{ "step": 0, "epoch": 0, "total_tokens": 0, "best_val": Infinity, "val_loss": 11.654257669448853, "mtp_val": 11.198893718719482, "fp8_mode": "torchao_rowwise", "seq_len": 4096, "config": { "size_name": "7b-moe", "vocab_size": 50304, "d_model": 1536, "n_layers": 30, "n_heads": 12, "n_kv_heads": 2, "head_dim": 128, "ffn_hidden": 6144, "use_moe": true, "n_dense_layers": 1, "n_routed_experts": 64, "n_shared_experts": 2, "moe_top_k": 6, "expert_hidden": 768, "routed_scale": 1.0, "bias_update_rate": 0.001, "aux_loss_weight": 0.001, "router_z_weight": 0.0001, "global_every": 4, "window": 1024, "rope_local": 10000.0, "rope_global": 500000.0, "use_qk_norm": true, "doc_masking": true, "tie_embeddings": false, "z_loss": 0.0001, "n_ce_chunks": 8, "use_mtp": true, "mtp_weight": 0.3, "resid_dropout": 0.0, "attn_dropout": 0.0, "use_fim": true, "fim_rate": 0.6, "fim_spm_rate": 0.5, "eos_id": 50268, "pad_id": 50272, "fim_pre_id": 50269, "fim_suf_id": 50270, "fim_mid_id": 50271, "fim_pad_id": -1, "param_dtype_str": "bf16", "stochastic_rounding": true, "ckpt_attn": true, "ckpt_ffn": true, "muon_lr": 0.0022, "adam_lr": 0.00045, "muon_momentum": 0.95, "muon_ns_steps": 5, "weight_decay": 0.1, "beta1": 0.9, "beta2": 0.95, "grad_clip": 1.0, "min_lr_ratio": 0.02, "warmup_tokens": 100000000, "decay_frac": 0.2, "target_total_tokens": 10000000000, "max_epochs": 4.0, "seq_len": 4096, "long_ctx_len": 16384, "long_ctx_frac": 0.12, "max_len": 16384, "micro_batch": 16, "grad_accum": 16, "batch_ramp_tokens": 2000000000, "log_every": 10, "eval_every_steps": 500, "eval_batches": 40, "ckpt_every_min": 45.0, "max_session_hours": 11.3, "overfit_patience": 5, "seed": 1337 } }