[2026-01-11 04:24:19,378] [WARNING] [axolotl.utils.trainer.prepare_optim_env:644] [PID:4935] P2P support not detected, setting `NCCL_P2P_DISABLE=1` [2026-01-11 04:24:19,378] [DEBUG] [axolotl.utils.config.resolve_dtype:66] [PID:4935] bf16 support detected, enabling for this configuration. [2026-01-11 04:24:19,702] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:4935] baseline 0.000GB () [2026-01-11 04:24:19,703] [INFO] [axolotl.cli.config.load_cfg:263] [PID:4935] config: { "activation_offloading": false, "adapter": "qlora", "axolotl_config_path": "olmo-stage1.yaml", "base_model": "allenai/Olmo-3.1-32B-Instruct", "base_model_config": "allenai/Olmo-3.1-32B-Instruct", "batch_size": 4, "bf16": true, "capabilities": { "bf16": true, "compute_capability": "sm_86", "fp8": false, "n_gpu": 2, "n_node": 1 }, "context_parallel_size": 2, "cut_cross_entropy": true, "dataloader_num_workers": 2, "dataloader_pin_memory": true, "dataloader_prefetch_factor": 256, "dataset_num_proc": 24, "dataset_prepared_path": "last_run_prepared", "datasets": [ { "message_property_mappings": { "content": "content", "role": "role" }, "path": "../marvin_no_anthologies.json", "trust_remote_code": false, "type": "completion" } ], "ddp": true, "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_sample_packing": true, "eval_steps": 0.2, "eval_table_size": 0, "evals_per_epoch": 5, "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": "Olmo3DecoderLayer", "use_orig_params": false }, "gc_steps": 10, "gradient_accumulation_steps": 2, "gradient_checkpointing": false, "group_by_length": false, "heads_k_stride": 1, "include_tkps": true, "learning_rate": 1e-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": true, "load_in_8bit": false, "local_rank": 0, "logging_steps": 1, "lora_alpha": 8, "lora_dropout": 0.05, "lora_r": 32, "lora_target_linear": true, "loraplus_lr_embedding": 1e-06, "lr_scheduler": "cosine", "max_grad_norm": 1.0, "mean_resizing_embeddings": false, "micro_batch_size": 1, "model_config_type": "olmo3", "num_epochs": 1.0, "optimizer": "adamw_torch_fused", "otel_metrics_host": "localhost", "otel_metrics_port": 8000, "output_dir": "ckpts-olmo-qlora/marvin-full", "pad_to_sequence_len": true, "peft_use_rslora": true, "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": "varlen_llama3", "sample_packing": true, "sample_packing_bin_size": 200, "sample_packing_group_size": 100000, "save_only_model": false, "save_safetensors": true, "save_total_limit": 2, "saves_per_epoch": 1, "seed": 69, "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": "allenai/Olmo-3.1-32B-Instruct", "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 }, "use_otel_metrics": false, "use_ray": false, "use_wandb": true, "val_set_size": 0.025, "vllm": { "device": "auto", "dtype": "auto", "gpu_memory_utilization": 0.9, "host": "0.0.0.0", "port": 8000 }, "wandb_name": "marvin-qlora-full", "wandb_project": "Olmo3", "warmup_ratio": 0.025, "weight_decay": 0.01, "world_size": 2 } [2026-01-11 04:24:20,543] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:4935] EOS: 100257 / <|endoftext|> [2026-01-11 04:24:20,543] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:4935] BOS: 100257 / <|endoftext|> [2026-01-11 04:24:20,543] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:282] [PID:4935] PAD: 100277 / <|pad|> [2026-01-11 04:24:20,543] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:283] [PID:4935] UNK: 100257 / <|endoftext|> [2026-01-11 04:24:38,731] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:475] [PID:4935] Loading prepared dataset from disk at last_run_prepared/8d6cb9376a109abc6ac61340266a9d2f... [2026-01-11 04:24:38,747] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:406] [PID:4935] total_num_tokens: 607_040 [2026-01-11 04:24:38,770] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:424] [PID:4935] `total_supervised_tokens: 607_040` [2026-01-11 04:24:38,851] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:39,552] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:39,778] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.22662973403930664 [2026-01-11 04:24:39,779] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:40,005] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.22648382186889648 [2026-01-11 04:24:40,005] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:40,232] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.22647571563720703 [2026-01-11 