Quentin Meeus
Finetune NER+ASR module for 5000 steps with LoRA
1e16abf
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 4.647006255585344,
"eval_steps": 200,
"global_step": 2600,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.18,
"learning_rate": 5e-05,
"loss": 0.4292,
"step": 100
},
{
"epoch": 0.36,
"learning_rate": 4.994863481875841e-05,
"loss": 0.2746,
"step": 200
},
{
"epoch": 0.36,
"eval_f1_score": 0.6564825695260478,
"eval_label_f1": 0.8343125734430082,
"eval_loss": 0.26023727655410767,
"eval_runtime": 344.2386,
"eval_samples_per_second": 2.905,
"eval_steps_per_second": 0.363,
"eval_wer": 0.10898676368139949,
"step": 200
},
{
"epoch": 0.54,
"learning_rate": 4.979475034558115e-05,
"loss": 0.2568,
"step": 300
},
{
"epoch": 0.71,
"learning_rate": 4.9538978924776634e-05,
"loss": 0.2481,
"step": 400
},
{
"epoch": 0.71,
"eval_f1_score": 0.6577916992952232,
"eval_label_f1": 0.8347689898198903,
"eval_loss": 0.246540829539299,
"eval_runtime": 341.9234,
"eval_samples_per_second": 2.925,
"eval_steps_per_second": 0.366,
"eval_wer": 0.10217509095131203,
"step": 400
},
{
"epoch": 0.89,
"learning_rate": 4.9182371575975736e-05,
"loss": 0.2412,
"step": 500
},
{
"epoch": 1.07,
"learning_rate": 4.8726393675266716e-05,
"loss": 0.2385,
"step": 600
},
{
"epoch": 1.07,
"eval_f1_score": 0.6684952978056427,
"eval_label_f1": 0.8322884012539185,
"eval_loss": 0.24104812741279602,
"eval_runtime": 342.1463,
"eval_samples_per_second": 2.923,
"eval_steps_per_second": 0.365,
"eval_wer": 0.10484557628299404,
"step": 600
},
{
"epoch": 1.25,
"learning_rate": 4.817291893365055e-05,
"loss": 0.2325,
"step": 700
},
{
"epoch": 1.43,
"learning_rate": 4.752422169756048e-05,
"loss": 0.2316,
"step": 800
},
{
"epoch": 1.43,
"eval_f1_score": 0.6724477729601892,
"eval_label_f1": 0.8316909735908553,
"eval_loss": 0.23740312457084656,
"eval_runtime": 344.4982,
"eval_samples_per_second": 2.903,
"eval_steps_per_second": 0.363,
"eval_wer": 0.10221379363727842,
"step": 800
},
{
"epoch": 1.61,
"learning_rate": 4.678296760308474e-05,
"loss": 0.2304,
"step": 900
},
{
"epoch": 1.79,
"learning_rate": 4.595220262229601e-05,
"loss": 0.2291,
"step": 1000
},
{
"epoch": 1.79,
"eval_f1_score": 0.6698076168040833,
"eval_label_f1": 0.8292108362779742,
"eval_loss": 0.2348490208387375,
"eval_runtime": 338.1818,
"eval_samples_per_second": 2.957,
"eval_steps_per_second": 0.37,
"eval_wer": 0.09683412028794798,
"step": 1000
},
{
"epoch": 1.97,
"learning_rate": 4.503534054669892e-05,
"loss": 0.2275,
"step": 1100
},
{
"epoch": 2.14,
"learning_rate": 4.4036148959228365e-05,
"loss": 0.2205,
"step": 1200
},
{
"epoch": 2.14,
"eval_f1_score": 0.6744822196170379,
"eval_label_f1": 0.8339194998046112,
"eval_loss": 0.2333754003047943,
"eval_runtime": 340.8077,
"eval_samples_per_second": 2.934,
"eval_steps_per_second": 0.367,
"eval_wer": 0.09636968805635111,
"step": 1200
},
{
"epoch": 2.32,
"learning_rate": 4.2958733752443195e-05,
"loss": 0.2224,
"step": 1300
},
{
"epoch": 2.5,
"learning_rate": 4.180752225653292e-05,
"loss": 0.2211,
"step": 1400
},
{
"epoch": 2.5,
"eval_f1_score": 0.672933803368586,
