | |
| "loss": 0.17, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13996057212352753, | |
| "step": 885, | |
| "valid_targets_mean": 2704.8, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 0.8725490196078431, | |
| "grad_norm": 0.5550186393534331, | |
| "learning_rate": 3.99268474491476e-05, | |
| "loss": 0.1658, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17755350470542908, | |
| "step": 890, | |
| "valid_targets_mean": 2633.4, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 0.8774509803921569, | |
| "grad_norm": 0.4441972349671847, | |
| "learning_rate": 3.9922610323423285e-05, | |
| "loss": 0.1572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14680537581443787, | |
| "step": 895, | |
| "valid_targets_mean": 3068.7, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 0.8823529411764706, | |
| "grad_norm": 0.5485344567306873, | |
| "learning_rate": 3.991825415456935e-05, | |
| "loss": 0.165, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1684146523475647, | |
| "step": 900, | |
| "valid_targets_mean": 2122.2, | |
| "valid_targets_min": 745 | |
| }, | |
| { | |
| "epoch": 0.8872549019607843, | |
| "grad_norm": 0.4650801011694969, | |
| "learning_rate": 3.991377896861513e-05, | |
| "loss": 0.1642, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15383949875831604, | |
| "step": 905, | |
| "valid_targets_mean": 2715.5, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 0.8921568627450981, | |
| "grad_norm": 0.5320235845602149, | |
| "learning_rate": 3.99091847923011e-05, | |
| "loss": 0.1541, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15561532974243164, | |
| "step": 910, | |
| "valid_targets_mean": 2538.2, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 0.8970588235294118, | |
| "grad_norm": 0.45876070962592214, | |
| "learning_rate": 3.990447165307873e-05, | |
| "loss": 0.1712, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.146046981215477, | |
| "step": 915, | |
| "valid_targets_mean": 2568.5, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 0.9019607843137255, | |
| "grad_norm": 0.5921830084761494, | |
| "learning_rate": 3.989963957911035e-05, | |
| "loss": 0.1888, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17852135002613068, | |
| "step": 920, | |
| "valid_targets_mean": 2035.5, | |
| "valid_targets_min": 784 | |
| }, | |
| { | |
| "epoch": 0.9068627450980392, | |
| "grad_norm": 0.5578416790380244, | |
| "learning_rate": 3.989468859926893e-05, | |
| "loss": 0.1648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15424934029579163, | |
| "step": 925, | |
| "valid_targets_mean": 2060.3, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 0.9117647058823529, | |
| "grad_norm": 0.61427402855988, | |
| "learning_rate": 3.9889618743137954e-05, | |
| "loss": 0.1921, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20084241032600403, | |
| "step": 930, | |
| "valid_targets_mean": 2377.2, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 0.9166666666666666, | |
| "grad_norm": 0.617879885813731, | |
| "learning_rate": 3.988443004101123e-05, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19452977180480957, | |
| "step": 935, | |
| "valid_targets_mean": 1994.8, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 0.9215686274509803, | |
| "grad_norm": 0.5918025094232898, | |
| "learning_rate": 3.9879122523892686e-05, | |
| "loss": 0.1624, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16469812393188477, | |
| "step": 940, | |
| "valid_targets_mean": 1928.2, | |
| "valid_targets_min": 671 | |
| }, | |
| { | |
| "epoch": 0.9264705882352942, | |
| "grad_norm": 0.47030820264667533, | |
| "learning_rate": 3.987369622349621e-05, | |
| "loss": 0.1625, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1485341638326645, | |
| "step": 945, | |
| "valid_targets_mean": 2903.6, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 0.9313725490196079, | |
| "grad_norm": 0.5720177758819267, | |
| "learning_rate": 3.986815117224546e-05, | |
| "loss": 0.1722, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.23535758256912231, | |
| "step": 950, | |
| "valid_targets_mean": 2154.2, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 0.9362745098039216, | |
| "grad_norm": 0.491844040588146, | |
| "learning_rate": 3.986248740327365e-05, | |
| "loss": 0.1639, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14158077538013458, | |
| "step": 955, | |
| "valid_targets_mean": 2549.4, | |
| "valid_targets_min": 787 | |
| }, | |
| { | |
| "epoch": 0.9411764705882353, | |
| "grad_norm": 0.6672714337590078, | |
| "learning_rate": 3.985670495042338e-05, | |
| "loss": 0.1757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19034962356090546, | |
| "step": 960, | |
| "valid_targets_mean": 1762.1, | |
| "valid_targets_min": 513 | |
| }, | |
| { | |
| "epoch": 0.946078431372549, | |
| "grad_norm": 0.5107414628675888, | |
| "learning_rate": 3.98508038482464e-05, | |
| "loss": 0.1749, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16910362243652344, | |
| "step": 965, | |
| "valid_targets_mean": 2209.5, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 0.9509803921568627, | |
| "grad_norm": 0.7159936896599701, | |
| "learning_rate": 3.984478413200345e-05, | |
| "loss": 0.1651, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18525683879852295, | |
| "step": 970, | |
| "valid_targets_mean": 2314.8, | |
| "valid_targets_min": 786 | |
| }, | |
| { | |
| "epoch": 0.9558823529411765, | |
| "grad_norm": 0.5069431684151768, | |
| "learning_rate": 3.983864583766399e-05, | |
| "loss": 0.1678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15477177500724792, | |
| "step": 975, | |
| "valid_targets_mean": 2671.3, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 0.9607843137254902, | |
| "grad_norm": 0.5442502847606222, | |
| "learning_rate": 3.983238900190604e-05, | |
| "loss": 0.1631, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1818421185016632, | |
| "step": 980, | |
| "valid_targets_mean": 2540.2, | |
| "valid_targets_min": 995 | |
| }, | |
| { | |
| "epoch": 0.9656862745098039, | |
| "grad_norm": 0.5846293701433558, | |
| "learning_rate": 3.9826013662115934e-05, | |
| "loss": 0.181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15935450792312622, | |
| "step": 985, | |
| "valid_targets_mean": 2114.6, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 0.9705882352941176, | |
| "grad_norm": 0.7518939257172016, | |
| "learning_rate": 3.981951985638811e-05, | |
| "loss": 0.1599, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16909798979759216, | |
| "step": 990, | |
| "valid_targets_mean": 1665.1, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 0.9754901960784313, | |
| "grad_norm": 0.4994396623985315, | |
| "learning_rate": 3.981290762352483e-05, | |
| "loss": 0.1554, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14983665943145752, | |
| "step": 995, | |
| "valid_targets_mean": 2305.2, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 0.9803921568627451, | |
| "grad_norm": 0.586025245179539, | |
| "learning_rate": 3.9806177003036046e-05, | |
| "loss": 0.1522, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16798822581768036, | |
| "step": 1000, | |
| "valid_targets_mean": 2330.8, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 0.9852941176470589, | |
| "grad_norm": 1.0406909868224459, | |
| "learning_rate": 3.979932803513908e-05, | |
| "loss": 0.1587, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18792645633220673, | |
| "step": 1005, | |
| "valid_targets_mean": 2235.0, | |
| "valid_targets_min": 821 | |
| }, | |
| { | |
| "epoch": 0.9901960784313726, | |
| "grad_norm": 0.6058346652555415, | |
| "learning_rate": 3.979236076075841e-05, | |
| "loss": 0.1556, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17952141165733337, | |
| "step": 1010, | |
| "valid_targets_mean": 1793.4, | |
| "valid_targets_min": 446 | |
| }, | |
| { | |
| "epoch": 0.9950980392156863, | |
| "grad_norm": 0.5394004450746253, | |
| "learning_rate": 3.9785275221525447e-05, | |
| "loss": 0.1584, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1644574999809265, | |
| "step": 1015, | |
| "valid_targets_mean": 1957.1, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 0.5123401147066629, | |
| "learning_rate": 3.977807145977824e-05, | |
| "loss": 0.1803, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17188802361488342, | |
| "step": 1020, | |
| "valid_targets_mean": 2460.4, | |
| "valid_targets_min": 979 | |
| }, | |
| { | |
| "epoch": 1.0049019607843137, | |
| "grad_norm": 0.5838212653448689, | |
| "learning_rate": 3.977074951856128e-05, | |
| "loss": 0.1523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16549797356128693, | |
| "step": 1025, | |
| "valid_targets_mean": 2017.8, | |
| "valid_targets_min": 559 | |
| }, | |
| { | |
| "epoch": 1.0098039215686274, | |
| "grad_norm": 0.6174829915961145, | |
| "learning_rate": 3.976330944162519e-05, | |
| "loss": 0.1659, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20157307386398315, | |
| "step": 1030, | |
| "valid_targets_mean": 2135.1, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 1.0147058823529411, | |
| "grad_norm": 0.4828068003627354, | |
| "learning_rate": 3.975575127342651e-05, | |
| "loss": 0.1523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17932741343975067, | |
| "step": 1035, | |
| "valid_targets_mean": 2793.8, | |
| "valid_targets_min": 960 | |
| }, | |
| { | |
| "epoch": 1.0196078431372548, | |
| "grad_norm": 0.62074325806201, | |
| "learning_rate": 3.974807505912737e-05, | |
| "loss": 0.167, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20777440071105957, | |
| "step": 1040, | |
| "valid_targets_mean": 2069.4, | |
| "valid_targets_min": 884 | |
| }, | |
| { | |
| "epoch": 1.0245098039215685, | |
| "grad_norm": 0.5157537624193805, | |
| "learning_rate": 3.974028084459531e-05, | |
| "loss": 0.1429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1442909985780716, | |
| "step": 1045, | |
| "valid_targets_mean": 2144.7, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 1.0294117647058822, | |
| "grad_norm": 0.5275408083767269, | |
| "learning_rate": 3.973236867640291e-05, | |
| "loss": 0.144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1591978371143341, | |
| "step": 1050, | |
| "valid_targets_mean": 2433.8, | |
| "valid_targets_min": 540 | |
| }, | |
| { | |
| "epoch": 1.0343137254901962, | |
| "grad_norm": 0.6054568333033143, | |
| "learning_rate": 3.972433860182757e-05, | |
| "loss": 0.1605, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1705915629863739, | |
| "step": 1055, | |
| "valid_targets_mean": 1799.0, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 1.0392156862745099, | |
| "grad_norm": 0.6782539679662744, | |
| "learning_rate": 3.971619066885122e-05, | |
| "loss": 0.1687, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16509351134300232, | |
| "step": 1060, | |
| "valid_targets_mean": 1800.1, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 1.0441176470588236, | |
| "grad_norm": 0.49281399947041626, | |
| "learning_rate": 3.9707924926160026e-05, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14969328045845032, | |
| "step": 1065, | |
| "valid_targets_mean": 2801.4, | |
| "valid_targets_min": 1152 | |
| }, | |
| { | |
| "epoch": 1.0490196078431373, | |
| "grad_norm": 0.588840531018445, | |
| "learning_rate": 3.96995414231441e-05, | |
| "loss": 0.1627, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1991569846868515, | |
| "step": 1070, | |
| "valid_targets_mean": 2264.0, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 1.053921568627451, | |
| "grad_norm": 0.6841222391097739, | |
| "learning_rate": 3.969104020989718e-05, | |
| "loss": 0.1672, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.21732574701309204, | |
| "step": 1075, | |
| "valid_targets_mean": 1794.9, | |
| "valid_targets_min": 590 | |
| }, | |
| { | |
| "epoch": 1.0588235294117647, | |
| "grad_norm": 0.515045459549045, | |
| "learning_rate": 3.96824213372164e-05, | |
| "loss": 0.1467, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1259552240371704, | |
| "step": 1080, | |
| "valid_targets_mean": 2122.6, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 1.0637254901960784, | |
| "grad_norm": 0.5616565763943626, | |
| "learning_rate": 3.9673684856601915e-05, | |
| "loss": 0.1446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14568236470222473, | |
| "step": 1085, | |
| "valid_targets_mean": 1993.7, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 1.0686274509803921, | |
| "grad_norm": 0.5261562423211233, | |
| "learning_rate": 3.9664830820256605e-05, | |
| "loss": 0.1587, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1599075198173523, | |
| "step": 1090, | |
| "valid_targets_mean": 2128.1, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 1.0735294117647058, | |
| "grad_norm": 0.43934072499401866, | |
| "learning_rate": 3.965585928108581e-05, | |
| "loss": 0.1519, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11759766191244125, | |
| "step": 1095, | |
| "valid_targets_mean": 2754.6, | |
| "valid_targets_min": 839 | |
| }, | |
| { | |
| "epoch": 1.0784313725490196, | |
| "grad_norm": 0.6033318921088762, | |
| "learning_rate": 3.964677029269697e-05, | |
| "loss": 0.1679, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13802024722099304, | |
| "step": 1100, | |
| "valid_targets_mean": 2244.1, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 1.0833333333333333, | |
| "grad_norm": 0.5046328382489379, | |
| "learning_rate": 3.963756390939931e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14507150650024414, | |
| "step": 1105, | |
| "valid_targets_mean": 2527.9, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 1.088235294117647, | |
| "grad_norm": 0.5022528514006354, | |
| "learning_rate": 3.962824018620353e-05, | |
| "loss": 0.1536, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.158194899559021, | |
| "step": 1110, | |
| "valid_targets_mean": 2257.4, | |
| "valid_targets_min": 786 | |
| }, | |
| { | |
| "epoch": 1.093137254901961, | |
| "grad_norm": 0.4675384543996388, | |
| "learning_rate": 3.9618799178821454e-05, | |
| "loss": 0.1572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16208568215370178, | |
| "step": 1115, | |
| "valid_targets_mean": 2411.1, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 1.0980392156862746, | |
| "grad_norm": 0.6462739777449577, | |
| "learning_rate": 3.960924094366574e-05, | |
| "loss": 0.1537, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17548783123493195, | |
| "step": 1120, | |
| "valid_targets_mean": 1877.2, | |
| "valid_targets_min": 559 | |
| }, | |
| { | |
| "epoch": 1.1029411764705883, | |
| "grad_norm": 0.5646237353311311, | |
| "learning_rate": 3.959956553784948e-05, | |
| "loss": 0.1339, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15153411030769348, | |
| "step": 1125, | |
| "valid_targets_mean": 2764.0, | |
| "valid_targets_min": 919 | |
| }, | |
| { | |
| "epoch": 1.107843137254902, | |
| "grad_norm": 0.4540975480652281, | |
| "learning_rate": 3.9589773019185924e-05, | |
| "loss": 0.1381, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12593434751033783, | |
| "step": 1130, | |
| "valid_targets_mean": 2523.4, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 1.1127450980392157, | |
| "grad_norm": 0.5547544466967875, | |
| "learning_rate": 3.957986344618809e-05, | |
| "loss": 0.1448, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14328861236572266, | |
| "step": 1135, | |
| "valid_targets_mean": 2087.6, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 1.1176470588235294, | |
| "grad_norm": 0.592476178518575, | |
| "learning_rate": 3.956983687806843e-05, | |
| "loss": 0.1607, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1588444709777832, | |
| "step": 1140, | |
| "valid_targets_mean": 1868.7, | |
| "valid_targets_min": 560 | |
| }, | |
| { | |
| "epoch": 1.1225490196078431, | |
| "grad_norm": 0.5970021551092038, | |
| "learning_rate": 3.955969337473846e-05, | |
| "loss": 0.1587, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15817104279994965, | |
| "step": 1145, | |
| "valid_targets_mean": 2292.9, | |
| "valid_targets_min": 764 | |
| }, | |
| { | |
| "epoch": 1.1274509803921569, | |
| "grad_norm": 0.5962854116508299, | |
| "learning_rate": 3.954943299680844e-05, | |
| "loss": 0.1578, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1732129156589508, | |
| "step": 1150, | |
| "valid_targets_mean": 2063.4, | |
| "valid_targets_min": 1125 | |
| }, | |
| { | |
| "epoch": 1.1323529411764706, | |
| "grad_norm": 0.6896713797977306, | |
| "learning_rate": 3.953905580558698e-05, | |
| "loss": 0.1539, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2036602646112442, | |
| "step": 1155, | |
| "valid_targets_mean": 1861.2, | |
| "valid_targets_min": 609 | |
| }, | |
| { | |
| "epoch": 1.1372549019607843, | |
| "grad_norm": 0.5073490806892655, | |
| "learning_rate": 3.952856186308068e-05, | |
| "loss": 0.1453, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15630340576171875, | |
| "step": 1160, | |
| "valid_targets_mean": 2235.9, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 1.142156862745098, | |
| "grad_norm": 0.45334538233451205, | |
| "learning_rate": 3.951795123199375e-05, | |
| "loss": 0.152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1358206868171692, | |
| "step": 1165, | |
| "valid_targets_mean": 2691.2, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 1.1470588235294117, | |
| "grad_norm": 0.4652006849941343, | |
| "learning_rate": 3.950722397572766e-05, | |
| "loss": 0.1572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14758041501045227, | |
| "step": 1170, | |
| "valid_targets_mean": 2904.9, | |
| "valid_targets_min": 1057 | |
| }, | |
| { | |
| "epoch": 1.1519607843137254, | |
| "grad_norm": 0.562043728061584, | |
| "learning_rate": 3.949638015838076e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1432192176580429, | |
| "step": 1175, | |
| "valid_targets_mean": 1881.5, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 1.156862745098039, | |
| "grad_norm": 0.7702082160549727, | |
| "learning_rate": 3.948541984474784e-05, | |
| "loss": 0.1439, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15741848945617676, | |
| "step": 1180, | |
| "valid_targets_mean": 2464.0, | |
| "valid_targets_min": 827 | |
| }, | |
| { | |
| "epoch": 1.161764705882353, | |
| "grad_norm": 0.5224013259456193, | |
| "learning_rate": 3.947434310031986e-05, | |
| "loss": 0.162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14922982454299927, | |
| "step": 1185, | |
| "valid_targets_mean": 2077.6, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 1.1666666666666667, | |
| "grad_norm": 0.5631555063513266, | |
| "learning_rate": 3.94631499912834e-05, | |
| "loss": 0.1566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15543070435523987, | |
| "step": 1190, | |
| "valid_targets_mean": 2052.9, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 1.1715686274509804, | |
| "grad_norm": 0.45985075769888245, | |
| "learning_rate": 3.945184058452044e-05, | |
| "loss": 0.1396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15431955456733704, | |
| "step": 1195, | |
| "valid_targets_mean": 2501.5, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 1.1764705882352942, | |
| "grad_norm": 0.544135678064552, | |
| "learning_rate": 3.944041494760779e-05, | |
| "loss": 0.1571, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14121049642562866, | |
| "step": 1200, | |
| "valid_targets_mean": 2084.9, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 1.1813725490196079, | |
| "grad_norm": 0.43080987042558744, | |
| "learning_rate": 3.9428873148816815e-05, | |
| "loss": 0.1577, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12293718755245209, | |
| "step": 1205, | |
| "valid_targets_mean": 2903.8, | |
| "valid_targets_min": 638 | |
| }, | |
| { | |
| "epoch": 1.1862745098039216, | |
| "grad_norm": 0.5346019417820428, | |
| "learning_rate": 3.9417215257112975e-05, | |
| "loss": 0.1427, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13670134544372559, | |
| "step": 1210, | |
| "valid_targets_mean": 2017.1, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 1.1911764705882353, | |
| "grad_norm": 0.5740555345382163, | |
| "learning_rate": 3.94054413421554e-05, | |
| "loss": 0.1467, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16550078988075256, | |
| "step": 1215, | |
| "valid_targets_mean": 2116.8, | |
| "valid_targets_min": 874 | |
| }, | |
| { | |
| "epoch": 1.196078431372549, | |
| "grad_norm": 0.5276636605184202, | |
| "learning_rate": 3.9393551474296506e-05, | |
| "loss": 0.1591, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13549819588661194, | |
| "step": 1220, | |
| "valid_targets_mean": 2193.8, | |
| "valid_targets_min": 1090 | |
| }, | |
| { | |
| "epoch": 1.2009803921568627, | |
| "grad_norm": 0.4329687250201491, | |
| "learning_rate": 3.938154572458156e-05, | |
| "loss": 0.143, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12151844054460526, | |
| "step": 1225, | |
| "valid_targets_mean": 2738.6, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 1.2058823529411764, | |
| "grad_norm": 0.7280188503712324, | |
| "learning_rate": 3.936942416474825e-05, | |
| "loss": 0.1681, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20381173491477966, | |
| "step": 1230, | |
| "valid_targets_mean": 2216.9, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 1.2107843137254901, | |
| "grad_norm": 0.6188709430418583, | |
| "learning_rate": 3.935718686722626e-05, | |
| "loss": 0.1721, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18018341064453125, | |
| "step": 1235, | |
| "valid_targets_mean": 1980.0, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 1.215686274509804, | |
| "grad_norm": 0.49994332511469786, | |
| "learning_rate": 3.934483390513683e-05, | |
| "loss": 0.155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13865551352500916, | |
| "step": 1240, | |
| "valid_targets_mean": 2634.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 1.2205882352941178, | |
| "grad_norm": 0.4609019249907111, | |
| "learning_rate": 3.933236535229236e-05, | |
| "loss": 0.1394, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13344664871692657, | |
| "step": 1245, | |
| "valid_targets_mean": 2502.8, | |
| "valid_targets_min": 767 | |
| }, | |
| { | |
| "epoch": 1.2254901960784315, | |
| "grad_norm": 0.6946779226729973, | |
| "learning_rate": 3.931978128319591e-05, | |
| "loss": 0.1616, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18459130823612213, | |
| "step": 1250, | |
| "valid_targets_mean": 2516.1, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 1.2303921568627452, | |
| "grad_norm": 0.3847744756653841, | |
| "learning_rate": 3.9307081773040774e-05, | |
| "loss": 0.1499, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11731822788715363, | |
| "step": 1255, | |
| "valid_targets_mean": 3076.5, | |
| "valid_targets_min": 802 | |
| }, | |
| { | |
| "epoch": 1.2352941176470589, | |
| "grad_norm": 0.4905255095526249, | |
| "learning_rate": 3.929426689771007e-05, | |
| "loss": 0.1476, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14215940237045288, | |
| "step": 1260, | |
| "valid_targets_mean": 2219.2, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 1.2401960784313726, | |
| "grad_norm": 0.5140168836960785, | |
| "learning_rate": 3.9281336733776224e-05, | |
| "loss": 0.1549, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1484440267086029, | |
| "step": 1265, | |
| "valid_targets_mean": 2392.3, | |
| "valid_targets_min": 833 | |
| }, | |
| { | |
| "epoch": 1.2450980392156863, | |
| "grad_norm": 0.510281749270732, | |
| "learning_rate": 3.926829135850056e-05, | |
| "loss": 0.1441, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13769188523292542, | |
| "step": 1270, | |
| "valid_targets_mean": 2149.2, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "grad_norm": 0.47694160804760755, | |
| "learning_rate": 3.925513084983282e-05, | |
| "loss": 0.1412, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1289670467376709, | |
| "step": 1275, | |
| "valid_targets_mean": 2160.6, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 1.2549019607843137, | |
| "grad_norm": 0.5769961746495987, | |
| "learning_rate": 3.924185528641071e-05, | |
| "loss": 0.1458, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16923269629478455, | |
| "step": 1280, | |
| "valid_targets_mean": 1923.4, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 1.2598039215686274, | |
| "grad_norm": 0.5097251595214417, | |
| "learning_rate": 3.9228464747559384e-05, | |
| "loss": 0.1535, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14006245136260986, | |
| "step": 1285, | |
| "valid_targets_mean": 2251.9, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 1.2647058823529411, | |
| "grad_norm": 0.48366970496366835, | |
| "learning_rate": 3.921495931329105e-05, | |
| "loss": 0.1635, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13093242049217224, | |
| "step": 1290, | |
| "valid_targets_mean": 2366.6, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 1.2696078431372548, | |
| "grad_norm": 0.5781181664431445, | |
| "learning_rate": 3.920133906430442e-05, | |
| "loss": 0.1559, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1567743420600891, | |
| "step": 1295, | |
| "valid_targets_mean": 2257.6, | |
| "valid_targets_min": 745 | |
| }, | |
| { | |
| "epoch": 1.2745098039215685, | |
| "grad_norm": 0.41978393338444187, | |
| "learning_rate": 3.918760408198426e-05, | |
| "loss": 0.1392, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14441536366939545, | |
| "step": 1300, | |
| "valid_targets_mean": 2817.1, | |
| "valid_targets_min": 1095 | |
| }, | |
| { | |
| "epoch": 1.2794117647058822, | |
| "grad_norm": 0.5308275909805295, | |
| "learning_rate": 3.9173754448400914e-05, | |
| "loss": 0.1674, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16451840102672577, | |
| "step": 1305, | |
| "valid_targets_mean": 2491.2, | |
| "valid_targets_min": 541 | |
| }, | |
| { | |
| "epoch": 1.284313725490196, | |
| "grad_norm": 0.5311164209246302, | |
| "learning_rate": 3.915979024630978e-05, | |
| "loss": 0.1588, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14257512986660004, | |
| "step": 1310, | |
| "valid_targets_mean": 2018.1, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 1.2892156862745099, | |
| "grad_norm": 0.4234229529058526, | |
| "learning_rate": 3.9145711559150854e-05, | |
| "loss": 0.1497, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1336820423603058, | |
| "step": 1315, | |
| "valid_targets_mean": 2803.6, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 1.2941176470588236, | |
| "grad_norm": 0.5124680921691537, | |
| "learning_rate": 3.9131518471048194e-05, | |
| "loss": 0.1595, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16409265995025635, | |
| "step": 1320, | |
| "valid_targets_mean": 2254.1, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 1.2990196078431373, | |
| "grad_norm": 0.5398508535353193, | |
| "learning_rate": 3.911721106680944e-05, | |
| "loss": 0.1489, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16770967841148376, | |
| "step": 1325, | |
| "valid_targets_mean": 2107.1, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 1.303921568627451, | |
| "grad_norm": 0.47066254116435274, | |
| "learning_rate": 3.910278943192531e-05, | |
| "loss": 0.1592, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1523783802986145, | |
| "step": 1330, | |
| "valid_targets_mean": 2441.0, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 1.3088235294117647, | |
| "grad_norm": 0.5302196796070335, | |
| "learning_rate": 3.9088253652569085e-05, | |
| "loss": 0.1489, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1501566767692566, | |
| "step": 1335, | |
| "valid_targets_mean": 2072.9, | |
| "valid_targets_min": 709 | |
| }, | |
| { | |
| "epoch": 1.3137254901960784, | |
| "grad_norm": 0.6412977475076532, | |
| "learning_rate": 3.907360381559608e-05, | |
| "loss": 0.1649, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14073000848293304, | |
| "step": 1340, | |
| "valid_targets_mean": 2276.2, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 1.3186274509803921, | |
| "grad_norm": 0.5172240165827114, | |
| "learning_rate": 3.9058840008543136e-05, | |
| "loss": 0.1505, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14915066957473755, | |
| "step": 1345, | |
| "valid_targets_mean": 2422.1, | |
| "valid_targets_min": 244 | |
| }, | |
| { | |
| "epoch": 1.3235294117647058, | |
| "grad_norm": 0.4079599036566721, | |
| "learning_rate": 3.9043962319628096e-05, | |
| "loss": 0.1512, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11990423500537872, | |
| "step": 1350, | |
| "valid_targets_mean": 2620.1, | |
| "valid_targets_min": 514 | |
| }, | |
| { | |
| "epoch": 1.3284313725490196, | |
| "grad_norm": 0.48696050356601855, | |
| "learning_rate": 3.902897083774929e-05, | |
| "loss": 0.15, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13719043135643005, | |
| "step": 1355, | |
| "valid_targets_mean": 2365.3, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 1.3333333333333333, | |
| "grad_norm": 0.5203167774413222, | |
| "learning_rate": 3.9013865652484984e-05, | |
| "loss": 0.1517, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15943309664726257, | |
| "step": 1360, | |
| "valid_targets_mean": 2323.3, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 1.3382352941176472, | |
| "grad_norm": 0.549425828566385, | |
| "learning_rate": 3.8998646854092854e-05, | |
| "loss": 0.1535, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15190866589546204, | |
| "step": 1365, | |
| "valid_targets_mean": 2337.4, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 1.343137254901961, | |
| "grad_norm": 0.600237727151271, | |
| "learning_rate": 3.898331453350944e-05, | |
| "loss": 0.1609, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20173123478889465, | |
| "step": 1370, | |
| "valid_targets_mean": 1845.9, | |
| "valid_targets_min": 520 | |
| }, | |
| { | |
| "epoch": 1.3480392156862746, | |
| "grad_norm": 0.5032072364639573, | |
| "learning_rate": 3.896786878234963e-05, | |
| "loss": 0.1488, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15287765860557556, | |
| "step": 1375, | |
| "valid_targets_mean": 2130.8, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 1.3529411764705883, | |
| "grad_norm": 0.6833370573017062, | |
| "learning_rate": 3.8952309692906074e-05, | |
| "loss": 0.1595, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18234968185424805, | |
| "step": 1380, | |
| "valid_targets_mean": 1924.2, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 1.357843137254902, | |
| "grad_norm": 0.41478093075036826, | |
| "learning_rate": 3.893663735814865e-05, | |
| "loss": 0.1559, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13855645060539246, | |
| "step": 1385, | |
| "valid_targets_mean": 2631.2, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 1.3627450980392157, | |
| "grad_norm": 0.5585388314174717, | |
| "learning_rate": 3.89208518717239e-05, | |
| "loss": 0.1616, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1809864044189453, | |
| "step": 1390, | |
| "valid_targets_mean": 1917.9, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 1.3676470588235294, | |
| "grad_norm": 0.4525025154823127, | |
| "learning_rate": 3.89049533279545e-05, | |
| "loss": 0.1508, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14793044328689575, | |
| "step": 1395, | |
| "valid_targets_mean": 2488.4, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 1.3725490196078431, | |
| "grad_norm": 0.4840178712448558, | |
| "learning_rate": 3.888894182183866e-05, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1626589596271515, | |
| "step": 1400, | |
| "valid_targets_mean": 2406.2, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 1.3774509803921569, | |
| "grad_norm": 0.47327404010363494, | |
| "learning_rate": 3.887281744904959e-05, | |
| "loss": 0.1531, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13927535712718964, | |
| "step": 1405, | |
| "valid_targets_mean": 2618.0, | |
| "valid_targets_min": 1047 | |
| }, | |
| { | |
| "epoch": 1.3823529411764706, | |
| "grad_norm": 0.5818804209128727, | |
| "learning_rate": 3.885658030593487e-05, | |
| "loss": 0.1415, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12791548669338226, | |
| "step": 1410, | |
| "valid_targets_mean": 2182.4, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 1.3872549019607843, | |
| "grad_norm": 0.5149356554417958, | |
| "learning_rate": 3.884023048951597e-05, | |
| "loss": 0.1585, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17590127885341644, | |
| "step": 1415, | |
| "valid_targets_mean": 2380.0, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 1.392156862745098, | |
| "grad_norm": 0.4027707139908952, | |
| "learning_rate": 3.8823768097487555e-05, | |
| "loss": 0.1473, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11872929334640503, | |
| "step": 1420, | |
| "valid_targets_mean": 2956.5, | |
| "valid_targets_min": 1389 | |
| }, | |
| { | |
| "epoch": 1.3970588235294117, | |
| "grad_norm": 0.5121711907073138, | |
| "learning_rate": 3.8807193228217006e-05, | |
| "loss": 0.1572, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15116044878959656, | |
| "step": 1425, | |
| "valid_targets_mean": 2139.8, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 1.4019607843137254, | |
| "grad_norm": 0.4615092163821721, | |
| "learning_rate": 3.879050598074377e-05, | |
| "loss": 0.1539, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16094964742660522, | |
| "step": 1430, | |
| "valid_targets_mean": 2579.1, | |
| "valid_targets_min": 944 | |
| }, | |
| { | |
| "epoch": 1.406862745098039, | |
| "grad_norm": 0.4837425625903746, | |
| "learning_rate": 3.877370645477878e-05, | |
| "loss": 0.138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1381511688232422, | |
| "step": 1435, | |
| "valid_targets_mean": 2226.8, | |
| "valid_targets_min": 363 | |
| }, | |
| { | |
| "epoch": 1.4117647058823528, | |
| "grad_norm": 0.49885604510183423, | |
| "learning_rate": 3.875679475070386e-05, | |
| "loss": 0.1446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13402841985225677, | |
| "step": 1440, | |
| "valid_targets_mean": 1817.2, | |
| "valid_targets_min": 357 | |
| }, | |
| { | |
| "epoch": 1.4166666666666667, | |
| "grad_norm": 0.5948084146879866, | |
| "learning_rate": 3.873977096957115e-05, | |
| "loss": 0.1562, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15349772572517395, | |
| "step": 1445, | |
| "valid_targets_mean": 2068.1, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 1.4215686274509804, | |
| "grad_norm": 0.43546114760012355, | |
| "learning_rate": 3.872263521310248e-05, | |
| "loss": 0.1581, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1208997517824173, | |
| "step": 1450, | |
| "valid_targets_mean": 2656.3, | |
| "valid_targets_min": 795 | |
| }, | |
| { | |
| "epoch": 1.4264705882352942, | |
| "grad_norm": 0.41007339804542614, | |
| "learning_rate": 3.870538758368874e-05, | |
| "loss": 0.1511, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12087947130203247, | |
| "step": 1455, | |
| "valid_targets_mean": 2850.0, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 1.4313725490196079, | |
| "grad_norm": 0.5477355070675948, | |
| "learning_rate": 3.868802818438931e-05, | |
| "loss": 0.1443, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14966945350170135, | |
| "step": 1460, | |
| "valid_targets_mean": 2159.5, | |
| "valid_targets_min": 887 | |
| }, | |
| { | |
| "epoch": 1.4362745098039216, | |
| "grad_norm": 0.5084422208686458, | |
| "learning_rate": 3.8670557118931406e-05, | |
| "loss": 0.1523, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15912967920303345, | |
| "step": 1465, | |
| "valid_targets_mean": 2255.6, | |
| "valid_targets_min": 890 | |
| }, | |
| { | |
| "epoch": 1.4411764705882353, | |
| "grad_norm": 0.5106364860275668, | |
| "learning_rate": 3.8652974491709524e-05, | |
| "loss": 0.1634, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1488266885280609, | |
| "step": 1470, | |
| "valid_targets_mean": 2155.7, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 1.446078431372549, | |
| "grad_norm": 0.5123306890644707, | |
| "learning_rate": 3.863528040778473e-05, | |
| "loss": 0.1528, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16354094445705414, | |
| "step": 1475, | |
| "valid_targets_mean": 2284.4, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 1.4509803921568627, | |
| "grad_norm": 0.59057508765795, | |
| "learning_rate": 3.861747497288409e-05, | |
| "loss": 0.1648, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1789345145225525, | |
| "step": 1480, | |
| "valid_targets_mean": 2036.8, | |
| "valid_targets_min": 890 | |
| }, | |
| { | |
| "epoch": 1.4558823529411764, | |
| "grad_norm": 0.4268181984593978, | |
| "learning_rate": 3.859955829340002e-05, | |
| "loss": 0.1412, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13590224087238312, | |
| "step": 1485, | |
| "valid_targets_mean": 2580.3, | |
| "valid_targets_min": 523 | |
| }, | |
| { | |
| "epoch": 1.4607843137254901, | |
| "grad_norm": 0.5925211684186468, | |
| "learning_rate": 3.8581530476389665e-05, | |
| "loss": 0.1436, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1535499542951584, | |
| "step": 1490, | |
| "valid_targets_mean": 2115.8, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 1.465686274509804, | |
| "grad_norm": 0.4175992000283523, | |
| "learning_rate": 3.8563391629574226e-05, | |
| "loss": 0.1494, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13089969754219055, | |
| "step": 1495, | |
| "valid_targets_mean": 2766.2, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 1.4705882352941178, | |
| "grad_norm": 0.4916655913807976, | |
| "learning_rate": 3.8545141861338355e-05, | |
| "loss": 0.1484, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15090090036392212, | |
| "step": 1500, | |
| "valid_targets_mean": 2067.1, | |
| "valid_targets_min": 309 | |
| }, | |
| { | |
| "epoch": 1.4754901960784315, | |
| "grad_norm": 0.5271031406287625, | |
| "learning_rate": 3.8526781280729494e-05, | |
| "loss": 0.1613, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1575801819562912, | |
| "step": 1505, | |
| "valid_targets_mean": 2504.5, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 1.4803921568627452, | |
| "grad_norm": 0.4701289225402125, | |
| "learning_rate": 3.85083099974572e-05, | |
| "loss": 0.1608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15561668574810028, | |
| "step": 1510, | |
| "valid_targets_mean": 2856.8, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 1.4852941176470589, | |
| "grad_norm": 0.5494676935044653, | |
| "learning_rate": 3.848972812189253e-05, | |
| "loss": 0.151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18085089325904846, | |
| "step": 1515, | |
| "valid_targets_mean": 2039.9, | |
| "valid_targets_min": 626 | |
| }, | |
| { | |
| "epoch": 1.4901960784313726, | |
| "grad_norm": 0.6263018300453485, | |
| "learning_rate": 3.847103576506734e-05, | |
| "loss": 0.1491, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15949448943138123, | |
| "step": 1520, | |
| "valid_targets_mean": 1484.6, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 1.4950980392156863, | |
| "grad_norm": 0.46999594845254383, | |
| "learning_rate": 3.845223303867366e-05, | |
| "loss": 0.1656, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13206514716148376, | |
| "step": 1525, | |
| "valid_targets_mean": 2437.2, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "grad_norm": 0.5255301328952996, | |
| "learning_rate": 3.8433320055063e-05, | |
| "loss": 0.1568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15027153491973877, | |
| "step": 1530, | |
| "valid_targets_mean": 1982.6, | |
| "valid_targets_min": 707 | |
| }, | |
| { | |
| "epoch": 1.5049019607843137, | |
| "grad_norm": 0.5347904846455934, | |
| "learning_rate": 3.84142969272457e-05, | |
| "loss": 0.1635, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19537004828453064, | |
| "step": 1535, | |
| "valid_targets_mean": 2316.7, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 1.5098039215686274, | |
| "grad_norm": 0.5646663834321809, | |
| "learning_rate": 3.8395163768890214e-05, | |
| "loss": 0.1566, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17573268711566925, | |
| "step": 1540, | |
| "valid_targets_mean": 1757.5, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 1.5147058823529411, | |
| "grad_norm": 0.5658551765278447, | |
| "learning_rate": 3.837592069432248e-05, | |
| "loss": 0.15, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15337808430194855, | |
| "step": 1545, | |
| "valid_targets_mean": 2001.4, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 1.5196078431372548, | |
| "grad_norm": 0.38422117950519047, | |
| "learning_rate": 3.8356567818525235e-05, | |
| "loss": 0.148, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11618193984031677, | |
| "step": 1550, | |
| "valid_targets_mean": 2760.7, | |
| "valid_targets_min": 853 | |
| }, | |
| { | |
| "epoch": 1.5245098039215685, | |
| "grad_norm": 0.5217212912328395, | |
| "learning_rate": 3.8337105257137264e-05, | |
| "loss": 0.1568, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15121859312057495, | |
| "step": 1555, | |
| "valid_targets_mean": 2225.5, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 1.5294117647058822, | |
| "grad_norm": 0.5285773256463285, | |
| "learning_rate": 3.8317533126452773e-05, | |
| "loss": 0.1389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15342125296592712, | |
| "step": 1560, | |
| "valid_targets_mean": 2033.9, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 1.534313725490196, | |
| "grad_norm": 0.45981631765341807, | |
| "learning_rate": 3.829785154342069e-05, | |
| "loss": 0.1521, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13747870922088623, | |
| "step": 1565, | |
| "valid_targets_mean": 2533.3, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 1.5392156862745097, | |
| "grad_norm": 0.4170627519498316, | |
| "learning_rate": 3.8278060625643945e-05, | |
| "loss": 0.1408, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1535811424255371, | |
| "step": 1570, | |
| "valid_targets_mean": 2868.6, | |
| "valid_targets_min": 943 | |
| }, | |
| { | |
| "epoch": 1.5441176470588234, | |
| "grad_norm": 0.3896454803749676, | |
| "learning_rate": 3.825816049137876e-05, | |
| "loss": 0.1478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11401885747909546, | |
| "step": 1575, | |
| "valid_targets_mean": 2787.1, | |
| "valid_targets_min": 817 | |
| }, | |
| { | |
| "epoch": 1.5490196078431373, | |
| "grad_norm": 0.5002979894193942, | |
| "learning_rate": 3.823815125953396e-05, | |
| "loss": 0.1451, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1406538188457489, | |
| "step": 1580, | |
| "valid_targets_mean": 2422.6, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 1.553921568627451, | |
| "grad_norm": 0.56850834145191, | |
| "learning_rate": 3.821803304967027e-05, | |
| "loss": 0.1425, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1491360068321228, | |
| "step": 1585, | |
| "valid_targets_mean": 1996.2, | |
| "valid_targets_min": 527 | |
| }, | |
| { | |
| "epoch": 1.5588235294117647, | |
| "grad_norm": 0.4998315996038415, | |
| "learning_rate": 3.819780598199958e-05, | |
| "loss": 0.1574, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15458178520202637, | |
| "step": 1590, | |
| "valid_targets_mean": 2790.4, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 1.5637254901960784, | |
| "grad_norm": 0.6662479740719508, | |
| "learning_rate": 3.8177470177384227e-05, | |
| "loss": 0.1631, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1752540022134781, | |
| "step": 1595, | |
| "valid_targets_mean": 1685.2, | |
| "valid_targets_min": 474 | |
| }, | |
| { | |
| "epoch": 1.5686274509803921, | |
| "grad_norm": 0.46836465561013363, | |
| "learning_rate": 3.81570257573363e-05, | |
| "loss": 0.1429, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14558619260787964, | |
| "step": 1600, | |
| "valid_targets_mean": 2609.4, | |
| "valid_targets_min": 890 | |
| }, | |
| { | |
| "epoch": 1.5735294117647058, | |
| "grad_norm": 0.7066979739662512, | |
| "learning_rate": 3.813647284401689e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1357910931110382, | |
| "step": 1605, | |
| "valid_targets_mean": 1793.6, | |
| "valid_targets_min": 574 | |
| }, | |
| { | |
| "epoch": 1.5784313725490198, | |
| "grad_norm": 0.49109212337132546, | |
| "learning_rate": 3.811581156023536e-05, | |
| "loss": 0.1577, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14518040418624878, | |
| "step": 1610, | |
| "valid_targets_mean": 2120.4, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 1.5833333333333335, | |
| "grad_norm": 0.5111836333604469, | |
| "learning_rate": 3.809504202944861e-05, | |
| "loss": 0.1464, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14256779849529266, | |
| "step": 1615, | |
| "valid_targets_mean": 1982.6, | |
| "valid_targets_min": 696 | |
| }, | |
| { | |
| "epoch": 1.5882352941176472, | |
| "grad_norm": 0.9023676798605845, | |
| "learning_rate": 3.807416437576036e-05, | |
| "loss": 0.1528, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1626691222190857, | |
| "step": 1620, | |
| "valid_targets_mean": 1895.9, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 1.593137254901961, | |
| "grad_norm": 0.5123255661123512, | |
| "learning_rate": 3.80531787239204e-05, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16807357966899872, | |
| "step": 1625, | |
| "valid_targets_mean": 2144.9, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 1.5980392156862746, | |
| "grad_norm": 0.491171773292401, | |
| "learning_rate": 3.803208519932381e-05, | |
| "loss": 0.1849, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2099742591381073, | |
| "step": 1630, | |
| "valid_targets_mean": 2904.7, | |
| "valid_targets_min": 950 | |
| }, | |
| { | |
| "epoch": 1.6029411764705883, | |
| "grad_norm": 0.3905325983675772, | |
| "learning_rate": 3.8010883928010265e-05, | |
| "loss": 0.152, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1118234246969223, | |
| "step": 1635, | |
| "valid_targets_mean": 2854.1, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 1.607843137254902, | |
| "grad_norm": 0.5356632738991045, | |
| "learning_rate": 3.798957503666325e-05, | |
| "loss": 0.1456, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17137929797172546, | |
| "step": 1640, | |
| "valid_targets_mean": 2408.6, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 1.6127450980392157, | |
| "grad_norm": 0.47756403160814764, | |
| "learning_rate": 3.7968158652609306e-05, | |
| "loss": 0.1545, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13389216363430023, | |
| "step": 1645, | |
| "valid_targets_mean": 2166.3, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 1.6176470588235294, | |
| "grad_norm": 0.6104217262915782, | |
| "learning_rate": 3.7946634903817284e-05, | |
| "loss": 0.1402, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17253968119621277, | |
| "step": 1650, | |
| "valid_targets_mean": 2078.8, | |
| "valid_targets_min": 529 | |
| }, | |
| { | |
| "epoch": 1.6225490196078431, | |
| "grad_norm": 0.7027710890124704, | |
| "learning_rate": 3.792500391889755e-05, | |
| "loss": 0.1412, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14878009259700775, | |
| "step": 1655, | |
| "valid_targets_mean": 2058.6, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 1.6274509803921569, | |
| "grad_norm": 0.5478596948797649, | |
| "learning_rate": 3.790326582710125e-05, | |
| "loss": 0.1353, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14182859659194946, | |
| "step": 1660, | |
| "valid_targets_mean": 2234.1, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 1.6323529411764706, | |
| "grad_norm": 0.5465599142774893, | |
| "learning_rate": 3.788142075831952e-05, | |
| "loss": 0.155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15382519364356995, | |
| "step": 1665, | |
| "valid_targets_mean": 1936.5, | |
| "valid_targets_min": 952 | |
| }, | |
| { | |
| "epoch": 1.6372549019607843, | |
| "grad_norm": 0.577878109070135, | |
| "learning_rate": 3.7859468843082716e-05, | |
| "loss": 0.1506, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17845892906188965, | |
| "step": 1670, | |
| "valid_targets_mean": 2106.3, | |
