Model: laion/Qwen3-8B_exp_tas_temp_0.25_traces_save-strategy_steps Source: Original Platform
16 lines
547 B
JSON
16 lines
547 B
JSON
{
|
|
"achieved_tflops_per_gpu": 0.0009559419924246866,
|
|
"achieved_tflops_per_gpu_theoretical": 220.21379658350594,
|
|
"epoch": 8.0,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.002475286368280649,
|
|
"mfu_percent": 0.0003063916642386816,
|
|
"mfu_percent_theoretical": 70.58134505881601,
|
|
"total_flos": 1139460138729472.0,
|
|
"train_loss": 0.08529083973322209,
|
|
"train_runtime": 37249.2574,
|
|
"train_samples_per_second": 2.073,
|
|
"train_steps_per_second": 0.065,
|
|
"valid_targets_mean": 2886.6,
|
|
"valid_targets_min": 328
|
|
} |