llama3-8b-full-pretrain-junk-tweet-1m-en-reproduce-bs8/all_results.json at main - llama3-8b-full-pretrain-junk-tweet-1m-en-reproduce-bs8 - Gitea: Git with a cup of tea

shuoxing/llama3-8b-full-pretrain-junk-tweet-1m-en-reproduce-bs8

Files

ModelHub XC 48a518be36 初始化项目，由ModelHub XC社区提供模型

Model: shuoxing/llama3-8b-full-pretrain-junk-tweet-1m-en-reproduce-bs8
Source: Original Platform

2026-05-28 05:08:17 +08:00

8 lines

200 B

JSON

Raw Permalink Blame History

 {
     "epoch": 3.0,
     "total_flos": 5863048888320.0,
     "train_loss": 2.8491841291374067,
     "train_runtime": 55958.28,
     "train_samples_per_second": 3.702,
     "train_steps_per_second": 0.463
 }