| "valid_targets_min": 877 | |
| }, | |
| { | |
| "epoch": 1.642156862745098, | |
| "grad_norm": 0.48117276564009087, | |
| "learning_rate": 3.7837410212559623e-05, | |
| "loss": 0.1592, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14203675091266632, | |
| "step": 1675, | |
| "valid_targets_mean": 2077.6, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 1.6470588235294117, | |
| "grad_norm": 0.5585520122750612, | |
| "learning_rate": 3.781524499855668e-05, | |
| "loss": 0.1463, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15015847980976105, | |
| "step": 1680, | |
| "valid_targets_mean": 1771.9, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 1.6519607843137254, | |
| "grad_norm": 0.4289941836302043, | |
| "learning_rate": 3.779297333351721e-05, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11602067947387695, | |
| "step": 1685, | |
| "valid_targets_mean": 2702.9, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 1.656862745098039, | |
| "grad_norm": 0.4886606859408909, | |
| "learning_rate": 3.777059535052059e-05, | |
| "loss": 0.1433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15592262148857117, | |
| "step": 1690, | |
| "valid_targets_mean": 2242.9, | |
| "valid_targets_min": 856 | |
| }, | |
| { | |
| "epoch": 1.6617647058823528, | |
| "grad_norm": 0.603285219664727, | |
| "learning_rate": 3.774811118328149e-05, | |
| "loss": 0.1524, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18172068893909454, | |
| "step": 1695, | |
| "valid_targets_mean": 1707.1, | |
| "valid_targets_min": 529 | |
| }, | |
| { | |
| "epoch": 1.6666666666666665, | |
| "grad_norm": 0.5302230844470484, | |
| "learning_rate": 3.772552096614904e-05, | |
| "loss": 0.1493, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15116459131240845, | |
| "step": 1700, | |
| "valid_targets_mean": 2325.4, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 1.6715686274509802, | |
| "grad_norm": 0.6006361801337328, | |
| "learning_rate": 3.7702824834106066e-05, | |
| "loss": 0.1456, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1827220320701599, | |
| "step": 1705, | |
| "valid_targets_mean": 2108.4, | |
| "valid_targets_min": 533 | |
| }, | |
| { | |
| "epoch": 1.6764705882352942, | |
| "grad_norm": 0.4592856246938828, | |
| "learning_rate": 3.768002292276826e-05, | |
| "loss": 0.1364, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12549643218517303, | |
| "step": 1710, | |
| "valid_targets_mean": 2249.1, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 1.6813725490196079, | |
| "grad_norm": 0.6134998129989979, | |
| "learning_rate": 3.765711536838338e-05, | |
| "loss": 0.1481, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18850260972976685, | |
| "step": 1715, | |
| "valid_targets_mean": 1851.9, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 1.6862745098039216, | |
| "grad_norm": 0.49997491291599816, | |
| "learning_rate": 3.763410230783042e-05, | |
| "loss": 0.1632, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.203414186835289, | |
| "step": 1720, | |
| "valid_targets_mean": 2292.8, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 1.6911764705882353, | |
| "grad_norm": 0.47966143153752505, | |
| "learning_rate": 3.7610983878618816e-05, | |
| "loss": 0.1439, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14274010062217712, | |
| "step": 1725, | |
| "valid_targets_mean": 2055.4, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 1.696078431372549, | |
| "grad_norm": 0.5041314204689977, | |
| "learning_rate": 3.7587760218887595e-05, | |
| "loss": 0.1662, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17796386778354645, | |
| "step": 1730, | |
| "valid_targets_mean": 2114.6, | |
| "valid_targets_min": 598 | |
| }, | |
| { | |
| "epoch": 1.7009803921568627, | |
| "grad_norm": 0.5496728194221527, | |
| "learning_rate": 3.756443146740457e-05, | |
| "loss": 0.1527, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1596316248178482, | |
| "step": 1735, | |
| "valid_targets_mean": 1827.0, | |
| "valid_targets_min": 829 | |
| }, | |
| { | |
| "epoch": 1.7058823529411766, | |
| "grad_norm": 0.5933172007849838, | |
| "learning_rate": 3.7540997763565525e-05, | |
| "loss": 0.1389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16515210270881653, | |
| "step": 1740, | |
| "valid_targets_mean": 1805.4, | |
| "valid_targets_min": 688 | |
| }, | |
| { | |
| "epoch": 1.7107843137254903, | |
| "grad_norm": 0.5533001908804063, | |
| "learning_rate": 3.751745924739333e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16086289286613464, | |
| "step": 1745, | |
| "valid_targets_mean": 2245.2, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 1.715686274509804, | |
| "grad_norm": 0.3625773782586867, | |
| "learning_rate": 3.7493816059537174e-05, | |
| "loss": 0.1389, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10024596750736237, | |
| "step": 1750, | |
| "valid_targets_mean": 2614.1, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 1.7205882352941178, | |
| "grad_norm": 0.4920024696196442, | |
| "learning_rate": 3.747006834127166e-05, | |
| "loss": 0.1479, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1591131091117859, | |
| "step": 1755, | |
| "valid_targets_mean": 2266.2, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 1.7254901960784315, | |
| "grad_norm": 0.40531676129734534, | |
| "learning_rate": 3.7446216234496e-05, | |
| "loss": 0.1618, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11736778914928436, | |
| "step": 1760, | |
| "valid_targets_mean": 2764.4, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 1.7303921568627452, | |
| "grad_norm": 0.5413087543165096, | |
| "learning_rate": 3.742225988173315e-05, | |
| "loss": 0.1432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16685521602630615, | |
| "step": 1765, | |
| "valid_targets_mean": 1968.8, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 1.7352941176470589, | |
| "grad_norm": 0.5458227730487121, | |
| "learning_rate": 3.7398199426128995e-05, | |
| "loss": 0.1518, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14602074027061462, | |
| "step": 1770, | |
| "valid_targets_mean": 1909.1, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 1.7401960784313726, | |
| "grad_norm": 0.4399415602242103, | |
| "learning_rate": 3.737403501145141e-05, | |
| "loss": 0.1442, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14302271604537964, | |
| "step": 1775, | |
| "valid_targets_mean": 2533.6, | |
| "valid_targets_min": 943 | |
| }, | |
| { | |
| "epoch": 1.7450980392156863, | |
| "grad_norm": 0.5607709048920859, | |
| "learning_rate": 3.7349766782089515e-05, | |
| "loss": 0.1496, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.18201008439064026, | |
| "step": 1780, | |
| "valid_targets_mean": 1960.6, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "grad_norm": 0.5414297127263566, | |
| "learning_rate": 3.732539488305269e-05, | |
| "loss": 0.1478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16086480021476746, | |
| "step": 1785, | |
| "valid_targets_mean": 2153.8, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 1.7549019607843137, | |
| "grad_norm": 0.5278807774304168, | |
| "learning_rate": 3.73009194599698e-05, | |
| "loss": 0.1601, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15399423241615295, | |
| "step": 1790, | |
| "valid_targets_mean": 1997.1, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 1.7598039215686274, | |
| "grad_norm": 0.5043590546147471, | |
| "learning_rate": 3.727634065908833e-05, | |
| "loss": 0.1663, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15289407968521118, | |
| "step": 1795, | |
| "valid_targets_mean": 2063.8, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 1.7647058823529411, | |
| "grad_norm": 0.44386672927984044, | |
| "learning_rate": 3.725165862727341e-05, | |
| "loss": 0.1471, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1439841389656067, | |
| "step": 1800, | |
| "valid_targets_mean": 2574.6, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 1.7696078431372548, | |
| "grad_norm": 0.6067341975360978, | |
| "learning_rate": 3.722687351200704e-05, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1851101666688919, | |
| "step": 1805, | |
| "valid_targets_mean": 1918.6, | |
| "valid_targets_min": 767 | |
| }, | |
| { | |
| "epoch": 1.7745098039215685, | |
| "grad_norm": 0.5472528253501442, | |
| "learning_rate": 3.720198546138718e-05, | |
| "loss": 0.1483, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1516643464565277, | |
| "step": 1810, | |
| "valid_targets_mean": 1931.3, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 1.7794117647058822, | |
| "grad_norm": 0.5273521005478816, | |
| "learning_rate": 3.717699462412683e-05, | |
| "loss": 0.1403, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12933941185474396, | |
| "step": 1815, | |
| "valid_targets_mean": 2039.8, | |
| "valid_targets_min": 525 | |
| }, | |
| { | |
| "epoch": 1.784313725490196, | |
| "grad_norm": 0.6077315011956368, | |
| "learning_rate": 3.715190114955319e-05, | |
| "loss": 0.1531, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17466555535793304, | |
| "step": 1820, | |
| "valid_targets_mean": 1669.6, | |
| "valid_targets_min": 494 | |
| }, | |
| { | |
| "epoch": 1.7892156862745097, | |
| "grad_norm": 0.512358696269543, | |
| "learning_rate": 3.712670518760674e-05, | |
| "loss": 0.1617, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1617322564125061, | |
| "step": 1825, | |
| "valid_targets_mean": 2220.2, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 1.7941176470588234, | |
| "grad_norm": 0.4336332041487301, | |
| "learning_rate": 3.7101406888840345e-05, | |
| "loss": 0.1485, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15320450067520142, | |
| "step": 1830, | |
| "valid_targets_mean": 2737.3, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 1.7990196078431373, | |
| "grad_norm": 0.4883048284620736, | |
| "learning_rate": 3.707600640441837e-05, | |
| "loss": 0.1462, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13818612694740295, | |
| "step": 1835, | |
| "valid_targets_mean": 2208.4, | |
| "valid_targets_min": 490 | |
| }, | |
| { | |
| "epoch": 1.803921568627451, | |
| "grad_norm": 0.43072754382140194, | |
| "learning_rate": 3.705050388611577e-05, | |
| "loss": 0.1485, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13516834378242493, | |
| "step": 1840, | |
| "valid_targets_mean": 2813.6, | |
| "valid_targets_min": 695 | |
| }, | |
| { | |
| "epoch": 1.8088235294117647, | |
| "grad_norm": 0.48514804711619286, | |
| "learning_rate": 3.702489948631716e-05, | |
| "loss": 0.157, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1558280736207962, | |
| "step": 1845, | |
| "valid_targets_mean": 2298.6, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 1.8137254901960784, | |
| "grad_norm": 0.43686024221397773, | |
| "learning_rate": 3.6999193358015955e-05, | |
| "loss": 0.1431, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13829517364501953, | |
| "step": 1850, | |
| "valid_targets_mean": 2430.9, | |
| "valid_targets_min": 769 | |
| }, | |
| { | |
| "epoch": 1.8186274509803921, | |
| "grad_norm": 0.41757476802350063, | |
| "learning_rate": 3.697338565481339e-05, | |
| "loss": 0.1385, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13145369291305542, | |
| "step": 1855, | |
| "valid_targets_mean": 2702.2, | |
| "valid_targets_min": 910 | |
| }, | |
| { | |
| "epoch": 1.8235294117647058, | |
| "grad_norm": 0.4474403426831047, | |
| "learning_rate": 3.694747653091768e-05, | |
| "loss": 0.1487, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1498376429080963, | |
| "step": 1860, | |
| "valid_targets_mean": 2116.8, | |
| "valid_targets_min": 1030 | |
| }, | |
| { | |
| "epoch": 1.8284313725490198, | |
| "grad_norm": 0.38358293329939164, | |
| "learning_rate": 3.692146614114303e-05, | |
| "loss": 0.1247, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10769416391849518, | |
| "step": 1865, | |
| "valid_targets_mean": 2927.0, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 1.8333333333333335, | |
| "grad_norm": 0.4665869269944144, | |
| "learning_rate": 3.689535464090873e-05, | |
| "loss": 0.139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1278778612613678, | |
| "step": 1870, | |
| "valid_targets_mean": 2084.3, | |
| "valid_targets_min": 459 | |
| }, | |
| { | |
| "epoch": 1.8382352941176472, | |
| "grad_norm": 0.5307097608707636, | |
| "learning_rate": 3.686914218623827e-05, | |
| "loss": 0.1549, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14258277416229248, | |
| "step": 1875, | |
| "valid_targets_mean": 2299.4, | |
| "valid_targets_min": 502 | |
| }, | |
| { | |
| "epoch": 1.843137254901961, | |
| "grad_norm": 0.6080976433925368, | |
| "learning_rate": 3.684282893375832e-05, | |
| "loss": 0.1608, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.19785189628601074, | |
| "step": 1880, | |
| "valid_targets_mean": 1875.2, | |
| "valid_targets_min": 540 | |
| }, | |
| { | |
| "epoch": 1.8480392156862746, | |
| "grad_norm": 0.5217574684853883, | |
| "learning_rate": 3.681641504069789e-05, | |
| "loss": 0.1482, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16252940893173218, | |
| "step": 1885, | |
| "valid_targets_mean": 1984.9, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 1.8529411764705883, | |
| "grad_norm": 0.47220196862435904, | |
| "learning_rate": 3.678990066488732e-05, | |
| "loss": 0.1465, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1338098645210266, | |
| "step": 1890, | |
| "valid_targets_mean": 2058.9, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 1.857843137254902, | |
| "grad_norm": 0.6141551360587747, | |
| "learning_rate": 3.676328596475737e-05, | |
| "loss": 0.177, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16199412941932678, | |
| "step": 1895, | |
| "valid_targets_mean": 1510.5, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 1.8627450980392157, | |
| "grad_norm": 0.5205786315285574, | |
| "learning_rate": 3.673657109933825e-05, | |
| "loss": 0.1374, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14440283179283142, | |
| "step": 1900, | |
| "valid_targets_mean": 1927.1, | |
| "valid_targets_min": 557 | |
| }, | |
| { | |
| "epoch": 1.8676470588235294, | |
| "grad_norm": 0.4953417684318436, | |
| "learning_rate": 3.6709756228258735e-05, | |
| "loss": 0.1514, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1402466744184494, | |
| "step": 1905, | |
| "valid_targets_mean": 2096.1, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 1.8725490196078431, | |
| "grad_norm": 0.4604839145265826, | |
| "learning_rate": 3.66828415117451e-05, | |
| "loss": 0.15, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1322975754737854, | |
| "step": 1910, | |
| "valid_targets_mean": 2135.5, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 1.8774509803921569, | |
| "grad_norm": 0.6455850290038503, | |
| "learning_rate": 3.665582711062025e-05, | |
| "loss": 0.1688, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.22754456102848053, | |
| "step": 1915, | |
| "valid_targets_mean": 1907.5, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 1.8823529411764706, | |
| "grad_norm": 0.4715702383406565, | |
| "learning_rate": 3.662871318630274e-05, | |
| "loss": 0.1414, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15306609869003296, | |
| "step": 1920, | |
| "valid_targets_mean": 2446.2, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 1.8872549019607843, | |
| "grad_norm": 0.41736403813308404, | |
| "learning_rate": 3.6601499900805806e-05, | |
| "loss": 0.136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13367821276187897, | |
| "step": 1925, | |
| "valid_targets_mean": 3024.9, | |
| "valid_targets_min": 1184 | |
| }, | |
| { | |
| "epoch": 1.892156862745098, | |
| "grad_norm": 0.5399907281651121, | |
| "learning_rate": 3.657418741673638e-05, | |
| "loss": 0.1473, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13740979135036469, | |
| "step": 1930, | |
| "valid_targets_mean": 1898.3, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 1.8970588235294117, | |
| "grad_norm": 0.5478245298908677, | |
| "learning_rate": 3.654677589729413e-05, | |
| "loss": 0.1374, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14879709482192993, | |
| "step": 1935, | |
| "valid_targets_mean": 2260.4, | |
| "valid_targets_min": 329 | |
| }, | |
| { | |
| "epoch": 1.9019607843137254, | |
| "grad_norm": 0.5065828787704405, | |
| "learning_rate": 3.6519265506270514e-05, | |
| "loss": 0.1487, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1565818339586258, | |
| "step": 1940, | |
| "valid_targets_mean": 2250.4, | |
| "valid_targets_min": 424 | |
| }, | |
| { | |
| "epoch": 1.906862745098039, | |
| "grad_norm": 0.36373265285176326, | |
| "learning_rate": 3.649165640804775e-05, | |
| "loss": 0.1393, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11532604694366455, | |
| "step": 1945, | |
| "valid_targets_mean": 2804.2, | |
| "valid_targets_min": 919 | |
| }, | |
| { | |
| "epoch": 1.9117647058823528, | |
| "grad_norm": 0.5378415324082915, | |
| "learning_rate": 3.6463948767597865e-05, | |
| "loss": 0.1594, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1640024185180664, | |
| "step": 1950, | |
| "valid_targets_mean": 1955.8, | |
| "valid_targets_min": 796 | |
| }, | |
| { | |
| "epoch": 1.9166666666666665, | |
| "grad_norm": 0.5486860769413845, | |
| "learning_rate": 3.643614275048172e-05, | |
| "loss": 0.146, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13714325428009033, | |
| "step": 1955, | |
| "valid_targets_mean": 1825.2, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 1.9215686274509802, | |
| "grad_norm": 0.4885348084048327, | |
| "learning_rate": 3.640823852284797e-05, | |
| "loss": 0.1615, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14006605744361877, | |
| "step": 1960, | |
| "valid_targets_mean": 2088.1, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 1.9264705882352942, | |
| "grad_norm": 0.43492961451338746, | |
| "learning_rate": 3.6380236251432135e-05, | |
| "loss": 0.1541, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1264418661594391, | |
| "step": 1965, | |
| "valid_targets_mean": 2195.3, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 1.9313725490196079, | |
| "grad_norm": 0.40264611933976796, | |
| "learning_rate": 3.635213610355556e-05, | |
| "loss": 0.1651, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13179033994674683, | |
| "step": 1970, | |
| "valid_targets_mean": 2733.8, | |
| "valid_targets_min": 568 | |
| }, | |
| { | |
| "epoch": 1.9362745098039216, | |
| "grad_norm": 0.36990460761813326, | |
| "learning_rate": 3.632393824712444e-05, | |
| "loss": 0.1279, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1107700914144516, | |
| "step": 1975, | |
| "valid_targets_mean": 2647.9, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 1.9411764705882353, | |
| "grad_norm": 0.41248718120084454, | |
| "learning_rate": 3.6295642850628785e-05, | |
| "loss": 0.1391, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10980524122714996, | |
| "step": 1980, | |
| "valid_targets_mean": 2564.8, | |
| "valid_targets_min": 513 | |
| }, | |
| { | |
| "epoch": 1.946078431372549, | |
| "grad_norm": 0.43158575489604406, | |
| "learning_rate": 3.626725008314146e-05, | |
| "loss": 0.1503, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1632903516292572, | |
| "step": 1985, | |
| "valid_targets_mean": 2678.8, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 1.9509803921568627, | |
| "grad_norm": 0.3745578920161323, | |
| "learning_rate": 3.623876011431714e-05, | |
| "loss": 0.1426, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11174066364765167, | |
| "step": 1990, | |
| "valid_targets_mean": 2612.4, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 1.9558823529411766, | |
| "grad_norm": 0.5970799142780278, | |
| "learning_rate": 3.621017311439127e-05, | |
| "loss": 0.1586, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15000393986701965, | |
| "step": 1995, | |
| "valid_targets_mean": 2207.6, | |
| "valid_targets_min": 575 | |
| }, | |
| { | |
| "epoch": 1.9607843137254903, | |
| "grad_norm": 0.6391894361261641, | |
| "learning_rate": 3.6181489254179154e-05, | |
| "loss": 0.1717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1867298036813736, | |
| "step": 2000, | |
| "valid_targets_mean": 1771.7, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 1.965686274509804, | |
| "grad_norm": 0.5067121980202994, | |
| "learning_rate": 3.6152708705074805e-05, | |
| "loss": 0.1761, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.2050846666097641, | |
| "step": 2005, | |
| "valid_targets_mean": 2280.9, | |
| "valid_targets_min": 793 | |
| }, | |
| { | |
| "epoch": 1.9705882352941178, | |
| "grad_norm": 0.42741514259257013, | |
| "learning_rate": 3.612383163904999e-05, | |
| "loss": 0.163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16011744737625122, | |
| "step": 2010, | |
| "valid_targets_mean": 2644.6, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 1.9754901960784315, | |
| "grad_norm": 0.4918328620919532, | |
| "learning_rate": 3.609485822865321e-05, | |
| "loss": 0.1345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15025535225868225, | |
| "step": 2015, | |
| "valid_targets_mean": 2447.8, | |
| "valid_targets_min": 946 | |
| }, | |
| { | |
| "epoch": 1.9803921568627452, | |
| "grad_norm": 0.4516353716793599, | |
| "learning_rate": 3.606578864700863e-05, | |
| "loss": 0.1478, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13741713762283325, | |
| "step": 2020, | |
| "valid_targets_mean": 2252.0, | |
| "valid_targets_min": 935 | |
| }, | |
| { | |
| "epoch": 1.9852941176470589, | |
| "grad_norm": 0.5604190301899437, | |
| "learning_rate": 3.603662306781507e-05, | |
| "loss": 0.1611, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16286993026733398, | |
| "step": 2025, | |
| "valid_targets_mean": 2178.6, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 1.9901960784313726, | |
| "grad_norm": 0.527379241441151, | |
| "learning_rate": 3.600736166534499e-05, | |
| "loss": 0.1461, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17896302044391632, | |
| "step": 2030, | |
| "valid_targets_mean": 1935.2, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 1.9950980392156863, | |
| "grad_norm": 0.5231350069083546, | |
| "learning_rate": 3.597800461444337e-05, | |
| "loss": 0.1464, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15988443791866302, | |
| "step": 2035, | |
| "valid_targets_mean": 1890.6, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.3655620271489274, | |
| "learning_rate": 3.5948552090526747e-05, | |
| "loss": 0.1388, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10807867348194122, | |
| "step": 2040, | |
| "valid_targets_mean": 2740.8, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 2.0049019607843137, | |
| "grad_norm": 0.44853090320448885, | |
| "learning_rate": 3.591900426958214e-05, | |
| "loss": 0.1252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12851010262966156, | |
| "step": 2045, | |
| "valid_targets_mean": 2482.2, | |
| "valid_targets_min": 1061 | |
| }, | |
| { | |
| "epoch": 2.0098039215686274, | |
| "grad_norm": 0.4583868678990238, | |
| "learning_rate": 3.5889361328165984e-05, | |
| "loss": 0.1255, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12361206114292145, | |
| "step": 2050, | |
| "valid_targets_mean": 2640.2, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 2.014705882352941, | |
| "grad_norm": 0.552121520532339, | |
| "learning_rate": 3.5859623443403073e-05, | |
| "loss": 0.1442, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17563346028327942, | |
| "step": 2055, | |
| "valid_targets_mean": 2000.2, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 2.019607843137255, | |
| "grad_norm": 0.5194438436937125, | |
| "learning_rate": 3.582979079298554e-05, | |
| "loss": 0.131, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14061996340751648, | |
| "step": 2060, | |
| "valid_targets_mean": 2127.3, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 2.0245098039215685, | |
| "grad_norm": 0.47306084034707796, | |
| "learning_rate": 3.579986355517176e-05, | |
| "loss": 0.1338, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1396002471446991, | |
| "step": 2065, | |
| "valid_targets_mean": 2682.8, | |
| "valid_targets_min": 707 | |
| }, | |
| { | |
| "epoch": 2.0294117647058822, | |
| "grad_norm": 0.46665365715691065, | |
| "learning_rate": 3.576984190878528e-05, | |
| "loss": 0.1431, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14326587319374084, | |
| "step": 2070, | |
| "valid_targets_mean": 2272.3, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 2.034313725490196, | |
| "grad_norm": 0.61689314251335, | |
| "learning_rate": 3.5739726033213785e-05, | |
| "loss": 0.1171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14959578216075897, | |
| "step": 2075, | |
| "valid_targets_mean": 1627.8, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 2.0392156862745097, | |
| "grad_norm": 0.5457065769996188, | |
| "learning_rate": 3.570951610840798e-05, | |
| "loss": 0.1377, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13586756587028503, | |
| "step": 2080, | |
| "valid_targets_mean": 2058.1, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 2.0441176470588234, | |
| "grad_norm": 0.4753398929172197, | |
| "learning_rate": 3.567921231488057e-05, | |
| "loss": 0.1238, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11476504802703857, | |
| "step": 2085, | |
| "valid_targets_mean": 2186.1, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 2.049019607843137, | |
| "grad_norm": 0.4759520403773523, | |
| "learning_rate": 3.564881483370512e-05, | |
| "loss": 0.1263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10093535482883453, | |
| "step": 2090, | |
| "valid_targets_mean": 2484.3, | |
| "valid_targets_min": 944 | |
| }, | |
| { | |
| "epoch": 2.053921568627451, | |
| "grad_norm": 0.47588801985114326, | |
| "learning_rate": 3.561832384651506e-05, | |
| "loss": 0.1239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12644708156585693, | |
| "step": 2095, | |
| "valid_targets_mean": 2357.6, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 2.0588235294117645, | |
| "grad_norm": 0.7125143297196016, | |
| "learning_rate": 3.5587739535502454e-05, | |
| "loss": 0.133, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1273224800825119, | |
| "step": 2100, | |
| "valid_targets_mean": 2272.1, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 2.063725490196078, | |
| "grad_norm": 0.5202077844278824, | |
| "learning_rate": 3.5557062083417094e-05, | |
| "loss": 0.1374, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1436024308204651, | |
| "step": 2105, | |
| "valid_targets_mean": 2256.0, | |
| "valid_targets_min": 805 | |
| }, | |
| { | |
| "epoch": 2.0686274509803924, | |
| "grad_norm": 0.6356593180787531, | |
| "learning_rate": 3.5526291673565256e-05, | |
| "loss": 0.134, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1912994682788849, | |
| "step": 2110, | |
| "valid_targets_mean": 2107.8, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 2.073529411764706, | |
| "grad_norm": 0.40074516453340914, | |
| "learning_rate": 3.5495428489808696e-05, | |
| "loss": 0.1247, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11367502808570862, | |
| "step": 2115, | |
| "valid_targets_mean": 2840.8, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 2.0784313725490198, | |
| "grad_norm": 0.592311502287731, | |
| "learning_rate": 3.546447271656351e-05, | |
| "loss": 0.126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1876107156276703, | |
| "step": 2120, | |
| "valid_targets_mean": 2378.8, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 2.0833333333333335, | |
| "grad_norm": 0.5815462989037039, | |
| "learning_rate": 3.5433424538799024e-05, | |
| "loss": 0.1498, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14821037650108337, | |
| "step": 2125, | |
| "valid_targets_mean": 2055.6, | |
| "valid_targets_min": 544 | |
| }, | |
| { | |
| "epoch": 2.088235294117647, | |
| "grad_norm": 0.5589277406530077, | |
| "learning_rate": 3.5402284142036733e-05, | |
| "loss": 0.1258, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1250324249267578, | |
| "step": 2130, | |
| "valid_targets_mean": 1668.1, | |
| "valid_targets_min": 545 | |
| }, | |
| { | |
| "epoch": 2.093137254901961, | |
| "grad_norm": 0.43853845260878704, | |
| "learning_rate": 3.5371051712349156e-05, | |
| "loss": 0.1324, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11538475751876831, | |
| "step": 2135, | |
| "valid_targets_mean": 2481.9, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 2.0980392156862746, | |
| "grad_norm": 0.5412188024421883, | |
| "learning_rate": 3.533972743635874e-05, | |
| "loss": 0.1235, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15065398812294006, | |
| "step": 2140, | |
| "valid_targets_mean": 2122.2, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 2.1029411764705883, | |
| "grad_norm": 0.511228541672383, | |
| "learning_rate": 3.530831150123672e-05, | |
| "loss": 0.13, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10731547325849533, | |
| "step": 2145, | |
| "valid_targets_mean": 2195.9, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 2.107843137254902, | |
| "grad_norm": 0.6186996765892316, | |
| "learning_rate": 3.527680409470205e-05, | |
| "loss": 0.1302, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1374223530292511, | |
| "step": 2150, | |
| "valid_targets_mean": 1655.9, | |
| "valid_targets_min": 502 | |
| }, | |
| { | |
| "epoch": 2.1127450980392157, | |
| "grad_norm": 0.5274842444330995, | |
| "learning_rate": 3.524520540502021e-05, | |
| "loss": 0.1361, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.125177264213562, | |
| "step": 2155, | |
| "valid_targets_mean": 1779.2, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 2.1176470588235294, | |
| "grad_norm": 0.39573551657318656, | |
| "learning_rate": 3.521351562100218e-05, | |
| "loss": 0.1381, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09408603608608246, | |
| "step": 2160, | |
| "valid_targets_mean": 2624.2, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 2.122549019607843, | |
| "grad_norm": 0.49074741867654176, | |
| "learning_rate": 3.518173493200318e-05, | |
| "loss": 0.1313, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12309744954109192, | |
| "step": 2165, | |
| "valid_targets_mean": 2309.3, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 2.127450980392157, | |
| "grad_norm": 0.528817977555379, | |
| "learning_rate": 3.514986352792169e-05, | |
| "loss": 0.129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11972697079181671, | |
| "step": 2170, | |
| "valid_targets_mean": 2068.5, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 2.1323529411764706, | |
| "grad_norm": 0.47728039921295046, | |
| "learning_rate": 3.511790159919818e-05, | |
| "loss": 0.1298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1214258000254631, | |
| "step": 2175, | |
| "valid_targets_mean": 2368.5, | |
| "valid_targets_min": 827 | |
| }, | |
| { | |
| "epoch": 2.1372549019607843, | |
| "grad_norm": 0.48897620372182415, | |
| "learning_rate": 3.508584933681406e-05, | |
| "loss": 0.119, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1263759583234787, | |
| "step": 2180, | |
| "valid_targets_mean": 2689.0, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 2.142156862745098, | |
| "grad_norm": 0.5061688289772812, | |
| "learning_rate": 3.5053706932290496e-05, | |
| "loss": 0.1207, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11041820049285889, | |
| "step": 2185, | |
| "valid_targets_mean": 2251.2, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 2.1470588235294117, | |
| "grad_norm": 0.43813308425959413, | |
| "learning_rate": 3.502147457768727e-05, | |
| "loss": 0.1227, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10830570757389069, | |
| "step": 2190, | |
| "valid_targets_mean": 2631.1, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 2.1519607843137254, | |
| "grad_norm": 0.968182961952041, | |
| "learning_rate": 3.4989152465601674e-05, | |
| "loss": 0.123, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12635761499404907, | |
| "step": 2195, | |
| "valid_targets_mean": 2246.6, | |
| "valid_targets_min": 497 | |
| }, | |
| { | |
| "epoch": 2.156862745098039, | |
| "grad_norm": 0.49739085063558636, | |
| "learning_rate": 3.4956740789167295e-05, | |
| "loss": 0.1411, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13866953551769257, | |
| "step": 2200, | |
| "valid_targets_mean": 2214.8, | |
| "valid_targets_min": 779 | |
| }, | |
| { | |
| "epoch": 2.161764705882353, | |
| "grad_norm": 0.5962862027967147, | |
| "learning_rate": 3.492423974205289e-05, | |
| "loss": 0.1248, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15598756074905396, | |
| "step": 2205, | |
| "valid_targets_mean": 1891.1, | |
| "valid_targets_min": 792 | |
| }, | |
| { | |
| "epoch": 2.1666666666666665, | |
| "grad_norm": 0.5592001479661793, | |
| "learning_rate": 3.489164951846126e-05, | |
| "loss": 0.1337, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13359807431697845, | |
| "step": 2210, | |
| "valid_targets_mean": 2060.6, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 2.1715686274509802, | |
| "grad_norm": 0.5689778264268539, | |
| "learning_rate": 3.485897031312803e-05, | |
| "loss": 0.1454, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15123270452022552, | |
| "step": 2215, | |
| "valid_targets_mean": 1901.0, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.176470588235294, | |
| "grad_norm": 0.5298808530933589, | |
| "learning_rate": 3.482620232132053e-05, | |
| "loss": 0.1329, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1346893608570099, | |
| "step": 2220, | |
| "valid_targets_mean": 2066.5, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 2.1813725490196076, | |
| "grad_norm": 0.49863038928798403, | |
| "learning_rate": 3.479334573883661e-05, | |
| "loss": 0.126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1366402953863144, | |
| "step": 2225, | |
| "valid_targets_mean": 2096.5, | |
| "valid_targets_min": 447 | |
| }, | |
| { | |
| "epoch": 2.186274509803922, | |
| "grad_norm": 0.5628304884727421, | |
| "learning_rate": 3.476040076200349e-05, | |
| "loss": 0.1263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1595756709575653, | |
| "step": 2230, | |
| "valid_targets_mean": 2224.7, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 2.1911764705882355, | |
| "grad_norm": 0.5623315824364589, | |
| "learning_rate": 3.4727367587676526e-05, | |
| "loss": 0.135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13027265667915344, | |
| "step": 2235, | |
| "valid_targets_mean": 1854.1, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 2.196078431372549, | |
| "grad_norm": 0.473141662338017, | |
| "learning_rate": 3.469424641323814e-05, | |
| "loss": 0.1331, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12759196758270264, | |
| "step": 2240, | |
| "valid_targets_mean": 2349.1, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 2.200980392156863, | |
| "grad_norm": 0.41221711934436744, | |
| "learning_rate": 3.4661037436596526e-05, | |
| "loss": 0.1241, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10086143016815186, | |
| "step": 2245, | |
| "valid_targets_mean": 2314.2, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.2058823529411766, | |
| "grad_norm": 0.46078471255102504, | |
| "learning_rate": 3.462774085618456e-05, | |
| "loss": 0.1365, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13009145855903625, | |
| "step": 2250, | |
| "valid_targets_mean": 2152.8, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 2.2107843137254903, | |
| "grad_norm": 0.4447788825171731, | |
| "learning_rate": 3.459435687095854e-05, | |
| "loss": 0.12, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1283532679080963, | |
| "step": 2255, | |
| "valid_targets_mean": 2686.1, | |
| "valid_targets_min": 564 | |
| }, | |
| { | |
| "epoch": 2.215686274509804, | |
| "grad_norm": 0.4495583706963151, | |
| "learning_rate": 3.456088568039706e-05, | |
| "loss": 0.1394, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12000022828578949, | |
| "step": 2260, | |
| "valid_targets_mean": 2626.5, | |
| "valid_targets_min": 887 | |
| }, | |
| { | |
| "epoch": 2.2205882352941178, | |
| "grad_norm": 0.4341070451837946, | |
| "learning_rate": 3.4527327484499774e-05, | |
| "loss": 0.1242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11233153194189072, | |
| "step": 2265, | |
| "valid_targets_mean": 2346.6, | |
| "valid_targets_min": 649 | |
| }, | |
| { | |
| "epoch": 2.2254901960784315, | |
| "grad_norm": 0.6511108625477997, | |
| "learning_rate": 3.4493682483786224e-05, | |
| "loss": 0.13, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1597345620393753, | |
| "step": 2270, | |
| "valid_targets_mean": 1768.8, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 2.230392156862745, | |
| "grad_norm": 0.5060816776168183, | |
| "learning_rate": 3.445995087929463e-05, | |
| "loss": 0.1306, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12492702901363373, | |
| "step": 2275, | |
| "valid_targets_mean": 1925.5, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 2.235294117647059, | |
| "grad_norm": 0.3940204045235387, | |
| "learning_rate": 3.44261328725807e-05, | |
| "loss": 0.1382, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12375809997320175, | |
| "step": 2280, | |
| "valid_targets_mean": 2956.6, | |
| "valid_targets_min": 1185 | |
| }, | |
| { | |
| "epoch": 2.2401960784313726, | |
| "grad_norm": 0.444465719098011, | |
| "learning_rate": 3.439222866571642e-05, | |
| "loss": 0.1115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12371980398893356, | |
| "step": 2285, | |
| "valid_targets_mean": 2072.8, | |
| "valid_targets_min": 598 | |
| }, | |
| { | |
| "epoch": 2.2450980392156863, | |
| "grad_norm": 0.4260395688487762, | |
| "learning_rate": 3.435823846128884e-05, | |
| "loss": 0.1414, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13873659074306488, | |
| "step": 2290, | |
| "valid_targets_mean": 2553.1, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "grad_norm": 0.43623375929776165, | |
| "learning_rate": 3.4324162462398866e-05, | |
| "loss": 0.1358, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11390223354101181, | |
| "step": 2295, | |
| "valid_targets_mean": 2437.5, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 2.2549019607843137, | |
| "grad_norm": 0.3515272676523296, | |
| "learning_rate": 3.429000087266006e-05, | |
| "loss": 0.1264, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09347573667764664, | |
| "step": 2300, | |
| "valid_targets_mean": 3243.8, | |
| "valid_targets_min": 1082 | |
| }, | |
| { | |
| "epoch": 2.2598039215686274, | |
| "grad_norm": 0.4408806721575717, | |
| "learning_rate": 3.425575389619741e-05, | |
| "loss": 0.1349, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11834917962551117, | |
| "step": 2305, | |
| "valid_targets_mean": 2635.5, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 2.264705882352941, | |
| "grad_norm": 0.5346131341964713, | |
| "learning_rate": 3.4221421737646104e-05, | |
| "loss": 0.1238, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12321878224611282, | |
| "step": 2310, | |
| "valid_targets_mean": 2153.0, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 2.269607843137255, | |
| "grad_norm": 0.5305132857296478, | |
| "learning_rate": 3.418700460215033e-05, | |
| "loss": 0.1413, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12815603613853455, | |
| "step": 2315, | |
| "valid_targets_mean": 2050.9, | |
| "valid_targets_min": 762 | |
| }, | |
| { | |
| "epoch": 2.2745098039215685, | |
| "grad_norm": 0.48597505457985135, | |
| "learning_rate": 3.415250269536203e-05, | |
| "loss": 0.1305, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1451520174741745, | |
| "step": 2320, | |
| "valid_targets_mean": 2444.1, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 2.2794117647058822, | |
| "grad_norm": 0.5662194806363534, | |
| "learning_rate": 3.411791622343967e-05, | |
| "loss": 0.1312, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13560031354427338, | |
| "step": 2325, | |
| "valid_targets_mean": 1847.0, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 2.284313725490196, | |
| "grad_norm": 0.451292753609537, | |
| "learning_rate": 3.408324539304704e-05, | |
| "loss": 0.1133, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11846590787172318, | |
| "step": 2330, | |
| "valid_targets_mean": 2227.3, | |
| "valid_targets_min": 749 | |
| }, | |
| { | |
| "epoch": 2.2892156862745097, | |
| "grad_norm": 0.4404840749164589, | |
| "learning_rate": 3.404849041135197e-05, | |
| "loss": 0.1349, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11944162845611572, | |
| "step": 2335, | |
| "valid_targets_mean": 2625.9, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 2.2941176470588234, | |
| "grad_norm": 0.4777756658786973, | |
| "learning_rate": 3.4013651486025135e-05, | |
| "loss": 0.1296, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12575772404670715, | |
| "step": 2340, | |
| "valid_targets_mean": 2447.4, | |
| "valid_targets_min": 666 | |
| }, | |
| { | |
| "epoch": 2.299019607843137, | |
| "grad_norm": 0.5519792965011824, | |
| "learning_rate": 3.3978728825238787e-05, | |
| "loss": 0.1432, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14594820141792297, | |
| "step": 2345, | |
| "valid_targets_mean": 2089.8, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 2.303921568627451, | |
| "grad_norm": 0.5812553672215698, | |
| "learning_rate": 3.394372263766552e-05, | |
| "loss": 0.1291, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10362367331981659, | |
| "step": 2350, | |
| "valid_targets_mean": 2284.7, | |
| "valid_targets_min": 817 | |
| }, | |
| { | |
| "epoch": 2.3088235294117645, | |
| "grad_norm": 0.4691399073192363, | |
| "learning_rate": 3.390863313247702e-05, | |
| "loss": 0.1202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11229017376899719, | |
| "step": 2355, | |
| "valid_targets_mean": 2364.8, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 2.313725490196078, | |
| "grad_norm": 0.49890381302134496, | |
| "learning_rate": 3.3873460519342847e-05, | |
| "loss": 0.1354, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1161687970161438, | |
| "step": 2360, | |
| "valid_targets_mean": 2548.6, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.318627450980392, | |
| "grad_norm": 0.48881421124675306, | |
| "learning_rate": 3.3838205008429114e-05, | |
| "loss": 0.1255, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10132157802581787, | |
| "step": 2365, | |
| "valid_targets_mean": 2347.0, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 2.323529411764706, | |
| "grad_norm": 0.5687640140377678, | |
| "learning_rate": 3.38028668103973e-05, | |
| "loss": 0.1317, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15966284275054932, | |
| "step": 2370, | |
| "valid_targets_mean": 2005.6, | |
| "valid_targets_min": 842 | |
| }, | |
| { | |
| "epoch": 2.3284313725490198, | |
| "grad_norm": 0.5241977540459799, | |
| "learning_rate": 3.376744613640294e-05, | |
| "loss": 0.1325, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12451574206352234, | |
| "step": 2375, | |
| "valid_targets_mean": 2067.6, | |
| "valid_targets_min": 970 | |
| }, | |
| { | |
| "epoch": 2.3333333333333335, | |
| "grad_norm": 0.4441320290066578, | |
| "learning_rate": 3.37319431980944e-05, | |
| "loss": 0.1186, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12139909714460373, | |
| "step": 2380, | |
| "valid_targets_mean": 2486.1, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 2.338235294117647, | |
| "grad_norm": 0.3958964877558289, | |
| "learning_rate": 3.369635820761162e-05, | |
| "loss": 0.1287, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10344143211841583, | |
| "step": 2385, | |
| "valid_targets_mean": 2639.9, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 2.343137254901961, | |
| "grad_norm": 0.4235899191085017, | |
| "learning_rate": 3.3660691377584774e-05, | |
| "loss": 0.1294, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11542128026485443, | |
| "step": 2390, | |
| "valid_targets_mean": 2604.4, | |
| "valid_targets_min": 645 | |
| }, | |
| { | |
| "epoch": 2.3480392156862746, | |
| "grad_norm": 0.4510881066136338, | |
| "learning_rate": 3.362494292113308e-05, | |
| "loss": 0.1301, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1214420273900032, | |
| "step": 2395, | |
| "valid_targets_mean": 2498.9, | |
| "valid_targets_min": 1012 | |
| }, | |
| { | |
| "epoch": 2.3529411764705883, | |
| "grad_norm": 0.5100119277747269, | |
| "learning_rate": 3.358911305186351e-05, | |
| "loss": 0.1323, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1164846122264862, | |
| "step": 2400, | |
| "valid_targets_mean": 1774.4, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 2.357843137254902, | |
| "grad_norm": 0.4567649270223125, | |
| "learning_rate": 3.355320198386947e-05, | |
| "loss": 0.1166, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10715487599372864, | |
| "step": 2405, | |
| "valid_targets_mean": 2297.6, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 2.3627450980392157, | |
| "grad_norm": 0.4873938724406906, | |
| "learning_rate": 3.351720993172957e-05, | |
| "loss": 0.1272, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11813487857580185, | |
| "step": 2410, | |
| "valid_targets_mean": 2144.9, | |
| "valid_targets_min": 766 | |
| }, | |
| { | |
| "epoch": 2.3676470588235294, | |
| "grad_norm": 0.4469656890585901, | |
| "learning_rate": 3.3481137110506305e-05, | |
| "loss": 0.1219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10170838981866837, | |
| "step": 2415, | |
| "valid_targets_mean": 2247.9, | |
| "valid_targets_min": 689 | |
| }, | |
| { | |
| "epoch": 2.372549019607843, | |
| "grad_norm": 0.47619819728984353, | |
| "learning_rate": 3.344498373574482e-05, | |
| "loss": 0.1317, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14894942939281464, | |
| "step": 2420, | |
| "valid_targets_mean": 2214.6, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 2.377450980392157, | |
| "grad_norm": 0.5003725526992597, | |
| "learning_rate": 3.340875002347156e-05, | |
| "loss": 0.1368, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13846082985401154, | |
| "step": 2425, | |
| "valid_targets_mean": 2310.4, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 2.3823529411764706, | |
| "grad_norm": 0.5491326217204983, | |
| "learning_rate": 3.337243619019301e-05, | |
| "loss": 0.1288, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1536971479654312, | |
| "step": 2430, | |
| "valid_targets_mean": 1858.8, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 2.3872549019607843, | |
| "grad_norm": 0.5328613699348111, | |
| "learning_rate": 3.3336042452894425e-05, | |
| "loss": 0.1144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11894248425960541, | |
| "step": 2435, | |
| "valid_targets_mean": 1888.1, | |
| "valid_targets_min": 658 | |
| }, | |
| { | |
| "epoch": 2.392156862745098, | |
| "grad_norm": 0.5532164249157558, | |
| "learning_rate": 3.329956902903848e-05, | |
| "loss": 0.126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14720863103866577, | |
| "step": 2440, | |
| "valid_targets_mean": 2460.6, | |
| "valid_targets_min": 523 | |
| }, | |
| { | |
| "epoch": 2.3970588235294117, | |
| "grad_norm": 0.6305635545547962, | |
| "learning_rate": 3.326301613656402e-05, | |
| "loss": 0.1345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16992811858654022, | |
| "step": 2445, | |
| "valid_targets_mean": 2158.1, | |
| "valid_targets_min": 834 | |
| }, | |
| { | |
| "epoch": 2.4019607843137254, | |
| "grad_norm": 0.47367213497868, | |
| "learning_rate": 3.322638399388472e-05, | |
| "loss": 0.1365, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12942060828208923, | |
| "step": 2450, | |
| "valid_targets_mean": 2176.0, | |
| "valid_targets_min": 913 | |
| }, | |
| { | |
| "epoch": 2.406862745098039, | |
| "grad_norm": 0.5708011022431659, | |
| "learning_rate": 3.318967281988782e-05, | |
| "loss": 0.1406, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1652604639530182, | |
| "step": 2455, | |
| "valid_targets_mean": 2101.6, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 2.411764705882353, | |
