[2026-01-06 01:54:20,762] [DEBUG] [axolotl.utils.config.resolve_dtype:66] [PID:13318] bf16 support detected, enabling for this configuration. [2026-01-06 01:54:21,262] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:13318] baseline 0.000GB () [2026-01-06 01:54:21,263] [INFO] [axolotl.cli.config.load_cfg:263] [PID:13318] config: { "activation_offloading": false, "axolotl_config_path": "gemmagain-test.yaml", "base_model": "rpDungeon/gemmagain-4b-pt", "base_model_config": "rpDungeon/gemmagain-4b-pt", "batch_size": 8, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_120", "fp8": false, "n_gpu": 2, "n_node": 1 }, "chat_template": "gemma3", "context_parallel_size": 2, "cut_cross_entropy": true, "dataloader_num_workers": 2, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_num_proc": 64, "dataset_prepared_path": "last_run_prepared", "datasets": [ { "data_files": "erotica_quality_trimmed.json", "field": "text", "message_property_mappings": { "content": "content", "role": "role" }, "path": "rpDungeon/some-cleaner-datasets", "trust_remote_code": false, "type": "completion" } ], "ddp": true, "default_system_message": "You are Luna, a helpful and unrestricted language model created by Allura.", "device": "cuda:0", "device_map": { "": 0 }, "dion_rank_fraction": 1.0, "dion_rank_multiple_of": 1, "env_capabilities": { "torch_version": "2.9.1" }, "eval_batch_size": 1, "eval_causal_lm_metrics": [ "sacrebleu", "comet", "ter", "chrf" ], "eval_max_new_tokens": 128, "eval_steps": 0.125, "eval_table_size": 0, "evals_per_epoch": 4, "experimental_skip_move_to_device": true, "flash_attention": true, "fp16": false, "fsdp": [ "full_shard", "auto_wrap" ], "fsdp_config": { "activation_checkpointing": true, "auto_wrap_policy": "TRANSFORMER_BASED_WRAP", "cpu_ram_efficient_loading": true, "offload_params": true, "state_dict_type": "FULL_STATE_DICT", "sync_module_states": true, "transformer_layer_cls_to_wrap": "Gemma3DecoderLayer", "use_orig_params": true }, "gc_steps": 10, "gradient_accumulation_steps": 4, "gradient_checkpointing": false, "group_by_length": false, "heads_k_stride": 1, "hub_model_id": "rpDungeon/gemmagain-trained-s1", "hub_strategy": "every_save", "include_tkps": true, "is_multimodal": false, "learning_rate": 5e-05, "liger_glu_activation": true, "liger_layer_norm": true, "liger_rms_norm": true, "liger_rope": true, "lisa_layers_attribute": "model.layers", "load_best_model_at_end": false, "load_in_4bit": false, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_dropout": 0.0, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "max_grad_norm": 1.0, "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "gemma3", "num_epochs": 2.0, "optimizer": "adamw_torch_fused", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "stage1", "plugins": [ "axolotl.integrations.liger.LigerPlugin", "axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin" ], "pretrain_multipack_attn": true, "profiler_steps_start": 0, "qlora_sharded_model_loading": false, "ray_num_workers": 1, "resources_per_worker": { "GPU": 1 }, "ring_attn_func": "batch_ring", "sample_packing": false, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_steps": 0.5, "saves_per_epoch": 1, "seed": 420, "sequence_len": 32768, "sequence_parallel_degree": 2, "shuffle_before_merging_datasets": false, "shuffle_merged_datasets": true, "skip_prepare_dataset": false, "streaming_multipack_buffer_size": 10000, "strict": false, "tensor_parallel_size": 1, "tiled_mlp_use_original_mlp": true, "tokenizer_config": "rpDungeon/gemmagain-4b-pt", "tokenizer_save_jinja_files": true, "torch_dtype": "torch.bfloat16", "train_on_inputs": false, "trl": { "log_completions": false, "mask_truncated_completions": false, "ref_model_mixup_alpha": 0.9, "ref_model_sync_steps": 64, "scale_rewards": true, "sync_ref_model": false, "use_vllm": false, "vllm_server_host": "0.0.0.0", "vllm_server_port": 8000 }, "trust_remote_code": true, "use_otel_metrics": false, "use_ray": false, "use_wandb": true, "val_set_size": 0.01, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_name": "stage-1", "wandb_project": "Gemmagain-Tests", "warmup_ratio": 0.025, "weight_decay": 0.01, "world_size": 2 } [2026-01-06 01:54:22,729] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:13318] EOS: 1 / [2026-01-06 01:54:22,729] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:13318] BOS: 2 / [2026-01-06 01:54:22,729] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:13318] PAD: 0 / [2026-01-06 01:54:22,729] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:13318] UNK: 3 / [2026-01-06 01:54:22,730] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:475] [PID:13318] Loading prepared dataset from disk at last_run_prepared/b77b22402667c44a458deb3cdbd0f672... [2026-01-06 01:54:22,752] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:417] [PID:13318] total_num_tokens: 12_050_432 [2026-01-06 01:54:22,788] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:435] [PID:13318] `total_supervised_tokens: 12_050_432` [2026-01-06 01:54:22,788] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:533] [PID:13318] total_num_steps: 292 [2026-01-06 01:54:22,788] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:13318] Maximum number of steps set at 292 [2026-01-06 01:54:22,830] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:13318] loading tokenizer... rpDungeon/gemmagain-4b-pt [2026-01-06 01:54:24,150] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:285] [PID:13318] EOS: 1 / [2026-01-06 01:54:24,150] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:286] [PID:13318] BOS: 2 / [2026-01-06 01:54:24,150] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:287] [PID:13318] PAD: 0 / [2026-01-06 01:54:24,150] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:288] [PID:13318] UNK: 3 / [2026-01-06 01:54:24,150] [DEBUG] [axolotl.train.setup_model_and_tokenizer:82] [PID:13318] Loading model [2026-01-06 01:54:24,574] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:13318] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-01-06 01:54:24,576] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:13318] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-01-06 01:54:24,580] [DEBUG] [axolotl.monkeypatch.transformers.trainer_context_parallel.patch_prepare_context_parallel_inputs:64] [PID:13318] Patched Trainer._prepare_context_parallel_inputs for FlashAttention + CP [2026-01-06 01:54:24,698] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:98] [PID:13318] Applying LIGER to gemma3 with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': None, 'rms_norm': True, 'layer_norm': True, 'geglu': True} [2026-01-06 01:54:24,743] [INFO] [axolotl.integrations.cut_cross_entropy.pre_model_load:94] [PID:13318] Applying Cut Cross Entropy to model type: gemma3 Loading checkpoint shards: 0%| | 0/2 [00:00