"eval_label_f1": 0.8327457892675283,
"eval_loss": 0.2319139689207077,
"eval_runtime": 341.1059,
"eval_samples_per_second": 2.932,
"eval_steps_per_second": 0.366,
"eval_wer": 0.09606006656861986,
"step": 1400
},
{
"epoch": 2.68,
"learning_rate": 4.058724504646834e-05,
"loss": 0.2196,
"step": 1500
},
{
"epoch": 2.86,
"learning_rate": 3.9302916503054246e-05,
"loss": 0.2163,
"step": 1600
},
{
"epoch": 2.86,
"eval_f1_score": 0.6731669266770671,
"eval_label_f1": 0.829953198127925,
"eval_loss": 0.2304619699716568,
"eval_runtime": 340.688,
"eval_samples_per_second": 2.935,
"eval_steps_per_second": 0.367,
"eval_wer": 0.09807260623887298,
"step": 1600
},
{
"epoch": 3.04,
"learning_rate": 3.7959814207763135e-05,
"loss": 0.219,
"step": 1700
},
{
"epoch": 3.22,
"learning_rate": 3.656345725602089e-05,
"loss": 0.2108,
"step": 1800
},
{
"epoch": 3.22,
"eval_f1_score": 0.6734375,
"eval_label_f1": 0.8328125000000001,
"eval_loss": 0.22986993193626404,
"eval_runtime": 338.4413,
"eval_samples_per_second": 2.955,
"eval_steps_per_second": 0.369,
"eval_wer": 0.09544082359315736,
"step": 1800
},
{
"epoch": 3.4,
"learning_rate": 3.5119583578059846e-05,
"loss": 0.2132,
"step": 1900
},
{
"epoch": 3.57,
"learning_rate": 3.363412636053269e-05,
"loss": 0.2104,
"step": 2000
},
{
"epoch": 3.57,
"eval_f1_score": 0.6792156862745098,
"eval_label_f1": 0.8368627450980393,
"eval_loss": 0.22974026203155518,
"eval_runtime": 343.1743,
"eval_samples_per_second": 2.914,
"eval_steps_per_second": 0.364,
"eval_wer": 0.09919498413189876,
"step": 2000
},
{
"epoch": 3.75,
"learning_rate": 3.211318966577581e-05,
"loss": 0.2128,
"step": 2100
},
{
"epoch": 3.93,
"learning_rate": 3.056302334890786e-05,
"loss": 0.2124,
"step": 2200
},
{
"epoch": 3.93,
"eval_f1_score": 0.6781925343811395,
"eval_label_f1": 0.8345776031434184,
"eval_loss": 0.2278670072555542,
"eval_runtime": 338.3142,
"eval_samples_per_second": 2.956,
"eval_steps_per_second": 0.369,
"eval_wer": 0.09447325644399722,
"step": 2200
},
{
"epoch": 4.11,
"learning_rate": 2.8989997375834482e-05,
"loss": 0.2077,
"step": 2300
},
{
"epoch": 4.29,
"learning_rate": 2.7400575647692046e-05,
"loss": 0.2027,
"step": 2400
},
{
"epoch": 4.29,
"eval_f1_score": 0.6789638932496076,
"eval_label_f1": 0.8335949764521193,
"eval_loss": 0.22793905436992645,
"eval_runtime": 338.197,
"eval_samples_per_second": 2.957,
"eval_steps_per_second": 0.37,
"eval_wer": 0.09443455375803081,
"step": 2400
},
{
"epoch": 4.47,
"learning_rate": 2.5801289439291388e-05,
"loss": 0.2077,
"step": 2500
},
{
"epoch": 4.65,
"learning_rate": 2.419871056070862e-05,
"loss": 0.2055,
"step": 2600
},
{
"epoch": 4.65,
"eval_f1_score": 0.6832347140039449,
"eval_label_f1": 0.8347140039447732,
"eval_loss": 0.2275087535381317,
"eval_runtime": 337.608,
"eval_samples_per_second": 2.962,
"eval_steps_per_second": 0.37,
"eval_wer": 0.09493768867559409,
"step": 2600
}
],
"logging_steps": 100,
"max_steps": 5000,
"num_input_tokens_seen": 0,
"num_train_epochs": 9,
"save_steps": 200,
"total_flos": 1.0235203668116929e+20,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}