| "grad_norm": 0.5099338224778304, | |
| "learning_rate": 3.3152882833932766e-05, | |
| "loss": 0.1178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.127591073513031, | |
| "step": 2460, | |
| "valid_targets_mean": 2128.4, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 2.4166666666666665, | |
| "grad_norm": 0.5003755514174549, | |
| "learning_rate": 3.311601425584996e-05, | |
| "loss": 0.138, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13985368609428406, | |
| "step": 2465, | |
| "valid_targets_mean": 2125.7, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 2.4215686274509802, | |
| "grad_norm": 0.5924950139777752, | |
| "learning_rate": 3.3079067305939366e-05, | |
| "loss": 0.1433, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14252334833145142, | |
| "step": 2470, | |
| "valid_targets_mean": 2268.9, | |
| "valid_targets_min": 1037 | |
| }, | |
| { | |
| "epoch": 2.426470588235294, | |
| "grad_norm": 0.5314224716211676, | |
| "learning_rate": 3.3042042204969296e-05, | |
| "loss": 0.1311, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14862433075904846, | |
| "step": 2475, | |
| "valid_targets_mean": 2261.7, | |
| "valid_targets_min": 560 | |
| }, | |
| { | |
| "epoch": 2.431372549019608, | |
| "grad_norm": 0.506747952151037, | |
| "learning_rate": 3.3004939174175e-05, | |
| "loss": 0.1357, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.139796644449234, | |
| "step": 2480, | |
| "valid_targets_mean": 2169.8, | |
| "valid_targets_min": 735 | |
| }, | |
| { | |
| "epoch": 2.436274509803922, | |
| "grad_norm": 0.5891764902816398, | |
| "learning_rate": 3.296775843525739e-05, | |
| "loss": 0.127, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12421401590108871, | |
| "step": 2485, | |
| "valid_targets_mean": 1692.9, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 2.4411764705882355, | |
| "grad_norm": 0.39384181544170843, | |
| "learning_rate": 3.293050021038171e-05, | |
| "loss": 0.1309, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12514618039131165, | |
| "step": 2490, | |
| "valid_targets_mean": 2835.9, | |
| "valid_targets_min": 890 | |
| }, | |
| { | |
| "epoch": 2.446078431372549, | |
| "grad_norm": 0.5466808342373171, | |
| "learning_rate": 3.289316472217621e-05, | |
| "loss": 0.1345, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15848739445209503, | |
| "step": 2495, | |
| "valid_targets_mean": 1778.9, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 2.450980392156863, | |
| "grad_norm": 0.4536221374191153, | |
| "learning_rate": 3.285575219373079e-05, | |
| "loss": 0.1279, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11861887574195862, | |
| "step": 2500, | |
| "valid_targets_mean": 2401.6, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 2.4558823529411766, | |
| "grad_norm": 0.39556844985138695, | |
| "learning_rate": 3.2818262848595696e-05, | |
| "loss": 0.1168, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09962082654237747, | |
| "step": 2505, | |
| "valid_targets_mean": 2596.6, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 2.4607843137254903, | |
| "grad_norm": 0.5563691562257486, | |
| "learning_rate": 3.2780696910780185e-05, | |
| "loss": 0.1209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14485809206962585, | |
| "step": 2510, | |
| "valid_targets_mean": 1906.2, | |
| "valid_targets_min": 696 | |
| }, | |
| { | |
| "epoch": 2.465686274509804, | |
| "grad_norm": 0.48374305437067666, | |
| "learning_rate": 3.274305460475116e-05, | |
| "loss": 0.1303, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1370125412940979, | |
| "step": 2515, | |
| "valid_targets_mean": 2272.5, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 2.4705882352941178, | |
| "grad_norm": 0.4042482270802887, | |
| "learning_rate": 3.270533615543186e-05, | |
| "loss": 0.1577, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11950810253620148, | |
| "step": 2520, | |
| "valid_targets_mean": 2691.9, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 2.4754901960784315, | |
| "grad_norm": 0.7546679032370863, | |
| "learning_rate": 3.2667541788200496e-05, | |
| "loss": 0.1359, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11613636463880539, | |
| "step": 2525, | |
| "valid_targets_mean": 2143.2, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 2.480392156862745, | |
| "grad_norm": 0.41794384750016256, | |
| "learning_rate": 3.26296717288889e-05, | |
| "loss": 0.136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1214665025472641, | |
| "step": 2530, | |
| "valid_targets_mean": 2751.1, | |
| "valid_targets_min": 1144 | |
| }, | |
| { | |
| "epoch": 2.485294117647059, | |
| "grad_norm": 0.5553263208758342, | |
| "learning_rate": 3.2591726203781205e-05, | |
| "loss": 0.1177, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13100360333919525, | |
| "step": 2535, | |
| "valid_targets_mean": 2431.9, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 2.4901960784313726, | |
| "grad_norm": 0.47914372549026196, | |
| "learning_rate": 3.255370543961245e-05, | |
| "loss": 0.1336, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1484772264957428, | |
| "step": 2540, | |
| "valid_targets_mean": 2592.8, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 2.4950980392156863, | |
| "grad_norm": 0.4662138254088204, | |
| "learning_rate": 3.2515609663567266e-05, | |
| "loss": 0.1361, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.16803815960884094, | |
| "step": 2545, | |
| "valid_targets_mean": 2494.6, | |
| "valid_targets_min": 633 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "grad_norm": 0.3930868423643063, | |
| "learning_rate": 3.247743910327849e-05, | |
| "loss": 0.1211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11918853223323822, | |
| "step": 2550, | |
| "valid_targets_mean": 2764.6, | |
| "valid_targets_min": 1225 | |
| }, | |
| { | |
| "epoch": 2.5049019607843137, | |
| "grad_norm": 0.5290514066877043, | |
| "learning_rate": 3.2439193986825835e-05, | |
| "loss": 0.1289, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1352325975894928, | |
| "step": 2555, | |
| "valid_targets_mean": 2115.9, | |
| "valid_targets_min": 519 | |
| }, | |
| { | |
| "epoch": 2.5098039215686274, | |
| "grad_norm": 0.46854808906106793, | |
| "learning_rate": 3.240087454273448e-05, | |
| "loss": 0.1295, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12803807854652405, | |
| "step": 2560, | |
| "valid_targets_mean": 2416.5, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 2.514705882352941, | |
| "grad_norm": 0.43005809391608746, | |
| "learning_rate": 3.2362480999973744e-05, | |
| "loss": 0.1274, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1205533966422081, | |
| "step": 2565, | |
| "valid_targets_mean": 2505.1, | |
| "valid_targets_min": 1080 | |
| }, | |
| { | |
| "epoch": 2.519607843137255, | |
| "grad_norm": 0.547921133473076, | |
| "learning_rate": 3.2324013587955716e-05, | |
| "loss": 0.1248, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1539292335510254, | |
| "step": 2570, | |
| "valid_targets_mean": 2068.6, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 2.5245098039215685, | |
| "grad_norm": 0.48177977616801043, | |
| "learning_rate": 3.228547253653386e-05, | |
| "loss": 0.1334, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13383543491363525, | |
| "step": 2575, | |
| "valid_targets_mean": 2834.5, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 2.5294117647058822, | |
| "grad_norm": 0.4326265292915774, | |
| "learning_rate": 3.2246858076001674e-05, | |
| "loss": 0.1287, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12128859758377075, | |
| "step": 2580, | |
| "valid_targets_mean": 2535.4, | |
| "valid_targets_min": 986 | |
| }, | |
| { | |
| "epoch": 2.534313725490196, | |
| "grad_norm": 0.5318647060540064, | |
| "learning_rate": 3.2208170437091267e-05, | |
| "loss": 0.1199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12436534464359283, | |
| "step": 2585, | |
| "valid_targets_mean": 2196.9, | |
| "valid_targets_min": 701 | |
| }, | |
| { | |
| "epoch": 2.5392156862745097, | |
| "grad_norm": 0.4846752026333883, | |
| "learning_rate": 3.216940985097203e-05, | |
| "loss": 0.1223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1433335542678833, | |
| "step": 2590, | |
| "valid_targets_mean": 2433.1, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 2.5441176470588234, | |
| "grad_norm": 0.4239896532044266, | |
| "learning_rate": 3.213057654924924e-05, | |
| "loss": 0.1257, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11380953341722488, | |
| "step": 2595, | |
| "valid_targets_mean": 2302.8, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 2.549019607843137, | |
| "grad_norm": 0.44078999516335776, | |
| "learning_rate": 3.2091670763962645e-05, | |
| "loss": 0.127, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12046624720096588, | |
| "step": 2600, | |
| "valid_targets_mean": 2539.6, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 2.553921568627451, | |
| "grad_norm": 0.5939898023952678, | |
| "learning_rate": 3.205269272758513e-05, | |
| "loss": 0.1252, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11827564239501953, | |
| "step": 2605, | |
| "valid_targets_mean": 2455.4, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 2.5588235294117645, | |
| "grad_norm": 0.5448767219249958, | |
| "learning_rate": 3.201364267302128e-05, | |
| "loss": 0.1242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13409531116485596, | |
| "step": 2610, | |
| "valid_targets_mean": 1850.7, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 2.563725490196078, | |
| "grad_norm": 0.47561028426644697, | |
| "learning_rate": 3.1974520833606016e-05, | |
| "loss": 0.1222, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11841538548469543, | |
| "step": 2615, | |
| "valid_targets_mean": 2128.8, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 2.568627450980392, | |
| "grad_norm": 0.5378413990458727, | |
| "learning_rate": 3.1935327443103195e-05, | |
| "loss": 0.1182, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12730035185813904, | |
| "step": 2620, | |
| "valid_targets_mean": 1888.8, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 2.5735294117647056, | |
| "grad_norm": 0.4922033292560799, | |
| "learning_rate": 3.189606273570422e-05, | |
| "loss": 0.1395, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13984598219394684, | |
| "step": 2625, | |
| "valid_targets_mean": 2094.8, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 2.5784313725490198, | |
| "grad_norm": 0.6332583070833103, | |
| "learning_rate": 3.18567269460266e-05, | |
| "loss": 0.1218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1428709626197815, | |
| "step": 2630, | |
| "valid_targets_mean": 1739.6, | |
| "valid_targets_min": 843 | |
| }, | |
| { | |
| "epoch": 2.5833333333333335, | |
| "grad_norm": 0.5513161663336141, | |
| "learning_rate": 3.1817320309112625e-05, | |
| "loss": 0.1401, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14765672385692596, | |
| "step": 2635, | |
| "valid_targets_mean": 1715.8, | |
| "valid_targets_min": 795 | |
| }, | |
| { | |
| "epoch": 2.588235294117647, | |
| "grad_norm": 0.6440133895872631, | |
| "learning_rate": 3.1777843060427876e-05, | |
| "loss": 0.1301, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15483424067497253, | |
| "step": 2640, | |
| "valid_targets_mean": 1774.8, | |
| "valid_targets_min": 514 | |
| }, | |
| { | |
| "epoch": 2.593137254901961, | |
| "grad_norm": 0.5336887900173594, | |
| "learning_rate": 3.173829543585989e-05, | |
| "loss": 0.1515, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.20561406016349792, | |
| "step": 2645, | |
| "valid_targets_mean": 2104.5, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 2.5980392156862746, | |
| "grad_norm": 0.4490637951442281, | |
| "learning_rate": 3.1698677671716716e-05, | |
| "loss": 0.1176, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1061968132853508, | |
| "step": 2650, | |
| "valid_targets_mean": 1995.6, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 2.6029411764705883, | |
| "grad_norm": 0.502444501852284, | |
| "learning_rate": 3.1658990004725464e-05, | |
| "loss": 0.136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11915481835603714, | |
| "step": 2655, | |
| "valid_targets_mean": 2343.6, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 2.607843137254902, | |
| "grad_norm": 0.5390839949044163, | |
| "learning_rate": 3.1619232672031e-05, | |
| "loss": 0.1382, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1582946479320526, | |
| "step": 2660, | |
| "valid_targets_mean": 2198.4, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 2.6127450980392157, | |
| "grad_norm": 0.5051852498871383, | |
| "learning_rate": 3.157940591119442e-05, | |
| "loss": 0.144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15259839594364166, | |
| "step": 2665, | |
| "valid_targets_mean": 1985.2, | |
| "valid_targets_min": 784 | |
| }, | |
| { | |
| "epoch": 2.6176470588235294, | |
| "grad_norm": 0.49993466913907697, | |
| "learning_rate": 3.153950996019166e-05, | |
| "loss": 0.1162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12850934267044067, | |
| "step": 2670, | |
| "valid_targets_mean": 2113.2, | |
| "valid_targets_min": 526 | |
| }, | |
| { | |
| "epoch": 2.622549019607843, | |
| "grad_norm": 0.5137848708127476, | |
| "learning_rate": 3.149954505741214e-05, | |
| "loss": 0.1219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1415955126285553, | |
| "step": 2675, | |
| "valid_targets_mean": 1886.8, | |
| "valid_targets_min": 806 | |
| }, | |
| { | |
| "epoch": 2.627450980392157, | |
| "grad_norm": 0.49292793100332416, | |
| "learning_rate": 3.1459511441657225e-05, | |
| "loss": 0.1382, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13631919026374817, | |
| "step": 2680, | |
| "valid_targets_mean": 2043.3, | |
| "valid_targets_min": 688 | |
| }, | |
| { | |
| "epoch": 2.6323529411764706, | |
| "grad_norm": 0.5216681230311012, | |
| "learning_rate": 3.141940935213891e-05, | |
| "loss": 0.1315, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12248043715953827, | |
| "step": 2685, | |
| "valid_targets_mean": 2078.9, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 2.6372549019607843, | |
| "grad_norm": 0.4707375189641917, | |
| "learning_rate": 3.13792390284783e-05, | |
| "loss": 0.1354, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13096097111701965, | |
| "step": 2690, | |
| "valid_targets_mean": 2617.3, | |
| "valid_targets_min": 825 | |
| }, | |
| { | |
| "epoch": 2.642156862745098, | |
| "grad_norm": 0.5264160275079668, | |
| "learning_rate": 3.1339000710704234e-05, | |
| "loss": 0.1327, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1339496374130249, | |
| "step": 2695, | |
| "valid_targets_mean": 1860.3, | |
| "valid_targets_min": 691 | |
| }, | |
| { | |
| "epoch": 2.6470588235294117, | |
| "grad_norm": 1.8516330663101164, | |
| "learning_rate": 3.129869463925185e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.128217414021492, | |
| "step": 2700, | |
| "valid_targets_mean": 2713.3, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 2.6519607843137254, | |
| "grad_norm": 0.5964001445350231, | |
| "learning_rate": 3.125832105496112e-05, | |
| "loss": 0.1194, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14027449488639832, | |
| "step": 2705, | |
| "valid_targets_mean": 1682.3, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 2.656862745098039, | |
| "grad_norm": 0.46865445803340855, | |
| "learning_rate": 3.12178801990754e-05, | |
| "loss": 0.133, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15145006775856018, | |
| "step": 2710, | |
| "valid_targets_mean": 2126.5, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 2.661764705882353, | |
| "grad_norm": 0.5266520204218963, | |
| "learning_rate": 3.1177372313240056e-05, | |
| "loss": 0.1408, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14159266650676727, | |
| "step": 2715, | |
| "valid_targets_mean": 2026.8, | |
| "valid_targets_min": 668 | |
| }, | |
| { | |
| "epoch": 2.6666666666666665, | |
| "grad_norm": 0.47431158065227663, | |
| "learning_rate": 3.1136797639500946e-05, | |
| "loss": 0.1327, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1264529824256897, | |
| "step": 2720, | |
| "valid_targets_mean": 2156.2, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 2.6715686274509802, | |
| "grad_norm": 0.46172333007882593, | |
| "learning_rate": 3.109615642030302e-05, | |
| "loss": 0.1186, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12161999940872192, | |
| "step": 2725, | |
| "valid_targets_mean": 2365.8, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 2.6764705882352944, | |
| "grad_norm": 0.5417612997694451, | |
| "learning_rate": 3.105544889848885e-05, | |
| "loss": 0.1422, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15983346104621887, | |
| "step": 2730, | |
| "valid_targets_mean": 1985.2, | |
| "valid_targets_min": 458 | |
| }, | |
| { | |
| "epoch": 2.681372549019608, | |
| "grad_norm": 0.5602903369720046, | |
| "learning_rate": 3.101467531729718e-05, | |
| "loss": 0.1274, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1274542212486267, | |
| "step": 2735, | |
| "valid_targets_mean": 1546.2, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 2.686274509803922, | |
| "grad_norm": 0.4499642056569709, | |
| "learning_rate": 3.0973835920361485e-05, | |
| "loss": 0.1319, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12221525609493256, | |
| "step": 2740, | |
| "valid_targets_mean": 2619.2, | |
| "valid_targets_min": 923 | |
| }, | |
| { | |
| "epoch": 2.6911764705882355, | |
| "grad_norm": 0.49926621820934225, | |
| "learning_rate": 3.093293095170849e-05, | |
| "loss": 0.1272, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14079585671424866, | |
| "step": 2745, | |
| "valid_targets_mean": 1941.2, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 2.696078431372549, | |
| "grad_norm": 0.47853732191376075, | |
| "learning_rate": 3.089196065575676e-05, | |
| "loss": 0.1178, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13084784150123596, | |
| "step": 2750, | |
| "valid_targets_mean": 2173.8, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 2.700980392156863, | |
| "grad_norm": 0.4604079974584735, | |
| "learning_rate": 3.0850925277315193e-05, | |
| "loss": 0.1236, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1272059977054596, | |
| "step": 2755, | |
| "valid_targets_mean": 2368.0, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 2.7058823529411766, | |
| "grad_norm": 0.5077703112679872, | |
| "learning_rate": 3.080982506158155e-05, | |
| "loss": 0.1288, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12848421931266785, | |
| "step": 2760, | |
| "valid_targets_mean": 1888.6, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 2.7107843137254903, | |
| "grad_norm": 0.4333748156398904, | |
| "learning_rate": 3.076866025414107e-05, | |
| "loss": 0.1163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10613423585891724, | |
| "step": 2765, | |
| "valid_targets_mean": 2228.8, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 2.715686274509804, | |
| "grad_norm": 0.5617367615142488, | |
| "learning_rate": 3.072743110096487e-05, | |
| "loss": 0.136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11530651152133942, | |
| "step": 2770, | |
| "valid_targets_mean": 2036.8, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 2.7205882352941178, | |
| "grad_norm": 0.5711011470753721, | |
| "learning_rate": 3.0686137848408616e-05, | |
| "loss": 0.1384, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14303429424762726, | |
| "step": 2775, | |
| "valid_targets_mean": 1693.8, | |
| "valid_targets_min": 324 | |
| }, | |
| { | |
| "epoch": 2.7254901960784315, | |
| "grad_norm": 0.5488360852479225, | |
| "learning_rate": 3.064478074321095e-05, | |
| "loss": 0.1452, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1471487134695053, | |
| "step": 2780, | |
| "valid_targets_mean": 1827.1, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 2.730392156862745, | |
| "grad_norm": 0.5149112512414382, | |
| "learning_rate": 3.060336003249206e-05, | |
| "loss": 0.137, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12844215333461761, | |
| "step": 2785, | |
| "valid_targets_mean": 1996.3, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 2.735294117647059, | |
| "grad_norm": 0.5573426653011521, | |
| "learning_rate": 3.056187596375221e-05, | |
| "loss": 0.1271, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1358461081981659, | |
| "step": 2790, | |
| "valid_targets_mean": 1847.9, | |
| "valid_targets_min": 781 | |
| }, | |
| { | |
| "epoch": 2.7401960784313726, | |
| "grad_norm": 0.4614772672771677, | |
| "learning_rate": 3.052032878487022e-05, | |
| "loss": 0.1298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13471747934818268, | |
| "step": 2795, | |
| "valid_targets_mean": 2591.0, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 2.7450980392156863, | |
| "grad_norm": 0.5562205728473466, | |
| "learning_rate": 3.047871874410201e-05, | |
| "loss": 0.1307, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13988906145095825, | |
| "step": 2800, | |
| "valid_targets_mean": 1747.5, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "grad_norm": 0.5350979428814614, | |
| "learning_rate": 3.0437046090079165e-05, | |
| "loss": 0.1219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12228747457265854, | |
| "step": 2805, | |
| "valid_targets_mean": 1716.3, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 2.7549019607843137, | |
| "grad_norm": 0.5068335768270578, | |
| "learning_rate": 3.039531107180733e-05, | |
| "loss": 0.1274, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12163315713405609, | |
| "step": 2810, | |
| "valid_targets_mean": 2055.1, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 2.7598039215686274, | |
| "grad_norm": 0.4353854673960122, | |
| "learning_rate": 3.0353513938664855e-05, | |
| "loss": 0.1309, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1248159259557724, | |
| "step": 2815, | |
| "valid_targets_mean": 2709.1, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 2.764705882352941, | |
| "grad_norm": 0.5757889526569985, | |
| "learning_rate": 3.03116549404012e-05, | |
| "loss": 0.1276, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12588320672512054, | |
| "step": 2820, | |
| "valid_targets_mean": 1643.4, | |
| "valid_targets_min": 665 | |
| }, | |
| { | |
| "epoch": 2.769607843137255, | |
| "grad_norm": 0.4571194123253969, | |
| "learning_rate": 3.0269734327135514e-05, | |
| "loss": 0.1571, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1337161362171173, | |
| "step": 2825, | |
| "valid_targets_mean": 2302.4, | |
| "valid_targets_min": 970 | |
| }, | |
| { | |
| "epoch": 2.7745098039215685, | |
| "grad_norm": 0.452298983086843, | |
| "learning_rate": 3.0227752349355105e-05, | |
| "loss": 0.1352, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11990043520927429, | |
| "step": 2830, | |
| "valid_targets_mean": 2196.6, | |
| "valid_targets_min": 702 | |
| }, | |
| { | |
| "epoch": 2.7794117647058822, | |
| "grad_norm": 0.5137398992266271, | |
| "learning_rate": 3.018570925791395e-05, | |
| "loss": 0.1223, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12331690639257431, | |
| "step": 2835, | |
| "valid_targets_mean": 1878.5, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 2.784313725490196, | |
| "grad_norm": 0.47457896935510097, | |
| "learning_rate": 3.0143605304031185e-05, | |
| "loss": 0.1212, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12229996919631958, | |
| "step": 2840, | |
| "valid_targets_mean": 2376.8, | |
| "valid_targets_min": 758 | |
| }, | |
| { | |
| "epoch": 2.7892156862745097, | |
| "grad_norm": 0.45690022114892714, | |
| "learning_rate": 3.0101440739289644e-05, | |
| "loss": 0.1185, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12366802990436554, | |
| "step": 2845, | |
| "valid_targets_mean": 2346.6, | |
| "valid_targets_min": 608 | |
| }, | |
| { | |
| "epoch": 2.7941176470588234, | |
| "grad_norm": 0.42759880074470524, | |
| "learning_rate": 3.00592158156343e-05, | |
| "loss": 0.1242, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10269932448863983, | |
| "step": 2850, | |
| "valid_targets_mean": 2301.4, | |
| "valid_targets_min": 839 | |
| }, | |
| { | |
| "epoch": 2.799019607843137, | |
| "grad_norm": 0.4304115392133804, | |
| "learning_rate": 3.00169307853708e-05, | |
| "loss": 0.1339, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11028525233268738, | |
| "step": 2855, | |
| "valid_targets_mean": 2272.4, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 2.803921568627451, | |
| "grad_norm": 0.4605724095449812, | |
| "learning_rate": 2.9974585901163955e-05, | |
| "loss": 0.1268, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12313034385442734, | |
| "step": 2860, | |
| "valid_targets_mean": 2610.8, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 2.8088235294117645, | |
| "grad_norm": 0.5421500878681277, | |
| "learning_rate": 2.9932181416036186e-05, | |
| "loss": 0.1313, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15357518196105957, | |
| "step": 2865, | |
| "valid_targets_mean": 2295.1, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 2.813725490196078, | |
| "grad_norm": 0.4288979220922448, | |
| "learning_rate": 2.988971758336609e-05, | |
| "loss": 0.1446, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10890544950962067, | |
| "step": 2870, | |
| "valid_targets_mean": 2487.8, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 2.818627450980392, | |
| "grad_norm": 0.5839474906818637, | |
| "learning_rate": 2.9847194656886846e-05, | |
| "loss": 0.1245, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1326325386762619, | |
| "step": 2875, | |
| "valid_targets_mean": 1586.2, | |
| "valid_targets_min": 523 | |
| }, | |
| { | |
| "epoch": 2.8235294117647056, | |
| "grad_norm": 0.3943595738679273, | |
| "learning_rate": 2.980461289068475e-05, | |
| "loss": 0.1172, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10610520839691162, | |
| "step": 2880, | |
| "valid_targets_mean": 2857.4, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 2.8284313725490198, | |
| "grad_norm": 0.3995014892923764, | |
| "learning_rate": 2.97619725391977e-05, | |
| "loss": 0.1304, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10689753293991089, | |
| "step": 2885, | |
| "valid_targets_mean": 2595.2, | |
| "valid_targets_min": 917 | |
| }, | |
| { | |
| "epoch": 2.8333333333333335, | |
| "grad_norm": 0.4339660498932304, | |
| "learning_rate": 2.9719273857213607e-05, | |
| "loss": 0.1405, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11175128817558289, | |
| "step": 2890, | |
| "valid_targets_mean": 2714.3, | |
| "valid_targets_min": 524 | |
| }, | |
| { | |
| "epoch": 2.838235294117647, | |
| "grad_norm": 0.4824549484105122, | |
| "learning_rate": 2.967651709986898e-05, | |
| "loss": 0.1396, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1368466168642044, | |
| "step": 2895, | |
| "valid_targets_mean": 2335.1, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 2.843137254901961, | |
| "grad_norm": 0.49320253448063933, | |
| "learning_rate": 2.9633702522647314e-05, | |
| "loss": 0.1218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1290884166955948, | |
| "step": 2900, | |
| "valid_targets_mean": 2013.7, | |
| "valid_targets_min": 495 | |
| }, | |
| { | |
| "epoch": 2.8480392156862746, | |
| "grad_norm": 0.4512954249823614, | |
| "learning_rate": 2.959083038137758e-05, | |
| "loss": 0.1309, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11821623891592026, | |
| "step": 2905, | |
| "valid_targets_mean": 2283.9, | |
| "valid_targets_min": 582 | |
| }, | |
| { | |
| "epoch": 2.8529411764705883, | |
| "grad_norm": 0.48511395512197114, | |
| "learning_rate": 2.9547900932232755e-05, | |
| "loss": 0.126, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12425636500120163, | |
| "step": 2910, | |
| "valid_targets_mean": 2048.6, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 2.857843137254902, | |
| "grad_norm": 0.5655839453858921, | |
| "learning_rate": 2.95049144317282e-05, | |
| "loss": 0.1413, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12589307129383087, | |
| "step": 2915, | |
| "valid_targets_mean": 1768.5, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 2.8627450980392157, | |
| "grad_norm": 0.4545048327238555, | |
| "learning_rate": 2.9461871136720205e-05, | |
| "loss": 0.1266, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13439792394638062, | |
| "step": 2920, | |
| "valid_targets_mean": 2661.6, | |
| "valid_targets_min": 580 | |
| }, | |
| { | |
| "epoch": 2.8676470588235294, | |
| "grad_norm": 0.4197095410493635, | |
| "learning_rate": 2.9418771304404408e-05, | |
| "loss": 0.1151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11430007219314575, | |
| "step": 2925, | |
| "valid_targets_mean": 2748.2, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 2.872549019607843, | |
| "grad_norm": 0.4762686149288785, | |
| "learning_rate": 2.9375615192314272e-05, | |
| "loss": 0.1319, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13634362816810608, | |
| "step": 2930, | |
| "valid_targets_mean": 2192.4, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 2.877450980392157, | |
| "grad_norm": 0.6573835179007171, | |
| "learning_rate": 2.9332403058319562e-05, | |
| "loss": 0.1423, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.17557133734226227, | |
| "step": 2935, | |
| "valid_targets_mean": 1823.9, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 2.8823529411764706, | |
| "grad_norm": 0.5697771049283034, | |
| "learning_rate": 2.928913516062477e-05, | |
| "loss": 0.1288, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13310866057872772, | |
| "step": 2940, | |
| "valid_targets_mean": 1937.6, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 2.8872549019607843, | |
| "grad_norm": 0.4815531389610093, | |
| "learning_rate": 2.9245811757767605e-05, | |
| "loss": 0.1249, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10856365412473679, | |
| "step": 2945, | |
| "valid_targets_mean": 2188.1, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 2.892156862745098, | |
| "grad_norm": 0.3525092757314404, | |
| "learning_rate": 2.920243310861743e-05, | |
| "loss": 0.1105, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08762288838624954, | |
| "step": 2950, | |
| "valid_targets_mean": 2877.1, | |
| "valid_targets_min": 633 | |
| }, | |
| { | |
| "epoch": 2.8970588235294117, | |
| "grad_norm": 0.44362286093916375, | |
| "learning_rate": 2.9158999472373727e-05, | |
| "loss": 0.1263, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1400848627090454, | |
| "step": 2955, | |
| "valid_targets_mean": 2650.9, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 2.9019607843137254, | |
| "grad_norm": 0.466533540036814, | |
| "learning_rate": 2.911551110856453e-05, | |
| "loss": 0.128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12361754477024078, | |
| "step": 2960, | |
| "valid_targets_mean": 2421.5, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 2.906862745098039, | |
| "grad_norm": 0.4275932951009806, | |
| "learning_rate": 2.9071968277044885e-05, | |
| "loss": 0.1325, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11012598872184753, | |
| "step": 2965, | |
| "valid_targets_mean": 2530.9, | |
| "valid_targets_min": 696 | |
| }, | |
| { | |
| "epoch": 2.911764705882353, | |
| "grad_norm": 0.47645750491140126, | |
| "learning_rate": 2.9028371237995315e-05, | |
| "loss": 0.1274, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11943289637565613, | |
| "step": 2970, | |
| "valid_targets_mean": 2157.4, | |
| "valid_targets_min": 787 | |
| }, | |
| { | |
| "epoch": 2.9166666666666665, | |
| "grad_norm": 0.5751641720681621, | |
| "learning_rate": 2.8984720251920228e-05, | |
| "loss": 0.1259, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.134115070104599, | |
| "step": 2975, | |
| "valid_targets_mean": 2491.7, | |
| "valid_targets_min": 855 | |
| }, | |
| { | |
| "epoch": 2.9215686274509802, | |
| "grad_norm": 0.5048594686042459, | |
| "learning_rate": 2.8941015579646388e-05, | |
| "loss": 0.1326, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1326657235622406, | |
| "step": 2980, | |
| "valid_targets_mean": 2276.6, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 2.9264705882352944, | |
| "grad_norm": 0.5218701473555022, | |
| "learning_rate": 2.8897257482321354e-05, | |
| "loss": 0.1256, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1356755495071411, | |
| "step": 2985, | |
| "valid_targets_mean": 2125.2, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 2.931372549019608, | |
| "grad_norm": 0.5719581951550526, | |
| "learning_rate": 2.88534462214119e-05, | |
| "loss": 0.1343, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13416975736618042, | |
| "step": 2990, | |
| "valid_targets_mean": 1582.3, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 2.936274509803922, | |
| "grad_norm": 0.4676682934177474, | |
| "learning_rate": 2.8809582058702497e-05, | |
| "loss": 0.129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13310152292251587, | |
| "step": 2995, | |
| "valid_targets_mean": 2478.7, | |
| "valid_targets_min": 738 | |
| }, | |
| { | |
| "epoch": 2.9411764705882355, | |
| "grad_norm": 0.5153112474110181, | |
| "learning_rate": 2.8765665256293683e-05, | |
| "loss": 0.121, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12872923910617828, | |
| "step": 3000, | |
| "valid_targets_mean": 2084.5, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 2.946078431372549, | |
| "grad_norm": 0.4820570169126928, | |
| "learning_rate": 2.8721696076600553e-05, | |
| "loss": 0.1338, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1477794647216797, | |
| "step": 3005, | |
| "valid_targets_mean": 2277.9, | |
| "valid_targets_min": 606 | |
| }, | |
| { | |
| "epoch": 2.950980392156863, | |
| "grad_norm": 0.5316651065874668, | |
| "learning_rate": 2.8677674782351164e-05, | |
| "loss": 0.1244, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13892558217048645, | |
| "step": 3010, | |
| "valid_targets_mean": 2058.4, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 2.9558823529411766, | |
| "grad_norm": 0.47285568304817926, | |
| "learning_rate": 2.863360163658498e-05, | |
| "loss": 0.1277, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10545437037944794, | |
| "step": 3015, | |
| "valid_targets_mean": 2418.2, | |
| "valid_targets_min": 1094 | |
| }, | |
| { | |
| "epoch": 2.9607843137254903, | |
| "grad_norm": 0.5576570344201479, | |
| "learning_rate": 2.8589476902651284e-05, | |
| "loss": 0.1265, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14480185508728027, | |
| "step": 3020, | |
| "valid_targets_mean": 2428.8, | |
| "valid_targets_min": 932 | |
| }, | |
| { | |
| "epoch": 2.965686274509804, | |
| "grad_norm": 0.3720697472883488, | |
| "learning_rate": 2.854530084420762e-05, | |
| "loss": 0.1085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0960945338010788, | |
| "step": 3025, | |
| "valid_targets_mean": 2547.1, | |
| "valid_targets_min": 516 | |
| }, | |
| { | |
| "epoch": 2.9705882352941178, | |
| "grad_norm": 0.3715873526257619, | |
| "learning_rate": 2.85010737252182e-05, | |
| "loss": 0.1091, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10219788551330566, | |
| "step": 3030, | |
| "valid_targets_mean": 2670.9, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 2.9754901960784315, | |
| "grad_norm": 0.6195399543085092, | |
| "learning_rate": 2.845679580995235e-05, | |
| "loss": 0.1561, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1750684678554535, | |
| "step": 3035, | |
| "valid_targets_mean": 1722.1, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 2.980392156862745, | |
| "grad_norm": 0.44368210641722294, | |
| "learning_rate": 2.84124673629829e-05, | |
| "loss": 0.1174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12278378009796143, | |
| "step": 3040, | |
| "valid_targets_mean": 2512.6, | |
| "valid_targets_min": 844 | |
| }, | |
| { | |
| "epoch": 2.985294117647059, | |
| "grad_norm": 0.5160554435392198, | |
| "learning_rate": 2.836808864918464e-05, | |
| "loss": 0.1325, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1273147016763687, | |
| "step": 3045, | |
| "valid_targets_mean": 2230.9, | |
| "valid_targets_min": 893 | |
| }, | |
| { | |
| "epoch": 2.9901960784313726, | |
| "grad_norm": 0.5018491404792775, | |
| "learning_rate": 2.8323659933732704e-05, | |
| "loss": 0.147, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14495849609375, | |
| "step": 3050, | |
| "valid_targets_mean": 2451.2, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 2.9950980392156863, | |
| "grad_norm": 0.47228194271588314, | |
| "learning_rate": 2.827918148210101e-05, | |
| "loss": 0.1219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12136131525039673, | |
| "step": 3055, | |
| "valid_targets_mean": 2365.8, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.41747992787256166, | |
| "learning_rate": 2.823465356006066e-05, | |
| "loss": 0.1198, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0959467664361, | |
| "step": 3060, | |
| "valid_targets_mean": 2515.4, | |
| "valid_targets_min": 653 | |
| }, | |
| { | |
| "epoch": 3.0049019607843137, | |
| "grad_norm": 0.45268172105117166, | |
| "learning_rate": 2.819007643367835e-05, | |
| "loss": 0.1141, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.107194684445858, | |
| "step": 3065, | |
| "valid_targets_mean": 2481.1, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 3.0098039215686274, | |
| "grad_norm": 0.48782268263068046, | |
| "learning_rate": 2.814545036931479e-05, | |
| "loss": 0.1128, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10597293078899384, | |
| "step": 3070, | |
| "valid_targets_mean": 2368.9, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 3.014705882352941, | |
| "grad_norm": 0.42013415537790194, | |
| "learning_rate": 2.810077563362312e-05, | |
| "loss": 0.0953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09026814997196198, | |
| "step": 3075, | |
| "valid_targets_mean": 2636.4, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 3.019607843137255, | |
| "grad_norm": 0.493550574092022, | |
| "learning_rate": 2.8056052493547277e-05, | |
| "loss": 0.1202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12190809100866318, | |
| "step": 3080, | |
| "valid_targets_mean": 2430.9, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 3.0245098039215685, | |
| "grad_norm": 0.5881502044936036, | |
| "learning_rate": 2.8011281216320446e-05, | |
| "loss": 0.1074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11721500754356384, | |
| "step": 3085, | |
| "valid_targets_mean": 1814.4, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 3.0294117647058822, | |
| "grad_norm": 0.5220105505984863, | |
| "learning_rate": 2.7966462069463454e-05, | |
| "loss": 0.1181, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11271218955516815, | |
| "step": 3090, | |
| "valid_targets_mean": 2519.4, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 3.034313725490196, | |
| "grad_norm": 0.6222515663568035, | |
| "learning_rate": 2.792159532078314e-05, | |
| "loss": 0.1151, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13791202008724213, | |
| "step": 3095, | |
| "valid_targets_mean": 1671.6, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 3.0392156862745097, | |
| "grad_norm": 0.5465811429304228, | |
| "learning_rate": 2.787668123837079e-05, | |
| "loss": 0.1144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12078981846570969, | |
| "step": 3100, | |
| "valid_targets_mean": 1913.5, | |
| "valid_targets_min": 758 | |
| }, | |
| { | |
| "epoch": 3.0441176470588234, | |
| "grad_norm": 0.5118828378244067, | |
| "learning_rate": 2.7831720090600526e-05, | |
| "loss": 0.111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11368224769830704, | |
| "step": 3105, | |
| "valid_targets_mean": 2430.1, | |
| "valid_targets_min": 974 | |
| }, | |
| { | |
| "epoch": 3.049019607843137, | |
| "grad_norm": 0.49206966044764583, | |
| "learning_rate": 2.778671214612768e-05, | |
| "loss": 0.1102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1097177192568779, | |
| "step": 3110, | |
| "valid_targets_mean": 2724.2, | |
| "valid_targets_min": 878 | |
| }, | |
| { | |
| "epoch": 3.053921568627451, | |
| "grad_norm": 0.542050378970244, | |
| "learning_rate": 2.7741657673887234e-05, | |
| "loss": 0.1079, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11170881241559982, | |
| "step": 3115, | |
| "valid_targets_mean": 1963.5, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 3.0588235294117645, | |
| "grad_norm": 0.4966603322376902, | |
| "learning_rate": 2.7696556943092163e-05, | |
| "loss": 0.1202, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10189454257488251, | |
| "step": 3120, | |
| "valid_targets_mean": 2127.2, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 3.063725490196078, | |
| "grad_norm": 0.4162238126926009, | |
| "learning_rate": 2.7651410223231855e-05, | |
| "loss": 0.1218, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09815062582492828, | |
| "step": 3125, | |
| "valid_targets_mean": 2666.4, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 3.0686274509803924, | |
| "grad_norm": 0.4999830676990576, | |
| "learning_rate": 2.7606217784070518e-05, | |
| "loss": 0.1132, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10285164415836334, | |
| "step": 3130, | |
| "valid_targets_mean": 2044.8, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 3.073529411764706, | |
| "grad_norm": 0.45664697329434784, | |
| "learning_rate": 2.7560979895645508e-05, | |
| "loss": 0.1174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1548200249671936, | |
| "step": 3135, | |
| "valid_targets_mean": 2774.8, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 3.0784313725490198, | |
| "grad_norm": 0.6033862754408237, | |
| "learning_rate": 2.7515696828265778e-05, | |
| "loss": 0.1103, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11691197752952576, | |
| "step": 3140, | |
| "valid_targets_mean": 1985.6, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 3.0833333333333335, | |
| "grad_norm": 0.5519436809793806, | |
| "learning_rate": 2.7470368852510235e-05, | |
| "loss": 0.1133, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12123852223157883, | |
| "step": 3145, | |
| "valid_targets_mean": 1779.9, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 3.088235294117647, | |
| "grad_norm": 0.6712077515102802, | |
| "learning_rate": 2.7424996239226118e-05, | |
| "loss": 0.125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13345475494861603, | |
| "step": 3150, | |
| "valid_targets_mean": 1769.3, | |
| "valid_targets_min": 511 | |
| }, | |
| { | |
| "epoch": 3.093137254901961, | |
| "grad_norm": 0.6947461203066284, | |
| "learning_rate": 2.73795792595274e-05, | |
| "loss": 0.1129, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.136090487241745, | |
| "step": 3155, | |
| "valid_targets_mean": 1951.0, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 3.0980392156862746, | |
| "grad_norm": 0.595632721442861, | |
| "learning_rate": 2.733411818479313e-05, | |
| "loss": 0.1198, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10106316208839417, | |
| "step": 3160, | |
| "valid_targets_mean": 2653.4, | |
| "valid_targets_min": 1135 | |
| }, | |
| { | |
| "epoch": 3.1029411764705883, | |
| "grad_norm": 0.5915082015158023, | |
| "learning_rate": 2.7288613286665877e-05, | |
| "loss": 0.1098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11444532871246338, | |
| "step": 3165, | |
| "valid_targets_mean": 1782.6, | |
| "valid_targets_min": 832 | |
| }, | |
| { | |
| "epoch": 3.107843137254902, | |
| "grad_norm": 0.4258687271969111, | |
| "learning_rate": 2.724306483705004e-05, | |
| "loss": 0.1058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10663177818059921, | |
| "step": 3170, | |
| "valid_targets_mean": 2679.2, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 3.1127450980392157, | |
| "grad_norm": 0.5434705921669944, | |
| "learning_rate": 2.719747310811023e-05, | |
| "loss": 0.1169, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10878196358680725, | |
| "step": 3175, | |
| "valid_targets_mean": 1852.3, | |
| "valid_targets_min": 712 | |
| }, | |
| { | |
| "epoch": 3.1176470588235294, | |
| "grad_norm": 0.518948863407975, | |
| "learning_rate": 2.7151838372269724e-05, | |
| "loss": 0.1277, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10524100065231323, | |
| "step": 3180, | |
| "valid_targets_mean": 2033.2, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 3.122549019607843, | |
| "grad_norm": 0.5670597404893448, | |
| "learning_rate": 2.7106160902208736e-05, | |
| "loss": 0.1139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11880999803543091, | |
| "step": 3185, | |
| "valid_targets_mean": 1744.6, | |
| "valid_targets_min": 524 | |
| }, | |
| { | |
| "epoch": 3.127450980392157, | |
| "grad_norm": 0.4593924657033769, | |
| "learning_rate": 2.706044097086281e-05, | |
| "loss": 0.1062, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0906146839261055, | |
| "step": 3190, | |
| "valid_targets_mean": 2278.1, | |
| "valid_targets_min": 651 | |
| }, | |
| { | |
| "epoch": 3.1323529411764706, | |
| "grad_norm": 0.5302919704743427, | |
| "learning_rate": 2.701467885142126e-05, | |
| "loss": 0.1007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11191104352474213, | |
| "step": 3195, | |
| "valid_targets_mean": 2203.2, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 3.1372549019607843, | |
| "grad_norm": 0.46770851652221684, | |
| "learning_rate": 2.696887481732545e-05, | |
| "loss": 0.1083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08962736278772354, | |
| "step": 3200, | |
| "valid_targets_mean": 2317.0, | |
| "valid_targets_min": 485 | |
| }, | |
| { | |
| "epoch": 3.142156862745098, | |
| "grad_norm": 0.461711248997322, | |
| "learning_rate": 2.6923029142267195e-05, | |
| "loss": 0.1086, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10215017199516296, | |
| "step": 3205, | |
| "valid_targets_mean": 2429.4, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 3.1470588235294117, | |
| "grad_norm": 0.5124333477538516, | |
| "learning_rate": 2.6877142100187157e-05, | |
| "loss": 0.1114, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10358607769012451, | |
| "step": 3210, | |
| "valid_targets_mean": 1864.0, | |
| "valid_targets_min": 536 | |
| }, | |
| { | |
| "epoch": 3.1519607843137254, | |
| "grad_norm": 0.3616554988301426, | |
| "learning_rate": 2.6831213965273133e-05, | |
| "loss": 0.1061, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08923672884702682, | |
| "step": 3215, | |
| "valid_targets_mean": 2760.3, | |
| "valid_targets_min": 464 | |
| }, | |
| { | |
| "epoch": 3.156862745098039, | |
| "grad_norm": 0.5006444640458854, | |
| "learning_rate": 2.6785245011958498e-05, | |
| "loss": 0.1088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10196967422962189, | |
| "step": 3220, | |
| "valid_targets_mean": 2452.2, | |
| "valid_targets_min": 894 | |
| }, | |
| { | |
| "epoch": 3.161764705882353, | |
| "grad_norm": 0.4732082212550895, | |
| "learning_rate": 2.6739235514920516e-05, | |
| "loss": 0.1089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10402283817529678, | |
| "step": 3225, | |
| "valid_targets_mean": 2514.3, | |
| "valid_targets_min": 992 | |
| }, | |
| { | |
| "epoch": 3.1666666666666665, | |
| "grad_norm": 0.39259653455459503, | |
| "learning_rate": 2.6693185749078696e-05, | |
| "loss": 0.0996, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09591707587242126, | |
| "step": 3230, | |
| "valid_targets_mean": 2907.6, | |
| "valid_targets_min": 702 | |
| }, | |
| { | |
| "epoch": 3.1715686274509802, | |
| "grad_norm": 0.5506349591543678, | |
| "learning_rate": 2.6647095989593194e-05, | |
| "loss": 0.1088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11052503436803818, | |
| "step": 3235, | |
| "valid_targets_mean": 2068.1, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 3.176470588235294, | |
| "grad_norm": 0.5110012783834129, | |
| "learning_rate": 2.660096651186312e-05, | |
| "loss": 0.1029, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10590462386608124, | |
| "step": 3240, | |
