{ "timestamp": "2026-08-15T13:32:23.333839", "checkpoint": "storage/models/bear_final.pt", "device": "cpu", "stages": { "pretrain": { "summary": { "avg_loss": 4.426291036605835, "avg_ppl": 83.62069494173322, "avg_bpb": 1.861745809924929, "task_count": 5 }, "details": [ { "id": "hf_pretrain_id_wiki_01", "source": "wikimedia/wikipedia (20231101.id)", "loss": 4.715993881225586, "ppl": 111.71979221759592, "bpb": 1.9835979548999556, "generated_sample": "Buddha di mana. Candi Borobudur adalah candi Buddha di mana di mana. C" }, { "id": "hf_pretrain_id_indommlu_01", "source": "indonlp/IndoMMLU", "loss": 6.034489154815674, "ppl": 417.5854336341011, "bpb": 2.5381713055250157, "generated_sample": "Alas yang benar-benar bekahata dan obaklah.\nJawaban: Alas yang" }, { "id": "hf_pretrain_en_mmlu_cs_01", "source": "cais/mmlu (computer_science)", "loss": 5.34050989151001, "ppl": 208.6190562599516, "bpb": 2.2462761330320546, "generated_sample": "To calculate the bandwidth of a packet, the bandwidth is then calculated using the value of the bandwidth of a packet. The" }, { "id": "hf_pretrain_en_arc_01", "source": "allenai/ai2_arc", "loss": 2.8477277755737305, "ppl": 17.248544715052773, "bpb": 1.1977850552833762, "generated_sample": "the normal force.\n\u2022 10.\nElijah, Martin M., Akbar, A. & Mits" }, { "id": "hf_pretrain_en_climbmix_01", "source": "karpathy/climbmix-400b-shuffle", "loss": 3.192734479904175, "ppl": 24.35493454957321, "bpb": 1.3428986008842438, "generated_sample": "probability of 1/2. In other words, the distribution of the sum of the given probability distribution is the probability" } ] } } }