Asap7772/Qwen3-4B-second-stage-DPO-lr-1e-6-beta-0.05-loss-sigmoid-rpo-1.0-ckpt-135

This model is a fine-tuned version of Asap7772/qwen3-4b-impabs-warmstart-sft-1e-5-1epoch-0828 using TRL (Transformer Reinforcement Learning).

Model Details

• Base Model: Asap7772/qwen3-4b-impabs-warmstart-sft-1e-5-1epoch-0828
• Checkpoint: checkpoint-135
• Fine-tuning Method: DPO (Direct Preference Optimization)
• Framework: TRL

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Asap7772/Qwen3-4B-second-stage-DPO-lr-1e-6-beta-0.05-loss-sigmoid-rpo-1.0-ckpt-135")
tokenizer = AutoTokenizer.from_pretrained("Asap7772/Qwen3-4B-second-stage-DPO-lr-1e-6-beta-0.05-loss-sigmoid-rpo-1.0-ckpt-135")

# Your inference code here

Training Details

This model was trained using the TRL library with DPO (Direct Preference Optimization).

Training Arguments

{
  "output_dir": "/iris/u/asap7772/trl/checkpoints/Qwen3-4B-second-stage-DPO-lr-1e-6-beta-0.05-loss-sigmoid-rpo-1.0",
  "overwrite_output_dir": false,
  "do_train": false,
  "do_eval": true,
  "do_predict": false,
  "eval_strategy": "epoch",
  "prediction_loss_only": false,
  "per_device_train_batch_size": 1,
  "per_device_eval_batch_size": 8,
  "per_gpu_train_batch_size": null,
  "per_gpu_eval_batch_size": null,
  "gradient_accumulation_steps": 16,
  "eval_accumulation_steps": null,
  "eval_delay": 0,
  "torch_empty_cache_steps": null,
  "learning_rate": 1e-06,
  "weight_decay": 0.01,
  "adam_beta1": 0.9,
  "adam_beta2": 0.999,
  "adam_epsilon": 1e-08,
  "max_grad_norm": 1.0,
  "num_train_epochs": 5,
  "max_steps": -1,
  "lr_scheduler_type": "linear",
  "lr_scheduler_kwargs": {},
  "warmup_ratio": 0.05,
  "warmup_steps": 0,
  "log_level": "passive",
  "log_level_replica": "warning",
  "log_on_each_node": true,
  "logging_dir": "/iris/u/asap7772/trl/checkpoints/Qwen3-4B-second-stage-DPO-lr-1e-6-beta-0.05-loss-sigmoid-rpo-1.0/runs/Sep15_02-56-05_iris-hgx-2.stanford.edu",
  "logging_strategy": "steps",
  "logging_first_step": true,
  "logging_steps": 1,
  "logging_nan_inf_filter": true,
  "save_strategy": "epoch",
  "save_steps": 500,
  "save_total_limit": null,
  "save_safetensors": true,
  "save_on_each_node": false,
  "save_only_model": false,
  "restore_callback_states_from_checkpoint": false,
  "no_cuda": false,
  "use_cpu": false,
  "use_mps_device": false,
  "seed": 42,
  "data_seed": null,
  "jit_mode_eval": false,
  "use_ipex": false,
  "bf16": true,
  "fp16": false,
  "fp16_opt_level": "O1",
  "half_precision_backend": "auto",
  "bf16_full_eval": false,
  "fp16_full_eval": false,
  "tf32": null,
  "local_rank": 0,
  "ddp_backend": null,
  "tpu_num_cores": null,
  "tpu_metrics_debug": false,
  "debug": [],
  "dataloader_drop_last": true,
  "eval_steps": 1,
  "dataloader_num_workers": 0,
  "dataloader_prefetch_factor": null,
  "past_index": -1,
  "run_name": "Qwen3-4B-second-stage-DPO-lr-1e-6-beta-0.05-loss-sigmoid-rpo-1.0",
  "disable_tqdm": false,
  "remove_unused_columns": false,
  "label_names": null,
  "load_best_model_at_end": false,
  "metric_for_best_model": null,
  "greater_is_better": null,
  "ignore_data_skip": false,
  "fsdp": [],
  "fsdp_min_num_params": 0,
  "fsdp_config": {
    "min_num_params": 0,
    "xla": false,
    "xla_fsdp_v2": false,
    "xla_fsdp_grad_ckpt": false
  },
  "fsdp_transformer_layer_cls_to_wrap": null,