| "valid_targets_mean": 2004.8, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 3.1813725490196076, | |
| "grad_norm": 0.48429980261496486, | |
| "learning_rate": 2.65547975915249e-05, | |
| "loss": 0.1171, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10138334333896637, | |
| "step": 3245, | |
| "valid_targets_mean": 2139.9, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 3.186274509803922, | |
| "grad_norm": 0.49960885032231134, | |
| "learning_rate": 2.6508589504450674e-05, | |
| "loss": 0.1131, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10819289833307266, | |
| "step": 3250, | |
| "valid_targets_mean": 1949.6, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 3.1911764705882355, | |
| "grad_norm": 0.41048885954549186, | |
| "learning_rate": 2.6462342526746582e-05, | |
| "loss": 0.0986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08373367786407471, | |
| "step": 3255, | |
| "valid_targets_mean": 2581.2, | |
| "valid_targets_min": 447 | |
| }, | |
| { | |
| "epoch": 3.196078431372549, | |
| "grad_norm": 0.5789520695615816, | |
| "learning_rate": 2.6416056934751163e-05, | |
| "loss": 0.1106, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09594295173883438, | |
| "step": 3260, | |
| "valid_targets_mean": 1791.2, | |
| "valid_targets_min": 503 | |
| }, | |
| { | |
| "epoch": 3.200980392156863, | |
| "grad_norm": 0.4565834458818512, | |
| "learning_rate": 2.6369733005033693e-05, | |
| "loss": 0.1097, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1069725751876831, | |
| "step": 3265, | |
| "valid_targets_mean": 2500.2, | |
| "valid_targets_min": 582 | |
| }, | |
| { | |
| "epoch": 3.2058823529411766, | |
| "grad_norm": 0.5371663721783211, | |
| "learning_rate": 2.6323371014392503e-05, | |
| "loss": 0.1136, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12323896586894989, | |
| "step": 3270, | |
| "valid_targets_mean": 2023.4, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 3.2107843137254903, | |
| "grad_norm": 0.46676874397280727, | |
| "learning_rate": 2.6276971239853368e-05, | |
| "loss": 0.1159, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10574575513601303, | |
| "step": 3275, | |
| "valid_targets_mean": 2582.6, | |
| "valid_targets_min": 674 | |
| }, | |
| { | |
| "epoch": 3.215686274509804, | |
| "grad_norm": 0.47929248550882253, | |
| "learning_rate": 2.6230533958667833e-05, | |
| "loss": 0.1079, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10941387712955475, | |
| "step": 3280, | |
| "valid_targets_mean": 2438.4, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 3.2205882352941178, | |
| "grad_norm": 0.4962450070888328, | |
| "learning_rate": 2.618405944831154e-05, | |
| "loss": 0.109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10915249586105347, | |
| "step": 3285, | |
| "valid_targets_mean": 2119.5, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 3.2254901960784315, | |
| "grad_norm": 0.4534435959489161, | |
| "learning_rate": 2.6137547986482607e-05, | |
| "loss": 0.1147, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09829148650169373, | |
| "step": 3290, | |
| "valid_targets_mean": 2296.5, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 3.230392156862745, | |
| "grad_norm": 0.5648115539160723, | |
| "learning_rate": 2.6090999851099924e-05, | |
| "loss": 0.1189, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1307239532470703, | |
| "step": 3295, | |
| "valid_targets_mean": 2227.4, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 3.235294117647059, | |
| "grad_norm": 0.5798437771314721, | |
| "learning_rate": 2.604441532030153e-05, | |
| "loss": 0.1023, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11284209787845612, | |
| "step": 3300, | |
| "valid_targets_mean": 1850.4, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 3.2401960784313726, | |
| "grad_norm": 0.45163997776172005, | |
| "learning_rate": 2.599779467244294e-05, | |
| "loss": 0.1065, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09802794456481934, | |
| "step": 3305, | |
| "valid_targets_mean": 2635.2, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 3.2450980392156863, | |
| "grad_norm": 0.5649182130019871, | |
| "learning_rate": 2.595113818609547e-05, | |
| "loss": 0.1104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1129395067691803, | |
| "step": 3310, | |
| "valid_targets_mean": 1722.3, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "grad_norm": 0.4884299890008094, | |
| "learning_rate": 2.590444614004458e-05, | |
| "loss": 0.1054, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11049424111843109, | |
| "step": 3315, | |
| "valid_targets_mean": 2286.6, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 3.2549019607843137, | |
| "grad_norm": 0.4484177235028694, | |
| "learning_rate": 2.585771881328821e-05, | |
| "loss": 0.1044, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08890993893146515, | |
| "step": 3320, | |
| "valid_targets_mean": 2585.2, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 3.2598039215686274, | |
| "grad_norm": 0.5045433489277796, | |
| "learning_rate": 2.5810956485035125e-05, | |
| "loss": 0.1101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12246538698673248, | |
| "step": 3325, | |
| "valid_targets_mean": 2569.7, | |
| "valid_targets_min": 927 | |
| }, | |
| { | |
| "epoch": 3.264705882352941, | |
| "grad_norm": 0.4652730472828903, | |
| "learning_rate": 2.5764159434703215e-05, | |
| "loss": 0.1251, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10294139385223389, | |
| "step": 3330, | |
| "valid_targets_mean": 2296.1, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 3.269607843137255, | |
| "grad_norm": 0.5793296301279914, | |
| "learning_rate": 2.5717327941917855e-05, | |
| "loss": 0.1108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09359988570213318, | |
| "step": 3335, | |
| "valid_targets_mean": 2318.1, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 3.2745098039215685, | |
| "grad_norm": 0.5084318900806167, | |
| "learning_rate": 2.567046228651022e-05, | |
| "loss": 0.102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11955564469099045, | |
| "step": 3340, | |
| "valid_targets_mean": 2201.9, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 3.2794117647058822, | |
| "grad_norm": 0.6829491331460958, | |
| "learning_rate": 2.5623562748515615e-05, | |
| "loss": 0.117, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14061784744262695, | |
| "step": 3345, | |
| "valid_targets_mean": 1760.6, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 3.284313725490196, | |
| "grad_norm": 0.5653936654283042, | |
| "learning_rate": 2.557662960817181e-05, | |
| "loss": 0.0989, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09769532829523087, | |
| "step": 3350, | |
| "valid_targets_mean": 2158.1, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 3.2892156862745097, | |
| "grad_norm": 0.4175355963316871, | |
| "learning_rate": 2.552966314591735e-05, | |
| "loss": 0.1172, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08977030217647552, | |
| "step": 3355, | |
| "valid_targets_mean": 2686.6, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 3.2941176470588234, | |
| "grad_norm": 0.4751300485635002, | |
| "learning_rate": 2.5482663642389883e-05, | |
| "loss": 0.1208, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08631863445043564, | |
| "step": 3360, | |
| "valid_targets_mean": 2316.3, | |
| "valid_targets_min": 627 | |
| }, | |
| { | |
| "epoch": 3.299019607843137, | |
| "grad_norm": 0.555883518092085, | |
| "learning_rate": 2.5435631378424504e-05, | |
| "loss": 0.1058, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10301592946052551, | |
| "step": 3365, | |
| "valid_targets_mean": 1677.7, | |
| "valid_targets_min": 688 | |
| }, | |
| { | |
| "epoch": 3.303921568627451, | |
| "grad_norm": 0.6179203271979932, | |
| "learning_rate": 2.5388566635052052e-05, | |
| "loss": 0.12, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1356745958328247, | |
| "step": 3370, | |
| "valid_targets_mean": 1944.9, | |
| "valid_targets_min": 747 | |
| }, | |
| { | |
| "epoch": 3.3088235294117645, | |
| "grad_norm": 0.4676890698107417, | |
| "learning_rate": 2.5341469693497434e-05, | |
| "loss": 0.1085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11229351162910461, | |
| "step": 3375, | |
| "valid_targets_mean": 2420.8, | |
| "valid_targets_min": 821 | |
| }, | |
| { | |
| "epoch": 3.313725490196078, | |
| "grad_norm": 0.48905784664475677, | |
| "learning_rate": 2.529434083517796e-05, | |
| "loss": 0.1134, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11473777890205383, | |
| "step": 3380, | |
| "valid_targets_mean": 2472.6, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 3.318627450980392, | |
| "grad_norm": 0.5941023968549567, | |
| "learning_rate": 2.5247180341701652e-05, | |
| "loss": 0.1125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12429068982601166, | |
| "step": 3385, | |
| "valid_targets_mean": 1746.0, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 3.323529411764706, | |
| "grad_norm": 0.4207492309955242, | |
| "learning_rate": 2.5199988494865545e-05, | |
| "loss": 0.119, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09677383303642273, | |
| "step": 3390, | |
| "valid_targets_mean": 2596.6, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 3.3284313725490198, | |
| "grad_norm": 0.6212127423535296, | |
| "learning_rate": 2.5152765576654035e-05, | |
| "loss": 0.1042, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09490111470222473, | |
| "step": 3395, | |
| "valid_targets_mean": 2402.4, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 3.3333333333333335, | |
| "grad_norm": 0.5428852197642803, | |
| "learning_rate": 2.5105511869237176e-05, | |
| "loss": 0.1155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11736579239368439, | |
| "step": 3400, | |
| "valid_targets_mean": 2132.8, | |
| "valid_targets_min": 903 | |
| }, | |
| { | |
| "epoch": 3.338235294117647, | |
| "grad_norm": 0.6068835771442634, | |
| "learning_rate": 2.5058227654968986e-05, | |
| "loss": 0.1211, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1142382025718689, | |
| "step": 3405, | |
| "valid_targets_mean": 2071.6, | |
| "valid_targets_min": 878 | |
| }, | |
| { | |
| "epoch": 3.343137254901961, | |
| "grad_norm": 0.3980675116079957, | |
| "learning_rate": 2.5010913216385778e-05, | |
| "loss": 0.1119, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08887724578380585, | |
| "step": 3410, | |
| "valid_targets_mean": 2649.5, | |
| "valid_targets_min": 696 | |
| }, | |
| { | |
| "epoch": 3.3480392156862746, | |
| "grad_norm": 0.5921220488695144, | |
| "learning_rate": 2.496356883620448e-05, | |
| "loss": 0.1182, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13174962997436523, | |
| "step": 3415, | |
| "valid_targets_mean": 1664.5, | |
| "valid_targets_min": 690 | |
| }, | |
| { | |
| "epoch": 3.3529411764705883, | |
| "grad_norm": 0.4555660361067227, | |
| "learning_rate": 2.491619479732089e-05, | |
| "loss": 0.1092, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09527282416820526, | |
| "step": 3420, | |
| "valid_targets_mean": 2478.8, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 3.357843137254902, | |
| "grad_norm": 0.6222613588669669, | |
| "learning_rate": 2.4868791382808053e-05, | |
| "loss": 0.1196, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12515193223953247, | |
| "step": 3425, | |
| "valid_targets_mean": 1625.1, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 3.3627450980392157, | |
| "grad_norm": 0.5514835585510608, | |
| "learning_rate": 2.4821358875914546e-05, | |
| "loss": 0.1044, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12040072679519653, | |
| "step": 3430, | |
| "valid_targets_mean": 2105.8, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 3.3676470588235294, | |
| "grad_norm": 0.5961299236034482, | |
| "learning_rate": 2.477389756006276e-05, | |
| "loss": 0.1153, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14823094010353088, | |
| "step": 3435, | |
| "valid_targets_mean": 2278.8, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 3.372549019607843, | |
| "grad_norm": 0.6437966485612172, | |
| "learning_rate": 2.4726407718847237e-05, | |
| "loss": 0.1134, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12897151708602905, | |
| "step": 3440, | |
| "valid_targets_mean": 1771.4, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 3.377450980392157, | |
| "grad_norm": 0.5267111874802913, | |
| "learning_rate": 2.4678889636032976e-05, | |
| "loss": 0.1072, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10938486456871033, | |
| "step": 3445, | |
| "valid_targets_mean": 1924.3, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 3.3823529411764706, | |
| "grad_norm": 0.496231210125758, | |
| "learning_rate": 2.4631343595553707e-05, | |
| "loss": 0.1081, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09718477725982666, | |
| "step": 3450, | |
| "valid_targets_mean": 2110.4, | |
| "valid_targets_min": 853 | |
| }, | |
| { | |
| "epoch": 3.3872549019607843, | |
| "grad_norm": 0.7417323818704427, | |
| "learning_rate": 2.4583769881510243e-05, | |
| "loss": 0.1209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12951020896434784, | |
| "step": 3455, | |
| "valid_targets_mean": 2396.2, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 3.392156862745098, | |
| "grad_norm": 0.686014144744929, | |
| "learning_rate": 2.453616877816872e-05, | |
| "loss": 0.1199, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10813941061496735, | |
| "step": 3460, | |
| "valid_targets_mean": 2572.1, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 3.3970588235294117, | |
| "grad_norm": 0.46956228200489375, | |
| "learning_rate": 2.4488540569958957e-05, | |
| "loss": 0.1034, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10174999386072159, | |
| "step": 3465, | |
| "valid_targets_mean": 2273.4, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 3.4019607843137254, | |
| "grad_norm": 0.47202061673863827, | |
| "learning_rate": 2.4440885541472734e-05, | |
| "loss": 0.1111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1065269336104393, | |
| "step": 3470, | |
| "valid_targets_mean": 2521.2, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 3.406862745098039, | |
| "grad_norm": 0.4710834442443773, | |
| "learning_rate": 2.439320397746207e-05, | |
| "loss": 0.1008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09578939527273178, | |
| "step": 3475, | |
| "valid_targets_mean": 2454.9, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 3.411764705882353, | |
| "grad_norm": 0.534990938857518, | |
| "learning_rate": 2.434549616283755e-05, | |
| "loss": 0.1144, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13655036687850952, | |
| "step": 3480, | |
| "valid_targets_mean": 2208.6, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 3.4166666666666665, | |
| "grad_norm": 0.5569406603515846, | |
| "learning_rate": 2.4297762382666628e-05, | |
| "loss": 0.1135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11844809353351593, | |
| "step": 3485, | |
| "valid_targets_mean": 2206.2, | |
| "valid_targets_min": 1016 | |
| }, | |
| { | |
| "epoch": 3.4215686274509802, | |
| "grad_norm": 0.573221969522095, | |
| "learning_rate": 2.4250002922171887e-05, | |
| "loss": 0.1166, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1180095300078392, | |
| "step": 3490, | |
| "valid_targets_mean": 2059.7, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 3.426470588235294, | |
| "grad_norm": 0.5023184061613964, | |
| "learning_rate": 2.4202218066729377e-05, | |
| "loss": 0.1244, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12228246033191681, | |
| "step": 3495, | |
| "valid_targets_mean": 2557.6, | |
| "valid_targets_min": 631 | |
| }, | |
| { | |
| "epoch": 3.431372549019608, | |
| "grad_norm": 0.4804109211903169, | |
| "learning_rate": 2.415440810186687e-05, | |
| "loss": 0.112, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12410437315702438, | |
| "step": 3500, | |
| "valid_targets_mean": 2642.3, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 3.436274509803922, | |
| "grad_norm": 0.7160334396984808, | |
| "learning_rate": 2.4106573313262186e-05, | |
| "loss": 0.1105, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1095886379480362, | |
| "step": 3505, | |
| "valid_targets_mean": 2399.1, | |
| "valid_targets_min": 606 | |
| }, | |
| { | |
| "epoch": 3.4411764705882355, | |
| "grad_norm": 0.573633678368746, | |
| "learning_rate": 2.405871398674148e-05, | |
| "loss": 0.1036, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12151598930358887, | |
| "step": 3510, | |
| "valid_targets_mean": 1821.2, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 3.446078431372549, | |
| "grad_norm": 0.47876480518730624, | |
| "learning_rate": 2.4010830408277507e-05, | |
| "loss": 0.1064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1134987398982048, | |
| "step": 3515, | |
| "valid_targets_mean": 2287.1, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 3.450980392156863, | |
| "grad_norm": 0.48121870038783876, | |
| "learning_rate": 2.3962922863987956e-05, | |
| "loss": 0.1132, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09954457730054855, | |
| "step": 3520, | |
| "valid_targets_mean": 2372.2, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 3.4558823529411766, | |
| "grad_norm": 0.5242680043426456, | |
| "learning_rate": 2.3914991640133707e-05, | |
| "loss": 0.12, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15709488093852997, | |
| "step": 3525, | |
| "valid_targets_mean": 2238.8, | |
| "valid_targets_min": 840 | |
| }, | |
| { | |
| "epoch": 3.4607843137254903, | |
| "grad_norm": 0.43408849677146805, | |
| "learning_rate": 2.3867037023117123e-05, | |
| "loss": 0.1001, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08314131200313568, | |
| "step": 3530, | |
| "valid_targets_mean": 2221.2, | |
| "valid_targets_min": 699 | |
| }, | |
| { | |
| "epoch": 3.465686274509804, | |
| "grad_norm": 0.4397584871724824, | |
| "learning_rate": 2.3819059299480375e-05, | |
| "loss": 0.1284, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1028500497341156, | |
| "step": 3535, | |
| "valid_targets_mean": 2543.4, | |
| "valid_targets_min": 649 | |
| }, | |
| { | |
| "epoch": 3.4705882352941178, | |
| "grad_norm": 0.49009687013344777, | |
| "learning_rate": 2.3771058755903686e-05, | |
| "loss": 0.1047, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10757431387901306, | |
| "step": 3540, | |
| "valid_targets_mean": 2223.6, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 3.4754901960784315, | |
| "grad_norm": 0.39557937893345524, | |
| "learning_rate": 2.372303567920362e-05, | |
| "loss": 0.1174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09384208172559738, | |
| "step": 3545, | |
| "valid_targets_mean": 2953.0, | |
| "valid_targets_min": 1049 | |
| }, | |
| { | |
| "epoch": 3.480392156862745, | |
| "grad_norm": 0.4895551988114163, | |
| "learning_rate": 2.367499035633141e-05, | |
| "loss": 0.1239, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10619791597127914, | |
| "step": 3550, | |
| "valid_targets_mean": 1790.4, | |
| "valid_targets_min": 670 | |
| }, | |
| { | |
| "epoch": 3.485294117647059, | |
| "grad_norm": 0.5343919639111113, | |
| "learning_rate": 2.3626923074371196e-05, | |
| "loss": 0.1118, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.112095907330513, | |
| "step": 3555, | |
| "valid_targets_mean": 2142.9, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 3.4901960784313726, | |
| "grad_norm": 0.39309956805603796, | |
| "learning_rate": 2.357883412053834e-05, | |
| "loss": 0.0982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07599136233329773, | |
| "step": 3560, | |
| "valid_targets_mean": 2629.5, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 3.4950980392156863, | |
| "grad_norm": 0.5743979416820177, | |
| "learning_rate": 2.3530723782177702e-05, | |
| "loss": 0.1125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11538234353065491, | |
| "step": 3565, | |
| "valid_targets_mean": 1909.1, | |
| "valid_targets_min": 580 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "grad_norm": 0.4267696073720374, | |
| "learning_rate": 2.3482592346761905e-05, | |
| "loss": 0.1125, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08938651531934738, | |
| "step": 3570, | |
| "valid_targets_mean": 2519.2, | |
| "valid_targets_min": 845 | |
| }, | |
| { | |
| "epoch": 3.5049019607843137, | |
| "grad_norm": 0.48972900333651787, | |
| "learning_rate": 2.3434440101889644e-05, | |
| "loss": 0.1036, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09942841529846191, | |
| "step": 3575, | |
| "valid_targets_mean": 2042.9, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 3.5098039215686274, | |
| "grad_norm": 0.5748476170765656, | |
| "learning_rate": 2.338626733528396e-05, | |
| "loss": 0.1182, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1080469936132431, | |
| "step": 3580, | |
| "valid_targets_mean": 1870.4, | |
| "valid_targets_min": 468 | |
| }, | |
| { | |
| "epoch": 3.514705882352941, | |
| "grad_norm": 0.5453346260954157, | |
| "learning_rate": 2.3338074334790513e-05, | |
| "loss": 0.1075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12305852770805359, | |
| "step": 3585, | |
| "valid_targets_mean": 2156.6, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 3.519607843137255, | |
| "grad_norm": 0.4046360880578505, | |
| "learning_rate": 2.3289861388375857e-05, | |
| "loss": 0.1108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08819185197353363, | |
| "step": 3590, | |
| "valid_targets_mean": 2738.9, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 3.5245098039215685, | |
| "grad_norm": 0.6118714832043691, | |
| "learning_rate": 2.3241628784125746e-05, | |
| "loss": 0.109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13428336381912231, | |
| "step": 3595, | |
| "valid_targets_mean": 1819.1, | |
| "valid_targets_min": 717 | |
| }, | |
| { | |
| "epoch": 3.5294117647058822, | |
| "grad_norm": 0.5313367384075776, | |
| "learning_rate": 2.319337681024338e-05, | |
| "loss": 0.1174, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13841737806797028, | |
| "step": 3600, | |
| "valid_targets_mean": 2279.8, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 3.534313725490196, | |
| "grad_norm": 0.5936013371744282, | |
| "learning_rate": 2.314510575504771e-05, | |
| "loss": 0.1141, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12145009636878967, | |
| "step": 3605, | |
| "valid_targets_mean": 2184.8, | |
| "valid_targets_min": 863 | |
| }, | |
| { | |
| "epoch": 3.5392156862745097, | |
| "grad_norm": 0.4501094327540122, | |
| "learning_rate": 2.3096815906971685e-05, | |
| "loss": 0.1034, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10091482102870941, | |
| "step": 3610, | |
| "valid_targets_mean": 2419.2, | |
| "valid_targets_min": 961 | |
| }, | |
| { | |
| "epoch": 3.5441176470588234, | |
| "grad_norm": 0.5917686920531069, | |
| "learning_rate": 2.3048507554560565e-05, | |
| "loss": 0.1099, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11512914299964905, | |
| "step": 3615, | |
| "valid_targets_mean": 1797.6, | |
| "valid_targets_min": 563 | |
| }, | |
| { | |
| "epoch": 3.549019607843137, | |
| "grad_norm": 0.6255611571989778, | |
| "learning_rate": 2.3000180986470175e-05, | |
| "loss": 0.107, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13142108917236328, | |
| "step": 3620, | |
| "valid_targets_mean": 1911.9, | |
| "valid_targets_min": 576 | |
| }, | |
| { | |
| "epoch": 3.553921568627451, | |
| "grad_norm": 0.5167486005612478, | |
| "learning_rate": 2.2951836491465177e-05, | |
| "loss": 0.1022, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1105874553322792, | |
| "step": 3625, | |
| "valid_targets_mean": 2156.2, | |
| "valid_targets_min": 329 | |
| }, | |
| { | |
| "epoch": 3.5588235294117645, | |
| "grad_norm": 0.5154655734331977, | |
| "learning_rate": 2.290347435841736e-05, | |
| "loss": 0.111, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12183243036270142, | |
| "step": 3630, | |
| "valid_targets_mean": 2346.5, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 3.563725490196078, | |
| "grad_norm": 0.5223303986667628, | |
| "learning_rate": 2.285509487630389e-05, | |
| "loss": 0.1198, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15623009204864502, | |
| "step": 3635, | |
| "valid_targets_mean": 2398.6, | |
| "valid_targets_min": 609 | |
| }, | |
| { | |
| "epoch": 3.568627450980392, | |
| "grad_norm": 0.4470634454069301, | |
| "learning_rate": 2.2806698334205612e-05, | |
| "loss": 0.1222, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11349236220121384, | |
| "step": 3640, | |
| "valid_targets_mean": 2423.0, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 3.5735294117647056, | |
| "grad_norm": 0.5404202109658872, | |
| "learning_rate": 2.2758285021305312e-05, | |
| "loss": 0.1025, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10349144786596298, | |
| "step": 3645, | |
| "valid_targets_mean": 1936.6, | |
| "valid_targets_min": 598 | |
| }, | |
| { | |
| "epoch": 3.5784313725490198, | |
| "grad_norm": 0.530293394551295, | |
| "learning_rate": 2.2709855226885975e-05, | |
| "loss": 0.1195, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11319459229707718, | |
| "step": 3650, | |
| "valid_targets_mean": 2097.7, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 3.5833333333333335, | |
| "grad_norm": 0.47036735332011803, | |
| "learning_rate": 2.266140924032908e-05, | |
| "loss": 0.1241, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10819728672504425, | |
| "step": 3655, | |
| "valid_targets_mean": 2694.6, | |
| "valid_targets_min": 1356 | |
| }, | |
| { | |
| "epoch": 3.588235294117647, | |
| "grad_norm": 0.5151780573315811, | |
| "learning_rate": 2.2612947351112836e-05, | |
| "loss": 0.1135, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1199716329574585, | |
| "step": 3660, | |
| "valid_targets_mean": 2438.4, | |
| "valid_targets_min": 924 | |
| }, | |
| { | |
| "epoch": 3.593137254901961, | |
| "grad_norm": 0.4298760911534975, | |
| "learning_rate": 2.2564469848810503e-05, | |
| "loss": 0.1049, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09393537044525146, | |
| "step": 3665, | |
| "valid_targets_mean": 2601.8, | |
| "valid_targets_min": 965 | |
| }, | |
| { | |
| "epoch": 3.5980392156862746, | |
| "grad_norm": 0.5502702790614998, | |
| "learning_rate": 2.2515977023088613e-05, | |
| "loss": 0.104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13896121084690094, | |
| "step": 3670, | |
| "valid_targets_mean": 2215.3, | |
| "valid_targets_min": 778 | |
| }, | |
| { | |
| "epoch": 3.6029411764705883, | |
| "grad_norm": 0.578499401998039, | |
| "learning_rate": 2.246746916370527e-05, | |
| "loss": 0.1155, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12631025910377502, | |
| "step": 3675, | |
| "valid_targets_mean": 1959.6, | |
| "valid_targets_min": 560 | |
| }, | |
| { | |
| "epoch": 3.607843137254902, | |
| "grad_norm": 0.623239876434199, | |
| "learning_rate": 2.241894656050841e-05, | |
| "loss": 0.1117, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13602277636528015, | |
| "step": 3680, | |
| "valid_targets_mean": 1764.9, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 3.6127450980392157, | |
| "grad_norm": 0.600465231043324, | |
| "learning_rate": 2.237040950343406e-05, | |
| "loss": 0.1194, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12396884709596634, | |
| "step": 3685, | |
| "valid_targets_mean": 1924.8, | |
| "valid_targets_min": 794 | |
| }, | |
| { | |
| "epoch": 3.6176470588235294, | |
| "grad_norm": 0.7319348639656394, | |
| "learning_rate": 2.2321858282504606e-05, | |
| "loss": 0.1325, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.15651845932006836, | |
| "step": 3690, | |
| "valid_targets_mean": 1656.2, | |
| "valid_targets_min": 795 | |
| }, | |
| { | |
| "epoch": 3.622549019607843, | |
| "grad_norm": 0.5398063229503928, | |
| "learning_rate": 2.2273293187827093e-05, | |
| "loss": 0.104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12730857729911804, | |
| "step": 3695, | |
| "valid_targets_mean": 1967.0, | |
| "valid_targets_min": 573 | |
| }, | |
| { | |
| "epoch": 3.627450980392157, | |
| "grad_norm": 0.7037019842887797, | |
| "learning_rate": 2.2224714509591436e-05, | |
| "loss": 0.115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14245694875717163, | |
| "step": 3700, | |
| "valid_targets_mean": 1353.9, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 3.6323529411764706, | |
| "grad_norm": 0.49584176697366805, | |
| "learning_rate": 2.2176122538068738e-05, | |
| "loss": 0.1088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0984085351228714, | |
| "step": 3705, | |
| "valid_targets_mean": 2135.8, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 3.6372549019607843, | |
| "grad_norm": 0.378093430237812, | |
| "learning_rate": 2.212751756360952e-05, | |
| "loss": 0.1005, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08473532646894455, | |
| "step": 3710, | |
| "valid_targets_mean": 2742.5, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 3.642156862745098, | |
| "grad_norm": 0.45671211592262223, | |
| "learning_rate": 2.2078899876642e-05, | |
| "loss": 0.1145, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09947746247053146, | |
| "step": 3715, | |
| "valid_targets_mean": 2311.5, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 3.6470588235294117, | |
| "grad_norm": 0.6022808568063974, | |
| "learning_rate": 2.2030269767670368e-05, | |
| "loss": 0.115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12485545873641968, | |
| "step": 3720, | |
| "valid_targets_mean": 1894.9, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 3.6519607843137254, | |
| "grad_norm": 0.4816296256677648, | |
| "learning_rate": 2.1981627527273023e-05, | |
| "loss": 0.1057, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1067422479391098, | |
| "step": 3725, | |
| "valid_targets_mean": 2501.2, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 3.656862745098039, | |
| "grad_norm": 0.48952979423754794, | |
| "learning_rate": 2.1932973446100863e-05, | |
| "loss": 0.1064, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10704466700553894, | |
| "step": 3730, | |
| "valid_targets_mean": 2467.1, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 3.661764705882353, | |
| "grad_norm": 0.5489743475632907, | |
| "learning_rate": 2.1884307814875537e-05, | |
| "loss": 0.1209, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1270291805267334, | |
| "step": 3735, | |
| "valid_targets_mean": 1769.5, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 3.6666666666666665, | |
| "grad_norm": 0.4766948278075901, | |
| "learning_rate": 2.18356309243877e-05, | |
| "loss": 0.1088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09539386630058289, | |
| "step": 3740, | |
| "valid_targets_mean": 2464.4, | |
| "valid_targets_min": 497 | |
| }, | |
| { | |
| "epoch": 3.6715686274509802, | |
| "grad_norm": 0.4847705657231007, | |
| "learning_rate": 2.1786943065495302e-05, | |
| "loss": 0.1113, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0939837098121643, | |
| "step": 3745, | |
| "valid_targets_mean": 2143.0, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 3.6764705882352944, | |
| "grad_norm": 0.48673504467946466, | |
| "learning_rate": 2.173824452912181e-05, | |
| "loss": 0.1086, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10983025282621384, | |
| "step": 3750, | |
| "valid_targets_mean": 2503.6, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 3.681372549019608, | |
| "grad_norm": 0.6394630567515267, | |
| "learning_rate": 2.1689535606254517e-05, | |
| "loss": 0.113, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13665525615215302, | |
| "step": 3755, | |
| "valid_targets_mean": 1859.1, | |
| "valid_targets_min": 424 | |
| }, | |
| { | |
| "epoch": 3.686274509803922, | |
| "grad_norm": 0.5404637568110688, | |
| "learning_rate": 2.164081658794276e-05, | |
| "loss": 0.1086, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11135978996753693, | |
| "step": 3760, | |
| "valid_targets_mean": 2087.4, | |
| "valid_targets_min": 784 | |
| }, | |
| { | |
| "epoch": 3.6911764705882355, | |
| "grad_norm": 0.5275619152100403, | |
| "learning_rate": 2.1592087765296186e-05, | |
| "loss": 0.1038, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1140526682138443, | |
| "step": 3765, | |
| "valid_targets_mean": 2225.5, | |
| "valid_targets_min": 1330 | |
| }, | |
| { | |
| "epoch": 3.696078431372549, | |
| "grad_norm": 0.4205808571949082, | |
| "learning_rate": 2.154334942948307e-05, | |
| "loss": 0.102, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09243638813495636, | |
| "step": 3770, | |
| "valid_targets_mean": 2574.8, | |
| "valid_targets_min": 789 | |
| }, | |
| { | |
| "epoch": 3.700980392156863, | |
| "grad_norm": 0.403760695863388, | |
| "learning_rate": 2.149460187172849e-05, | |
| "loss": 0.0959, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08560940623283386, | |
| "step": 3775, | |
| "valid_targets_mean": 2659.8, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 3.7058823529411766, | |
| "grad_norm": 0.4755601372114485, | |
| "learning_rate": 2.144584538331264e-05, | |
| "loss": 0.1093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12424452602863312, | |
| "step": 3780, | |
| "valid_targets_mean": 2388.5, | |
| "valid_targets_min": 609 | |
| }, | |
| { | |
| "epoch": 3.7107843137254903, | |
| "grad_norm": 0.5410875160432338, | |
| "learning_rate": 2.139708025556909e-05, | |
| "loss": 0.1185, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1277793049812317, | |
| "step": 3785, | |
| "valid_targets_mean": 2157.8, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 3.715686274509804, | |
| "grad_norm": 0.5018258824660917, | |
| "learning_rate": 2.1348306779883004e-05, | |
| "loss": 0.1162, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11679060012102127, | |
| "step": 3790, | |
| "valid_targets_mean": 2386.1, | |
| "valid_targets_min": 674 | |
| }, | |
| { | |
| "epoch": 3.7205882352941178, | |
| "grad_norm": 0.4285080068758283, | |
| "learning_rate": 2.1299525247689466e-05, | |
| "loss": 0.1083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08835404366254807, | |
| "step": 3795, | |
| "valid_targets_mean": 2637.9, | |
| "valid_targets_min": 929 | |
| }, | |
| { | |
| "epoch": 3.7254901960784315, | |
| "grad_norm": 0.530318250431301, | |
| "learning_rate": 2.1250735950471664e-05, | |
| "loss": 0.1225, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11783511936664581, | |
| "step": 3800, | |
| "valid_targets_mean": 2486.4, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 3.730392156862745, | |
| "grad_norm": 0.44985130125257894, | |
| "learning_rate": 2.12019391797592e-05, | |
| "loss": 0.103, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08755841851234436, | |
| "step": 3805, | |
| "valid_targets_mean": 2712.6, | |
| "valid_targets_min": 939 | |
| }, | |
| { | |
| "epoch": 3.735294117647059, | |
| "grad_norm": 0.5286935035251132, | |
| "learning_rate": 2.1153135227126337e-05, | |
| "loss": 0.1053, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10353609919548035, | |
| "step": 3810, | |
| "valid_targets_mean": 2064.2, | |
| "valid_targets_min": 337 | |
| }, | |
| { | |
| "epoch": 3.7401960784313726, | |
| "grad_norm": 0.4601036655576194, | |
| "learning_rate": 2.1104324384190234e-05, | |
| "loss": 0.098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10336270183324814, | |
| "step": 3815, | |
| "valid_targets_mean": 2523.7, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 3.7450980392156863, | |
| "grad_norm": 0.561715253391712, | |
| "learning_rate": 2.105550694260925e-05, | |
| "loss": 0.1085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10617645829916, | |
| "step": 3820, | |
| "valid_targets_mean": 1696.9, | |
| "valid_targets_min": 543 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "grad_norm": 0.508039672374628, | |
| "learning_rate": 2.1006683194081144e-05, | |
| "loss": 0.1106, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12362570315599442, | |
| "step": 3825, | |
| "valid_targets_mean": 2148.6, | |
| "valid_targets_min": 544 | |
| }, | |
| { | |
| "epoch": 3.7549019607843137, | |
| "grad_norm": 0.5062010089032453, | |
| "learning_rate": 2.0957853430341373e-05, | |
| "loss": 0.1065, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10452181100845337, | |
| "step": 3830, | |
| "valid_targets_mean": 2122.1, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 3.7598039215686274, | |
| "grad_norm": 0.447000980086604, | |
| "learning_rate": 2.090901794316135e-05, | |
| "loss": 0.104, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09081129729747772, | |
| "step": 3835, | |
| "valid_targets_mean": 2808.3, | |
| "valid_targets_min": 922 | |
| }, | |
| { | |
| "epoch": 3.764705882352941, | |
| "grad_norm": 0.49795009536099594, | |
| "learning_rate": 2.086017702434666e-05, | |
| "loss": 0.1163, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12170379608869553, | |
| "step": 3840, | |
| "valid_targets_mean": 2495.4, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 3.769607843137255, | |
| "grad_norm": 0.44563295566531946, | |
| "learning_rate": 2.081133096573535e-05, | |
| "loss": 0.1231, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1004648432135582, | |
| "step": 3845, | |
| "valid_targets_mean": 2524.3, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 3.7745098039215685, | |
| "grad_norm": 0.47095104017094225, | |
| "learning_rate": 2.0762480059196207e-05, | |
| "loss": 0.1091, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09739089012145996, | |
| "step": 3850, | |
| "valid_targets_mean": 2408.9, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 3.7794117647058822, | |
| "grad_norm": 0.4882628996629055, | |
| "learning_rate": 2.0713624596626945e-05, | |
| "loss": 0.1145, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10591503232717514, | |
| "step": 3855, | |
| "valid_targets_mean": 2017.6, | |
| "valid_targets_min": 564 | |
| }, | |
| { | |
| "epoch": 3.784313725490196, | |
| "grad_norm": 0.49785785642908204, | |
| "learning_rate": 2.0664764869952535e-05, | |
| "loss": 0.1098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12032024562358856, | |
| "step": 3860, | |
| "valid_targets_mean": 2626.7, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 3.7892156862745097, | |
| "grad_norm": 0.40453258551835614, | |
| "learning_rate": 2.061590117112341e-05, | |
| "loss": 0.1052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1063418835401535, | |
| "step": 3865, | |
| "valid_targets_mean": 2805.0, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 3.7941176470588234, | |
| "grad_norm": 0.4375416998410625, | |
| "learning_rate": 2.056703379211373e-05, | |
| "loss": 0.1078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11425399035215378, | |
| "step": 3870, | |
| "valid_targets_mean": 2723.6, | |
| "valid_targets_min": 539 | |
| }, | |
| { | |
| "epoch": 3.799019607843137, | |
| "grad_norm": 0.45739452140837245, | |
| "learning_rate": 2.0518163024919677e-05, | |
| "loss": 0.1022, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09378018975257874, | |
| "step": 3875, | |
| "valid_targets_mean": 2357.4, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 3.803921568627451, | |
| "grad_norm": 0.5672950655638906, | |
| "learning_rate": 2.0469289161557637e-05, | |
| "loss": 0.1059, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1068982183933258, | |
| "step": 3880, | |
| "valid_targets_mean": 2123.6, | |
| "valid_targets_min": 782 | |
| }, | |
| { | |
| "epoch": 3.8088235294117645, | |
| "grad_norm": 0.6099769627480358, | |
| "learning_rate": 2.042041249406253e-05, | |
| "loss": 0.1041, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1426199972629547, | |
| "step": 3885, | |
| "valid_targets_mean": 2168.8, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 3.813725490196078, | |
| "grad_norm": 0.5004067455512964, | |
| "learning_rate": 2.0371533314486017e-05, | |
| "loss": 0.1033, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11419737339019775, | |
| "step": 3890, | |
| "valid_targets_mean": 2435.4, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 3.818627450980392, | |
| "grad_norm": 0.5661678936881082, | |
| "learning_rate": 2.0322651914894754e-05, | |
| "loss": 0.109, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14499229192733765, | |
| "step": 3895, | |
| "valid_targets_mean": 2460.9, | |
| "valid_targets_min": 730 | |
| }, | |
| { | |
| "epoch": 3.8235294117647056, | |
| "grad_norm": 0.49732546990907106, | |
| "learning_rate": 2.02737685873687e-05, | |
| "loss": 0.1071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11961804330348969, | |
| "step": 3900, | |
| "valid_targets_mean": 2059.0, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 3.8284313725490198, | |
| "grad_norm": 0.4993481866260662, | |
| "learning_rate": 2.0224883623999313e-05, | |
| "loss": 0.114, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10117202252149582, | |
| "step": 3905, | |
| "valid_targets_mean": 2220.4, | |
| "valid_targets_min": 893 | |
| }, | |
| { | |
| "epoch": 3.8333333333333335, | |
| "grad_norm": 0.5538681599370652, | |
| "learning_rate": 2.0175997316887802e-05, | |
| "loss": 0.1139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.109494149684906, | |
| "step": 3910, | |
| "valid_targets_mean": 1981.0, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 3.838235294117647, | |
| "grad_norm": 0.45218493019291356, | |
| "learning_rate": 2.012710995814345e-05, | |
| "loss": 0.1074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09603475034236908, | |
| "step": 3915, | |
| "valid_targets_mean": 2731.0, | |
| "valid_targets_min": 862 | |
| }, | |
| { | |
| "epoch": 3.843137254901961, | |
| "grad_norm": 0.44119940995141127, | |
| "learning_rate": 2.0078221839881805e-05, | |
| "loss": 0.1203, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10505522787570953, | |
| "step": 3920, | |
| "valid_targets_mean": 2638.4, | |
| "valid_targets_min": 909 | |
| }, | |
| { | |
| "epoch": 3.8480392156862746, | |
| "grad_norm": 0.4960608782273173, | |
| "learning_rate": 2.0029333254222934e-05, | |
| "loss": 0.1212, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11102761328220367, | |
| "step": 3925, | |
| "valid_targets_mean": 1909.0, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 3.8529411764705883, | |
| "grad_norm": 0.5586535920480638, | |
| "learning_rate": 1.9980444493289725e-05, | |
| "loss": 0.108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12268592417240143, | |
| "step": 3930, | |
| "valid_targets_mean": 2376.0, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 3.857843137254902, | |
| "grad_norm": 0.5393773385617813, | |
| "learning_rate": 1.9931555849206106e-05, | |
| "loss": 0.1101, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11982365697622299, | |
| "step": 3935, | |
| "valid_targets_mean": 2329.8, | |
| "valid_targets_min": 616 | |
| }, | |
| { | |
| "epoch": 3.8627450980392157, | |
| "grad_norm": 0.4725284612750934, | |
| "learning_rate": 1.9882667614095287e-05, | |
| "loss": 0.1088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10902339220046997, | |
| "step": 3940, | |
| "valid_targets_mean": 2420.7, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 3.8676470588235294, | |
| "grad_norm": 0.4171871943502758, | |
| "learning_rate": 1.9833780080078063e-05, | |
| "loss": 0.0977, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0916203111410141, | |
| "step": 3945, | |
| "valid_targets_mean": 2523.8, | |
| "valid_targets_min": 745 | |
| }, | |
| { | |
| "epoch": 3.872549019607843, | |
| "grad_norm": 0.5163398783108317, | |
| "learning_rate": 1.9784893539271026e-05, | |
| "loss": 0.1099, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10017865151166916, | |
| "step": 3950, | |
| "valid_targets_mean": 1958.1, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 3.877450980392157, | |
| "grad_norm": 0.7754438443029995, | |
| "learning_rate": 1.9736008283784824e-05, | |
| "loss": 0.1298, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1341777741909027, | |
| "step": 3955, | |
| "valid_targets_mean": 1683.8, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 3.8823529411764706, | |
| "grad_norm": 0.5273279125239405, | |
| "learning_rate": 1.9687124605722448e-05, | |
| "loss": 0.12, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12256628274917603, | |
| "step": 3960, | |
| "valid_targets_mean": 2042.1, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 3.8872549019607843, | |
| "grad_norm": 0.4144565504865219, | |
| "learning_rate": 1.963824279717744e-05, | |
| "loss": 0.0989, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0842713862657547, | |
| "step": 3965, | |
| "valid_targets_mean": 2443.1, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 3.892156862745098, | |
| "grad_norm": 0.5679541187571767, | |
| "learning_rate": 1.9589363150232194e-05, | |
| "loss": 0.1084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11258754879236221, | |
| "step": 3970, | |
| "valid_targets_mean": 1869.7, | |
| "valid_targets_min": 844 | |
| }, | |
| { | |
| "epoch": 3.8970588235294117, | |
| "grad_norm": 0.5272383246437135, | |
| "learning_rate": 1.9540485956956166e-05, | |
| "loss": 0.1208, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12583494186401367, | |
| "step": 3975, | |
| "valid_targets_mean": 2390.6, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 3.9019607843137254, | |
| "grad_norm": 0.49411797125684626, | |
| "learning_rate": 1.9491611509404156e-05, | |
| "loss": 0.1092, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11986801028251648, | |
| "step": 3980, | |
| "valid_targets_mean": 2076.8, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 3.906862745098039, | |
| "grad_norm": 0.5683723121360503, | |
| "learning_rate": 1.9442740099614573e-05, | |
| "loss": 0.1286, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12163309752941132, | |
| "step": 3985, | |
| "valid_targets_mean": 2084.8, | |
| "valid_targets_min": 949 | |
| }, | |
| { | |
| "epoch": 3.911764705882353, | |
| "grad_norm": 0.4366403979214397, | |
| "learning_rate": 1.9393872019607658e-05, | |
| "loss": 0.1139, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10083067417144775, | |
| "step": 3990, | |
| "valid_targets_mean": 2892.9, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 3.9166666666666665, | |
| "grad_norm": 0.47443132807924027, | |
| "learning_rate": 1.9345007561383755e-05, | |
| "loss": 0.0997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09945064783096313, | |
| "step": 3995, | |
| "valid_targets_mean": 2261.6, | |
| "valid_targets_min": 998 | |
| }, | |
| { | |
| "epoch": 3.9215686274509802, | |
| "grad_norm": 0.6756799450871543, | |
| "learning_rate": 1.929614701692159e-05, | |
| "loss": 0.1219, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1440548300743103, | |
| "step": 4000, | |
| "valid_targets_mean": 1480.9, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 3.9264705882352944, | |
| "grad_norm": 0.6431235449873466, | |
| "learning_rate": 1.924729067817646e-05, | |
| "loss": 0.1093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12017422914505005, | |
| "step": 4005, | |
| "valid_targets_mean": 1699.6, | |
| "valid_targets_min": 521 | |
| }, | |
| { | |
| "epoch": 3.931372549019608, | |
| "grad_norm": 0.5361794880222236, | |
| "learning_rate": 1.919843883707858e-05, | |
| "loss": 0.1043, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12894302606582642, | |
| "step": 4010, | |
| "valid_targets_mean": 2346.9, | |
| "valid_targets_min": 810 | |
| }, | |
| { | |
| "epoch": 3.936274509803922, | |
| "grad_norm": 0.5846150875543148, | |
| "learning_rate": 1.9149591785531256e-05, | |
| "loss": 0.1127, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12245102971792221, | |
| "step": 4015, | |
| "valid_targets_mean": 1880.8, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 3.9411764705882355, | |
| "grad_norm": 0.6251138407178463, | |
| "learning_rate": 1.9100749815409185e-05, | |
| "loss": 0.1075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1449219137430191, | |
| "step": 4020, | |
| "valid_targets_mean": 1870.9, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 3.946078431372549, | |
| "grad_norm": 0.4889900134198115, | |