04:24:40,232] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:40,458] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.22629952430725098 [2026-01-11 04:24:41,068] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:438] [PID:4935] gather_len_batches: [19, 19] [2026-01-11 04:24:41,131] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:483] [PID:4935] data_loader_len: 4 [2026-01-11 04:24:41,144] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:499] [PID:4935] sample_packing_eff_est across ranks: [0.9750205874443054, 0.9750205874443054] [2026-01-11 04:24:41,144] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:511] [PID:4935] sample_packing_eff_est: None [2026-01-11 04:24:41,144] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:522] [PID:4935] total_num_steps: 8 [2026-01-11 04:24:41,207] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:406] [PID:4935] total_num_tokens: 21_945_632 [2026-01-11 04:24:41,350] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:424] [PID:4935] `total_supervised_tokens: 21_945_632` [2026-01-11 04:24:41,359] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:41,585] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:41,811] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.2260892391204834 [2026-01-11 04:24:41,811] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:42,037] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.22601652145385742 [2026-01-11 04:24:42,037] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:42,263] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.2259969711303711 [2026-01-11 04:24:42,264] [DEBUG] [axolotl.utils.samplers.multipack.pack_parallel:177] [PID:4935] Using single process for pack_parallel, running sequentially. [2026-01-11 04:24:42,490] [DEBUG] [axolotl.utils.samplers.multipack.__len__:462] [PID:4935] generate_batches time: 0.22606921195983887 [2026-01-11 04:24:42,490] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:438] [PID:4935] gather_len_batches: [677, 678] [2026-01-11 04:24:42,491] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:483] [PID:4935] data_loader_len: 169 [2026-01-11 04:24:42,491] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:499] [PID:4935] sample_packing_eff_est across ranks: [0.9877987504005432, 0.9877987504005432] [2026-01-11 04:24:42,492] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:511] [PID:4935] sample_packing_eff_est: 0.99 [2026-01-11 04:24:42,492] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:522] [PID:4935] total_num_steps: 338 [2026-01-11 04:24:42,492] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:4935] Maximum number of steps set at 338 [2026-01-11 04:24:42,505] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:4935] loading tokenizer... allenai/Olmo-3.1-32B-Instruct [2026-01-11 04:24:43,213] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:280] [PID:4935] EOS: 100257 / <|endoftext|> [2026-01-11 04:24:43,213] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:281] [PID:4935] BOS: 100257 / <|endoftext|> [2026-01-11 04:24:43,213] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:282] [PID:4935] PAD: 100277 / <|pad|> [2026-01-11 04:24:43,213] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:283] [PID:4935] UNK: 100257 / <|endoftext|> [2026-01-11 04:24:43,213] [DEBUG] [axolotl.train.setup_model_and_tokenizer:82] [PID:4935] Loading model [2026-01-11 04:24:43,408] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:87] [PID:4935] Patched Trainer.evaluation_loop with nanmean loss calculation [2026-01-11 04:24:43,409] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:138] [PID:4935] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation [2026-01-11 04:24:43,411] [DEBUG] [axolotl.monkeypatch.transformers.trainer_context_parallel.patch_prepare_context_parallel_inputs:64] [PID:4935] Patched Trainer._prepare_context_parallel_inputs for FlashAttention + CP [2026-01-11 04:24:43,413] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:301] [PID:4935] Applying multipack dataloader patch for sample packing... [2026-01-11 04:24:43,508] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:98] [PID:4935] Applying LIGER to olmo3 with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': None, 'rms_norm': True, 'swiglu': True} [2026-01-11 04:24:43,601] [INFO] [axolotl.integrations.cut_cross_entropy.pre_model_load:94] [PID:4935] Applying Cut Cross Entropy to model type: olmo3 Loading checkpoint shards: 0%| | 0/14 [00:00