  "accelerator_config": "AcceleratorConfig(split_batches=False, dispatch_batches=None, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False)",
  "parallelism_config": null,
  "deepspeed": null,
  "label_smoothing_factor": 0.0,
  "optim": "adamw_torch",
  "optim_args": null,
  "adafactor": false,
  "group_by_length": false,
  "length_column_name": "length",
  "report_to": [
    "wandb"
  ],
  "ddp_find_unused_parameters": null,
  "ddp_bucket_cap_mb": null,
  "ddp_broadcast_buffers": null,
  "dataloader_pin_memory": false,
  "dataloader_persistent_workers": false,
  "skip_memory_metrics": true,
  "use_legacy_prediction_loop": false,
  "push_to_hub": false,
  "resume_from_checkpoint": null,
  "hub_model_id": null,
  "hub_strategy": "every_save",
  "hub_token": null,
  "hub_private_repo": null,
  "hub_always_push": false,
  "hub_revision": null,
  "gradient_checkpointing": true,
  "gradient_checkpointing_kwargs": null,
  "include_inputs_for_metrics": false,
  "include_for_metrics": [],
  "eval_do_concat_batches": true,
  "fp16_backend": "auto",
  "push_to_hub_model_id": null,
  "push_to_hub_organization": null,
  "push_to_hub_token": null,
  "mp_parameters": "",
  "auto_find_batch_size": false,
  "full_determinism": false,
  "torchdynamo": null,
  "ray_scope": "last",
  "ddp_timeout": 1800,
  "torch_compile": false,
  "torch_compile_backend": null,
  "torch_compile_mode": null,
  "include_tokens_per_second": false,
  "include_num_input_tokens_seen": false,
  "neftune_noise_alpha": null,
  "optim_target_modules": null,
  "batch_eval_metrics": false,
  "eval_on_start": false,
  "use_liger_kernel": false,
  "liger_kernel_config": null,
  "eval_use_gather_object": false,
  "average_tokens_across_devices": true,
  "model_init_kwargs": null,
  "ref_model_init_kwargs": null,
  "model_adapter_name": null,
  "ref_adapter_name": null,
  "force_use_ref_model": false,
  "disable_dropout": true,
  "use_logits_to_keep": false,
  "dataset_num_proc": null,
  "padding_value": null,
  "label_pad_token_id": -100,
  "max_prompt_length": 8192,
  "max_completion_length": 1024,
  "max_length": 9216,
  "truncation_mode": "keep_end",
  "padding_free": false,
  "precompute_ref_log_probs": false,
  "precompute_ref_batch_size": null,
  "tools": null,
  "loss_type": "sigmoid",
  "use_liger_loss": false,
  "base_model_attribute_name": "model",
  "beta": 0.05,
  "f_divergence_type": "FDivergenceType.REVERSE_KL",
  "f_alpha_divergence_coef": 1.0,
  "reference_free": false,
  "label_smoothing": 0.0,
  "use_weighting": false,
  "rpo_alpha": 1.0,
  "ld_alpha": null,
  "discopop_tau": 0.05,
  "loss_weights": null,
  "sync_ref_model": false,
  "ref_model_mixup_alpha": 0.6,
  "ref_model_sync_steps": 512,
  "generate_during_eval": false,
  "distributed_state": "Distributed environment: DEEPSPEED  Backend: nccl\nNum processes: 8\nProcess index: 0\nLocal process index: 0\nDevice: cuda:0\n",
  "_n_gpu": 1,
  "__cached__setup_devices": "cuda:0",
  "deepspeed_plugin": "DeepSpeedPlugin(hf_ds_config=<transformers.integrations.deepspeed.HfTrainerDeepSpeedConfig object at 0x7fefd5bdc9d0>, gradient_accumulation_steps='auto', gradient_clipping='auto', zero_stage=3, is_train_batch_min=True, offload_optimizer_device='cpu', offload_param_device='cpu', offload_optimizer_nvme_path='none', offload_param_nvme_path='none', zero3_init_flag=True, zero3_save_16bit_model=True, transformer_moe_cls_names=None, enable_msamp=False, msamp_opt_level='O1')"
}