| "learning_rate": 1.905191321855671e-05, | |
| "loss": 0.1081, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10330167412757874, | |
| "step": 4025, | |
| "valid_targets_mean": 2078.4, | |
| "valid_targets_min": 695 | |
| }, | |
| { | |
| "epoch": 3.950980392156863, | |
| "grad_norm": 0.4651475351048309, | |
| "learning_rate": 1.9003082286786056e-05, | |
| "loss": 0.1115, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1023440957069397, | |
| "step": 4030, | |
| "valid_targets_mean": 2225.9, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 3.9558823529411766, | |
| "grad_norm": 0.4021607738658494, | |
| "learning_rate": 1.895425731187561e-05, | |
| "loss": 0.0951, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08778274804353714, | |
| "step": 4035, | |
| "valid_targets_mean": 2479.3, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 3.9607843137254903, | |
| "grad_norm": 0.6217464371869531, | |
| "learning_rate": 1.8905438585568148e-05, | |
| "loss": 0.1233, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12868231534957886, | |
| "step": 4040, | |
| "valid_targets_mean": 1891.4, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 3.965686274509804, | |
| "grad_norm": 0.4856130983795896, | |
| "learning_rate": 1.8856626399569114e-05, | |
| "loss": 0.1046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.112669438123703, | |
| "step": 4045, | |
| "valid_targets_mean": 2243.1, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 3.9705882352941178, | |
| "grad_norm": 0.4750769123731014, | |
| "learning_rate": 1.880782104554489e-05, | |
| "loss": 0.098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09736385941505432, | |
| "step": 4050, | |
| "valid_targets_mean": 2243.9, | |
| "valid_targets_min": 482 | |
| }, | |
| { | |
| "epoch": 3.9754901960784315, | |
| "grad_norm": 0.405301707427997, | |
| "learning_rate": 1.875902281512101e-05, | |
| "loss": 0.1037, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08424034714698792, | |
| "step": 4055, | |
| "valid_targets_mean": 2575.7, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 3.980392156862745, | |
| "grad_norm": 0.4561960422936826, | |
| "learning_rate": 1.8710231999880446e-05, | |
| "loss": 0.1052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09456402063369751, | |
| "step": 4060, | |
| "valid_targets_mean": 2422.7, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 3.985294117647059, | |
| "grad_norm": 0.3696563343388035, | |
| "learning_rate": 1.866144889136189e-05, | |
| "loss": 0.0976, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07512452453374863, | |
| "step": 4065, | |
| "valid_targets_mean": 2838.2, | |
| "valid_targets_min": 823 | |
| }, | |
| { | |
| "epoch": 3.9901960784313726, | |
| "grad_norm": 0.46963438097328625, | |
| "learning_rate": 1.861267378105795e-05, | |
| "loss": 0.1116, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12728336453437805, | |
| "step": 4070, | |
| "valid_targets_mean": 2365.9, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 3.9950980392156863, | |
| "grad_norm": 0.4803268550916603, | |
| "learning_rate": 1.856390696041345e-05, | |
| "loss": 0.1158, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10719768702983856, | |
| "step": 4075, | |
| "valid_targets_mean": 2434.1, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.5293457714014153, | |
| "learning_rate": 1.8515148720823703e-05, | |
| "loss": 0.1031, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10042376816272736, | |
| "step": 4080, | |
| "valid_targets_mean": 1962.6, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 4.004901960784314, | |
| "grad_norm": 0.5644338107149101, | |
| "learning_rate": 1.846639935363272e-05, | |
| "loss": 0.0947, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09875704348087311, | |
| "step": 4085, | |
| "valid_targets_mean": 1990.8, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 4.009803921568627, | |
| "grad_norm": 0.4444963540738002, | |
| "learning_rate": 1.8417659150131507e-05, | |
| "loss": 0.0906, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07949922233819962, | |
| "step": 4090, | |
| "valid_targets_mean": 2462.4, | |
| "valid_targets_min": 472 | |
| }, | |
| { | |
| "epoch": 4.014705882352941, | |
| "grad_norm": 0.5230285801588905, | |
| "learning_rate": 1.8368928401556334e-05, | |
| "loss": 0.0913, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09187965095043182, | |
| "step": 4095, | |
| "valid_targets_mean": 2216.2, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 4.019607843137255, | |
| "grad_norm": 0.444478652714691, | |
| "learning_rate": 1.8320207399086936e-05, | |
| "loss": 0.091, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08924181759357452, | |
| "step": 4100, | |
| "valid_targets_mean": 2502.8, | |
| "valid_targets_min": 520 | |
| }, | |
| { | |
| "epoch": 4.0245098039215685, | |
| "grad_norm": 0.4467580920948847, | |
| "learning_rate": 1.827149643384484e-05, | |
| "loss": 0.0838, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0863981693983078, | |
| "step": 4105, | |
| "valid_targets_mean": 3037.9, | |
| "valid_targets_min": 960 | |
| }, | |
| { | |
| "epoch": 4.029411764705882, | |
| "grad_norm": 0.5074124961478061, | |
| "learning_rate": 1.8222795796891596e-05, | |
| "loss": 0.093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09525004774332047, | |
| "step": 4110, | |
| "valid_targets_mean": 2213.2, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 4.034313725490196, | |
| "grad_norm": 0.4688138171504505, | |
| "learning_rate": 1.8174105779227038e-05, | |
| "loss": 0.0834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07826852798461914, | |
| "step": 4115, | |
| "valid_targets_mean": 2524.2, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 4.03921568627451, | |
| "grad_norm": 0.5495964133592056, | |
| "learning_rate": 1.8125426671787535e-05, | |
| "loss": 0.0829, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08855336904525757, | |
| "step": 4120, | |
| "valid_targets_mean": 2184.7, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 4.044117647058823, | |
| "grad_norm": 0.6649719549369908, | |
| "learning_rate": 1.8076758765444272e-05, | |
| "loss": 0.0954, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10872084647417068, | |
| "step": 4125, | |
| "valid_targets_mean": 1982.1, | |
| "valid_targets_min": 905 | |
| }, | |
| { | |
| "epoch": 4.049019607843137, | |
| "grad_norm": 0.558854630746821, | |
| "learning_rate": 1.8028102351001515e-05, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07840010523796082, | |
| "step": 4130, | |
| "valid_targets_mean": 2404.1, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 4.053921568627451, | |
| "grad_norm": 0.5709714861960441, | |
| "learning_rate": 1.7979457719194847e-05, | |
| "loss": 0.1013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09838423132896423, | |
| "step": 4135, | |
| "valid_targets_mean": 2030.4, | |
| "valid_targets_min": 715 | |
| }, | |
| { | |
| "epoch": 4.0588235294117645, | |
| "grad_norm": 0.6188829630512815, | |
| "learning_rate": 1.7930825160689448e-05, | |
| "loss": 0.1066, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14148059487342834, | |
| "step": 4140, | |
| "valid_targets_mean": 1973.7, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 4.063725490196078, | |
| "grad_norm": 0.5426801567468893, | |
| "learning_rate": 1.7882204966078377e-05, | |
| "loss": 0.1098, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09950780123472214, | |
| "step": 4145, | |
| "valid_targets_mean": 2231.3, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 4.068627450980392, | |
| "grad_norm": 0.47654840956847533, | |
| "learning_rate": 1.7833597425880786e-05, | |
| "loss": 0.1008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0965999960899353, | |
| "step": 4150, | |
| "valid_targets_mean": 2403.3, | |
| "valid_targets_min": 712 | |
| }, | |
| { | |
| "epoch": 4.073529411764706, | |
| "grad_norm": 0.5293696212843991, | |
| "learning_rate": 1.7785002830540225e-05, | |
| "loss": 0.0871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08954203128814697, | |
| "step": 4155, | |
| "valid_targets_mean": 2344.8, | |
| "valid_targets_min": 538 | |
| }, | |
| { | |
| "epoch": 4.078431372549019, | |
| "grad_norm": 0.5975080969452089, | |
| "learning_rate": 1.7736421470422915e-05, | |
| "loss": 0.0865, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09741682559251785, | |
| "step": 4160, | |
| "valid_targets_mean": 2047.7, | |
| "valid_targets_min": 660 | |
| }, | |
| { | |
| "epoch": 4.083333333333333, | |
| "grad_norm": 0.48604759979340684, | |
| "learning_rate": 1.7687853635815953e-05, | |
| "loss": 0.0943, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08825595676898956, | |
| "step": 4165, | |
| "valid_targets_mean": 2451.2, | |
| "valid_targets_min": 1195 | |
| }, | |
| { | |
| "epoch": 4.088235294117647, | |
| "grad_norm": 0.5268849125233499, | |
| "learning_rate": 1.7639299616925645e-05, | |
| "loss": 0.0897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.13385260105133057, | |
| "step": 4170, | |
| "valid_targets_mean": 2277.0, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 4.0931372549019605, | |
| "grad_norm": 0.492794433426126, | |
| "learning_rate": 1.7590759703875752e-05, | |
| "loss": 0.0926, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10988749563694, | |
| "step": 4175, | |
| "valid_targets_mean": 2417.8, | |
| "valid_targets_min": 847 | |
| }, | |
| { | |
| "epoch": 4.098039215686274, | |
| "grad_norm": 0.6013133531068834, | |
| "learning_rate": 1.7542234186705722e-05, | |
| "loss": 0.0941, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09224649518728256, | |
| "step": 4180, | |
| "valid_targets_mean": 1765.2, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 4.102941176470588, | |
| "grad_norm": 0.7495395672799343, | |
| "learning_rate": 1.7493723355368995e-05, | |
| "loss": 0.1011, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10524407029151917, | |
| "step": 4185, | |
| "valid_targets_mean": 1362.1, | |
| "valid_targets_min": 595 | |
| }, | |
| { | |
| "epoch": 4.107843137254902, | |
| "grad_norm": 0.6364621332679006, | |
| "learning_rate": 1.744522749973127e-05, | |
| "loss": 0.1011, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08376060426235199, | |
| "step": 4190, | |
| "valid_targets_mean": 1726.0, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 4.112745098039215, | |
| "grad_norm": 0.5201856732191488, | |
| "learning_rate": 1.7396746909568748e-05, | |
| "loss": 0.0934, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0875372588634491, | |
| "step": 4195, | |
| "valid_targets_mean": 2352.5, | |
| "valid_targets_min": 541 | |
| }, | |
| { | |
| "epoch": 4.117647058823529, | |
| "grad_norm": 0.49596492927230623, | |
| "learning_rate": 1.7348281874566417e-05, | |
| "loss": 0.0909, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08408987522125244, | |
| "step": 4200, | |
| "valid_targets_mean": 2508.0, | |
| "valid_targets_min": 891 | |
| }, | |
| { | |
| "epoch": 4.122549019607843, | |
| "grad_norm": 0.5693396360569704, | |
| "learning_rate": 1.7299832684316323e-05, | |
| "loss": 0.085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09746822714805603, | |
| "step": 4205, | |
| "valid_targets_mean": 2194.4, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 4.127450980392156, | |
| "grad_norm": 0.43092782426008985, | |
| "learning_rate": 1.725139962831584e-05, | |
| "loss": 0.0917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09253295511007309, | |
| "step": 4210, | |
| "valid_targets_mean": 2626.4, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 4.132352941176471, | |
| "grad_norm": 0.4724890600327457, | |
| "learning_rate": 1.7202982995965912e-05, | |
| "loss": 0.0948, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0815335363149643, | |
| "step": 4215, | |
| "valid_targets_mean": 2930.8, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 4.137254901960785, | |
| "grad_norm": 0.5689305179379935, | |
| "learning_rate": 1.7154583076569376e-05, | |
| "loss": 0.0986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08635899424552917, | |
| "step": 4220, | |
| "valid_targets_mean": 2238.6, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 4.142156862745098, | |
| "grad_norm": 0.47177819632108414, | |
| "learning_rate": 1.7106200159329197e-05, | |
| "loss": 0.0872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08042912185192108, | |
| "step": 4225, | |
| "valid_targets_mean": 2839.6, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 4.147058823529412, | |
| "grad_norm": 0.48065078368458986, | |
| "learning_rate": 1.705783453334673e-05, | |
| "loss": 0.0943, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08426964282989502, | |
| "step": 4230, | |
| "valid_targets_mean": 2610.4, | |
| "valid_targets_min": 950 | |
| }, | |
| { | |
| "epoch": 4.151960784313726, | |
| "grad_norm": 0.6553478415791882, | |
| "learning_rate": 1.7009486487620024e-05, | |
| "loss": 0.1046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09982205182313919, | |
| "step": 4235, | |
| "valid_targets_mean": 1791.4, | |
| "valid_targets_min": 662 | |
| }, | |
| { | |
| "epoch": 4.1568627450980395, | |
| "grad_norm": 0.6041270719163585, | |
| "learning_rate": 1.6961156311042085e-05, | |
| "loss": 0.0959, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11420536041259766, | |
| "step": 4240, | |
| "valid_targets_mean": 2226.8, | |
| "valid_targets_min": 801 | |
| }, | |
| { | |
| "epoch": 4.161764705882353, | |
| "grad_norm": 0.5619416375343212, | |
| "learning_rate": 1.6912844292399143e-05, | |
| "loss": 0.0955, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0868932455778122, | |
| "step": 4245, | |
| "valid_targets_mean": 1945.8, | |
| "valid_targets_min": 829 | |
| }, | |
| { | |
| "epoch": 4.166666666666667, | |
| "grad_norm": 0.6065314115589832, | |
| "learning_rate": 1.6864550720368915e-05, | |
| "loss": 0.0953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11492280662059784, | |
| "step": 4250, | |
| "valid_targets_mean": 2105.5, | |
| "valid_targets_min": 907 | |
| }, | |
| { | |
| "epoch": 4.171568627450981, | |
| "grad_norm": 0.5544175260955679, | |
| "learning_rate": 1.6816275883518917e-05, | |
| "loss": 0.0886, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08502060174942017, | |
| "step": 4255, | |
| "valid_targets_mean": 2211.4, | |
| "valid_targets_min": 515 | |
| }, | |
| { | |
| "epoch": 4.176470588235294, | |
| "grad_norm": 0.5310513730362705, | |
| "learning_rate": 1.6768020070304704e-05, | |
| "loss": 0.0905, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09640628099441528, | |
| "step": 4260, | |
| "valid_targets_mean": 2702.8, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 4.181372549019608, | |
| "grad_norm": 0.7288794320718324, | |
| "learning_rate": 1.6719783569068154e-05, | |
| "loss": 0.0955, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10868661850690842, | |
| "step": 4265, | |
| "valid_targets_mean": 1660.6, | |
| "valid_targets_min": 524 | |
| }, | |
| { | |
| "epoch": 4.186274509803922, | |
| "grad_norm": 0.49490583836739976, | |
| "learning_rate": 1.6671566668035765e-05, | |
| "loss": 0.094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08493231236934662, | |
| "step": 4270, | |
| "valid_targets_mean": 2460.1, | |
| "valid_targets_min": 485 | |
| }, | |
| { | |
| "epoch": 4.1911764705882355, | |
| "grad_norm": 0.548822021790285, | |
| "learning_rate": 1.6623369655316915e-05, | |
| "loss": 0.0975, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09106888622045517, | |
| "step": 4275, | |
| "valid_targets_mean": 2058.1, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 4.196078431372549, | |
| "grad_norm": 0.5580145934153147, | |
| "learning_rate": 1.657519281890213e-05, | |
| "loss": 0.0871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08765994012355804, | |
| "step": 4280, | |
| "valid_targets_mean": 1777.9, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 4.200980392156863, | |
| "grad_norm": 0.45822454413438735, | |
| "learning_rate": 1.6527036446661396e-05, | |
| "loss": 0.0929, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08106301724910736, | |
| "step": 4285, | |
| "valid_targets_mean": 3128.1, | |
| "valid_targets_min": 988 | |
| }, | |
| { | |
| "epoch": 4.205882352941177, | |
| "grad_norm": 0.6180110752258359, | |
| "learning_rate": 1.647890082634241e-05, | |
| "loss": 0.096, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1506294459104538, | |
| "step": 4290, | |
| "valid_targets_mean": 2282.9, | |
| "valid_targets_min": 867 | |
| }, | |
| { | |
| "epoch": 4.21078431372549, | |
| "grad_norm": 0.5445669104444656, | |
| "learning_rate": 1.6430786245568865e-05, | |
| "loss": 0.0844, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09008940309286118, | |
| "step": 4295, | |
| "valid_targets_mean": 2389.7, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 4.215686274509804, | |
| "grad_norm": 0.5055418184306785, | |
| "learning_rate": 1.638269299183875e-05, | |
| "loss": 0.0932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08286339044570923, | |
| "step": 4300, | |
| "valid_targets_mean": 2237.7, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 4.220588235294118, | |
| "grad_norm": 0.453249556971749, | |
| "learning_rate": 1.6334621352522606e-05, | |
| "loss": 0.0912, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09033799916505814, | |
| "step": 4305, | |
| "valid_targets_mean": 2639.6, | |
| "valid_targets_min": 695 | |
| }, | |
| { | |
| "epoch": 4.2254901960784315, | |
| "grad_norm": 0.6613125276955629, | |
| "learning_rate": 1.6286571614861845e-05, | |
| "loss": 0.1013, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09440892934799194, | |
| "step": 4310, | |
| "valid_targets_mean": 1669.6, | |
| "valid_targets_min": 625 | |
| }, | |
| { | |
| "epoch": 4.230392156862745, | |
| "grad_norm": 0.5980358240831692, | |
| "learning_rate": 1.6238544065966974e-05, | |
| "loss": 0.0972, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09913182258605957, | |
| "step": 4315, | |
| "valid_targets_mean": 1975.6, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 4.235294117647059, | |
| "grad_norm": 0.511127190936502, | |
| "learning_rate": 1.6190538992815945e-05, | |
| "loss": 0.1012, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11423664540052414, | |
| "step": 4320, | |
| "valid_targets_mean": 2052.5, | |
| "valid_targets_min": 541 | |
| }, | |
| { | |
| "epoch": 4.240196078431373, | |
| "grad_norm": 0.5626942407060462, | |
| "learning_rate": 1.6142556682252404e-05, | |
| "loss": 0.0953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09917297214269638, | |
| "step": 4325, | |
| "valid_targets_mean": 2183.7, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 4.245098039215686, | |
| "grad_norm": 0.6911199917714397, | |
| "learning_rate": 1.609459742098398e-05, | |
| "loss": 0.1068, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.14647285640239716, | |
| "step": 4330, | |
| "valid_targets_mean": 1848.5, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "grad_norm": 0.531542801596997, | |
| "learning_rate": 1.604666149558057e-05, | |
| "loss": 0.0868, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09663405269384384, | |
| "step": 4335, | |
| "valid_targets_mean": 2882.6, | |
| "valid_targets_min": 1324 | |
| }, | |
| { | |
| "epoch": 4.254901960784314, | |
| "grad_norm": 0.5773829854027778, | |
| "learning_rate": 1.5998749192472666e-05, | |
| "loss": 0.1095, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11062102019786835, | |
| "step": 4340, | |
| "valid_targets_mean": 1834.8, | |
| "valid_targets_min": 447 | |
| }, | |
| { | |
| "epoch": 4.259803921568627, | |
| "grad_norm": 0.5691933965635515, | |
| "learning_rate": 1.595086079794955e-05, | |
| "loss": 0.0996, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08927673101425171, | |
| "step": 4345, | |
| "valid_targets_mean": 1940.2, | |
| "valid_targets_min": 542 | |
| }, | |
| { | |
| "epoch": 4.264705882352941, | |
| "grad_norm": 0.46575860504404326, | |
| "learning_rate": 1.5902996598157705e-05, | |
| "loss": 0.0976, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09074290096759796, | |
| "step": 4350, | |
| "valid_targets_mean": 2670.6, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 4.269607843137255, | |
| "grad_norm": 0.4873669002421505, | |
| "learning_rate": 1.5855156879099012e-05, | |
| "loss": 0.093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08654345571994781, | |
| "step": 4355, | |
| "valid_targets_mean": 2558.1, | |
| "valid_targets_min": 963 | |
| }, | |
| { | |
| "epoch": 4.2745098039215685, | |
| "grad_norm": 0.57544806936859, | |
| "learning_rate": 1.5807341926629073e-05, | |
| "loss": 0.0922, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10240624845027924, | |
| "step": 4360, | |
| "valid_targets_mean": 2204.3, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 4.279411764705882, | |
| "grad_norm": 0.667247621859043, | |
| "learning_rate": 1.575955202645551e-05, | |
| "loss": 0.1035, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11562317609786987, | |
| "step": 4365, | |
| "valid_targets_mean": 2166.3, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 4.284313725490196, | |
| "grad_norm": 0.5322428960008451, | |
| "learning_rate": 1.571178746413625e-05, | |
| "loss": 0.0957, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09002898633480072, | |
| "step": 4370, | |
| "valid_targets_mean": 2169.4, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 4.28921568627451, | |
| "grad_norm": 0.44379666480899166, | |
| "learning_rate": 1.5664048525077827e-05, | |
| "loss": 0.0948, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07158397883176804, | |
| "step": 4375, | |
| "valid_targets_mean": 3081.1, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 4.294117647058823, | |
| "grad_norm": 0.5134899363308195, | |
| "learning_rate": 1.5616335494533644e-05, | |
| "loss": 0.1019, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.085150808095932, | |
| "step": 4380, | |
| "valid_targets_mean": 2649.4, | |
| "valid_targets_min": 910 | |
| }, | |
| { | |
| "epoch": 4.299019607843137, | |
| "grad_norm": 0.5633333659303986, | |
| "learning_rate": 1.5568648657602316e-05, | |
| "loss": 0.0902, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08845022320747375, | |
| "step": 4385, | |
| "valid_targets_mean": 1813.4, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 4.303921568627451, | |
| "grad_norm": 0.4207910793387121, | |
| "learning_rate": 1.5520988299225942e-05, | |
| "loss": 0.0994, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08506350219249725, | |
| "step": 4390, | |
| "valid_targets_mean": 3005.1, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 4.3088235294117645, | |
| "grad_norm": 0.5975251690244275, | |
| "learning_rate": 1.5473354704188397e-05, | |
| "loss": 0.1012, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08913823962211609, | |
| "step": 4395, | |
| "valid_targets_mean": 1923.8, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 4.313725490196078, | |
| "grad_norm": 0.5423697805466529, | |
| "learning_rate": 1.5425748157113632e-05, | |
| "loss": 0.1007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1019233986735344, | |
| "step": 4400, | |
| "valid_targets_mean": 2368.1, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 4.318627450980392, | |
| "grad_norm": 0.5387628160624569, | |
| "learning_rate": 1.5378168942464003e-05, | |
| "loss": 0.0852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09121586382389069, | |
| "step": 4405, | |
| "valid_targets_mean": 2608.4, | |
| "valid_targets_min": 596 | |
| }, | |
| { | |
| "epoch": 4.323529411764706, | |
| "grad_norm": 0.5633768847500502, | |
| "learning_rate": 1.5330617344538517e-05, | |
| "loss": 0.0932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08939993381500244, | |
| "step": 4410, | |
| "valid_targets_mean": 1958.5, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 4.328431372549019, | |
| "grad_norm": 0.7836118578665316, | |
| "learning_rate": 1.528309364747118e-05, | |
| "loss": 0.0949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12606091797351837, | |
| "step": 4415, | |
| "valid_targets_mean": 1545.8, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 4.333333333333333, | |
| "grad_norm": 0.7139630949750333, | |
| "learning_rate": 1.5235598135229291e-05, | |
| "loss": 0.0998, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10998935997486115, | |
| "step": 4420, | |
| "valid_targets_mean": 2041.0, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 4.338235294117647, | |
| "grad_norm": 0.6463858784540754, | |
| "learning_rate": 1.5188131091611715e-05, | |
| "loss": 0.0944, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11081689596176147, | |
| "step": 4425, | |
| "valid_targets_mean": 1975.4, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 4.3431372549019605, | |
| "grad_norm": 0.6052231565253691, | |
| "learning_rate": 1.5140692800247218e-05, | |
| "loss": 0.1046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09583979099988937, | |
| "step": 4430, | |
| "valid_targets_mean": 1846.4, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 4.348039215686274, | |
| "grad_norm": 0.46120014193765946, | |
| "learning_rate": 1.5093283544592776e-05, | |
| "loss": 0.0908, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08437848836183548, | |
| "step": 4435, | |
| "valid_targets_mean": 2861.5, | |
| "valid_targets_min": 511 | |
| }, | |
| { | |
| "epoch": 4.352941176470588, | |
| "grad_norm": 0.4357769914711236, | |
| "learning_rate": 1.504590360793186e-05, | |
| "loss": 0.0875, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07600148022174835, | |
| "step": 4440, | |
| "valid_targets_mean": 2830.0, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 4.357843137254902, | |
| "grad_norm": 0.7829148140428298, | |
| "learning_rate": 1.4998553273372732e-05, | |
| "loss": 0.0873, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10139364749193192, | |
| "step": 4445, | |
| "valid_targets_mean": 1361.4, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 4.362745098039215, | |
| "grad_norm": 0.5434292008208409, | |
| "learning_rate": 1.4951232823846807e-05, | |
| "loss": 0.0965, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07670040428638458, | |
| "step": 4450, | |
| "valid_targets_mean": 2044.3, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 4.367647058823529, | |
| "grad_norm": 0.45939612712149325, | |
| "learning_rate": 1.490394254210691e-05, | |
| "loss": 0.0988, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07639499008655548, | |
| "step": 4455, | |
| "valid_targets_mean": 2279.1, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 4.372549019607844, | |
| "grad_norm": 0.5477638681309787, | |
| "learning_rate": 1.4856682710725605e-05, | |
| "loss": 0.0937, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09940779954195023, | |
| "step": 4460, | |
| "valid_targets_mean": 2314.7, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 4.377450980392156, | |
| "grad_norm": 0.6193221420104923, | |
| "learning_rate": 1.4809453612093507e-05, | |
| "loss": 0.1, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12221089005470276, | |
| "step": 4465, | |
| "valid_targets_mean": 2122.3, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 4.382352941176471, | |
| "grad_norm": 0.5363699703866529, | |
| "learning_rate": 1.4762255528417609e-05, | |
| "loss": 0.0823, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0714312195777893, | |
| "step": 4470, | |
| "valid_targets_mean": 2175.7, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 4.387254901960785, | |
| "grad_norm": 0.3973947214228722, | |
| "learning_rate": 1.4715088741719555e-05, | |
| "loss": 0.0802, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06857871264219284, | |
| "step": 4475, | |
| "valid_targets_mean": 3006.4, | |
| "valid_targets_min": 739 | |
| }, | |
| { | |
| "epoch": 4.392156862745098, | |
| "grad_norm": 0.6548559347394757, | |
| "learning_rate": 1.4667953533833994e-05, | |
| "loss": 0.093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10961537063121796, | |
| "step": 4480, | |
| "valid_targets_mean": 1610.0, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 4.397058823529412, | |
| "grad_norm": 0.6143204977429464, | |
| "learning_rate": 1.46208501864069e-05, | |
| "loss": 0.094, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12123645842075348, | |
| "step": 4485, | |
| "valid_targets_mean": 2003.9, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 4.401960784313726, | |
| "grad_norm": 0.5182077364235051, | |
| "learning_rate": 1.4573778980893842e-05, | |
| "loss": 0.0843, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08992286771535873, | |
| "step": 4490, | |
| "valid_targets_mean": 2517.8, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 4.4068627450980395, | |
| "grad_norm": 0.48022364706715087, | |
| "learning_rate": 1.4526740198558345e-05, | |
| "loss": 0.095, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07461203634738922, | |
| "step": 4495, | |
| "valid_targets_mean": 2504.1, | |
| "valid_targets_min": 387 | |
| }, | |
| { | |
| "epoch": 4.411764705882353, | |
| "grad_norm": 0.6520442662505085, | |
| "learning_rate": 1.4479734120470202e-05, | |
| "loss": 0.0959, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.089952252805233, | |
| "step": 4500, | |
| "valid_targets_mean": 2090.4, | |
| "valid_targets_min": 898 | |
| }, | |
| { | |
| "epoch": 4.416666666666667, | |
| "grad_norm": 0.6404826866372381, | |
| "learning_rate": 1.4432761027503788e-05, | |
| "loss": 0.0964, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10904596745967865, | |
| "step": 4505, | |
| "valid_targets_mean": 1745.3, | |
| "valid_targets_min": 350 | |
| }, | |
| { | |
| "epoch": 4.421568627450981, | |
| "grad_norm": 0.5056249456016008, | |
| "learning_rate": 1.4385821200336368e-05, | |
| "loss": 0.0866, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08835747838020325, | |
| "step": 4510, | |
| "valid_targets_mean": 2380.8, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 4.426470588235294, | |
| "grad_norm": 0.6402655121728962, | |
| "learning_rate": 1.4338914919446452e-05, | |
| "loss": 0.0957, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11426950991153717, | |
| "step": 4515, | |
| "valid_targets_mean": 1765.1, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 4.431372549019608, | |
| "grad_norm": 0.48332863889117145, | |
| "learning_rate": 1.42920424651121e-05, | |
| "loss": 0.0891, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07937314361333847, | |
| "step": 4520, | |
| "valid_targets_mean": 2311.2, | |
| "valid_targets_min": 539 | |
| }, | |
| { | |
| "epoch": 4.436274509803922, | |
| "grad_norm": 0.583535718159293, | |
| "learning_rate": 1.424520411740923e-05, | |
| "loss": 0.0922, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08863136917352676, | |
| "step": 4525, | |
| "valid_targets_mean": 2048.4, | |
| "valid_targets_min": 907 | |
| }, | |
| { | |
| "epoch": 4.4411764705882355, | |
| "grad_norm": 0.4337974903941218, | |
| "learning_rate": 1.419840015620999e-05, | |
| "loss": 0.0938, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09409506618976593, | |
| "step": 4530, | |
| "valid_targets_mean": 2672.8, | |
| "valid_targets_min": 694 | |
| }, | |
| { | |
| "epoch": 4.446078431372549, | |
| "grad_norm": 0.5747359814706138, | |
| "learning_rate": 1.415163086118105e-05, | |
| "loss": 0.1089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09305723011493683, | |
| "step": 4535, | |
| "valid_targets_mean": 2288.3, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 4.450980392156863, | |
| "grad_norm": 0.5368082314568046, | |
| "learning_rate": 1.4104896511781916e-05, | |
| "loss": 0.0956, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08980883657932281, | |
| "step": 4540, | |
| "valid_targets_mean": 2108.5, | |
| "valid_targets_min": 679 | |
| }, | |
| { | |
| "epoch": 4.455882352941177, | |
| "grad_norm": 0.4427323912880307, | |
| "learning_rate": 1.4058197387263327e-05, | |
| "loss": 0.0827, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07779604941606522, | |
| "step": 4545, | |
| "valid_targets_mean": 2846.7, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 4.46078431372549, | |
| "grad_norm": 0.5754395884162968, | |
| "learning_rate": 1.4011533766665512e-05, | |
| "loss": 0.0932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11345474421977997, | |
| "step": 4550, | |
| "valid_targets_mean": 2048.6, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 4.465686274509804, | |
| "grad_norm": 0.7338185572895771, | |
| "learning_rate": 1.396490592881656e-05, | |
| "loss": 0.0909, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09632228314876556, | |
| "step": 4555, | |
| "valid_targets_mean": 1559.9, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 4.470588235294118, | |
| "grad_norm": 0.6497231280991327, | |
| "learning_rate": 1.3918314152330758e-05, | |
| "loss": 0.0973, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09701061248779297, | |
| "step": 4560, | |
| "valid_targets_mean": 1984.2, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 4.4754901960784315, | |
| "grad_norm": 0.4926967911541651, | |
| "learning_rate": 1.3871758715606905e-05, | |
| "loss": 0.0932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0813327506184578, | |
| "step": 4565, | |
| "valid_targets_mean": 2670.1, | |
| "valid_targets_min": 762 | |
| }, | |
| { | |
| "epoch": 4.480392156862745, | |
| "grad_norm": 0.6155351420308762, | |
| "learning_rate": 1.3825239896826678e-05, | |
| "loss": 0.0938, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09319113940000534, | |
| "step": 4570, | |
| "valid_targets_mean": 1925.0, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 4.485294117647059, | |
| "grad_norm": 0.5286966094801919, | |
| "learning_rate": 1.3778757973952916e-05, | |
| "loss": 0.0976, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10444878786802292, | |
| "step": 4575, | |
| "valid_targets_mean": 2229.3, | |
| "valid_targets_min": 616 | |
| }, | |
| { | |
| "epoch": 4.490196078431373, | |
| "grad_norm": 0.6034568137100172, | |
| "learning_rate": 1.3732313224728034e-05, | |
| "loss": 0.0967, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08842505514621735, | |
| "step": 4580, | |
| "valid_targets_mean": 2529.2, | |
| "valid_targets_min": 823 | |
| }, | |
| { | |
| "epoch": 4.495098039215686, | |
| "grad_norm": 0.49921650284434393, | |
| "learning_rate": 1.3685905926672307e-05, | |
| "loss": 0.0918, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09927654266357422, | |
| "step": 4585, | |
| "valid_targets_mean": 2443.6, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "grad_norm": 0.5430297012879352, | |
| "learning_rate": 1.3639536357082213e-05, | |
| "loss": 0.1082, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07649785280227661, | |
| "step": 4590, | |
| "valid_targets_mean": 2006.4, | |
| "valid_targets_min": 664 | |
| }, | |
| { | |
| "epoch": 4.504901960784314, | |
| "grad_norm": 0.5544828568038234, | |
| "learning_rate": 1.3593204793028827e-05, | |
| "loss": 0.0939, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10276834666728973, | |
| "step": 4595, | |
| "valid_targets_mean": 2263.1, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 4.509803921568627, | |
| "grad_norm": 0.4257567508568936, | |
| "learning_rate": 1.3546911511356109e-05, | |
| "loss": 0.0873, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06598569452762604, | |
| "step": 4600, | |
| "valid_targets_mean": 2500.6, | |
| "valid_targets_min": 620 | |
| }, | |
| { | |
| "epoch": 4.514705882352941, | |
| "grad_norm": 0.5350814926710116, | |
| "learning_rate": 1.3500656788679259e-05, | |
| "loss": 0.0939, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08269689232110977, | |
| "step": 4605, | |
| "valid_targets_mean": 2247.2, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 4.519607843137255, | |
| "grad_norm": 0.5071553325687435, | |
| "learning_rate": 1.3454440901383105e-05, | |
| "loss": 0.0993, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08212989568710327, | |
| "step": 4610, | |
| "valid_targets_mean": 2416.6, | |
| "valid_targets_min": 752 | |
| }, | |
| { | |
| "epoch": 4.5245098039215685, | |
| "grad_norm": 0.48521769183282765, | |
| "learning_rate": 1.3408264125620409e-05, | |
| "loss": 0.0962, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09017767012119293, | |
| "step": 4615, | |
| "valid_targets_mean": 2546.1, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 4.529411764705882, | |
| "grad_norm": 0.5302744572378841, | |
| "learning_rate": 1.3362126737310218e-05, | |
| "loss": 0.0854, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07743710279464722, | |
| "step": 4620, | |
| "valid_targets_mean": 2307.6, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 4.534313725490196, | |
| "grad_norm": 0.5819255012958557, | |
| "learning_rate": 1.3316029012136251e-05, | |
| "loss": 0.0945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09885326027870178, | |
| "step": 4625, | |
| "valid_targets_mean": 2172.9, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 4.53921568627451, | |
| "grad_norm": 0.49414618949329153, | |
| "learning_rate": 1.3269971225545222e-05, | |
| "loss": 0.0852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09445025026798248, | |
| "step": 4630, | |
| "valid_targets_mean": 2684.0, | |
| "valid_targets_min": 762 | |
| }, | |
| { | |
| "epoch": 4.544117647058823, | |
| "grad_norm": 0.5097040673012775, | |
| "learning_rate": 1.322395365274518e-05, | |
| "loss": 0.0887, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09663759171962738, | |
| "step": 4635, | |
| "valid_targets_mean": 2331.1, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 4.549019607843137, | |
| "grad_norm": 0.5251723752326636, | |
| "learning_rate": 1.317797656870392e-05, | |
| "loss": 0.0918, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08932511508464813, | |
| "step": 4640, | |
| "valid_targets_mean": 2332.5, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 4.553921568627451, | |
| "grad_norm": 0.6185932099393986, | |
| "learning_rate": 1.3132040248147275e-05, | |
| "loss": 0.0883, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0976918414235115, | |
| "step": 4645, | |
| "valid_targets_mean": 2019.3, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 4.5588235294117645, | |
| "grad_norm": 0.6806589296776067, | |
| "learning_rate": 1.3086144965557533e-05, | |
| "loss": 0.0959, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11448152363300323, | |
| "step": 4650, | |
| "valid_targets_mean": 1729.6, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 4.563725490196078, | |
| "grad_norm": 0.5779608518902201, | |
| "learning_rate": 1.3040290995171741e-05, | |
| "loss": 0.0924, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07267005741596222, | |
| "step": 4655, | |
| "valid_targets_mean": 2079.3, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 4.568627450980392, | |
| "grad_norm": 0.6828527890624431, | |
| "learning_rate": 1.2994478610980105e-05, | |
| "loss": 0.1063, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11123380064964294, | |
| "step": 4660, | |
| "valid_targets_mean": 1486.1, | |
| "valid_targets_min": 807 | |
| }, | |
| { | |
| "epoch": 4.573529411764706, | |
| "grad_norm": 0.5564317445708631, | |
| "learning_rate": 1.2948708086724358e-05, | |
| "loss": 0.0981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09014254063367844, | |
| "step": 4665, | |
| "valid_targets_mean": 2018.9, | |
| "valid_targets_min": 459 | |
| }, | |
| { | |
| "epoch": 4.578431372549019, | |
| "grad_norm": 0.48228755801063306, | |
| "learning_rate": 1.2902979695896078e-05, | |
| "loss": 0.085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07801198959350586, | |
| "step": 4670, | |
| "valid_targets_mean": 2877.0, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 4.583333333333333, | |
| "grad_norm": 0.5410741347026093, | |
| "learning_rate": 1.2857293711735106e-05, | |
| "loss": 0.0936, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10315775126218796, | |
| "step": 4675, | |
| "valid_targets_mean": 2320.1, | |
| "valid_targets_min": 959 | |
| }, | |
| { | |
| "epoch": 4.588235294117647, | |
| "grad_norm": 0.6455727403969245, | |
| "learning_rate": 1.2811650407227896e-05, | |
| "loss": 0.1052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10891681909561157, | |
| "step": 4680, | |
| "valid_targets_mean": 1875.6, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 4.5931372549019605, | |
| "grad_norm": 0.6265275777081649, | |
| "learning_rate": 1.2766050055105845e-05, | |
| "loss": 0.0926, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0980343222618103, | |
| "step": 4685, | |
| "valid_targets_mean": 2058.0, | |
| "valid_targets_min": 661 | |
| }, | |
| { | |
| "epoch": 4.598039215686274, | |
| "grad_norm": 0.5365945342201405, | |
| "learning_rate": 1.2720492927843742e-05, | |
| "loss": 0.1006, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08697868138551712, | |
| "step": 4690, | |
| "valid_targets_mean": 2090.6, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 4.602941176470588, | |
| "grad_norm": 0.3810046962990753, | |
| "learning_rate": 1.267497929765807e-05, | |
| "loss": 0.0982, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0712059885263443, | |
| "step": 4695, | |
| "valid_targets_mean": 3317.7, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 4.607843137254902, | |
| "grad_norm": 0.5918911867571459, | |
| "learning_rate": 1.2629509436505396e-05, | |
| "loss": 0.0927, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0884181559085846, | |
| "step": 4700, | |
| "valid_targets_mean": 2422.4, | |
| "valid_targets_min": 566 | |
| }, | |
| { | |
| "epoch": 4.612745098039216, | |
| "grad_norm": 0.5213457877084579, | |
| "learning_rate": 1.2584083616080787e-05, | |
| "loss": 0.1024, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0884481742978096, | |
| "step": 4705, | |
| "valid_targets_mean": 2052.0, | |
| "valid_targets_min": 502 | |
| }, | |
| { | |
| "epoch": 4.617647058823529, | |
| "grad_norm": 0.567330475510558, | |
| "learning_rate": 1.2538702107816121e-05, | |
| "loss": 0.0936, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11953169852495193, | |
| "step": 4710, | |
| "valid_targets_mean": 2700.6, | |
| "valid_targets_min": 626 | |
| }, | |
| { | |
| "epoch": 4.622549019607844, | |
| "grad_norm": 0.6026523328666679, | |
| "learning_rate": 1.2493365182878527e-05, | |
| "loss": 0.0943, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09880198538303375, | |
| "step": 4715, | |
| "valid_targets_mean": 1875.8, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 4.627450980392156, | |
| "grad_norm": 0.5441784314363494, | |
| "learning_rate": 1.2448073112168716e-05, | |
| "loss": 0.0868, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09551490843296051, | |
| "step": 4720, | |
| "valid_targets_mean": 2144.7, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 4.632352941176471, | |
| "grad_norm": 0.5798838689286685, | |
| "learning_rate": 1.2402826166319382e-05, | |
| "loss": 0.0949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1071678102016449, | |
| "step": 4725, | |
| "valid_targets_mean": 2335.9, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 4.637254901960784, | |
| "grad_norm": 0.6052688653087752, | |
| "learning_rate": 1.2357624615693605e-05, | |
| "loss": 0.0942, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10750146210193634, | |
| "step": 4730, | |
| "valid_targets_mean": 2203.2, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 4.642156862745098, | |
| "grad_norm": 0.5546544496485989, | |
| "learning_rate": 1.2312468730383195e-05, | |
| "loss": 0.0969, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10274793952703476, | |
| "step": 4735, | |
| "valid_targets_mean": 2163.4, | |
| "valid_targets_min": 523 | |
| }, | |
| { | |
| "epoch": 4.647058823529412, | |
| "grad_norm": 0.5533444220155, | |
| "learning_rate": 1.2267358780207105e-05, | |
| "loss": 0.1015, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1120515763759613, | |
| "step": 4740, | |
| "valid_targets_mean": 2334.5, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 4.651960784313726, | |
| "grad_norm": 0.5676611492306864, | |
| "learning_rate": 1.2222295034709827e-05, | |
| "loss": 0.1005, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09991104900836945, | |
| "step": 4745, | |
| "valid_targets_mean": 2311.3, | |
| "valid_targets_min": 638 | |
| }, | |
| { | |
| "epoch": 4.6568627450980395, | |
| "grad_norm": 0.48304589572378387, | |
| "learning_rate": 1.2177277763159746e-05, | |
| "loss": 0.0897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07975605130195618, | |
| "step": 4750, | |
| "valid_targets_mean": 2585.1, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 4.661764705882353, | |
| "grad_norm": 0.4820375384596627, | |
| "learning_rate": 1.2132307234547565e-05, | |
| "loss": 0.1059, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12615613639354706, | |
| "step": 4755, | |
| "valid_targets_mean": 2636.7, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 4.666666666666667, | |
| "grad_norm": 0.6077582225101309, | |
| "learning_rate": 1.2087383717584684e-05, | |
| "loss": 0.0981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0969112291932106, | |
| "step": 4760, | |
| "valid_targets_mean": 2067.4, | |
| "valid_targets_min": 780 | |
| }, | |
| { | |
| "epoch": 4.671568627450981, | |
| "grad_norm": 0.5582574957902202, | |
| "learning_rate": 1.2042507480701587e-05, | |
| "loss": 0.0865, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07808205485343933, | |
| "step": 4765, | |
| "valid_targets_mean": 2271.1, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 4.676470588235294, | |
| "grad_norm": 0.5228146595740085, | |
| "learning_rate": 1.1997678792046252e-05, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10070804506540298, | |
| "step": 4770, | |
| "valid_targets_mean": 2396.7, | |
| "valid_targets_min": 768 | |
| }, | |
| { | |
| "epoch": 4.681372549019608, | |
| "grad_norm": 0.8866670649854181, | |
| "learning_rate": 1.1952897919482549e-05, | |
| "loss": 0.0996, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11867621541023254, | |
| "step": 4775, | |
| "valid_targets_mean": 1673.6, | |
| "valid_targets_min": 665 | |
| }, | |
| { | |
| "epoch": 4.686274509803922, | |
| "grad_norm": 0.7094125157085205, | |
| "learning_rate": 1.1908165130588633e-05, | |
| "loss": 0.1008, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1672782450914383, | |
| "step": 4780, | |
| "valid_targets_mean": 1552.0, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 4.6911764705882355, | |
| "grad_norm": 0.5321239377206641, | |
| "learning_rate": 1.1863480692655326e-05, | |
| "loss": 0.1093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09801702201366425, | |
| "step": 4785, | |
| "valid_targets_mean": 2317.7, | |
| "valid_targets_min": 1126 | |
| }, | |
| { | |
| "epoch": 4.696078431372549, | |
| "grad_norm": 0.5302224100951789, | |
| "learning_rate": 1.1818844872684566e-05, | |
| "loss": 0.0903, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0928771048784256, | |
| "step": 4790, | |
| "valid_targets_mean": 2279.2, | |
| "valid_targets_min": 690 | |
| }, | |
| { | |
| "epoch": 4.700980392156863, | |
| "grad_norm": 0.6024447446598006, | |
| "learning_rate": 1.1774257937387774e-05, | |
| "loss": 0.0798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08303597569465637, | |
| "step": 4795, | |
| "valid_targets_mean": 2043.1, | |
| "valid_targets_min": 342 | |
| }, | |
| { | |
| "epoch": 4.705882352941177, | |
| "grad_norm": 0.5604418830899587, | |
| "learning_rate": 1.1729720153184264e-05, | |
| "loss": 0.0945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10681148618459702, | |
| "step": 4800, | |
| "valid_targets_mean": 2133.1, | |
| "valid_targets_min": 753 | |
| }, | |
| { | |
| "epoch": 4.71078431372549, | |
| "grad_norm": 0.49573708522463433, | |
| "learning_rate": 1.1685231786199664e-05, | |
| "loss": 0.1036, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08099532872438431, | |
| "step": 4805, | |
| "valid_targets_mean": 2240.6, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 4.715686274509804, | |
| "grad_norm": 0.5709774627312741, | |
| "learning_rate": 1.164079310226434e-05, | |
| "loss": 0.089, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10106971859931946, | |
| "step": 4810, | |
| "valid_targets_mean": 2072.1, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 4.720588235294118, | |
| "grad_norm": 0.5301681363435597, | |
| "learning_rate": 1.1596404366911744e-05, | |
| "loss": 0.1003, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10889250785112381, | |
| "step": 4815, | |
| "valid_targets_mean": 2291.0, | |
| "valid_targets_min": 588 | |
| }, | |
| { | |
| "epoch": 4.7254901960784315, | |
| "grad_norm": 0.5065174062499794, | |
| "learning_rate": 1.1552065845376915e-05, | |
| "loss": 0.0987, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09046753495931625, | |
| "step": 4820, | |
| "valid_targets_mean": 2729.2, | |
| "valid_targets_min": 909 | |
| }, | |
| { | |
| "epoch": 4.730392156862745, | |
| "grad_norm": 0.46010715663863505, | |
| "learning_rate": 1.1507777802594831e-05, | |
| "loss": 0.0907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0826692283153534, | |
| "step": 4825, | |
| "valid_targets_mean": 2386.8, | |
| "valid_targets_min": 570 | |
| }, | |
| { | |
| "epoch": 4.735294117647059, | |
| "grad_norm": 0.48984283135483603, | |
| "learning_rate": 1.146354050319884e-05, | |
| "loss": 0.085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0765889436006546, | |
| "step": 4830, | |
| "valid_targets_mean": 2352.8, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 4.740196078431373, | |
| "grad_norm": 0.6000883358087854, | |
| "learning_rate": 1.1419354211519104e-05, | |
| "loss": 0.0993, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09945641458034515, | |
| "step": 4835, | |
| "valid_targets_mean": 2212.0, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 4.745098039215686, | |
| "grad_norm": 0.4978854046536584, | |
| "learning_rate": 1.1375219191580973e-05, | |
| "loss": 0.0881, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08206206560134888, | |
| "step": 4840, | |
| "valid_targets_mean": 2067.2, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "grad_norm": 0.5315947439144812, | |
| "learning_rate": 1.133113570710346e-05, | |
| "loss": 0.0984, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0920734703540802, | |
| "step": 4845, | |
| "valid_targets_mean": 1935.8, | |
| "valid_targets_min": 518 | |
| }, | |
| { | |
| "epoch": 4.754901960784314, | |
| "grad_norm": 0.4716458449566336, | |
| "learning_rate": 1.1287104021497617e-05, | |
| "loss": 0.1046, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08418993651866913, | |
| "step": 4850, | |
| "valid_targets_mean": 2616.8, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 4.759803921568627, | |
| "grad_norm": 0.5301504010187645, | |
| "learning_rate": 1.1243124397864999e-05, | |
| "loss": 0.1052, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10015548765659332, | |
| "step": 4855, | |
| "valid_targets_mean": 2534.8, | |
| "valid_targets_min": 1040 | |
| }, | |
| { | |
| "epoch": 4.764705882352941, | |
| "grad_norm": 0.4463389854552508, | |
| "learning_rate": 1.1199197098996079e-05, | |
| "loss": 0.0907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0668463408946991, | |
| "step": 4860, | |
| "valid_targets_mean": 2604.6, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 4.769607843137255, | |
| "grad_norm": 0.5763709238343063, | |
| "learning_rate": 1.1155322387368663e-05, | |
| "loss": 0.0905, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08877779543399811, | |
| "step": 4865, | |
| "valid_targets_mean": 2057.5, | |
| "valid_targets_min": 609 | |
| }, | |
| { | |
| "epoch": 4.7745098039215685, | |
| "grad_norm": 0.4666198250671742, | |
| "learning_rate": 1.1111500525146326e-05, | |
| "loss": 0.0874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07614760100841522, | |
| "step": 4870, | |
| "valid_targets_mean": 2428.4, | |
| "valid_targets_min": 903 | |
| }, | |
| { | |
| "epoch": 4.779411764705882, | |
| "grad_norm": 0.6826323097336842, | |
| "learning_rate": 1.1067731774176883e-05, | |
| "loss": 0.1004, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09942520409822464, | |
| "step": 4875, | |
| "valid_targets_mean": 1747.4, | |
| "valid_targets_min": 726 | |
| }, | |
| { | |
| "epoch": 4.784313725490196, | |
| "grad_norm": 0.7934705417827008, | |
| "learning_rate": 1.1024016395990758e-05, | |
| "loss": 0.1029, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10979147255420685, | |
| "step": 4880, | |
| "valid_targets_mean": 1707.7, | |
| "valid_targets_min": 755 | |
| }, | |
| { | |
| "epoch": 4.78921568627451, | |
| "grad_norm": 0.5251535439314821, | |
| "learning_rate": 1.0980354651799488e-05, | |
| "loss": 0.0866, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09340649098157883, | |
| "step": 4885, | |
| "valid_targets_mean": 2131.6, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 4.794117647058823, | |
| "grad_norm": 0.5444234009344121, | |
| "learning_rate": 1.0936746802494124e-05, | |
| "loss": 0.0932, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10292315483093262, | |
| "step": 4890, | |
| "valid_targets_mean": 2253.0, | |
| "valid_targets_min": 514 | |
| }, | |
| { | |
| "epoch": 4.799019607843137, | |
| "grad_norm": 0.4878077628758911, | |
| "learning_rate": 1.0893193108643671e-05, | |
| "loss": 0.0889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09365487098693848, | |
| "step": 4895, | |
| "valid_targets_mean": 2397.1, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 4.803921568627451, | |
| "grad_norm": 0.5545755624860067, | |
| "learning_rate": 1.084969383049354e-05, | |
| "loss": 0.0907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09213241934776306, | |
| "step": 4900, | |
| "valid_targets_mean": 2270.0, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 4.8088235294117645, | |
| "grad_norm": 0.6017361382666815, | |
| "learning_rate": 1.0806249227964e-05, | |
| "loss": 0.1001, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09305038303136826, | |
| "step": 4905, | |
| "valid_targets_mean": 1651.3, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 4.813725490196078, | |
| "grad_norm": 0.48831061998554787, | |
| "learning_rate": 1.0762859560648634e-05, | |
| "loss": 0.1014, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08866013586521149, | |
| "step": 4910, | |
| "valid_targets_mean": 2533.8, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 4.818627450980392, | |
| "grad_norm": 0.5436658478885745, | |
| "learning_rate": 1.0719525087812737e-05, | |
| "loss": 0.1074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10058622062206268, | |
| "step": 4915, | |
| "valid_targets_mean": 2305.4, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 4.823529411764706, | |
| "grad_norm": 0.5264234660847347, | |
| "learning_rate": 1.0676246068391816e-05, | |
| "loss": 0.0919, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09133768826723099, | |
| "step": 4920, | |
| "valid_targets_mean": 2342.2, | |
| "valid_targets_min": 691 | |
| }, | |
| { | |
| "epoch": 4.828431372549019, | |
| "grad_norm": 0.5865401723602275, | |
| "learning_rate": 1.0633022760990046e-05, | |
| "loss": 0.1001, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0870981216430664, | |
| "step": 4925, | |
| "valid_targets_mean": 2040.4, | |
| "valid_targets_min": 523 | |
| }, | |
| { | |
| "epoch": 4.833333333333333, | |
| "grad_norm": 0.5066015819283176, | |
| "learning_rate": 1.0589855423878679e-05, | |
| "loss": 0.0936, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09252853691577911, | |
| "step": 4930, | |
| "valid_targets_mean": 2513.6, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 4.838235294117647, | |
| "grad_norm": 0.5163264858489794, | |
| "learning_rate": 1.0546744314994549e-05, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08159564435482025, | |
| "step": 4935, | |
| "valid_targets_mean": 1908.4, | |
| "valid_targets_min": 559 | |
| }, | |
| { | |
| "epoch": 4.8431372549019605, | |
| "grad_norm": 0.542313951486232, | |
| "learning_rate": 1.0503689691938512e-05, | |
| "loss": 0.0913, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1150393933057785, | |
| "step": 4940, | |
| "valid_targets_mean": 2331.6, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 4.848039215686274, | |
| "grad_norm": 0.5435833906831937, | |
| "learning_rate": 1.046069181197389e-05, | |
| "loss": 0.1007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09760939329862595, | |
| "step": 4945, | |
| "valid_targets_mean": 2145.4, | |
| "valid_targets_min": 773 | |
| }, | |
| { | |
| "epoch": 4.852941176470588, | |
| "grad_norm": 0.4369663282479705, | |
| "learning_rate": 1.041775093202495e-05, | |
| "loss": 0.0996, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08269096910953522, | |
| "step": 4950, | |
| "valid_targets_mean": 3061.4, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 4.857843137254902, | |
| "grad_norm": 0.4758291844439797, | |
| "learning_rate": 1.0374867308675388e-05, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09057286381721497, | |
| "step": 4955, | |
| "valid_targets_mean": 2415.4, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 4.862745098039216, | |
| "grad_norm": 0.6272118961769553, | |
| "learning_rate": 1.0332041198166754e-05, | |
| "loss": 0.0778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09164794534444809, | |
| "step": 4960, | |
| "valid_targets_mean": 1584.6, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 4.867647058823529, | |
| "grad_norm": 0.5576322781572296, | |
| "learning_rate": 1.0289272856396954e-05, | |
| "loss": 0.1, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1521686166524887, | |
| "step": 4965, | |
| "valid_targets_mean": 2158.5, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 4.872549019607844, | |
| "grad_norm": 0.49324132743081855, | |
| "learning_rate": 1.0246562538918721e-05, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09149819612503052, | |
| "step": 4970, | |
| "valid_targets_mean": 2297.7, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 4.877450980392156, | |
| "grad_norm": 0.5383819550781447, | |
| "learning_rate": 1.020391050093805e-05, | |
| "loss": 0.0908, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10303015261888504, | |
| "step": 4975, | |
| "valid_targets_mean": 2042.1, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 4.882352941176471, | |
| "grad_norm": 0.527883183046873, | |
| "learning_rate": 1.0161316997312709e-05, | |
| "loss": 0.0859, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08816616982221603, | |
| "step": 4980, | |
| "valid_targets_mean": 2140.2, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 4.887254901960784, | |
| "grad_norm": 0.4867613644619435, | |
| "learning_rate": 1.0118782282550716e-05, | |
| "loss": 0.0828, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10141561925411224, | |
| "step": 4985, | |
| "valid_targets_mean": 2612.1, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 4.892156862745098, | |
| "grad_norm": 0.5079271131229582, | |
| "learning_rate": 1.0076306610808814e-05, | |
| "loss": 0.0915, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11670317500829697, | |
| "step": 4990, | |
| "valid_targets_mean": 2409.9, | |
| "valid_targets_min": 671 | |
| }, | |
| { | |
| "epoch": 4.897058823529412, | |
| "grad_norm": 0.40829801958468664, | |
| "learning_rate": 1.0033890235890927e-05, | |
| "loss": 0.0921, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08440542221069336, | |
| "step": 4995, | |
| "valid_targets_mean": 3008.8, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 4.901960784313726, | |
| "grad_norm": 0.5894884163587373, | |
| "learning_rate": 9.99153341124668e-06, | |
| "loss": 0.0934, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11752729117870331, | |
| "step": 5000, | |
| "valid_targets_mean": 1987.6, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 4.9068627450980395, | |
| "grad_norm": 0.6283199481005249, | |
| "learning_rate": 9.949236389969867e-06, | |
| "loss": 0.1045, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09560035169124603, | |
| "step": 5005, | |
| "valid_targets_mean": 1691.4, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 4.911764705882353, | |
| "grad_norm": 0.7708788939045854, | |
| "learning_rate": 9.906999424796924e-06, | |
| "loss": 0.0993, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12335900217294693, | |
| "step": 5010, | |
| "valid_targets_mean": 1527.1, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 4.916666666666667, | |
| "grad_norm": 0.5051261469161019, | |
| "learning_rate": 9.864822768105456e-06, | |
| "loss": 0.0995, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09845969825983047, | |
| "step": 5015, | |
| "valid_targets_mean": 2510.3, | |
| "valid_targets_min": 650 | |
| }, | |
| { | |
| "epoch": 4.921568627450981, | |
| "grad_norm": 0.6859191386588651, | |
| "learning_rate": 9.822706671912705e-06, | |
| "loss": 0.0933, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11246706545352936, | |
| "step": 5020, | |
| "valid_targets_mean": 1509.5, | |
| "valid_targets_min": 608 | |
| }, | |
| { | |
| "epoch": 4.926470588235294, | |
| "grad_norm": 0.6250735945680871, | |
| "learning_rate": 9.780651387874039e-06, | |
| "loss": 0.0953, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09886276721954346, | |
| "step": 5025, | |
| "valid_targets_mean": 1722.8, | |
| "valid_targets_min": 569 | |
| }, | |
| { | |
| "epoch": 4.931372549019608, | |
| "grad_norm": 0.6608007328070579, | |
| "learning_rate": 9.738657167281451e-06, | |
| "loss": 0.1108, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10095824301242828, | |
| "step": 5030, | |
| "valid_targets_mean": 1996.3, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 4.936274509803922, | |
| "grad_norm": 0.5050703755720812, | |
| "learning_rate": 9.696724261062079e-06, | |
| "loss": 0.0893, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09285195171833038, | |
| "step": 5035, | |
| "valid_targets_mean": 2361.6, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 4.9411764705882355, | |
| "grad_norm": 0.5243779388155254, | |
| "learning_rate": 9.654852919776689e-06, | |
| "loss": 0.0897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07373321056365967, | |
| "step": 5040, | |
| "valid_targets_mean": 2309.0, | |
| "valid_targets_min": 606 | |
| }, | |
| { | |
| "epoch": 4.946078431372549, | |
| "grad_norm": 0.5484866741806738, | |
| "learning_rate": 9.613043393618162e-06, | |
| "loss": 0.0941, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08428633958101273, | |
| "step": 5045, | |
| "valid_targets_mean": 2396.1, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 4.950980392156863, | |
| "grad_norm": 0.4760131574819217, | |
| "learning_rate": 9.57129593241004e-06, | |
| "loss": 0.0968, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09283538162708282, | |
| "step": 5050, | |
| "valid_targets_mean": 2947.1, | |
| "valid_targets_min": 1134 | |
| }, | |
| { | |
| "epoch": 4.955882352941177, | |
| "grad_norm": 0.7049917083821713, | |
| "learning_rate": 9.529610785604995e-06, | |
| "loss": 0.0949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10550940036773682, | |
| "step": 5055, | |
| "valid_targets_mean": 1514.4, | |
| "valid_targets_min": 672 | |
| }, | |
| { | |
| "epoch": 4.96078431372549, | |
| "grad_norm": 0.48941292868914, | |
| "learning_rate": 9.487988202283345e-06, | |
| "loss": 0.0902, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0809384286403656, | |
| "step": 5060, | |
| "valid_targets_mean": 2195.8, | |
| "valid_targets_min": 512 | |
| }, | |
| { | |
| "epoch": 4.965686274509804, | |
| "grad_norm": 0.5931382077234513, | |
| "learning_rate": 9.446428431151591e-06, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08493739366531372, | |
| "step": 5065, | |
| "valid_targets_mean": 2018.6, | |
| "valid_targets_min": 457 | |
| }, | |
| { | |
| "epoch": 4.970588235294118, | |
| "grad_norm": 0.5893891050217172, | |
| "learning_rate": 9.404931720540918e-06, | |
| "loss": 0.0924, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11142192035913467, | |
| "step": 5070, | |
| "valid_targets_mean": 2016.1, | |
| "valid_targets_min": 459 | |
| }, | |
| { | |
| "epoch": 4.9754901960784315, | |
| "grad_norm": 0.6663388117976916, | |
| "learning_rate": 9.363498318405679e-06, | |
| "loss": 0.0921, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10338947176933289, | |
| "step": 5075, | |
| "valid_targets_mean": 1906.9, | |
| "valid_targets_min": 762 | |
| }, | |
| { | |
| "epoch": 4.980392156862745, | |
| "grad_norm": 0.5702917919323601, | |
| "learning_rate": 9.322128472321973e-06, | |
| "loss": 0.093, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09254737198352814, | |
| "step": 5080, | |
| "valid_targets_mean": 2369.6, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 4.985294117647059, | |
| "grad_norm": 0.6006989977907731, | |
| "learning_rate": 9.280822429486117e-06, | |
| "loss": 0.0863, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10931229591369629, | |
| "step": 5085, | |
| "valid_targets_mean": 1962.5, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 4.990196078431373, | |
| "grad_norm": 0.6041195773866957, | |
| "learning_rate": 9.239580436713178e-06, | |
| "loss": 0.1032, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09572377055883408, | |
| "step": 5090, | |
| "valid_targets_mean": 2157.7, | |
| "valid_targets_min": 507 | |
| }, | |
| { | |
| "epoch": 4.995098039215686, | |
| "grad_norm": 0.6617547843586967, | |
| "learning_rate": 9.198402740435527e-06, | |
| "loss": 0.0981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.132956400513649, | |
| "step": 5095, | |
| "valid_targets_mean": 2033.6, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.47759667778213716, | |
| "learning_rate": 9.157289586701328e-06, | |
| "loss": 0.0901, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08661957085132599, | |
| "step": 5100, | |
| "valid_targets_mean": 2665.0, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 5.004901960784314, | |
| "grad_norm": 0.4604848433671045, | |
| "learning_rate": 9.116241221173103e-06, | |
| "loss": 0.0867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07293270528316498, | |
| "step": 5105, | |
| "valid_targets_mean": 2602.8, | |
| "valid_targets_min": 818 | |
| }, | |
| { | |
| "epoch": 5.009803921568627, | |
| "grad_norm": 0.5911007179196364, | |
| "learning_rate": 9.075257889126234e-06, | |
| "loss": 0.0851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08851397782564163, | |
| "step": 5110, | |
| "valid_targets_mean": 2199.7, | |
| "valid_targets_min": 652 | |
| }, | |
| { | |
| "epoch": 5.014705882352941, | |
| "grad_norm": 0.5015517934829464, | |
| "learning_rate": 9.0343398354475e-06, | |
| "loss": 0.0741, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0782584622502327, | |
| "step": 5115, | |
| "valid_targets_mean": 2538.9, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 5.019607843137255, | |
| "grad_norm": 0.5541843585633855, | |
| "learning_rate": 8.993487304633648e-06, | |
| "loss": 0.0819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07696060091257095, | |
| "step": 5120, | |
| "valid_targets_mean": 2012.0, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 5.0245098039215685, | |
| "grad_norm": 0.5085417438896622, | |
| "learning_rate": 8.952700540789884e-06, | |
| "loss": 0.0894, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07348785549402237, | |
| "step": 5125, | |
| "valid_targets_mean": 2494.4, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 5.029411764705882, | |
| "grad_norm": 0.5189146300290645, | |
| "learning_rate": 8.911979787628456e-06, | |
| "loss": 0.0736, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07135240733623505, | |
| "step": 5130, | |
| "valid_targets_mean": 2331.8, | |
| "valid_targets_min": 785 | |
| }, | |
| { | |
| "epoch": 5.034313725490196, | |
| "grad_norm": 0.6293919997645581, | |
| "learning_rate": 8.871325288467188e-06, | |
| "loss": 0.0986, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10753931850194931, | |
| "step": 5135, | |
| "valid_targets_mean": 2485.3, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 5.03921568627451, | |
| "grad_norm": 0.5845395385864605, | |
| "learning_rate": 8.830737286227972e-06, | |
| "loss": 0.0819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09616108238697052, | |
| "step": 5140, | |
| "valid_targets_mean": 2392.1, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 5.044117647058823, | |
| "grad_norm": 0.5407442298797527, | |
| "learning_rate": 8.790216023435411e-06, | |
| "loss": 0.0937, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08268275856971741, | |
| "step": 5145, | |
| "valid_targets_mean": 2277.2, | |
| "valid_targets_min": 619 | |
| }, | |
| { | |
| "epoch": 5.049019607843137, | |
| "grad_norm": 0.5343504627780364, | |
| "learning_rate": 8.74976174221531e-06, | |
| "loss": 0.0755, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07687253504991531, | |
| "step": 5150, | |
| "valid_targets_mean": 2395.2, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 5.053921568627451, | |
| "grad_norm": 0.503203022422961, | |
| "learning_rate": 8.709374684293221e-06, | |
| "loss": 0.0795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07528114318847656, | |
| "step": 5155, | |
| "valid_targets_mean": 2509.9, | |
| "valid_targets_min": 424 | |
| }, | |
| { | |
| "epoch": 5.0588235294117645, | |
| "grad_norm": 0.4820740870162874, | |
| "learning_rate": 8.669055090993041e-06, | |
| "loss": 0.0726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06485399603843689, | |
| "step": 5160, | |
| "valid_targets_mean": 2117.0, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 5.063725490196078, | |
| "grad_norm": 0.6534311563959244, | |
| "learning_rate": 8.628803203235536e-06, | |
| "loss": 0.0957, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11405531316995621, | |
| "step": 5165, | |
| "valid_targets_mean": 2078.2, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 5.068627450980392, | |
| "grad_norm": 0.5159605033551381, | |
| "learning_rate": 8.588619261536904e-06, | |
| "loss": 0.0823, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07087569683790207, | |
| "step": 5170, | |
| "valid_targets_mean": 2643.9, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 5.073529411764706, | |
| "grad_norm": 0.6120277300768532, | |
| "learning_rate": 8.548503506007363e-06, | |
| "loss": 0.0877, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08435557782649994, | |
| "step": 5175, | |
| "valid_targets_mean": 1904.4, | |
| "valid_targets_min": 203 | |
| }, | |
| { | |
| "epoch": 5.078431372549019, | |
| "grad_norm": 0.5750915500710682, | |
| "learning_rate": 8.508456176349698e-06, | |
| "loss": 0.0868, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07057558000087738, | |
| "step": 5180, | |
| "valid_targets_mean": 2130.0, | |
| "valid_targets_min": 902 | |
| }, | |
| { | |
| "epoch": 5.083333333333333, | |
| "grad_norm": 0.5929571350769494, | |
| "learning_rate": 8.468477511857826e-06, | |
| "loss": 0.0792, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08237534761428833, | |
| "step": 5185, | |
| "valid_targets_mean": 2016.9, | |
| "valid_targets_min": 712 | |
| }, | |
| { | |
| "epoch": 5.088235294117647, | |
| "grad_norm": 0.49073426404229986, | |
| "learning_rate": 8.42856775141537e-06, | |
| "loss": 0.0818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07128225266933441, | |
| "step": 5190, | |
| "valid_targets_mean": 2563.1, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 5.0931372549019605, | |
| "grad_norm": 0.5348493318460731, | |
| "learning_rate": 8.388727133494219e-06, | |
| "loss": 0.0783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06887640058994293, | |
| "step": 5195, | |
| "valid_targets_mean": 2482.9, | |
| "valid_targets_min": 623 | |
| }, | |
| { | |
| "epoch": 5.098039215686274, | |
| "grad_norm": 0.5146171818297789, | |
| "learning_rate": 8.348955896153146e-06, | |
| "loss": 0.0807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07764531672000885, | |
| "step": 5200, | |
| "valid_targets_mean": 2469.7, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 5.102941176470588, | |
| "grad_norm": 0.5561557225636563, | |
| "learning_rate": 8.30925427703633e-06, | |
| "loss": 0.0919, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11974679678678513, | |
| "step": 5205, | |
| "valid_targets_mean": 2453.7, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 5.107843137254902, | |
| "grad_norm": 0.6539916720138971, | |
| "learning_rate": 8.269622513371971e-06, | |
| "loss": 0.0768, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09084124863147736, | |
| "step": 5210, | |
| "valid_targets_mean": 1766.9, | |
| "valid_targets_min": 342 | |
| }, | |
| { | |
| "epoch": 5.112745098039215, | |
| "grad_norm": 0.5429121808535396, | |
| "learning_rate": 8.230060841970875e-06, | |
| "loss": 0.0754, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07802797853946686, | |
| "step": 5215, | |
| "valid_targets_mean": 2213.3, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 5.117647058823529, | |
| "grad_norm": 0.5083279394937158, | |
| "learning_rate": 8.190569499225008e-06, | |
| "loss": 0.0814, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06983786821365356, | |
| "step": 5220, | |
| "valid_targets_mean": 2548.4, | |
| "valid_targets_min": 837 | |
| }, | |
| { | |
| "epoch": 5.122549019607843, | |
| "grad_norm": 0.5465649932109973, | |
| "learning_rate": 8.1511487211061e-06, | |
| "loss": 0.0892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07368411868810654, | |
| "step": 5225, | |
| "valid_targets_mean": 2058.0, | |
| "valid_targets_min": 539 | |
| }, | |
| { | |
| "epoch": 5.127450980392156, | |
| "grad_norm": 0.558225572156942, | |
| "learning_rate": 8.111798743164267e-06, | |
| "loss": 0.0911, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08299843966960907, | |
| "step": 5230, | |
| "valid_targets_mean": 1902.3, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 5.132352941176471, | |
| "grad_norm": 0.5136237271349839, | |
| "learning_rate": 8.072519800526542e-06, | |
| "loss": 0.0763, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07817132771015167, | |
| "step": 5235, | |
| "valid_targets_mean": 2501.4, | |
| "valid_targets_min": 643 | |
| }, | |
| { | |
| "epoch": 5.137254901960785, | |
| "grad_norm": 0.498321689726259, | |
| "learning_rate": 8.033312127895518e-06, | |
| "loss": 0.0759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09710898995399475, | |
| "step": 5240, | |
| "valid_targets_mean": 2590.6, | |
| "valid_targets_min": 543 | |
| }, | |
| { | |
| "epoch": 5.142156862745098, | |
| "grad_norm": 0.7408036748383444, | |
| "learning_rate": 7.994175959547939e-06, | |
| "loss": 0.0836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11459003388881683, | |
| "step": 5245, | |
| "valid_targets_mean": 1611.8, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 5.147058823529412, | |
| "grad_norm": 0.4975297290259382, | |
| "learning_rate": 7.95511152933327e-06, | |
| "loss": 0.0802, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06459469348192215, | |
| "step": 5250, | |
| "valid_targets_mean": 2624.0, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 5.151960784313726, | |
| "grad_norm": 0.7561461760461909, | |
| "learning_rate": 7.916119070672325e-06, | |
| "loss": 0.0861, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0870676040649414, | |
| "step": 5255, | |
| "valid_targets_mean": 1697.4, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 5.1568627450980395, | |
| "grad_norm": 0.5210675746878989, | |
| "learning_rate": 7.87719881655588e-06, | |
| "loss": 0.0844, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06843873858451843, | |
| "step": 5260, | |
| "valid_targets_mean": 2269.1, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 5.161764705882353, | |
| "grad_norm": 0.5338645649878168, | |
| "learning_rate": 7.83835099954327e-06, | |
| "loss": 0.0981, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0681375116109848, | |
| "step": 5265, | |
| "valid_targets_mean": 2572.6, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 5.166666666666667, | |
| "grad_norm": 0.651416544867339, | |
| "learning_rate": 7.799575851760982e-06, | |
| "loss": 0.0864, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08005984127521515, | |
| "step": 5270, | |
| "valid_targets_mean": 2171.5, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 5.171568627450981, | |
| "grad_norm": 0.6330539334076882, | |
| "learning_rate": 7.76087360490129e-06, | |
| "loss": 0.0861, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09047305583953857, | |
| "step": 5275, | |
| "valid_targets_mean": 1923.2, | |
| "valid_targets_min": 548 | |
| }, | |
| { | |
| "epoch": 5.176470588235294, | |
| "grad_norm": 0.44154350607425974, | |
| "learning_rate": 7.722244490220873e-06, | |
| "loss": 0.0681, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06011586636304855, | |
| "step": 5280, | |
| "valid_targets_mean": 3123.1, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 5.181372549019608, | |
| "grad_norm": 0.5534899671335772, | |
| "learning_rate": 7.683688738539414e-06, | |
| "loss": 0.0845, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0851341038942337, | |
| "step": 5285, | |
| "valid_targets_mean": 2315.2, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 5.186274509803922, | |
| "grad_norm": 0.568120445090173, | |
| "learning_rate": 7.64520658023824e-06, | |
| "loss": 0.0915, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0778503492474556, | |
| "step": 5290, | |
| "valid_targets_mean": 2170.7, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 5.1911764705882355, | |
| "grad_norm": 0.4820906324235686, | |
| "learning_rate": 7.606798245258941e-06, | |
| "loss": 0.0949, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11274714022874832, | |
| "step": 5295, | |
| "valid_targets_mean": 2448.4, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.196078431372549, | |
| "grad_norm": 0.5565416986397015, | |
| "learning_rate": 7.568463963101982e-06, | |
| "loss": 0.0858, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07629026472568512, | |
| "step": 5300, | |
| "valid_targets_mean": 2519.6, | |
| "valid_targets_min": 564 | |
| }, | |
| { | |
| "epoch": 5.200980392156863, | |
| "grad_norm": 0.5875632797684447, | |
| "learning_rate": 7.530203962825331e-06, | |
| "loss": 0.0767, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07195107638835907, | |
| "step": 5305, | |
| "valid_targets_mean": 2161.2, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 5.205882352941177, | |
| "grad_norm": 0.5878351907497511, | |
| "learning_rate": 7.4920184730431275e-06, | |
| "loss": 0.0776, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06824355572462082, | |
| "step": 5310, | |
| "valid_targets_mean": 2125.2, | |
| "valid_targets_min": 792 | |
| }, | |
| { | |
| "epoch": 5.21078431372549, | |
| "grad_norm": 0.6921615760995266, | |
| "learning_rate": 7.453907721924285e-06, | |
| "loss": 0.0773, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08252941071987152, | |
| "step": 5315, | |
| "valid_targets_mean": 1936.6, | |
| "valid_targets_min": 538 | |
| }, | |
| { | |
| "epoch": 5.215686274509804, | |
| "grad_norm": 0.6435003909309579, | |
| "learning_rate": 7.415871937191111e-06, | |
| "loss": 0.0737, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07561911642551422, | |
| "step": 5320, | |
| "valid_targets_mean": 2016.7, | |
| "valid_targets_min": 635 | |
| }, | |
| { | |
| "epoch": 5.220588235294118, | |
| "grad_norm": 0.5023546345333582, | |
| "learning_rate": 7.377911346117992e-06, | |
| "loss": 0.0799, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06859388202428818, | |
| "step": 5325, | |
| "valid_targets_mean": 2208.1, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 5.2254901960784315, | |
| "grad_norm": 0.6998975765045649, | |
| "learning_rate": 7.340026175530002e-06, | |
| "loss": 0.0997, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1062772274017334, | |
| "step": 5330, | |
| "valid_targets_mean": 1918.6, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 5.230392156862745, | |
| "grad_norm": 0.62750845180046, | |
| "learning_rate": 7.302216651801546e-06, | |
| "loss": 0.0872, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07626760005950928, | |
| "step": 5335, | |
| "valid_targets_mean": 1799.3, | |
| "valid_targets_min": 718 | |
| }, | |
| { | |
| "epoch": 5.235294117647059, | |
| "grad_norm": 0.5543171693917915, | |
| "learning_rate": 7.264483000855036e-06, | |
| "loss": 0.0839, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07348915934562683, | |
| "step": 5340, | |
| "valid_targets_mean": 2069.6, | |
| "valid_targets_min": 754 | |
| }, | |
| { | |
| "epoch": 5.240196078431373, | |
| "grad_norm": 0.7983146319280329, | |
| "learning_rate": 7.226825448159524e-06, | |
| "loss": 0.0791, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08151835203170776, | |
| "step": 5345, | |
| "valid_targets_mean": 1926.6, | |
| "valid_targets_min": 783 | |
| }, | |
| { | |
| "epoch": 5.245098039215686, | |
| "grad_norm": 0.5799482989527313, | |
| "learning_rate": 7.1892442187293324e-06, | |
| "loss": 0.0869, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12980447709560394, | |
| "step": 5350, | |
| "valid_targets_mean": 2295.6, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "grad_norm": 0.4881711367941992, | |
| "learning_rate": 7.151739537122755e-06, | |
| "loss": 0.083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06775844097137451, | |
| "step": 5355, | |
| "valid_targets_mean": 2253.4, | |
| "valid_targets_min": 621 | |
| }, | |
| { | |
| "epoch": 5.254901960784314, | |
| "grad_norm": 0.5353490999682617, | |
| "learning_rate": 7.114311627440675e-06, | |
| "loss": 0.071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0793384313583374, | |
| "step": 5360, | |
| "valid_targets_mean": 2420.6, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 5.259803921568627, | |
| "grad_norm": 0.538659559297651, | |
| "learning_rate": 7.076960713325242e-06, | |
| "loss": 0.0783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0664932131767273, | |
| "step": 5365, | |
| "valid_targets_mean": 2518.3, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 5.264705882352941, | |
| "grad_norm": 0.6568125968077196, | |
| "learning_rate": 7.039687017958545e-06, | |
| "loss": 0.0854, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0847780704498291, | |
| "step": 5370, | |
| "valid_targets_mean": 1899.9, | |
| "valid_targets_min": 706 | |
| }, | |
| { | |
| "epoch": 5.269607843137255, | |
| "grad_norm": 0.6122298995911225, | |
| "learning_rate": 7.002490764061274e-06, | |
| "loss": 0.0914, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09091790020465851, | |
| "step": 5375, | |
| "valid_targets_mean": 1925.0, | |
| "valid_targets_min": 609 | |
| }, | |
| { | |
| "epoch": 5.2745098039215685, | |
| "grad_norm": 0.922364301866522, | |
| "learning_rate": 6.965372173891369e-06, | |
| "loss": 0.0744, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07770726084709167, | |
| "step": 5380, | |
| "valid_targets_mean": 1841.9, | |
| "valid_targets_min": 543 | |
| }, | |
| { | |
| "epoch": 5.279411764705882, | |
| "grad_norm": 0.5027301818761994, | |
| "learning_rate": 6.928331469242709e-06, | |
| "loss": 0.074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07609498500823975, | |
| "step": 5385, | |
| "valid_targets_mean": 2420.9, | |
| "valid_targets_min": 797 | |
| }, | |
| { | |
| "epoch": 5.284313725490196, | |
| "grad_norm": 0.5073890103240528, | |
| "learning_rate": 6.891368871443795e-06, | |
| "loss": 0.0756, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0688600093126297, | |
| "step": 5390, | |
| "valid_targets_mean": 2477.0, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 5.28921568627451, | |
| "grad_norm": 0.5298744083568486, | |
| "learning_rate": 6.854484601356426e-06, | |
| "loss": 0.0845, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07411527633666992, | |
| "step": 5395, | |
| "valid_targets_mean": 2140.9, | |
| "valid_targets_min": 363 | |
| }, | |
| { | |
| "epoch": 5.294117647058823, | |
| "grad_norm": 0.5242195295040767, | |
| "learning_rate": 6.817678879374341e-06, | |
| "loss": 0.0687, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06534264236688614, | |
| "step": 5400, | |
| "valid_targets_mean": 2309.1, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 5.299019607843137, | |
| "grad_norm": 0.4536802947760404, | |
| "learning_rate": 6.7809519254219655e-06, | |
| "loss": 0.0815, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06536039710044861, | |
| "step": 5405, | |
| "valid_targets_mean": 2977.7, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 5.303921568627451, | |
| "grad_norm": 0.5065375359596163, | |
| "learning_rate": 6.744303958953043e-06, | |
| "loss": 0.0842, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07473203539848328, | |
| "step": 5410, | |
| "valid_targets_mean": 2966.2, | |
| "valid_targets_min": 1232 | |
| }, | |
| { | |
| "epoch": 5.3088235294117645, | |
| "grad_norm": 0.5342380006283294, | |
| "learning_rate": 6.707735198949344e-06, | |
| "loss": 0.0744, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07432928681373596, | |
| "step": 5415, | |
| "valid_targets_mean": 2556.4, | |
| "valid_targets_min": 594 | |
| }, | |
| { | |
| "epoch": 5.313725490196078, | |
| "grad_norm": 0.6116252161599829, | |
| "learning_rate": 6.671245863919374e-06, | |
| "loss": 0.0862, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09420572221279144, | |
| "step": 5420, | |
| "valid_targets_mean": 1872.7, | |
| "valid_targets_min": 663 | |
| }, | |
| { | |
| "epoch": 5.318627450980392, | |
| "grad_norm": 0.38722492088167965, | |
| "learning_rate": 6.634836171897048e-06, | |
| "loss": 0.0739, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06971578299999237, | |
| "step": 5425, | |
| "valid_targets_mean": 3064.0, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 5.323529411764706, | |
| "grad_norm": 0.48575352597613497, | |
| "learning_rate": 6.598506340440373e-06, | |
| "loss": 0.081, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07496002316474915, | |
| "step": 5430, | |
| "valid_targets_mean": 2613.2, | |
| "valid_targets_min": 771 | |
| }, | |
| { | |
| "epoch": 5.328431372549019, | |
| "grad_norm": 0.5766518334984045, | |
| "learning_rate": 6.562256586630198e-06, | |
| "loss": 0.0797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07142172753810883, | |
| "step": 5435, | |
| "valid_targets_mean": 2121.6, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 5.333333333333333, | |
| "grad_norm": 0.5616562487814789, | |
| "learning_rate": 6.526087127068857e-06, | |
| "loss": 0.0811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07984551787376404, | |
| "step": 5440, | |
| "valid_targets_mean": 2259.9, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 5.338235294117647, | |
| "grad_norm": 0.5297286772555554, | |
| "learning_rate": 6.489998177878925e-06, | |
| "loss": 0.0907, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07350513339042664, | |
| "step": 5445, | |
| "valid_targets_mean": 2587.9, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.3431372549019605, | |
| "grad_norm": 0.4643067993227384, | |
| "learning_rate": 6.453989954701882e-06, | |
| "loss": 0.0892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0627140998840332, | |
| "step": 5450, | |
| "valid_targets_mean": 2341.8, | |
| "valid_targets_min": 665 | |
| }, | |
| { | |
| "epoch": 5.348039215686274, | |
| "grad_norm": 0.7694605309291938, | |
| "learning_rate": 6.418062672696877e-06, | |
| "loss": 0.0958, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10127004981040955, | |
| "step": 5455, | |
| "valid_targets_mean": 1646.6, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 5.352941176470588, | |
| "grad_norm": 0.6921060023364909, | |
| "learning_rate": 6.382216546539395e-06, | |
| "loss": 0.0877, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08449102193117142, | |
| "step": 5460, | |
| "valid_targets_mean": 2167.2, | |
| "valid_targets_min": 485 | |
| }, | |
| { | |
| "epoch": 5.357843137254902, | |
| "grad_norm": 0.6955216056484889, | |
| "learning_rate": 6.346451790419992e-06, | |
| "loss": 0.08, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07503059506416321, | |
| "step": 5465, | |
| "valid_targets_mean": 1879.7, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 5.362745098039215, | |
| "grad_norm": 0.42726578395620934, | |
| "learning_rate": 6.310768618043011e-06, | |
| "loss": 0.0667, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06386620551347733, | |
| "step": 5470, | |
| "valid_targets_mean": 2749.9, | |
| "valid_targets_min": 933 | |
| }, | |
| { | |
| "epoch": 5.367647058823529, | |
| "grad_norm": 0.5327119640641524, | |
| "learning_rate": 6.275167242625331e-06, | |
| "loss": 0.0859, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06920159608125687, | |
| "step": 5475, | |
| "valid_targets_mean": 2259.7, | |
| "valid_targets_min": 781 | |
| }, | |
| { | |
| "epoch": 5.372549019607844, | |
| "grad_norm": 0.5250883694648594, | |
| "learning_rate": 6.239647876895045e-06, | |
| "loss": 0.0753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07292656600475311, | |
| "step": 5480, | |
| "valid_targets_mean": 2220.8, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 5.377450980392156, | |
| "grad_norm": 0.6092967365208375, | |
| "learning_rate": 6.204210733090241e-06, | |
| "loss": 0.0823, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09393665939569473, | |
| "step": 5485, | |
| "valid_targets_mean": 2369.6, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 5.382352941176471, | |
| "grad_norm": 0.6038113201133489, | |
| "learning_rate": 6.168856022957685e-06, | |
| "loss": 0.0869, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09418021887540817, | |
| "step": 5490, | |
| "valid_targets_mean": 2383.8, | |
| "valid_targets_min": 751 | |
| }, | |
| { | |
| "epoch": 5.387254901960785, | |
| "grad_norm": 0.5057934505834986, | |
| "learning_rate": 6.133583957751585e-06, | |
| "loss": 0.0816, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07620443403720856, | |
| "step": 5495, | |
| "valid_targets_mean": 2849.1, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 5.392156862745098, | |
| "grad_norm": 0.5776768115860323, | |
| "learning_rate": 6.098394748232333e-06, | |
| "loss": 0.0874, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08147083967924118, | |
| "step": 5500, | |
| "valid_targets_mean": 2674.8, | |
| "valid_targets_min": 622 | |
| }, | |
| { | |
| "epoch": 5.397058823529412, | |
| "grad_norm": 0.6709503859844336, | |
| "learning_rate": 6.063288604665227e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10320451855659485, | |
| "step": 5505, | |
| "valid_targets_mean": 1907.4, | |
| "valid_targets_min": 474 | |
| }, | |
| { | |
| "epoch": 5.401960784313726, | |
| "grad_norm": 0.5332746140305541, | |
| "learning_rate": 6.028265736819234e-06, | |
| "loss": 0.0945, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08802657574415207, | |
| "step": 5510, | |
| "valid_targets_mean": 2443.4, | |
| "valid_targets_min": 676 | |
| }, | |
| { | |
| "epoch": 5.4068627450980395, | |
| "grad_norm": 0.5303408521919405, | |
| "learning_rate": 5.993326353965705e-06, | |
| "loss": 0.0852, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07660799473524094, | |
| "step": 5515, | |
| "valid_targets_mean": 2294.6, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 5.411764705882353, | |
| "grad_norm": 0.5504797907176615, | |
| "learning_rate": 5.958470664877154e-06, | |
| "loss": 0.0775, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0719451829791069, | |
| "step": 5520, | |
| "valid_targets_mean": 2223.8, | |
| "valid_targets_min": 759 | |
| }, | |
| { | |
| "epoch": 5.416666666666667, | |
| "grad_norm": 0.5048702357817372, | |
| "learning_rate": 5.9236988778260144e-06, | |
| "loss": 0.0805, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07212197780609131, | |
| "step": 5525, | |
| "valid_targets_mean": 2394.9, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 5.421568627450981, | |
| "grad_norm": 0.5582414610550838, | |
| "learning_rate": 5.8890112005833565e-06, | |
| "loss": 0.0811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1013210117816925, | |
| "step": 5530, | |
| "valid_targets_mean": 2547.2, | |
| "valid_targets_min": 535 | |
| }, | |
| { | |
| "epoch": 5.426470588235294, | |
| "grad_norm": 0.5997664325867221, | |
| "learning_rate": 5.854407840417693e-06, | |
| "loss": 0.0925, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11854314804077148, | |
| "step": 5535, | |
| "valid_targets_mean": 2110.2, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 5.431372549019608, | |
| "grad_norm": 0.5953673256263169, | |
| "learning_rate": 5.819889004093719e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09450195729732513, | |
| "step": 5540, | |
| "valid_targets_mean": 2220.2, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 5.436274509803922, | |
| "grad_norm": 0.6808012844229521, | |
| "learning_rate": 5.785454897871059e-06, | |
| "loss": 0.0917, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12832430005073547, | |
| "step": 5545, | |
| "valid_targets_mean": 2311.3, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 5.4411764705882355, | |
| "grad_norm": 0.5726522699115283, | |
| "learning_rate": 5.751105727503057e-06, | |
| "loss": 0.0787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0782998651266098, | |
| "step": 5550, | |
| "valid_targets_mean": 2264.9, | |
| "valid_targets_min": 485 | |
| }, | |
| { | |
| "epoch": 5.446078431372549, | |
| "grad_norm": 0.5868036175605869, | |
| "learning_rate": 5.716841698235553e-06, | |
| "loss": 0.0899, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07577871531248093, | |
| "step": 5555, | |
| "valid_targets_mean": 1881.0, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 5.450980392156863, | |
| "grad_norm": 0.5728814564653399, | |
| "learning_rate": 5.682663014805631e-06, | |
| "loss": 0.088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08744935691356659, | |
| "step": 5560, | |
| "valid_targets_mean": 2001.8, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 5.455882352941177, | |
| "grad_norm": 0.5131924143447337, | |
| "learning_rate": 5.648569881440422e-06, | |
| "loss": 0.0871, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0721743106842041, | |
| "step": 5565, | |
| "valid_targets_mean": 2405.6, | |
| "valid_targets_min": 604 | |
| }, | |
| { | |
| "epoch": 5.46078431372549, | |
| "grad_norm": 0.635324527204431, | |
| "learning_rate": 5.6145625018558715e-06, | |
| "loss": 0.0831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10122652351856232, | |
| "step": 5570, | |
| "valid_targets_mean": 2350.2, | |
| "valid_targets_min": 598 | |
| }, | |
| { | |
| "epoch": 5.465686274509804, | |
| "grad_norm": 0.6267729572441716, | |
| "learning_rate": 5.580641079255509e-06, | |
| "loss": 0.0961, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10425372421741486, | |
| "step": 5575, | |
| "valid_targets_mean": 1768.8, | |
| "valid_targets_min": 790 | |
| }, | |
| { | |
| "epoch": 5.470588235294118, | |
| "grad_norm": 0.5930960268540582, | |
| "learning_rate": 5.54680581632925e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07789880037307739, | |
| "step": 5580, | |
| "valid_targets_mean": 2379.4, | |
| "valid_targets_min": 580 | |
| }, | |
| { | |
| "epoch": 5.4754901960784315, | |
| "grad_norm": 0.5411247803074882, | |
| "learning_rate": 5.513056915252195e-06, | |
| "loss": 0.084, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07947784662246704, | |
| "step": 5585, | |
| "valid_targets_mean": 2909.1, | |
| "valid_targets_min": 852 | |
| }, | |
| { | |
| "epoch": 5.480392156862745, | |
| "grad_norm": 0.5846627085120992, | |
| "learning_rate": 5.479394577683404e-06, | |
| "loss": 0.0797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07893018424510956, | |
| "step": 5590, | |
| "valid_targets_mean": 2389.6, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 5.485294117647059, | |
| "grad_norm": 0.5935404654101385, | |
| "learning_rate": 5.44581900476469e-06, | |
| "loss": 0.0799, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08584611862897873, | |
| "step": 5595, | |
| "valid_targets_mean": 2056.1, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 5.490196078431373, | |
| "grad_norm": 0.5868161180417224, | |
| "learning_rate": 5.412330397119419e-06, | |
| "loss": 0.0742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07897976040840149, | |
| "step": 5600, | |
| "valid_targets_mean": 2170.9, | |
| "valid_targets_min": 708 | |
| }, | |
| { | |
| "epoch": 5.495098039215686, | |
| "grad_norm": 0.5805575731061238, | |
| "learning_rate": 5.3789289548513326e-06, | |
| "loss": 0.0785, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09302123636007309, | |
| "step": 5605, | |
| "valid_targets_mean": 2172.2, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "grad_norm": 0.6347183322088364, | |
| "learning_rate": 5.345614877543313e-06, | |
| "loss": 0.0817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07956665754318237, | |
| "step": 5610, | |
| "valid_targets_mean": 1914.4, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 5.504901960784314, | |
| "grad_norm": 0.7293893954848631, | |
| "learning_rate": 5.312388364256229e-06, | |
| "loss": 0.1007, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10558699816465378, | |
| "step": 5615, | |
| "valid_targets_mean": 1635.8, | |
| "valid_targets_min": 688 | |
| }, | |
| { | |
| "epoch": 5.509803921568627, | |
| "grad_norm": 0.5270248184938987, | |
| "learning_rate": 5.279249613527735e-06, | |
| "loss": 0.0742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06831567734479904, | |
| "step": 5620, | |
| "valid_targets_mean": 2452.7, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 5.514705882352941, | |
| "grad_norm": 0.5350392884902185, | |
| "learning_rate": 5.246198823371063e-06, | |
| "loss": 0.0736, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06904664635658264, | |
| "step": 5625, | |
| "valid_targets_mean": 2162.8, | |
| "valid_targets_min": 808 | |
| }, | |
| { | |
| "epoch": 5.519607843137255, | |
| "grad_norm": 0.6344941061658974, | |
| "learning_rate": 5.213236191273854e-06, | |
| "loss": 0.0884, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09255999326705933, | |
| "step": 5630, | |
| "valid_targets_mean": 2347.1, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 5.5245098039215685, | |
| "grad_norm": 0.6020792880176296, | |
| "learning_rate": 5.180361914196996e-06, | |
| "loss": 0.0879, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09349136054515839, | |
| "step": 5635, | |
| "valid_targets_mean": 2380.8, | |
| "valid_targets_min": 607 | |
| }, | |
| { | |
| "epoch": 5.529411764705882, | |
| "grad_norm": 0.5956397945862633, | |
| "learning_rate": 5.147576188573433e-06, | |
| "loss": 0.0836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09244567155838013, | |
| "step": 5640, | |
| "valid_targets_mean": 2050.1, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 5.534313725490196, | |
| "grad_norm": 0.5153802600162333, | |
| "learning_rate": 5.114879210306967e-06, | |
| "loss": 0.0772, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06410102546215057, | |
| "step": 5645, | |
| "valid_targets_mean": 2220.6, | |
| "valid_targets_min": 560 | |
| }, | |
| { | |
| "epoch": 5.53921568627451, | |
| "grad_norm": 0.5903655078820536, | |
| "learning_rate": 5.0822711747711365e-06, | |
| "loss": 0.0802, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07721748948097229, | |
| "step": 5650, | |
| "valid_targets_mean": 2455.4, | |
| "valid_targets_min": 521 | |
| }, | |
| { | |
| "epoch": 5.544117647058823, | |
| "grad_norm": 0.5429688822860307, | |
| "learning_rate": 5.049752276808002e-06, | |
| "loss": 0.0841, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0777660384774208, | |
| "step": 5655, | |
| "valid_targets_mean": 2515.0, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 5.549019607843137, | |
| "grad_norm": 0.5604176522917697, | |
| "learning_rate": 5.0173227107270065e-06, | |
| "loss": 0.0901, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08657435327768326, | |
| "step": 5660, | |
| "valid_targets_mean": 2180.7, | |
| "valid_targets_min": 733 | |
| }, | |
| { | |
| "epoch": 5.553921568627451, | |
| "grad_norm": 0.5281339651549892, | |
| "learning_rate": 4.984982670303818e-06, | |
| "loss": 0.0733, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06807395815849304, | |
| "step": 5665, | |
| "valid_targets_mean": 2235.6, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 5.5588235294117645, | |
| "grad_norm": 0.5690739702658926, | |
| "learning_rate": 4.952732348779165e-06, | |
| "loss": 0.0743, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06617975234985352, | |
| "step": 5670, | |
| "valid_targets_mean": 2433.0, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 5.563725490196078, | |
| "grad_norm": 0.5656611047345302, | |
| "learning_rate": 4.920571938857665e-06, | |
| "loss": 0.0716, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08199770003557205, | |
| "step": 5675, | |
| "valid_targets_mean": 2496.0, | |
| "valid_targets_min": 742 | |
| }, | |
| { | |
| "epoch": 5.568627450980392, | |
| "grad_norm": 0.5265721948588862, | |
| "learning_rate": 4.88850163270671e-06, | |
| "loss": 0.0753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07681460678577423, | |
| "step": 5680, | |
| "valid_targets_mean": 2276.1, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 5.573529411764706, | |
| "grad_norm": 0.43922395927036295, | |
| "learning_rate": 4.856521621955282e-06, | |
| "loss": 0.0772, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06098077446222305, | |
| "step": 5685, | |
| "valid_targets_mean": 2503.2, | |
| "valid_targets_min": 816 | |
| }, | |
| { | |
| "epoch": 5.578431372549019, | |
| "grad_norm": 0.5517412995836757, | |
| "learning_rate": 4.824632097692816e-06, | |
| "loss": 0.0739, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07269728183746338, | |
| "step": 5690, | |
| "valid_targets_mean": 2103.2, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 5.583333333333333, | |
| "grad_norm": 0.5648727695267294, | |
| "learning_rate": 4.7928332504680876e-06, | |
| "loss": 0.0759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07408298552036285, | |
| "step": 5695, | |
| "valid_targets_mean": 2224.1, | |
| "valid_targets_min": 725 | |
| }, | |
| { | |
| "epoch": 5.588235294117647, | |
| "grad_norm": 0.4929550418851452, | |
| "learning_rate": 4.761125270288041e-06, | |
| "loss": 0.0772, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.074101522564888, | |
| "step": 5700, | |
| "valid_targets_mean": 2437.9, | |
| "valid_targets_min": 533 | |
| }, | |
| { | |
| "epoch": 5.5931372549019605, | |
| "grad_norm": 0.6146098190528329, | |
| "learning_rate": 4.72950834661666e-06, | |
| "loss": 0.0815, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09067805856466293, | |
| "step": 5705, | |
| "valid_targets_mean": 2192.0, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 5.598039215686274, | |
| "grad_norm": 0.5713767381680502, | |
| "learning_rate": 4.69798266837384e-06, | |
| "loss": 0.0752, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0734923854470253, | |
| "step": 5710, | |
| "valid_targets_mean": 2141.4, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 5.602941176470588, | |
| "grad_norm": 0.530929712629617, | |
| "learning_rate": 4.666548423934265e-06, | |
| "loss": 0.0735, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07157990336418152, | |
| "step": 5715, | |
| "valid_targets_mean": 2109.9, | |
| "valid_targets_min": 595 | |
| }, | |
| { | |
| "epoch": 5.607843137254902, | |
| "grad_norm": 0.6962615608499858, | |
| "learning_rate": 4.635205801126286e-06, | |
| "loss": 0.0831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11094634234905243, | |
| "step": 5720, | |
| "valid_targets_mean": 1813.5, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 5.612745098039216, | |
| "grad_norm": 0.4782088252652507, | |
| "learning_rate": 4.603954987230765e-06, | |
| "loss": 0.0783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07033517956733704, | |
| "step": 5725, | |
| "valid_targets_mean": 2793.8, | |
| "valid_targets_min": 573 | |
| }, | |
| { | |
| "epoch": 5.617647058823529, | |
| "grad_norm": 0.5951363588950894, | |
| "learning_rate": 4.572796168980007e-06, | |
| "loss": 0.0818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07607996463775635, | |
| "step": 5730, | |
| "valid_targets_mean": 2104.0, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 5.622549019607844, | |
| "grad_norm": 0.4987162833335496, | |
| "learning_rate": 4.541729532556598e-06, | |
| "loss": 0.0795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07763554155826569, | |
| "step": 5735, | |
| "valid_targets_mean": 2146.9, | |
| "valid_targets_min": 526 | |
| }, | |
| { | |
| "epoch": 5.627450980392156, | |
| "grad_norm": 0.5669056837913017, | |
| "learning_rate": 4.5107552635923125e-06, | |
| "loss": 0.0798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07391584664583206, | |
| "step": 5740, | |
| "valid_targets_mean": 2252.2, | |
| "valid_targets_min": 761 | |
| }, | |
| { | |
| "epoch": 5.632352941176471, | |
| "grad_norm": 0.4535207599733533, | |
| "learning_rate": 4.479873547167015e-06, | |
| "loss": 0.0784, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07132834196090698, | |
| "step": 5745, | |
| "valid_targets_mean": 2706.5, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 5.637254901960784, | |
| "grad_norm": 0.5964000308881082, | |
| "learning_rate": 4.449084567807544e-06, | |
| "loss": 0.0703, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07358132302761078, | |
| "step": 5750, | |
| "valid_targets_mean": 1825.4, | |
| "valid_targets_min": 688 | |
| }, | |
| { | |
| "epoch": 5.642156862745098, | |
| "grad_norm": 0.551048381068549, | |
| "learning_rate": 4.41838850948659e-06, | |
| "loss": 0.0895, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0815672054886818, | |
| "step": 5755, | |
| "valid_targets_mean": 2226.1, | |
| "valid_targets_min": 721 | |
| }, | |
| { | |
| "epoch": 5.647058823529412, | |
| "grad_norm": 0.6183042975984667, | |
| "learning_rate": 4.38778555562164e-06, | |
| "loss": 0.0877, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08897422254085541, | |
| "step": 5760, | |
| "valid_targets_mean": 2168.8, | |
| "valid_targets_min": 640 | |
| }, | |
| { | |
| "epoch": 5.651960784313726, | |
| "grad_norm": 0.516411332728326, | |
| "learning_rate": 4.357275889073833e-06, | |
| "loss": 0.0747, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07211611419916153, | |
| "step": 5765, | |
| "valid_targets_mean": 2329.8, | |
| "valid_targets_min": 954 | |
| }, | |
| { | |
| "epoch": 5.6568627450980395, | |
| "grad_norm": 0.6726342729830591, | |
| "learning_rate": 4.326859692146898e-06, | |
| "loss": 0.0823, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07956632971763611, | |
| "step": 5770, | |
| "valid_targets_mean": 1732.2, | |
| "valid_targets_min": 574 | |
| }, | |
| { | |
| "epoch": 5.661764705882353, | |
| "grad_norm": 0.6069399571516791, | |
| "learning_rate": 4.296537146586061e-06, | |
| "loss": 0.0809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08645060658454895, | |
| "step": 5775, | |
| "valid_targets_mean": 2165.1, | |
| "valid_targets_min": 617 | |
| }, | |
| { | |
| "epoch": 5.666666666666667, | |
| "grad_norm": 0.6122057076391019, | |
| "learning_rate": 4.266308433576955e-06, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07423560321331024, | |
| "step": 5780, | |
| "valid_targets_mean": 2185.4, | |
| "valid_targets_min": 817 | |
| }, | |
| { | |
| "epoch": 5.671568627450981, | |
| "grad_norm": 0.5318258105059224, | |
| "learning_rate": 4.236173733744542e-06, | |
| "loss": 0.0829, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07637228816747665, | |
| "step": 5785, | |
| "valid_targets_mean": 2447.4, | |
| "valid_targets_min": 667 | |
| }, | |
| { | |
| "epoch": 5.676470588235294, | |
| "grad_norm": 0.5769559464124637, | |
| "learning_rate": 4.206133227152014e-06, | |
| "loss": 0.0765, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06808263808488846, | |
| "step": 5790, | |
| "valid_targets_mean": 2259.2, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 5.681372549019608, | |
| "grad_norm": 0.5541240828900388, | |
| "learning_rate": 4.1761870932997374e-06, | |
| "loss": 0.0882, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09163584560155869, | |
| "step": 5795, | |
| "valid_targets_mean": 2225.0, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 5.686274509803922, | |
| "grad_norm": 0.7151472521157439, | |
| "learning_rate": 4.146335511124186e-06, | |
| "loss": 0.0894, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07315774261951447, | |
| "step": 5800, | |
| "valid_targets_mean": 2137.9, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 5.6911764705882355, | |
| "grad_norm": 0.5293164265887055, | |
| "learning_rate": 4.11657865899685e-06, | |
| "loss": 0.0815, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08617638051509857, | |
| "step": 5805, | |
| "valid_targets_mean": 2464.4, | |
| "valid_targets_min": 813 | |
| }, | |
| { | |
| "epoch": 5.696078431372549, | |
| "grad_norm": 1.0791096159670817, | |
| "learning_rate": 4.086916714723183e-06, | |
| "loss": 0.0768, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10144323855638504, | |
| "step": 5810, | |
| "valid_targets_mean": 1979.8, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 5.700980392156863, | |
| "grad_norm": 0.5517469532362859, | |
| "learning_rate": 4.057349855541557e-06, | |
| "loss": 0.0753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07608313858509064, | |
| "step": 5815, | |
| "valid_targets_mean": 2220.6, | |
| "valid_targets_min": 740 | |
| }, | |
| { | |
| "epoch": 5.705882352941177, | |
| "grad_norm": 0.4856089404958147, | |
| "learning_rate": 4.02787825812214e-06, | |
| "loss": 0.0786, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06653439998626709, | |
| "step": 5820, | |
| "valid_targets_mean": 2393.6, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 5.71078431372549, | |
| "grad_norm": 0.6100738421759316, | |
| "learning_rate": 3.998502098565924e-06, | |
| "loss": 0.0839, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08721679449081421, | |
| "step": 5825, | |
| "valid_targets_mean": 2000.5, | |
| "valid_targets_min": 599 | |
| }, | |
| { | |
| "epoch": 5.715686274509804, | |
| "grad_norm": 0.4645738101717131, | |
| "learning_rate": 3.969221552403626e-06, | |
| "loss": 0.086, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06543716788291931, | |
| "step": 5830, | |
| "valid_targets_mean": 2722.8, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 5.720588235294118, | |
| "grad_norm": 0.6813920591255378, | |
| "learning_rate": 3.940036794594633e-06, | |
| "loss": 0.0794, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0911283940076828, | |
| "step": 5835, | |
| "valid_targets_mean": 1840.8, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 5.7254901960784315, | |
| "grad_norm": 0.7151713477093936, | |
| "learning_rate": 3.910947999525985e-06, | |
| "loss": 0.0884, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11087524890899658, | |
| "step": 5840, | |
| "valid_targets_mean": 1622.9, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 5.730392156862745, | |
| "grad_norm": 0.5463288856443642, | |
| "learning_rate": 3.881955341011303e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07577422261238098, | |
| "step": 5845, | |
| "valid_targets_mean": 2336.6, | |
| "valid_targets_min": 778 | |
| }, | |
| { | |
| "epoch": 5.735294117647059, | |
| "grad_norm": 0.6439262012281611, | |
| "learning_rate": 3.853058992289782e-06, | |
| "loss": 0.085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09771288931369781, | |
| "step": 5850, | |
| "valid_targets_mean": 1823.0, | |
| "valid_targets_min": 526 | |
| }, | |
| { | |
| "epoch": 5.740196078431373, | |
| "grad_norm": 0.7531604446568076, | |
| "learning_rate": 3.824259126025123e-06, | |
| "loss": 0.0889, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0781392902135849, | |
| "step": 5855, | |
| "valid_targets_mean": 1648.1, | |
| "valid_targets_min": 246 | |
| }, | |
| { | |
| "epoch": 5.745098039215686, | |
| "grad_norm": 0.5736859529863001, | |
| "learning_rate": 3.7955559143045296e-06, | |
| "loss": 0.0757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07947244495153427, | |
| "step": 5860, | |
| "valid_targets_mean": 2443.7, | |
| "valid_targets_min": 557 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "grad_norm": 0.543099315036524, | |
| "learning_rate": 3.7669495286376667e-06, | |
| "loss": 0.0854, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08552545309066772, | |
| "step": 5865, | |
| "valid_targets_mean": 2266.9, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 5.754901960784314, | |
| "grad_norm": 0.4491128637598562, | |
| "learning_rate": 3.738440139955628e-06, | |
| "loss": 0.0836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06782278418540955, | |
| "step": 5870, | |
| "valid_targets_mean": 2791.9, | |
| "valid_targets_min": 357 | |
| }, | |
| { | |
| "epoch": 5.759803921568627, | |
| "grad_norm": 0.5223104359168639, | |
| "learning_rate": 3.7100279186099263e-06, | |
| "loss": 0.0726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06796294450759888, | |
| "step": 5875, | |
| "valid_targets_mean": 1959.1, | |
| "valid_targets_min": 711 | |
| }, | |
| { | |
| "epoch": 5.764705882352941, | |
| "grad_norm": 0.49720653904211687, | |
| "learning_rate": 3.6817130343714814e-06, | |
| "loss": 0.0819, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07060331106185913, | |
| "step": 5880, | |
| "valid_targets_mean": 2528.8, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 5.769607843137255, | |
| "grad_norm": 0.5702908165008694, | |
| "learning_rate": 3.6534956564295844e-06, | |
| "loss": 0.0773, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07758490741252899, | |
| "step": 5885, | |
| "valid_targets_mean": 2392.8, | |
| "valid_targets_min": 820 | |
| }, | |
| { | |
| "epoch": 5.7745098039215685, | |
| "grad_norm": 0.5124808107408252, | |
| "learning_rate": 3.625375953390906e-06, | |
| "loss": 0.0804, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08320645987987518, | |
| "step": 5890, | |
| "valid_targets_mean": 2681.4, | |
| "valid_targets_min": 975 | |
| }, | |
| { | |
| "epoch": 5.779411764705882, | |
| "grad_norm": 0.535163800457717, | |
| "learning_rate": 3.597354093278489e-06, | |
| "loss": 0.0836, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08327630162239075, | |
| "step": 5895, | |
| "valid_targets_mean": 2386.8, | |
| "valid_targets_min": 777 | |
| }, | |
| { | |
| "epoch": 5.784313725490196, | |
| "grad_norm": 0.5751280657945049, | |
| "learning_rate": 3.5694302435307248e-06, | |
| "loss": 0.0815, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.083884596824646, | |
| "step": 5900, | |
| "valid_targets_mean": 2702.4, | |
| "valid_targets_min": 691 | |
| }, | |
| { | |
| "epoch": 5.78921568627451, | |
| "grad_norm": 0.47314846704263797, | |
| "learning_rate": 3.5416045710003632e-06, | |
| "loss": 0.0843, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09700559079647064, | |
| "step": 5905, | |
| "valid_targets_mean": 2672.6, | |
| "valid_targets_min": 512 | |
| }, | |
| { | |
| "epoch": 5.794117647058823, | |
| "grad_norm": 0.47095171471968333, | |
| "learning_rate": 3.5138772419535315e-06, | |
| "loss": 0.0858, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06798610091209412, | |
| "step": 5910, | |
| "valid_targets_mean": 2714.8, | |
| "valid_targets_min": 693 | |
| }, | |
| { | |
| "epoch": 5.799019607843137, | |
| "grad_norm": 0.49993363026960297, | |
| "learning_rate": 3.4862484220687277e-06, | |
| "loss": 0.0751, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0669768750667572, | |
| "step": 5915, | |
| "valid_targets_mean": 2422.4, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 5.803921568627451, | |
| "grad_norm": 0.5832018709883363, | |
| "learning_rate": 3.458718276435813e-06, | |
| "loss": 0.0842, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08323369920253754, | |
| "step": 5920, | |
| "valid_targets_mean": 2593.4, | |
| "valid_targets_min": 804 | |
| }, | |
| { | |
| "epoch": 5.8088235294117645, | |
| "grad_norm": 0.60658841552699, | |
| "learning_rate": 3.4312869695550634e-06, | |
| "loss": 0.0761, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0748073011636734, | |
| "step": 5925, | |
| "valid_targets_mean": 1869.9, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 5.813725490196078, | |
| "grad_norm": 0.48601904801057055, | |
| "learning_rate": 3.403954665336149e-06, | |
| "loss": 0.0669, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06104360893368721, | |
| "step": 5930, | |
| "valid_targets_mean": 2344.1, | |
| "valid_targets_min": 812 | |
| }, | |
| { | |
| "epoch": 5.818627450980392, | |
| "grad_norm": 0.5235780006216862, | |
| "learning_rate": 3.3767215270971755e-06, | |
| "loss": 0.0867, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11922559142112732, | |
| "step": 5935, | |
| "valid_targets_mean": 2289.8, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 5.823529411764706, | |
| "grad_norm": 0.6324375338401088, | |
| "learning_rate": 3.349587717563709e-06, | |
| "loss": 0.0802, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08978225290775299, | |
| "step": 5940, | |
| "valid_targets_mean": 2146.6, | |
| "valid_targets_min": 618 | |
| }, | |
| { | |
| "epoch": 5.828431372549019, | |
| "grad_norm": 0.6014931507395612, | |
| "learning_rate": 3.322553398867807e-06, | |
| "loss": 0.0853, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07696583867073059, | |
| "step": 5945, | |
| "valid_targets_mean": 1989.6, | |
| "valid_targets_min": 559 | |
| }, | |
| { | |
| "epoch": 5.833333333333333, | |
| "grad_norm": 0.6125932724999763, | |
| "learning_rate": 3.295618732547021e-06, | |
| "loss": 0.0897, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.12296289205551147, | |
| "step": 5950, | |
| "valid_targets_mean": 2366.6, | |
| "valid_targets_min": 387 | |
| }, | |
| { | |
| "epoch": 5.838235294117647, | |
| "grad_norm": 0.58654531214764, | |
| "learning_rate": 3.2687838795434623e-06, | |
| "loss": 0.0846, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07440722733736038, | |
| "step": 5955, | |
| "valid_targets_mean": 2124.7, | |
| "valid_targets_min": 464 | |
| }, | |
| { | |
| "epoch": 5.8431372549019605, | |
| "grad_norm": 0.6024776452686876, | |
| "learning_rate": 3.242049000202838e-06, | |
| "loss": 0.0843, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08400903642177582, | |
| "step": 5960, | |
| "valid_targets_mean": 2285.9, | |
| "valid_targets_min": 458 | |
| }, | |
| { | |
| "epoch": 5.848039215686274, | |
| "grad_norm": 0.6422151134830493, | |
| "learning_rate": 3.215414254273468e-06, | |
| "loss": 0.088, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11055001616477966, | |
| "step": 5965, | |
| "valid_targets_mean": 2167.0, | |
| "valid_targets_min": 786 | |
| }, | |
| { | |
| "epoch": 5.852941176470588, | |
| "grad_norm": 0.6620300611104084, | |
| "learning_rate": 3.18887980090536e-06, | |
| "loss": 0.0847, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08255288749933243, | |
| "step": 5970, | |
| "valid_targets_mean": 1708.4, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 5.857843137254902, | |
| "grad_norm": 0.5930788009232768, | |
| "learning_rate": 3.16244579864925e-06, | |
| "loss": 0.0841, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08282943069934845, | |
| "step": 5975, | |
| "valid_targets_mean": 2136.5, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 5.862745098039216, | |
| "grad_norm": 0.6669457051379303, | |
| "learning_rate": 3.1361124054556312e-06, | |
| "loss": 0.0832, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10319022089242935, | |
| "step": 5980, | |
| "valid_targets_mean": 2080.8, | |
| "valid_targets_min": 985 | |
| }, | |
| { | |
| "epoch": 5.867647058823529, | |
| "grad_norm": 0.7614162676195864, | |
| "learning_rate": 3.1098797786738433e-06, | |
| "loss": 0.085, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09599629044532776, | |
| "step": 5985, | |
| "valid_targets_mean": 1428.2, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 5.872549019607844, | |
| "grad_norm": 0.5860425697430323, | |
| "learning_rate": 3.0837480750511182e-06, | |
| "loss": 0.078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10159435868263245, | |
| "step": 5990, | |
| "valid_targets_mean": 2192.9, | |
| "valid_targets_min": 940 | |
| }, | |
| { | |
| "epoch": 5.877450980392156, | |
| "grad_norm": 0.5054169362663372, | |
| "learning_rate": 3.0577174507316496e-06, | |
| "loss": 0.0742, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06875654309988022, | |
| "step": 5995, | |
| "valid_targets_mean": 2596.9, | |
| "valid_targets_min": 536 | |
| }, | |
| { | |
| "epoch": 5.882352941176471, | |
| "grad_norm": 0.5199111577493412, | |
| "learning_rate": 3.031788061255634e-06, | |
| "loss": 0.0847, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07488842308521271, | |
| "step": 6000, | |
| "valid_targets_mean": 2414.6, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 5.887254901960784, | |
| "grad_norm": 0.563203147883309, | |
| "learning_rate": 3.005960061558386e-06, | |
| "loss": 0.0824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07726642489433289, | |
| "step": 6005, | |
| "valid_targets_mean": 2126.5, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 5.892156862745098, | |
| "grad_norm": 0.8120718693179341, | |
| "learning_rate": 2.980233605969374e-06, | |
| "loss": 0.0697, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09062586724758148, | |
| "step": 6010, | |
| "valid_targets_mean": 2541.3, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 5.897058823529412, | |
| "grad_norm": 0.6852211896553415, | |
| "learning_rate": 2.9546088482113046e-06, | |
| "loss": 0.0781, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07951194047927856, | |
| "step": 6015, | |
| "valid_targets_mean": 1838.3, | |
| "valid_targets_min": 892 | |
| }, | |
| { | |
| "epoch": 5.901960784313726, | |
| "grad_norm": 0.5527062870733078, | |
| "learning_rate": 2.9290859413992322e-06, | |
| "loss": 0.074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07983888685703278, | |
| "step": 6020, | |
| "valid_targets_mean": 2328.2, | |
| "valid_targets_min": 555 | |
| }, | |
| { | |
| "epoch": 5.9068627450980395, | |
| "grad_norm": 0.5587593159739197, | |
| "learning_rate": 2.903665038039616e-06, | |
| "loss": 0.0884, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07825113087892532, | |
| "step": 6025, | |
| "valid_targets_mean": 2058.4, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 5.911764705882353, | |
| "grad_norm": 0.6497842414090926, | |
| "learning_rate": 2.8783462900294035e-06, | |
| "loss": 0.0882, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09731444716453552, | |
| "step": 6030, | |
| "valid_targets_mean": 1918.4, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 5.916666666666667, | |
| "grad_norm": 0.5685987854033889, | |
| "learning_rate": 2.8531298486551564e-06, | |
| "loss": 0.0769, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07977601885795593, | |
| "step": 6035, | |
| "valid_targets_mean": 2517.5, | |
| "valid_targets_min": 695 | |
| }, | |
| { | |
| "epoch": 5.921568627450981, | |
| "grad_norm": 0.6558953205650121, | |
| "learning_rate": 2.8280158645921042e-06, | |
| "loss": 0.0834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07963621616363525, | |
| "step": 6040, | |
| "valid_targets_mean": 2447.0, | |
| "valid_targets_min": 900 | |
| }, | |
| { | |
| "epoch": 5.926470588235294, | |
| "grad_norm": 0.5166248894196765, | |
| "learning_rate": 2.803004487903287e-06, | |
| "loss": 0.0739, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07894115149974823, | |
| "step": 6045, | |
| "valid_targets_mean": 2478.1, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 5.931372549019608, | |
| "grad_norm": 0.6029700113130794, | |
| "learning_rate": 2.778095868038617e-06, | |
| "loss": 0.075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0775521844625473, | |
| "step": 6050, | |
| "valid_targets_mean": 1711.4, | |
| "valid_targets_min": 509 | |
| }, | |
| { | |
| "epoch": 5.936274509803922, | |
| "grad_norm": 0.6580105250928281, | |
| "learning_rate": 2.753290153834023e-06, | |
| "loss": 0.0851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08928344398736954, | |
| "step": 6055, | |
| "valid_targets_mean": 1875.9, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 5.9411764705882355, | |
| "grad_norm": 0.5678302569504097, | |
| "learning_rate": 2.728587493510535e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07457900792360306, | |
| "step": 6060, | |
| "valid_targets_mean": 2254.5, | |
| "valid_targets_min": 865 | |
| }, | |
| { | |
| "epoch": 5.946078431372549, | |
| "grad_norm": 0.6494590024071005, | |
| "learning_rate": 2.7039880346734016e-06, | |
| "loss": 0.0771, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09324674308300018, | |
| "step": 6065, | |
| "valid_targets_mean": 1873.6, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 5.950980392156863, | |
| "grad_norm": 0.4673173745041048, | |
| "learning_rate": 2.679491924311226e-06, | |
| "loss": 0.0851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07084433734416962, | |
| "step": 6070, | |
| "valid_targets_mean": 2818.4, | |
| "valid_targets_min": 957 | |
| }, | |
| { | |
| "epoch": 5.955882352941177, | |
| "grad_norm": 0.6025288740140903, | |
| "learning_rate": 2.655099308795077e-06, | |
| "loss": 0.0776, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08057056367397308, | |
| "step": 6075, | |
| "valid_targets_mean": 1935.4, | |
| "valid_targets_min": 656 | |
| }, | |
| { | |
| "epoch": 5.96078431372549, | |
| "grad_norm": 0.6488585540480777, | |
| "learning_rate": 2.630810333877598e-06, | |
| "loss": 0.087, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08397980779409409, | |
| "step": 6080, | |
| "valid_targets_mean": 1951.4, | |
| "valid_targets_min": 848 | |
| }, | |
| { | |
| "epoch": 5.965686274509804, | |
| "grad_norm": 0.5375473700344824, | |
| "learning_rate": 2.606625144692165e-06, | |
| "loss": 0.0813, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07249542325735092, | |
| "step": 6085, | |
| "valid_targets_mean": 2204.8, | |
| "valid_targets_min": 798 | |
| }, | |
| { | |
| "epoch": 5.970588235294118, | |
| "grad_norm": 0.5006793601077577, | |
| "learning_rate": 2.582543885752e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06924687325954437, | |
| "step": 6090, | |
| "valid_targets_mean": 2698.9, | |
| "valid_targets_min": 923 | |
| }, | |
| { | |
| "epoch": 5.9754901960784315, | |
| "grad_norm": 0.5050408076456617, | |
| "learning_rate": 2.5585667009493053e-06, | |
| "loss": 0.0813, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06595446169376373, | |
| "step": 6095, | |
| "valid_targets_mean": 2226.9, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 5.980392156862745, | |
| "grad_norm": 0.5315306645380031, | |
| "learning_rate": 2.534693733554425e-06, | |
| "loss": 0.0964, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08023549616336823, | |
| "step": 6100, | |
| "valid_targets_mean": 2730.5, | |
| "valid_targets_min": 720 | |
| }, | |
| { | |
| "epoch": 5.985294117647059, | |
| "grad_norm": 0.703760578715545, | |
| "learning_rate": 2.510925126214969e-06, | |
| "loss": 0.0803, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09702526032924652, | |
| "step": 6105, | |
| "valid_targets_mean": 1739.1, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 5.990196078431373, | |
| "grad_norm": 0.4674576209361368, | |
| "learning_rate": 2.487261020954972e-06, | |
| "loss": 0.0879, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06910908222198486, | |
| "step": 6110, | |
| "valid_targets_mean": 2932.8, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 5.995098039215686, | |
| "grad_norm": 0.5152485984762009, | |
| "learning_rate": 2.4637015591740276e-06, | |
| "loss": 0.0721, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06391312181949615, | |
| "step": 6115, | |
| "valid_targets_mean": 2238.2, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.6130671169794297, | |
| "learning_rate": 2.440246881646464e-06, | |
| "loss": 0.0788, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08633176982402802, | |
| "step": 6120, | |
| "valid_targets_mean": 1896.4, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 6.004901960784314, | |
| "grad_norm": 0.5430984365740257, | |
| "learning_rate": 2.416897128520499e-06, | |
| "loss": 0.0688, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07396151125431061, | |
| "step": 6125, | |
| "valid_targets_mean": 2240.2, | |
| "valid_targets_min": 685 | |
| }, | |
| { | |
| "epoch": 6.009803921568627, | |
| "grad_norm": 0.5763052421927343, | |
| "learning_rate": 2.393652439317382e-06, | |
| "loss": 0.066, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07698511332273483, | |
| "step": 6130, | |
| "valid_targets_mean": 1928.5, | |
| "valid_targets_min": 572 | |
| }, | |
| { | |
| "epoch": 6.014705882352941, | |
| "grad_norm": 0.48819703817877225, | |
| "learning_rate": 2.370512952930595e-06, | |
| "loss": 0.0718, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06292486190795898, | |
| "step": 6135, | |
| "valid_targets_mean": 2492.7, | |
| "valid_targets_min": 543 | |
| }, | |
| { | |
| "epoch": 6.019607843137255, | |
| "grad_norm": 0.5671829047793151, | |
| "learning_rate": 2.3474788076249988e-06, | |
| "loss": 0.0798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07058997452259064, | |
| "step": 6140, | |
| "valid_targets_mean": 2339.8, | |
| "valid_targets_min": 888 | |
| }, | |
| { | |
| "epoch": 6.0245098039215685, | |
| "grad_norm": 0.5496278977860942, | |
| "learning_rate": 2.3245501410359972e-06, | |
| "loss": 0.0809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07255090028047562, | |
| "step": 6145, | |
| "valid_targets_mean": 2154.3, | |
| "valid_targets_min": 482 | |
| }, | |
| { | |
| "epoch": 6.029411764705882, | |
| "grad_norm": 0.4937862359973531, | |
| "learning_rate": 2.301727090168744e-06, | |
| "loss": 0.0676, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0625627338886261, | |
| "step": 6150, | |
| "valid_targets_mean": 2450.2, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 6.034313725490196, | |
| "grad_norm": 0.5717628053168052, | |
| "learning_rate": 2.2790097913973154e-06, | |
| "loss": 0.0713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07228903472423553, | |
| "step": 6155, | |
| "valid_targets_mean": 2367.8, | |
| "valid_targets_min": 534 | |
| }, | |
| { | |
| "epoch": 6.03921568627451, | |
| "grad_norm": 0.5785020160716992, | |
| "learning_rate": 2.256398380463869e-06, | |
| "loss": 0.075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09864409267902374, | |
| "step": 6160, | |
| "valid_targets_mean": 2456.7, | |
| "valid_targets_min": 794 | |
| }, | |
| { | |
| "epoch": 6.044117647058823, | |
| "grad_norm": 0.804246675079279, | |
| "learning_rate": 2.233892992477873e-06, | |
| "loss": 0.0809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0953570306301117, | |
| "step": 6165, | |
| "valid_targets_mean": 1702.3, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 6.049019607843137, | |
| "grad_norm": 0.5222243433110547, | |
| "learning_rate": 2.211493761915271e-06, | |
| "loss": 0.0658, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06962548196315765, | |
| "step": 6170, | |
| "valid_targets_mean": 2213.9, | |
| "valid_targets_min": 490 | |
| }, | |
| { | |
| "epoch": 6.053921568627451, | |
| "grad_norm": 0.7161229915529039, | |
| "learning_rate": 2.1892008226176763e-06, | |
| "loss": 0.0898, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10996219515800476, | |
| "step": 6175, | |
| "valid_targets_mean": 1936.4, | |
| "valid_targets_min": 687 | |
| }, | |
| { | |
| "epoch": 6.0588235294117645, | |
| "grad_norm": 0.5057245594944937, | |
| "learning_rate": 2.1670143077916015e-06, | |
| "loss": 0.0729, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06661894172430038, | |
| "step": 6180, | |
| "valid_targets_mean": 2843.9, | |
| "valid_targets_min": 776 | |
| }, | |
| { | |
| "epoch": 6.063725490196078, | |
| "grad_norm": 0.6052163077723632, | |
| "learning_rate": 2.1449343500076304e-06, | |
| "loss": 0.077, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06724628806114197, | |
| "step": 6185, | |
| "valid_targets_mean": 2032.4, | |
| "valid_targets_min": 594 | |
| }, | |
| { | |
| "epoch": 6.068627450980392, | |
| "grad_norm": 0.7024005601958995, | |
| "learning_rate": 2.1229610811996503e-06, | |
| "loss": 0.0803, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07796479761600494, | |
| "step": 6190, | |
| "valid_targets_mean": 1835.6, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 6.073529411764706, | |
| "grad_norm": 0.45302602525860136, | |
| "learning_rate": 2.101094632664038e-06, | |
| "loss": 0.074, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06258255243301392, | |
| "step": 6195, | |
| "valid_targets_mean": 2617.1, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.078431372549019, | |
| "grad_norm": 0.5491802706652505, | |
| "learning_rate": 2.0793351350588953e-06, | |
| "loss": 0.0817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06602051109075546, | |
| "step": 6200, | |
| "valid_targets_mean": 2098.4, | |
| "valid_targets_min": 716 | |
| }, | |
| { | |
| "epoch": 6.083333333333333, | |
| "grad_norm": 0.6543025395149007, | |
| "learning_rate": 2.057682718403269e-06, | |
| "loss": 0.0804, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0722709521651268, | |
| "step": 6205, | |
| "valid_targets_mean": 1912.4, | |
| "valid_targets_min": 509 | |
| }, | |
| { | |
| "epoch": 6.088235294117647, | |
| "grad_norm": 0.7439931048575309, | |
| "learning_rate": 2.036137512076353e-06, | |
| "loss": 0.077, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09067465364933014, | |
| "step": 6210, | |
| "valid_targets_mean": 1648.8, | |
| "valid_targets_min": 684 | |
| }, | |
| { | |
| "epoch": 6.0931372549019605, | |
| "grad_norm": 0.4120655244655399, | |
| "learning_rate": 2.0146996448167443e-06, | |
| "loss": 0.0834, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.051044829189777374, | |
| "step": 6215, | |
| "valid_targets_mean": 2614.7, | |
| "valid_targets_min": 878 | |
| }, | |
| { | |
| "epoch": 6.098039215686274, | |
| "grad_norm": 0.5495070413200112, | |
| "learning_rate": 1.993369244721659e-06, | |
| "loss": 0.0775, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0813751369714737, | |
| "step": 6220, | |
| "valid_targets_mean": 2856.7, | |
| "valid_targets_min": 841 | |
| }, | |
| { | |
| "epoch": 6.102941176470588, | |
| "grad_norm": 0.5365374420440991, | |
| "learning_rate": 1.9721464392461564e-06, | |
| "loss": 0.0673, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06650418788194656, | |
| "step": 6225, | |
| "valid_targets_mean": 2202.2, | |
| "valid_targets_min": 387 | |
| }, | |
| { | |
| "epoch": 6.107843137254902, | |
| "grad_norm": 0.7655061230649243, | |
| "learning_rate": 1.951031355202386e-06, | |
| "loss": 0.0755, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09154021739959717, | |
| "step": 6230, | |
| "valid_targets_mean": 2112.3, | |
| "valid_targets_min": 669 | |
| }, | |
| { | |
| "epoch": 6.112745098039215, | |
| "grad_norm": 0.6562468194054213, | |
| "learning_rate": 1.930024118758853e-06, | |
| "loss": 0.0743, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08058568090200424, | |
| "step": 6235, | |
| "valid_targets_mean": 2077.5, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 6.117647058823529, | |
| "grad_norm": 0.6095516260099263, | |
| "learning_rate": 1.909124855439617e-06, | |
| "loss": 0.0771, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09607884287834167, | |
| "step": 6240, | |
| "valid_targets_mean": 2540.3, | |
| "valid_targets_min": 557 | |
| }, | |
| { | |
| "epoch": 6.122549019607843, | |
| "grad_norm": 0.5325419229723192, | |
| "learning_rate": 1.8883336901235849e-06, | |
| "loss": 0.0753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06588293612003326, | |
| "step": 6245, | |
| "valid_targets_mean": 2171.6, | |
| "valid_targets_min": 611 | |
| }, | |
| { | |
| "epoch": 6.127450980392156, | |
| "grad_norm": 0.5276564470921136, | |
| "learning_rate": 1.86765074704375e-06, | |
| "loss": 0.0795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06421571224927902, | |
| "step": 6250, | |
| "valid_targets_mean": 2687.8, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 6.132352941176471, | |
| "grad_norm": 0.5575953855054443, | |
| "learning_rate": 1.847076149786433e-06, | |
| "loss": 0.0696, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06234298646450043, | |
| "step": 6255, | |
| "valid_targets_mean": 2139.8, | |
| "valid_targets_min": 799 | |
| }, | |
| { | |
| "epoch": 6.137254901960785, | |
| "grad_norm": 0.49891548427790866, | |
| "learning_rate": 1.826610021290569e-06, | |
| "loss": 0.0716, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06937284022569656, | |
| "step": 6260, | |
| "valid_targets_mean": 2843.1, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 6.142156862745098, | |
| "grad_norm": 0.6665462095056259, | |
| "learning_rate": 1.8062524838469574e-06, | |
| "loss": 0.0762, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08383327722549438, | |
| "step": 6265, | |
| "valid_targets_mean": 2028.3, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.147058823529412, | |
| "grad_norm": 0.6691471419132663, | |
| "learning_rate": 1.7860036590975439e-06, | |
| "loss": 0.0776, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08335322141647339, | |
| "step": 6270, | |
| "valid_targets_mean": 1612.1, | |
| "valid_targets_min": 593 | |
| }, | |
| { | |
| "epoch": 6.151960784313726, | |
| "grad_norm": 0.6285610224128853, | |
| "learning_rate": 1.7658636680346776e-06, | |
| "loss": 0.0681, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0754515677690506, | |
| "step": 6275, | |
| "valid_targets_mean": 2081.2, | |
| "valid_targets_min": 784 | |
| }, | |
| { | |
| "epoch": 6.1568627450980395, | |
| "grad_norm": 0.572293098108352, | |
| "learning_rate": 1.745832631000395e-06, | |
| "loss": 0.0739, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06705275923013687, | |
| "step": 6280, | |
| "valid_targets_mean": 2276.4, | |
| "valid_targets_min": 927 | |
| }, | |
| { | |
| "epoch": 6.161764705882353, | |
| "grad_norm": 0.45124349299968974, | |
| "learning_rate": 1.7259106676857084e-06, | |
| "loss": 0.0695, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0617854930460453, | |
| "step": 6285, | |
| "valid_targets_mean": 2758.0, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 6.166666666666667, | |
| "grad_norm": 0.628955391242213, | |
| "learning_rate": 1.7060978971298791e-06, | |
| "loss": 0.071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06164342910051346, | |
| "step": 6290, | |
| "valid_targets_mean": 2637.5, | |
| "valid_targets_min": 958 | |
| }, | |
| { | |
| "epoch": 6.171568627450981, | |
| "grad_norm": 0.5621167700542326, | |
| "learning_rate": 1.6863944377197162e-06, | |
| "loss": 0.0662, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06263579428195953, | |
| "step": 6295, | |
| "valid_targets_mean": 2171.8, | |
| "valid_targets_min": 703 | |
| }, | |
| { | |
| "epoch": 6.176470588235294, | |
| "grad_norm": 0.5736760579865152, | |
| "learning_rate": 1.666800407188862e-06, | |
| "loss": 0.0831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.11034345626831055, | |
| "step": 6300, | |
| "valid_targets_mean": 2621.9, | |
| "valid_targets_min": 811 | |
| }, | |
| { | |
| "epoch": 6.181372549019608, | |
| "grad_norm": 0.5192244127522521, | |
| "learning_rate": 1.6473159226170898e-06, | |
| "loss": 0.0739, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06342881917953491, | |
| "step": 6305, | |
| "valid_targets_mean": 2492.2, | |
| "valid_targets_min": 760 | |
| }, | |
| { | |
| "epoch": 6.186274509803922, | |
| "grad_norm": 0.5645893985791716, | |
| "learning_rate": 1.6279411004296042e-06, | |
| "loss": 0.0741, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07318265736103058, | |
| "step": 6310, | |
| "valid_targets_mean": 2655.1, | |
| "valid_targets_min": 770 | |
| }, | |
| { | |
| "epoch": 6.1911764705882355, | |
| "grad_norm": 0.4889759389855647, | |
| "learning_rate": 1.6086760563963478e-06, | |
| "loss": 0.0726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.068635493516922, | |
| "step": 6315, | |
| "valid_targets_mean": 2335.9, | |
| "valid_targets_min": 750 | |
| }, | |
| { | |
| "epoch": 6.196078431372549, | |
| "grad_norm": 0.4874921438984983, | |
| "learning_rate": 1.5895209056313166e-06, | |
| "loss": 0.068, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06178947165608406, | |
| "step": 6320, | |
| "valid_targets_mean": 2636.8, | |
| "valid_targets_min": 765 | |
| }, | |
| { | |
| "epoch": 6.200980392156863, | |
| "grad_norm": 0.4891287624222315, | |
| "learning_rate": 1.5704757625918454e-06, | |
| "loss": 0.0691, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06538208574056625, | |
| "step": 6325, | |
| "valid_targets_mean": 3014.9, | |
| "valid_targets_min": 1074 | |
| }, | |
| { | |
| "epoch": 6.205882352941177, | |
| "grad_norm": 0.5999748822792935, | |
| "learning_rate": 1.5515407410779636e-06, | |
| "loss": 0.0717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08500402420759201, | |
| "step": 6330, | |
| "valid_targets_mean": 2314.1, | |
| "valid_targets_min": 859 | |
| }, | |
| { | |
| "epoch": 6.21078431372549, | |
| "grad_norm": 0.49689354235234345, | |
| "learning_rate": 1.5327159542316784e-06, | |
| "loss": 0.0705, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07124780118465424, | |
| "step": 6335, | |
| "valid_targets_mean": 2509.1, | |
| "valid_targets_min": 775 | |
| }, | |
| { | |
| "epoch": 6.215686274509804, | |
| "grad_norm": 0.4661213868566254, | |
| "learning_rate": 1.5140015145363163e-06, | |
| "loss": 0.0822, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05759389325976372, | |
| "step": 6340, | |
| "valid_targets_mean": 2664.4, | |
| "valid_targets_min": 647 | |
| }, | |
| { | |
| "epoch": 6.220588235294118, | |
| "grad_norm": 0.6441063247249776, | |
| "learning_rate": 1.4953975338158545e-06, | |
| "loss": 0.0678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.059634268283843994, | |
| "step": 6345, | |
| "valid_targets_mean": 1688.2, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 6.2254901960784315, | |
| "grad_norm": 0.6443076433004111, | |
| "learning_rate": 1.4769041232342464e-06, | |
| "loss": 0.0795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10399801284074783, | |
| "step": 6350, | |
| "valid_targets_mean": 2111.4, | |
| "valid_targets_min": 830 | |
| }, | |
| { | |
| "epoch": 6.230392156862745, | |
| "grad_norm": 0.5315458798893512, | |
| "learning_rate": 1.458521393294754e-06, | |
| "loss": 0.0704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.061626315116882324, | |
| "step": 6355, | |
| "valid_targets_mean": 2154.8, | |
| "valid_targets_min": 734 | |
| }, | |
| { | |
| "epoch": 6.235294117647059, | |
| "grad_norm": 0.5879437478837025, | |
| "learning_rate": 1.4402494538392975e-06, | |
| "loss": 0.0736, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06649856269359589, | |
| "step": 6360, | |
| "valid_targets_mean": 1821.4, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 6.240196078431373, | |
| "grad_norm": 0.4657404472137014, | |
| "learning_rate": 1.4220884140477841e-06, | |
| "loss": 0.0663, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05414208024740219, | |
| "step": 6365, | |
| "valid_targets_mean": 2473.9, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 6.245098039215686, | |
| "grad_norm": 0.5742416534799186, | |
| "learning_rate": 1.4040383824374738e-06, | |
| "loss": 0.0734, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07009696960449219, | |
| "step": 6370, | |
| "valid_targets_mean": 2067.3, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "grad_norm": 0.6270014269516866, | |
| "learning_rate": 1.386099466862314e-06, | |
| "loss": 0.0792, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07653535902500153, | |
| "step": 6375, | |
| "valid_targets_mean": 1960.4, | |
| "valid_targets_min": 722 | |
| }, | |
| { | |
| "epoch": 6.254901960784314, | |
| "grad_norm": 0.566807267759067, | |
| "learning_rate": 1.3682717745123154e-06, | |
| "loss": 0.0733, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08027191460132599, | |
| "step": 6380, | |
| "valid_targets_mean": 2312.1, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 6.259803921568627, | |
| "grad_norm": 0.5502588217275843, | |
| "learning_rate": 1.3505554119128861e-06, | |
| "loss": 0.071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0736890435218811, | |
| "step": 6385, | |
| "valid_targets_mean": 1971.5, | |
| "valid_targets_min": 723 | |
| }, | |
| { | |
| "epoch": 6.264705882352941, | |
| "grad_norm": 0.5131370822262248, | |
| "learning_rate": 1.3329504849242114e-06, | |
| "loss": 0.0679, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06139993295073509, | |
| "step": 6390, | |
| "valid_targets_mean": 2139.0, | |
| "valid_targets_min": 697 | |
| }, | |
| { | |
| "epoch": 6.269607843137255, | |
| "grad_norm": 0.6748155034034762, | |
| "learning_rate": 1.3154570987406202e-06, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06903311610221863, | |
| "step": 6395, | |
| "valid_targets_mean": 1693.1, | |
| "valid_targets_min": 629 | |
| }, | |
| { | |
| "epoch": 6.2745098039215685, | |
| "grad_norm": 0.5579527424856937, | |
| "learning_rate": 1.2980753578899563e-06, | |
| "loss": 0.0733, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07270924001932144, | |
| "step": 6400, | |
| "valid_targets_mean": 2242.1, | |
| "valid_targets_min": 526 | |
| }, | |
| { | |
| "epoch": 6.279411764705882, | |
| "grad_norm": 0.6649096720503214, | |
| "learning_rate": 1.2808053662329422e-06, | |
| "loss": 0.0693, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07944712042808533, | |
| "step": 6405, | |
| "valid_targets_mean": 1749.5, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 6.284313725490196, | |
| "grad_norm": 0.4422205331028691, | |
| "learning_rate": 1.2636472269625833e-06, | |
| "loss": 0.0733, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05389444902539253, | |
| "step": 6410, | |
| "valid_targets_mean": 2626.2, | |
| "valid_targets_min": 539 | |
| }, | |
| { | |
| "epoch": 6.28921568627451, | |
| "grad_norm": 0.5066639539553149, | |
| "learning_rate": 1.2466010426035214e-06, | |
| "loss": 0.0794, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06264401972293854, | |
| "step": 6415, | |
| "valid_targets_mean": 2138.4, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 6.294117647058823, | |
| "grad_norm": 0.7673642339162735, | |
| "learning_rate": 1.2296669150114381e-06, | |
| "loss": 0.0781, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08284231275320053, | |
| "step": 6420, | |
| "valid_targets_mean": 1619.0, | |
| "valid_targets_min": 826 | |
| }, | |
| { | |
| "epoch": 6.299019607843137, | |
| "grad_norm": 0.5811878185494508, | |
| "learning_rate": 1.2128449453724556e-06, | |
| "loss": 0.0738, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07080619782209396, | |
| "step": 6425, | |
| "valid_targets_mean": 2293.2, | |
| "valid_targets_min": 671 | |
| }, | |
| { | |
| "epoch": 6.303921568627451, | |
| "grad_norm": 0.4888633277532672, | |
| "learning_rate": 1.1961352342025135e-06, | |
| "loss": 0.066, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05815008282661438, | |
| "step": 6430, | |
| "valid_targets_mean": 2700.2, | |
| "valid_targets_min": 712 | |
| }, | |
| { | |
| "epoch": 6.3088235294117645, | |
| "grad_norm": 0.6944638611542142, | |
| "learning_rate": 1.1795378813467727e-06, | |
| "loss": 0.075, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08229673653841019, | |
| "step": 6435, | |
| "valid_targets_mean": 1831.1, | |
| "valid_targets_min": 757 | |
| }, | |
| { | |
| "epoch": 6.313725490196078, | |
| "grad_norm": 0.5615966039821809, | |
| "learning_rate": 1.1630529859790318e-06, | |
| "loss": 0.0926, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.061017826199531555, | |
| "step": 6440, | |
| "valid_targets_mean": 2502.1, | |
| "valid_targets_min": 657 | |
| }, | |
| { | |
| "epoch": 6.318627450980392, | |
| "grad_norm": 0.6763422320122082, | |
| "learning_rate": 1.1466806466011127e-06, | |
| "loss": 0.0882, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08773969858884811, | |
| "step": 6445, | |
| "valid_targets_mean": 1737.6, | |
| "valid_targets_min": 357 | |
| }, | |
| { | |
| "epoch": 6.323529411764706, | |
| "grad_norm": 0.5777497347614018, | |
| "learning_rate": 1.1304209610422933e-06, | |
| "loss": 0.0675, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06521066278219223, | |
| "step": 6450, | |
| "valid_targets_mean": 2161.2, | |
| "valid_targets_min": 522 | |
| }, | |
| { | |
| "epoch": 6.328431372549019, | |
| "grad_norm": 0.5679910722880794, | |
| "learning_rate": 1.1142740264587104e-06, | |
| "loss": 0.0654, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05448193848133087, | |
| "step": 6455, | |
| "valid_targets_mean": 2527.9, | |
| "valid_targets_min": 710 | |
| }, | |
| { | |
| "epoch": 6.333333333333333, | |
| "grad_norm": 0.509042676333728, | |
| "learning_rate": 1.0982399393327813e-06, | |
| "loss": 0.0736, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07492335885763168, | |
| "step": 6460, | |
| "valid_targets_mean": 2692.8, | |
| "valid_targets_min": 612 | |
| }, | |
| { | |
| "epoch": 6.338235294117647, | |
| "grad_norm": 0.6072611147130427, | |
| "learning_rate": 1.0823187954726322e-06, | |
| "loss": 0.0786, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07300375401973724, | |
| "step": 6465, | |
| "valid_targets_mean": 2096.9, | |
| "valid_targets_min": 592 | |
| }, | |
| { | |
| "epoch": 6.3431372549019605, | |
| "grad_norm": 1.5730165234423314, | |
| "learning_rate": 1.0665106900115174e-06, | |
| "loss": 0.0706, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07045087218284607, | |
| "step": 6470, | |
| "valid_targets_mean": 2194.9, | |
| "valid_targets_min": 677 | |
| }, | |
| { | |
| "epoch": 6.348039215686274, | |
| "grad_norm": 0.5320050036216825, | |
| "learning_rate": 1.0508157174072563e-06, | |
| "loss": 0.072, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06787510216236115, | |
| "step": 6475, | |
| "valid_targets_mean": 2446.6, | |
| "valid_targets_min": 683 | |
| }, | |
| { | |
| "epoch": 6.352941176470588, | |
| "grad_norm": 0.597125652728074, | |
| "learning_rate": 1.0352339714416715e-06, | |
| "loss": 0.0792, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08531296253204346, | |
| "step": 6480, | |
| "valid_targets_mean": 2121.9, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 6.357843137254902, | |
| "grad_norm": 0.7163258924570245, | |
| "learning_rate": 1.0197655452200172e-06, | |
| "loss": 0.0916, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08984912186861038, | |
| "step": 6485, | |
| "valid_targets_mean": 1824.8, | |
| "valid_targets_min": 641 | |
| }, | |
| { | |
| "epoch": 6.362745098039215, | |
| "grad_norm": 0.5664824321655607, | |
| "learning_rate": 1.0044105311704455e-06, | |
| "loss": 0.0762, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07289232313632965, | |
| "step": 6490, | |
| "valid_targets_mean": 2176.4, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 6.367647058823529, | |
| "grad_norm": 0.6202502575591109, | |
| "learning_rate": 9.891690210434235e-07, | |
| "loss": 0.0756, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07717530429363251, | |
| "step": 6495, | |
| "valid_targets_mean": 1831.9, | |
| "valid_targets_min": 616 | |
| }, | |
| { | |
| "epoch": 6.372549019607844, | |
| "grad_norm": 0.5595520685931048, | |
| "learning_rate": 9.740411059112031e-07, | |
| "loss": 0.0996, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06521070003509521, | |
| "step": 6500, | |
| "valid_targets_mean": 1985.8, | |
| "valid_targets_min": 638 | |
| }, | |
| { | |
| "epoch": 6.377450980392156, | |
| "grad_norm": 0.6126975781808517, | |
| "learning_rate": 9.590268761672838e-07, | |
| "loss": 0.0713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08089681714773178, | |
| "step": 6505, | |
| "valid_targets_mean": 1971.1, | |
| "valid_targets_min": 680 | |
| }, | |
| { | |
| "epoch": 6.382352941176471, | |
| "grad_norm": 0.5778356925814528, | |
| "learning_rate": 9.441264215258595e-07, | |
| "loss": 0.0798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07549488544464111, | |
| "step": 6510, | |
| "valid_targets_mean": 1980.7, | |
| "valid_targets_min": 619 | |
| }, | |
| { | |
| "epoch": 6.387254901960785, | |
| "grad_norm": 0.5593674597722778, | |
| "learning_rate": 9.293398310212831e-07, | |
| "loss": 0.0781, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0599190816283226, | |
| "step": 6515, | |
| "valid_targets_mean": 2093.8, | |
| "valid_targets_min": 675 | |
| }, | |
| { | |
| "epoch": 6.392156862745098, | |
| "grad_norm": 0.6399167415852633, | |
| "learning_rate": 9.146671930075412e-07, | |
| "loss": 0.0792, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10022995620965958, | |
| "step": 6520, | |
| "valid_targets_mean": 1830.9, | |
| "valid_targets_min": 727 | |
| }, | |
| { | |
| "epoch": 6.397058823529412, | |
| "grad_norm": 0.6140652768928999, | |
| "learning_rate": 9.00108595157716e-07, | |
| "loss": 0.0713, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07537797093391418, | |
| "step": 6525, | |
| "valid_targets_mean": 2011.1, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 6.401960784313726, | |
| "grad_norm": 0.7285222625014569, | |
| "learning_rate": 8.856641244634767e-07, | |
| "loss": 0.0764, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09332075715065002, | |
| "step": 6530, | |
| "valid_targets_mean": 1789.8, | |
| "valid_targets_min": 930 | |
| }, | |
| { | |
| "epoch": 6.4068627450980395, | |
| "grad_norm": 0.542847290160346, | |
| "learning_rate": 8.713338672345451e-07, | |
| "loss": 0.0709, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06750819832086563, | |
| "step": 6535, | |
| "valid_targets_mean": 2273.8, | |
| "valid_targets_min": 731 | |
| }, | |
| { | |
| "epoch": 6.411764705882353, | |
| "grad_norm": 0.5668560043361112, | |
| "learning_rate": 8.571179090981885e-07, | |
| "loss": 0.0751, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07465088367462158, | |
| "step": 6540, | |
| "valid_targets_mean": 2430.4, | |
| "valid_targets_min": 847 | |
| }, | |
| { | |
| "epoch": 6.416666666666667, | |
| "grad_norm": 0.6375632023114209, | |
| "learning_rate": 8.430163349987075e-07, | |
| "loss": 0.0851, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07439914345741272, | |
| "step": 6545, | |
| "valid_targets_mean": 1622.2, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 6.421568627450981, | |
| "grad_norm": 0.5292873557345532, | |
| "learning_rate": 8.290292291969226e-07, | |
| "loss": 0.0824, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0702093094587326, | |
| "step": 6550, | |
| "valid_targets_mean": 2615.8, | |
| "valid_targets_min": 1126 | |
| }, | |
| { | |
| "epoch": 6.426470588235294, | |
| "grad_norm": 0.5216899096730249, | |
| "learning_rate": 8.151566752696704e-07, | |
| "loss": 0.0776, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08521877229213715, | |
| "step": 6555, | |
| "valid_targets_mean": 2622.1, | |
| "valid_targets_min": 564 | |
| }, | |
| { | |
| "epoch": 6.431372549019608, | |
| "grad_norm": 0.6567140287942022, | |
| "learning_rate": 8.013987561093217e-07, | |
| "loss": 0.0795, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07485121488571167, | |
| "step": 6560, | |
| "valid_targets_mean": 1729.0, | |
| "valid_targets_min": 513 | |
| }, | |
| { | |
| "epoch": 6.436274509803922, | |
| "grad_norm": 0.601647391705643, | |
| "learning_rate": 7.877555539232551e-07, | |
| "loss": 0.0777, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08071115612983704, | |
| "step": 6565, | |
| "valid_targets_mean": 2257.5, | |
| "valid_targets_min": 745 | |
| }, | |
| { | |
| "epoch": 6.4411764705882355, | |
| "grad_norm": 0.562631688655108, | |
| "learning_rate": 7.742271502333998e-07, | |
| "loss": 0.0738, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09080608189105988, | |
| "step": 6570, | |
| "valid_targets_mean": 2499.1, | |
| "valid_targets_min": 698 | |
| }, | |
| { | |
| "epoch": 6.446078431372549, | |
| "grad_norm": 0.5329236768943743, | |
| "learning_rate": 7.608136258757248e-07, | |
| "loss": 0.0748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06047336757183075, | |
| "step": 6575, | |
| "valid_targets_mean": 2170.6, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 6.450980392156863, | |
| "grad_norm": 0.5275499144973336, | |
| "learning_rate": 7.475150609997595e-07, | |
| "loss": 0.0766, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0709349513053894, | |
| "step": 6580, | |
| "valid_targets_mean": 2400.8, | |
| "valid_targets_min": 648 | |
| }, | |
| { | |
| "epoch": 6.455882352941177, | |
| "grad_norm": 0.5101600270500972, | |
| "learning_rate": 7.343315350681246e-07, | |
| "loss": 0.0809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07443030178546906, | |
| "step": 6585, | |
| "valid_targets_mean": 2751.6, | |
| "valid_targets_min": 1062 | |
| }, | |
| { | |
| "epoch": 6.46078431372549, | |
| "grad_norm": 0.592728343239668, | |
| "learning_rate": 7.212631268560466e-07, | |
| "loss": 0.0737, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06636979430913925, | |
| "step": 6590, | |
| "valid_targets_mean": 2612.5, | |
| "valid_targets_min": 809 | |
| }, | |
| { | |
| "epoch": 6.465686274509804, | |
| "grad_norm": 0.5073605489538774, | |
| "learning_rate": 7.083099144508976e-07, | |
| "loss": 0.07, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06348920613527298, | |
| "step": 6595, | |
| "valid_targets_mean": 2609.2, | |
| "valid_targets_min": 692 | |
| }, | |
| { | |
| "epoch": 6.470588235294118, | |
| "grad_norm": 0.4888244064732885, | |
| "learning_rate": 6.95471975251718e-07, | |
| "loss": 0.0676, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07031454145908356, | |
| "step": 6600, | |
| "valid_targets_mean": 2671.5, | |
| "valid_targets_min": 803 | |
| }, | |
| { | |
| "epoch": 6.4754901960784315, | |
| "grad_norm": 0.48986840780382246, | |
| "learning_rate": 6.827493859687572e-07, | |
| "loss": 0.0685, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05794578790664673, | |
| "step": 6605, | |
| "valid_targets_mean": 2573.7, | |
| "valid_targets_min": 864 | |
| }, | |
| { | |
| "epoch": 6.480392156862745, | |
| "grad_norm": 0.5823150173486473, | |
| "learning_rate": 6.701422226230225e-07, | |
| "loss": 0.0721, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0833081379532814, | |
| "step": 6610, | |
| "valid_targets_mean": 2362.8, | |
| "valid_targets_min": 610 | |
| }, | |
| { | |
| "epoch": 6.485294117647059, | |
| "grad_norm": 0.5286454997343215, | |
| "learning_rate": 6.57650560545815e-07, | |
| "loss": 0.0744, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06765834242105484, | |
| "step": 6615, | |
| "valid_targets_mean": 2444.2, | |
| "valid_targets_min": 637 | |
| }, | |
| { | |
| "epoch": 6.490196078431373, | |
| "grad_norm": 0.6399951864881893, | |
| "learning_rate": 6.452744743782835e-07, | |
| "loss": 0.0733, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08042435348033905, | |
| "step": 6620, | |
| "valid_targets_mean": 1860.4, | |
| "valid_targets_min": 705 | |
| }, | |
| { | |
| "epoch": 6.495098039215686, | |
| "grad_norm": 0.5739546748120099, | |
| "learning_rate": 6.330140380709826e-07, | |
| "loss": 0.0712, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07454740256071091, | |
| "step": 6625, | |
| "valid_targets_mean": 2359.0, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "grad_norm": 0.5924429822824983, | |
| "learning_rate": 6.208693248834241e-07, | |
| "loss": 0.0828, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10847502946853638, | |
| "step": 6630, | |
| "valid_targets_mean": 1880.0, | |
| "valid_targets_min": 594 | |
| }, | |
| { | |
| "epoch": 6.504901960784314, | |
| "grad_norm": 0.618365600846528, | |
| "learning_rate": 6.088404073836374e-07, | |
| "loss": 0.0875, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07501372694969177, | |
| "step": 6635, | |
| "valid_targets_mean": 2034.2, | |
| "valid_targets_min": 642 | |
| }, | |
| { | |
| "epoch": 6.509803921568627, | |
| "grad_norm": 0.7514817551450208, | |
| "learning_rate": 5.969273574477519e-07, | |
| "loss": 0.0748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09385409951210022, | |
| "step": 6640, | |
| "valid_targets_mean": 1769.9, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 6.514705882352941, | |
| "grad_norm": 0.6156340459167169, | |
| "learning_rate": 5.851302462595465e-07, | |
| "loss": 0.0747, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06911460310220718, | |
| "step": 6645, | |
| "valid_targets_mean": 2101.6, | |
| "valid_targets_min": 605 | |
| }, | |
| { | |
| "epoch": 6.519607843137255, | |
| "grad_norm": 0.6492120191653313, | |
| "learning_rate": 5.734491443100342e-07, | |
| "loss": 0.0689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07696253806352615, | |
| "step": 6650, | |
| "valid_targets_mean": 1681.4, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 6.5245098039215685, | |
| "grad_norm": 0.6462025577850812, | |
| "learning_rate": 5.618841213970472e-07, | |
| "loss": 0.0787, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08429364860057831, | |
| "step": 6655, | |
| "valid_targets_mean": 1914.6, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 6.529411764705882, | |
| "grad_norm": 0.6186638590533011, | |
| "learning_rate": 5.504352466248097e-07, | |
| "loss": 0.0748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08027391880750656, | |
| "step": 6660, | |
| "valid_targets_mean": 1882.0, | |
| "valid_targets_min": 719 | |
| }, | |
| { | |
| "epoch": 6.534313725490196, | |
| "grad_norm": 0.5970585117284097, | |
| "learning_rate": 5.391025884035239e-07, | |
| "loss": 0.0735, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07223717123270035, | |
| "step": 6665, | |
| "valid_targets_mean": 1795.7, | |
| "valid_targets_min": 614 | |
| }, | |
| { | |
| "epoch": 6.53921568627451, | |
| "grad_norm": 0.6393704732605341, | |
| "learning_rate": 5.278862144489716e-07, | |
| "loss": 0.0807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07152163982391357, | |
| "step": 6670, | |
| "valid_targets_mean": 2575.4, | |
| "valid_targets_min": 741 | |
| }, | |
| { | |
| "epoch": 6.544117647058823, | |
| "grad_norm": 0.6110025268089397, | |
| "learning_rate": 5.16786191782106e-07, | |
| "loss": 0.0689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07727265357971191, | |
| "step": 6675, | |
| "valid_targets_mean": 2013.5, | |
| "valid_targets_min": 632 | |
| }, | |
| { | |
| "epoch": 6.549019607843137, | |
| "grad_norm": 0.6890547681596108, | |
| "learning_rate": 5.058025867286387e-07, | |
| "loss": 0.076, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08940169960260391, | |
| "step": 6680, | |
| "valid_targets_mean": 1832.7, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 6.553921568627451, | |
| "grad_norm": 0.526821095063165, | |
| "learning_rate": 4.949354649186666e-07, | |
| "loss": 0.0726, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09530247747898102, | |
| "step": 6685, | |
| "valid_targets_mean": 2345.0, | |
| "valid_targets_min": 468 | |
| }, | |
| { | |
| "epoch": 6.5588235294117645, | |
| "grad_norm": 0.5638933046634943, | |
| "learning_rate": 4.841848912862568e-07, | |
| "loss": 0.0842, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06429076939821243, | |
| "step": 6690, | |
| "valid_targets_mean": 2157.4, | |
| "valid_targets_min": 962 | |
| }, | |
| { | |
| "epoch": 6.563725490196078, | |
| "grad_norm": 0.7151405526912445, | |
| "learning_rate": 4.7355093006907325e-07, | |
| "loss": 0.0837, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06609296798706055, | |
| "step": 6695, | |
| "valid_targets_mean": 1982.6, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 6.568627450980392, | |
| "grad_norm": 0.5698233570409097, | |
| "learning_rate": 4.630336448079864e-07, | |
| "loss": 0.0698, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07055316120386124, | |
| "step": 6700, | |
| "valid_targets_mean": 2221.2, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.573529411764706, | |
| "grad_norm": 0.6362640401880392, | |
| "learning_rate": 4.5263309834669753e-07, | |
| "loss": 0.0728, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07047757506370544, | |
| "step": 6705, | |
| "valid_targets_mean": 2399.4, | |
| "valid_targets_min": 651 | |
| }, | |
| { | |
| "epoch": 6.578431372549019, | |
| "grad_norm": 0.624490274090889, | |
| "learning_rate": 4.423493528313638e-07, | |
| "loss": 0.0738, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07188694179058075, | |
| "step": 6710, | |
| "valid_targets_mean": 2270.2, | |
| "valid_targets_min": 602 | |
| }, | |
| { | |
| "epoch": 6.583333333333333, | |
| "grad_norm": 0.6588545635409063, | |
| "learning_rate": 4.3218246971021394e-07, | |
| "loss": 0.0763, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10030744224786758, | |
| "step": 6715, | |
| "valid_targets_mean": 1812.2, | |
| "valid_targets_min": 613 | |
| }, | |
| { | |
| "epoch": 6.588235294117647, | |
| "grad_norm": 0.6143564999883911, | |
| "learning_rate": 4.2213250973319964e-07, | |
| "loss": 0.0759, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.1096772700548172, | |
| "step": 6720, | |
| "valid_targets_mean": 2134.1, | |
| "valid_targets_min": 700 | |
| }, | |
| { | |
| "epoch": 6.5931372549019605, | |
| "grad_norm": 0.6373064405570183, | |
| "learning_rate": 4.121995329516293e-07, | |
| "loss": 0.0757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.075977623462677, | |
| "step": 6725, | |
| "valid_targets_mean": 1834.6, | |
| "valid_targets_min": 598 | |
| }, | |
| { | |
| "epoch": 6.598039215686274, | |
| "grad_norm": 0.5775137974021963, | |
| "learning_rate": 4.0238359871779045e-07, | |
| "loss": 0.0763, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07746519148349762, | |
| "step": 6730, | |
| "valid_targets_mean": 2040.2, | |
| "valid_targets_min": 729 | |
| }, | |
| { | |
| "epoch": 6.602941176470588, | |
| "grad_norm": 0.6154801243115351, | |
| "learning_rate": 3.926847656846189e-07, | |
| "loss": 0.0811, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07492032647132874, | |
| "step": 6735, | |
| "valid_targets_mean": 1690.9, | |
| "valid_targets_min": 516 | |
| }, | |
| { | |
| "epoch": 6.607843137254902, | |
| "grad_norm": 0.5285933942636529, | |
| "learning_rate": 3.831030918053369e-07, | |
| "loss": 0.068, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06620896607637405, | |
| "step": 6740, | |
| "valid_targets_mean": 2294.8, | |
| "valid_targets_min": 955 | |
| }, | |
| { | |
| "epoch": 6.612745098039216, | |
| "grad_norm": 0.5893823543019614, | |
| "learning_rate": 3.736386343331044e-07, | |
| "loss": 0.0818, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09671109914779663, | |
| "step": 6745, | |
| "valid_targets_mean": 2318.4, | |
| "valid_targets_min": 853 | |
| }, | |
| { | |
| "epoch": 6.617647058823529, | |
| "grad_norm": 0.5462674467458982, | |
| "learning_rate": 3.64291449820684e-07, | |
| "loss": 0.0748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.061107758432626724, | |
| "step": 6750, | |
| "valid_targets_mean": 2305.8, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 6.622549019607844, | |
| "grad_norm": 0.6074116112610899, | |
| "learning_rate": 3.5506159412010076e-07, | |
| "loss": 0.0731, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06659223139286041, | |
| "step": 6755, | |
| "valid_targets_mean": 2211.5, | |
| "valid_targets_min": 756 | |
| }, | |
| { | |
| "epoch": 6.627450980392156, | |
| "grad_norm": 0.5632099448887126, | |
| "learning_rate": 3.4594912238230307e-07, | |
| "loss": 0.0715, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0652645155787468, | |
| "step": 6760, | |
| "valid_targets_mean": 2370.2, | |
| "valid_targets_min": 502 | |
| }, | |
| { | |
| "epoch": 6.632352941176471, | |
| "grad_norm": 0.4944591686617319, | |
| "learning_rate": 3.3695408905684456e-07, | |
| "loss": 0.0817, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.054533012211322784, | |
| "step": 6765, | |
| "valid_targets_mean": 2398.9, | |
| "valid_targets_min": 649 | |
| }, | |
| { | |
| "epoch": 6.637254901960784, | |
| "grad_norm": 0.5915700018309602, | |
| "learning_rate": 3.280765478915449e-07, | |
| "loss": 0.0704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07783752679824829, | |
| "step": 6770, | |
| "valid_targets_mean": 2240.8, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 6.642156862745098, | |
| "grad_norm": 0.49047093994848395, | |
| "learning_rate": 3.193165519321739e-07, | |
| "loss": 0.0704, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0692506730556488, | |
| "step": 6775, | |
| "valid_targets_mean": 2598.6, | |
| "valid_targets_min": 1083 | |
| }, | |
| { | |
| "epoch": 6.647058823529412, | |
| "grad_norm": 0.5773913930901008, | |
| "learning_rate": 3.1067415352214583e-07, | |
| "loss": 0.0798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07195072621107101, | |
| "step": 6780, | |
| "valid_targets_mean": 2246.1, | |
| "valid_targets_min": 654 | |
| }, | |
| { | |
| "epoch": 6.651960784313726, | |
| "grad_norm": 0.6854796204536878, | |
| "learning_rate": 3.0214940430218996e-07, | |
| "loss": 0.078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08964002132415771, | |
| "step": 6785, | |
| "valid_targets_mean": 1977.6, | |
| "valid_targets_min": 636 | |
| }, | |
| { | |
| "epoch": 6.6568627450980395, | |
| "grad_norm": 0.589375571510179, | |
| "learning_rate": 2.9374235521005157e-07, | |
| "loss": 0.0697, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0778995007276535, | |
| "step": 6790, | |
| "valid_targets_mean": 2066.0, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 6.661764705882353, | |
| "grad_norm": 0.5131243023890815, | |
| "learning_rate": 2.8545305648018274e-07, | |
| "loss": 0.0691, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0669882595539093, | |
| "step": 6795, | |
| "valid_targets_mean": 2585.2, | |
| "valid_targets_min": 910 | |
| }, | |
| { | |
| "epoch": 6.666666666666667, | |
| "grad_norm": 0.57976654440454, | |
| "learning_rate": 2.772815576434429e-07, | |
| "loss": 0.0732, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07049468159675598, | |
| "step": 6800, | |
| "valid_targets_mean": 2076.2, | |
| "valid_targets_min": 923 | |
| }, | |
| { | |
| "epoch": 6.671568627450981, | |
| "grad_norm": 0.6333760051695486, | |
| "learning_rate": 2.6922790752681004e-07, | |
| "loss": 0.0723, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08557892590761185, | |
| "step": 6805, | |
| "valid_targets_mean": 2019.9, | |
| "valid_targets_min": 615 | |
| }, | |
| { | |
| "epoch": 6.676470588235294, | |
| "grad_norm": 0.5500510919093051, | |
| "learning_rate": 2.6129215425307884e-07, | |
| "loss": 0.0748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07296749949455261, | |
| "step": 6810, | |
| "valid_targets_mean": 2380.2, | |
| "valid_targets_min": 961 | |
| }, | |
| { | |
| "epoch": 6.681372549019608, | |
| "grad_norm": 0.5284267171327073, | |
| "learning_rate": 2.534743452405786e-07, | |
| "loss": 0.0783, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07741603255271912, | |
| "step": 6815, | |
| "valid_targets_mean": 2762.5, | |
| "valid_targets_min": 821 | |
| }, | |
| { | |
| "epoch": 6.686274509803922, | |
| "grad_norm": 0.5208764125308674, | |
| "learning_rate": 2.4577452720289554e-07, | |
| "loss": 0.073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06157371401786804, | |
| "step": 6820, | |
| "valid_targets_mean": 2351.6, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 6.6911764705882355, | |
| "grad_norm": 0.7033957323624661, | |
| "learning_rate": 2.3819274614857114e-07, | |
| "loss": 0.0731, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06887654215097427, | |
| "step": 6825, | |
| "valid_targets_mean": 1957.0, | |
| "valid_targets_min": 717 | |
| }, | |
| { | |
| "epoch": 6.696078431372549, | |
| "grad_norm": 0.5497440299921376, | |
| "learning_rate": 2.3072904738085766e-07, | |
| "loss": 0.0799, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06968720257282257, | |
| "step": 6830, | |
| "valid_targets_mean": 2151.9, | |
| "valid_targets_min": 649 | |
| }, | |
| { | |
| "epoch": 6.700980392156863, | |
| "grad_norm": 0.5027531856286584, | |
| "learning_rate": 2.2338347549742956e-07, | |
| "loss": 0.0729, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07394878566265106, | |
| "step": 6835, | |
| "valid_targets_mean": 2225.8, | |
| "valid_targets_min": 824 | |
| }, | |
| { | |
| "epoch": 6.705882352941177, | |
| "grad_norm": 0.4273091436307248, | |
| "learning_rate": 2.1615607439011698e-07, | |
| "loss": 0.0653, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05667303502559662, | |
| "step": 6840, | |
| "valid_targets_mean": 3133.9, | |
| "valid_targets_min": 852 | |
| }, | |
| { | |
| "epoch": 6.71078431372549, | |
| "grad_norm": 0.5815477377429077, | |
| "learning_rate": 2.0904688724464828e-07, | |
| "loss": 0.0845, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08845975250005722, | |
| "step": 6845, | |
| "valid_targets_mean": 2000.3, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 6.715686274509804, | |
| "grad_norm": 1.2981100793881648, | |
| "learning_rate": 2.0205595654039012e-07, | |
| "loss": 0.0655, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06515184044837952, | |
| "step": 6850, | |
| "valid_targets_mean": 2063.4, | |
| "valid_targets_min": 1143 | |
| }, | |
| { | |
| "epoch": 6.720588235294118, | |
| "grad_norm": 0.6841758509817507, | |
| "learning_rate": 1.9518332405009887e-07, | |
| "loss": 0.0798, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07749398797750473, | |
| "step": 6855, | |
| "valid_targets_mean": 1602.4, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 6.7254901960784315, | |
| "grad_norm": 0.4924527381520498, | |
| "learning_rate": 1.8842903083965857e-07, | |
| "loss": 0.0694, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06516847014427185, | |
| "step": 6860, | |
| "valid_targets_mean": 2914.9, | |
| "valid_targets_min": 591 | |
| }, | |
| { | |
| "epoch": 6.730392156862745, | |
| "grad_norm": 0.6303188289451982, | |
| "learning_rate": 1.8179311726784777e-07, | |
| "loss": 0.0757, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07469521462917328, | |
| "step": 6865, | |
| "valid_targets_mean": 2210.1, | |
| "valid_targets_min": 811 | |
| }, | |
| { | |
| "epoch": 6.735294117647059, | |
| "grad_norm": 0.8053913454744731, | |
| "learning_rate": 1.7527562298609524e-07, | |
| "loss": 0.0666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06828522682189941, | |
| "step": 6870, | |
| "valid_targets_mean": 1897.0, | |
| "valid_targets_min": 582 | |
| }, | |
| { | |
| "epoch": 6.740196078431373, | |
| "grad_norm": 0.5568843332211943, | |
| "learning_rate": 1.6887658693823583e-07, | |
| "loss": 0.0678, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07886748015880585, | |
| "step": 6875, | |
| "valid_targets_mean": 2553.4, | |
| "valid_targets_min": 916 | |
| }, | |
| { | |
| "epoch": 6.745098039215686, | |
| "grad_norm": 0.691870344299773, | |
| "learning_rate": 1.6259604736029057e-07, | |
| "loss": 0.0666, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0789848044514656, | |
| "step": 6880, | |
| "valid_targets_mean": 1709.3, | |
| "valid_targets_min": 664 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "grad_norm": 0.5896811602843932, | |
| "learning_rate": 1.564340417802246e-07, | |
| "loss": 0.0717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07563844323158264, | |
| "step": 6885, | |
| "valid_targets_mean": 2423.1, | |
| "valid_targets_min": 424 | |
| }, | |
| { | |
| "epoch": 6.754901960784314, | |
| "grad_norm": 0.5474320570604836, | |
| "learning_rate": 1.503906070177341e-07, | |
| "loss": 0.0947, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08105449378490448, | |
| "step": 6890, | |
| "valid_targets_mean": 2342.6, | |
| "valid_targets_min": 791 | |
| }, | |
| { | |
| "epoch": 6.759803921568627, | |
| "grad_norm": 0.6121453772298724, | |
| "learning_rate": 1.4446577918401982e-07, | |
| "loss": 0.073, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07447556406259537, | |
| "step": 6895, | |
| "valid_targets_mean": 2341.2, | |
| "valid_targets_min": 549 | |
| }, | |
| { | |
| "epoch": 6.764705882352941, | |
| "grad_norm": 0.6114492732117995, | |
| "learning_rate": 1.38659593681576e-07, | |
| "loss": 0.083, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07287880033254623, | |
| "step": 6900, | |
| "valid_targets_mean": 2059.4, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 6.769607843137255, | |
| "grad_norm": 0.6588140419762275, | |
| "learning_rate": 1.3297208520397064e-07, | |
| "loss": 0.0797, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07726430892944336, | |
| "step": 6905, | |
| "valid_targets_mean": 1786.8, | |
| "valid_targets_min": 246 | |
| }, | |
| { | |
| "epoch": 6.7745098039215685, | |
| "grad_norm": 0.5657506011735933, | |
| "learning_rate": 1.274032877356457e-07, | |
| "loss": 0.0778, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07481957226991653, | |
| "step": 6910, | |
| "valid_targets_mean": 2344.2, | |
| "valid_targets_min": 655 | |
| }, | |
| { | |
| "epoch": 6.779411764705882, | |
| "grad_norm": 0.5928810950417512, | |
| "learning_rate": 1.219532345517127e-07, | |
| "loss": 0.0709, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06960079818964005, | |
| "step": 6915, | |
| "valid_targets_mean": 1776.4, | |
| "valid_targets_min": 503 | |
| }, | |
| { | |
| "epoch": 6.784313725490196, | |
| "grad_norm": 0.6762501537064162, | |
| "learning_rate": 1.1662195821775746e-07, | |
| "loss": 0.0776, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0917593464255333, | |
| "step": 6920, | |
| "valid_targets_mean": 1760.2, | |
| "valid_targets_min": 839 | |
| }, | |
| { | |
| "epoch": 6.78921568627451, | |
| "grad_norm": 0.5864555760924591, | |
| "learning_rate": 1.1140949058963125e-07, | |
| "loss": 0.0892, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08153297007083893, | |
| "step": 6925, | |
| "valid_targets_mean": 2450.8, | |
| "valid_targets_min": 1001 | |
| }, | |
| { | |
| "epoch": 6.794117647058823, | |
| "grad_norm": 0.6232735656011807, | |
| "learning_rate": 1.0631586281327988e-07, | |
| "loss": 0.0891, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08557818084955215, | |
| "step": 6930, | |
| "valid_targets_mean": 1834.9, | |
| "valid_targets_min": 766 | |
| }, | |
| { | |
| "epoch": 6.799019607843137, | |
| "grad_norm": 0.5980549584459026, | |
| "learning_rate": 1.0134110532454167e-07, | |
| "loss": 0.0659, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06608110666275024, | |
| "step": 6935, | |
| "valid_targets_mean": 2118.2, | |
| "valid_targets_min": 646 | |
| }, | |
| { | |
| "epoch": 6.803921568627451, | |
| "grad_norm": 0.5574126476293245, | |
| "learning_rate": 9.648524784897418e-08, | |
| "loss": 0.0839, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07937115430831909, | |
| "step": 6940, | |
| "valid_targets_mean": 2194.7, | |
| "valid_targets_min": 746 | |
| }, | |
| { | |
| "epoch": 6.8088235294117645, | |
| "grad_norm": 0.5488303476404753, | |
| "learning_rate": 9.17483194016744e-08, | |
| "loss": 0.0794, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.09899978339672089, | |
| "step": 6945, | |
| "valid_targets_mean": 2651.8, | |
| "valid_targets_min": 732 | |
| }, | |
| { | |
| "epoch": 6.813725490196078, | |
| "grad_norm": 0.5878960633853861, | |
| "learning_rate": 8.713034828710554e-08, | |
| "loss": 0.0807, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07570293545722961, | |
| "step": 6950, | |
| "valid_targets_mean": 2299.9, | |
| "valid_targets_min": 774 | |
| }, | |
| { | |
| "epoch": 6.818627450980392, | |
| "grad_norm": 0.5334957139961155, | |
| "learning_rate": 8.26313620989283e-08, | |
| "loss": 0.082, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07348759472370148, | |
| "step": 6955, | |
| "valid_targets_mean": 2329.1, | |
| "valid_targets_min": 858 | |
| }, | |
| { | |
| "epoch": 6.823529411764706, | |
| "grad_norm": 0.5061147065315302, | |
| "learning_rate": 7.825138771983432e-08, | |
| "loss": 0.069, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05788544565439224, | |
| "step": 6960, | |
| "valid_targets_mean": 2415.1, | |
| "valid_targets_min": 714 | |
| }, | |
| { | |
| "epoch": 6.828431372549019, | |
| "grad_norm": 0.5150049175490301, | |
| "learning_rate": 7.399045132138627e-08, | |
| "loss": 0.0792, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06460592150688171, | |
| "step": 6965, | |
| "valid_targets_mean": 2752.2, | |
| "valid_targets_min": 713 | |
| }, | |
| { | |
| "epoch": 6.833333333333333, | |
| "grad_norm": 0.4705212783129525, | |
| "learning_rate": 6.98485783638625e-08, | |
| "loss": 0.0744, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0671989843249321, | |
| "step": 6970, | |
| "valid_targets_mean": 2491.8, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 6.838235294117647, | |
| "grad_norm": 0.5857825766534208, | |
| "learning_rate": 6.582579359610597e-08, | |
| "loss": 0.0754, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07692444324493408, | |
| "step": 6975, | |
| "valid_targets_mean": 2063.6, | |
| "valid_targets_min": 567 | |
| }, | |
| { | |
| "epoch": 6.8431372549019605, | |
| "grad_norm": 0.5021496002955151, | |
| "learning_rate": 6.192212105537109e-08, | |
| "loss": 0.078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08223918080329895, | |
| "step": 6980, | |
| "valid_targets_mean": 2688.7, | |
| "valid_targets_min": 686 | |
| }, | |
| { | |
| "epoch": 6.848039215686274, | |
| "grad_norm": 0.4348685643630877, | |
| "learning_rate": 5.813758406718606e-08, | |
| "loss": 0.0689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05415298044681549, | |
| "step": 6985, | |
| "valid_targets_mean": 2850.8, | |
| "valid_targets_min": 597 | |
| }, | |
| { | |
| "epoch": 6.852941176470588, | |
| "grad_norm": 0.5560387400479142, | |
| "learning_rate": 5.4472205245208463e-08, | |
| "loss": 0.0813, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06583650410175323, | |
| "step": 6990, | |
| "valid_targets_mean": 2302.1, | |
| "valid_targets_min": 758 | |
| }, | |
| { | |
| "epoch": 6.857843137254902, | |
| "grad_norm": 0.565742943832882, | |
| "learning_rate": 5.092600649109658e-08, | |
| "loss": 0.0784, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06734960526227951, | |
| "step": 6995, | |
| "valid_targets_mean": 2259.1, | |
| "valid_targets_min": 763 | |
| }, | |
| { | |
| "epoch": 6.862745098039216, | |
| "grad_norm": 0.5171417741345915, | |
| "learning_rate": 4.7499008994373875e-08, | |
| "loss": 0.0727, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06790250539779663, | |
| "step": 7000, | |
| "valid_targets_mean": 2513.7, | |
| "valid_targets_min": 953 | |
| }, | |
| { | |
| "epoch": 6.867647058823529, | |
| "grad_norm": 0.503597251844863, | |
| "learning_rate": 4.4191233232300235e-08, | |
| "loss": 0.0685, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06478115916252136, | |
| "step": 7005, | |
| "valid_targets_mean": 2516.2, | |
| "valid_targets_min": 530 | |
| }, | |
| { | |
| "epoch": 6.872549019607844, | |
| "grad_norm": 0.6054050789758687, | |
| "learning_rate": 4.10026989697565e-08, | |
| "loss": 0.0707, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06746941804885864, | |
| "step": 7010, | |
| "valid_targets_mean": 2090.3, | |
| "valid_targets_min": 1144 | |
| }, | |
| { | |
| "epoch": 6.877450980392156, | |
| "grad_norm": 0.5606095020900572, | |
| "learning_rate": 3.793342525911792e-08, | |
| "loss": 0.0857, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0673239454627037, | |
| "step": 7015, | |
| "valid_targets_mean": 2314.0, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 6.882352941176471, | |
| "grad_norm": 0.6021982902697057, | |
| "learning_rate": 3.4983430440149733e-08, | |
| "loss": 0.0831, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.10573314130306244, | |
| "step": 7020, | |
| "valid_targets_mean": 2055.6, | |
| "valid_targets_min": 681 | |
| }, | |
| { | |
| "epoch": 6.887254901960784, | |
| "grad_norm": 0.5568757053589237, | |
| "learning_rate": 3.215273213988734e-08, | |
| "loss": 0.0764, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07062698900699615, | |
| "step": 7025, | |
| "valid_targets_mean": 2395.8, | |
| "valid_targets_min": 724 | |
| }, | |
| { | |
| "epoch": 6.892156862745098, | |
| "grad_norm": 0.7240829790900177, | |
| "learning_rate": 2.9441347272543e-08, | |
| "loss": 0.0719, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0858493223786354, | |
| "step": 7030, | |
| "valid_targets_mean": 1650.8, | |
| "valid_targets_min": 748 | |
| }, | |
| { | |
| "epoch": 6.897058823529412, | |
| "grad_norm": 0.6349762124135037, | |
| "learning_rate": 2.6849292039390352e-08, | |
| "loss": 0.0685, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0749552771449089, | |
| "step": 7035, | |
| "valid_targets_mean": 1839.1, | |
| "valid_targets_min": 544 | |
| }, | |
| { | |
| "epoch": 6.901960784313726, | |
| "grad_norm": 0.5676422569750085, | |
| "learning_rate": 2.4376581928682307e-08, | |
| "loss": 0.0931, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06373101472854614, | |
| "step": 7040, | |
| "valid_targets_mean": 2116.4, | |
| "valid_targets_min": 678 | |
| }, | |
| { | |
| "epoch": 6.9068627450980395, | |
| "grad_norm": 0.5744939628095065, | |
| "learning_rate": 2.2023231715546655e-08, | |
| "loss": 0.0721, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06783310323953629, | |
| "step": 7045, | |
| "valid_targets_mean": 2597.4, | |
| "valid_targets_min": 813 | |
| }, | |
| { | |
| "epoch": 6.911764705882353, | |
| "grad_norm": 0.6284874290227057, | |
| "learning_rate": 1.9789255461906132e-08, | |
| "loss": 0.0809, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08478754758834839, | |
| "step": 7050, | |
| "valid_targets_mean": 2029.5, | |
| "valid_targets_min": 531 | |
| }, | |
| { | |
| "epoch": 6.916666666666667, | |
| "grad_norm": 0.5411737246105554, | |
| "learning_rate": 1.7674666516387383e-08, | |
| "loss": 0.0718, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06868691742420197, | |
| "step": 7055, | |
| "valid_targets_mean": 2305.8, | |
| "valid_targets_min": 673 | |
| }, | |
| { | |
| "epoch": 6.921568627450981, | |
| "grad_norm": 0.49737197499649116, | |
| "learning_rate": 1.56794775142477e-08, | |
| "loss": 0.0735, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06329074501991272, | |
| "step": 7060, | |
| "valid_targets_mean": 2280.2, | |
| "valid_targets_min": 589 | |
| }, | |
| { | |
| "epoch": 6.926470588235294, | |
| "grad_norm": 0.6751026916886007, | |
| "learning_rate": 1.380370037729728e-08, | |
| "loss": 0.0717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08442050218582153, | |
| "step": 7065, | |
| "valid_targets_mean": 2083.4, | |
| "valid_targets_min": 659 | |
| }, | |
| { | |
| "epoch": 6.931372549019608, | |
| "grad_norm": 0.5030081784740777, | |
| "learning_rate": 1.204734631382154e-08, | |
| "loss": 0.0862, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06494186818599701, | |
| "step": 7070, | |
| "valid_targets_mean": 2563.1, | |
| "valid_targets_min": 869 | |
| }, | |
| { | |
| "epoch": 6.936274509803922, | |
| "grad_norm": 0.5113385667698365, | |
| "learning_rate": 1.0410425818525582e-08, | |
| "loss": 0.0749, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07213737070560455, | |
| "step": 7075, | |
| "valid_targets_mean": 2695.8, | |
| "valid_targets_min": 800 | |
| }, | |
| { | |
| "epoch": 6.9411764705882355, | |
| "grad_norm": 0.6583346196023455, | |
| "learning_rate": 8.892948672465373e-09, | |
| "loss": 0.078, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07703264057636261, | |
| "step": 7080, | |
| "valid_targets_mean": 2112.5, | |
| "valid_targets_min": 728 | |
| }, | |
| { | |
| "epoch": 6.946078431372549, | |
| "grad_norm": 0.4106397687578483, | |
| "learning_rate": 7.494923942985566e-09, | |
| "loss": 0.0689, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.05719529092311859, | |
| "step": 7085, | |
| "valid_targets_mean": 3062.9, | |
| "valid_targets_min": 634 | |
| }, | |
| { | |
| "epoch": 6.950980392156863, | |
| "grad_norm": 0.5361202378127926, | |
| "learning_rate": 6.216359983675091e-09, | |
| "loss": 0.0725, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.0644589364528656, | |
| "step": 7090, | |
| "valid_targets_mean": 2359.8, | |
| "valid_targets_min": 682 | |
| }, | |
| { | |
| "epoch": 6.955882352941177, | |
| "grad_norm": 0.5868554124771347, | |
| "learning_rate": 5.057264434307208e-09, | |
| "loss": 0.0747, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06894619762897491, | |
| "step": 7095, | |
| "valid_targets_mean": 2037.6, | |
| "valid_targets_min": 835 | |
| }, | |
| { | |
| "epoch": 6.96078431372549, | |
| "grad_norm": 0.5026454645152694, | |
| "learning_rate": 4.017644220797312e-09, | |
| "loss": 0.0685, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06370195746421814, | |
| "step": 7100, | |
| "valid_targets_mean": 2236.1, | |
| "valid_targets_min": 822 | |
| }, | |
| { | |
| "epoch": 6.965686274509804, | |
| "grad_norm": 0.5172626012676429, | |
| "learning_rate": 3.097505555169633e-09, | |
| "loss": 0.071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07044585049152374, | |
| "step": 7105, | |
| "valid_targets_mean": 2222.4, | |
| "valid_targets_min": 536 | |
| }, | |
| { | |
| "epoch": 6.970588235294118, | |
| "grad_norm": 0.48563914103910366, | |
| "learning_rate": 2.2968539355061603e-09, | |
| "loss": 0.0641, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.060300637036561966, | |
| "step": 7110, | |
| "valid_targets_mean": 2759.5, | |
| "valid_targets_min": 819 | |
| }, | |
| { | |
| "epoch": 6.9754901960784315, | |
| "grad_norm": 0.5009744586716859, | |
| "learning_rate": 1.6156941459222197e-09, | |
| "loss": 0.0748, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08746712654829025, | |
| "step": 7115, | |
| "valid_targets_mean": 2686.8, | |
| "valid_targets_min": 928 | |
| }, | |
| { | |
| "epoch": 6.980392156862745, | |
| "grad_norm": 0.6099074515269169, | |
| "learning_rate": 1.0540302565376081e-09, | |
| "loss": 0.0655, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06961031258106232, | |
| "step": 7120, | |
| "valid_targets_mean": 2117.1, | |
| "valid_targets_min": 788 | |
| }, | |
| { | |
| "epoch": 6.985294117647059, | |
| "grad_norm": 0.5865242816170935, | |
| "learning_rate": 6.118656234499476e-10, | |
| "loss": 0.071, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.07421500980854034, | |
| "step": 7125, | |
| "valid_targets_mean": 2357.5, | |
| "valid_targets_min": 744 | |
| }, | |
| { | |
| "epoch": 6.990196078431373, | |
| "grad_norm": 0.5149688957802767, | |
| "learning_rate": 2.892028887147014e-10, | |
| "loss": 0.0717, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.06031516194343567, | |
| "step": 7130, | |
| "valid_targets_mean": 2506.4, | |
| "valid_targets_min": 704 | |
| }, | |
| { | |
| "epoch": 6.995098039215686, | |
| "grad_norm": 0.6304851349269752, | |
| "learning_rate": 8.604398032963091e-11, | |
| "loss": 0.0967, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.08709007501602173, | |
| "step": 7135, | |
| "valid_targets_mean": 2069.6, | |
| "valid_targets_min": 737 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.48718293954798514, | |
| "learning_rate": 2.3901122303549018e-12, | |
| "loss": 0.0753, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.059055425226688385, | |
| "step": 7140, | |
| "valid_targets_mean": 2420.6, | |
| "valid_targets_min": 743 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "loss_nan_ranks": 0, | |
| "loss_rank_avg": 0.059055425226688385, | |
| "step": 7140, | |
| "total_flos": 1483460788355072.0, | |
| "train_loss": 0.12303723783469668, | |
| "train_runtime": 34233.4373, | |
| "train_samples_per_second": 3.334, | |
| "train_steps_per_second": 0.209, | |
| "valid_targets_mean": 2420.6, | |
| "valid_targets_min": 743 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 7140, | |
| "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": 1483460788355072.0, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |