| { |
| "step": 179, |
| "epoch": 4, |
| "config": { |
| "model": { |
| "pretrained_name": "/llm_reco/dehua/model/Qwen2.5-VL-7B-Instruct", |
| "num_labels": 92 |
| }, |
| "loss": { |
| "type": "label_smoothing", |
| "smoothing": 0.1, |
| "temperature": 1.0 |
| }, |
| "datasets": { |
| "dataset_configs": { |
| "fru92": { |
| "num_classes": 92, |
| "description": "fru92 dataset", |
| "eval_ratio": 1 |
| } |
| }, |
| "enable_logits_masking": false, |
| "shuffle_datasets": false |
| }, |
| "data": { |
| "train_jsonl": "/llm_reco/dehua/code/qwen_classify/data/dataset_food_label/fru92_train.jsonl", |
| "val_jsonl": "/llm_reco/dehua/code/qwen_classify/data/dataset_food_label/fru92_test.jsonl" |
| }, |
| "training": { |
| "epochs": 5, |
| "lr": "5e-6", |
| "weight_decay": 0.01, |
| "warmup_steps": 0.1, |
| "output_dir": "/mmu_mllm_hdd_2/madehua/model/qwen_classify/fru92_ls_5e_6", |
| "logging_steps": 10000, |
| "save_steps": 50, |
| "eval_steps": 50, |
| "save_hf_format": true, |
| "save_deepspeed_format": false, |
| "num_workers": 32, |
| "use_fused_adam": true, |
| "lr_scheduler": { |
| "type": "cosine", |
| "final_lr_ratio": 0.1, |
| "num_cycles": 0.5 |
| }, |
| "best_model_tracking": { |
| "enabled": true, |
| "metric": "overall_accuracy", |
| "mode": "max", |
| "save_best_only": true |
| }, |
| "evaluation": { |
| "partial_eval_during_training": true, |
| "full_eval_at_end": true, |
| "eval_best_model_only": true |
| }, |
| "num_epochs": 5, |
| "learning_rate": "5e-6" |
| }, |
| "deepspeed": "configs/ds_s2.json", |
| "wandb": { |
| "enabled": true, |
| "project": "qwen_classification", |
| "run_name": "fru92_cosine_5e_6_ls", |
| "tags": [ |
| "qwen", |
| "deepspeed", |
| "classification", |
| "cosine_hold", |
| "single_dataset" |
| ], |
| "notes": "Qwen2.5-VL Food101 classification with cosine+hold scheduler", |
| "log_dataset_metrics": false, |
| "log_overall_metrics": true |
| }, |
| "accelerate": { |
| "bf16": true |
| }, |
| "output_dir": "/mmu_mllm_hdd_2/madehua/model/qwen_classify/fru92_ls_5e_6", |
| "logging_steps": 10000, |
| "save_steps": 50, |
| "eval_steps": 50, |
| "save_hf_format": true, |
| "save_deepspeed_format": false |
| }, |
| "dataset_metrics": {}, |
| "is_best_model": true, |
| "best_metric_value": 0.9421112982405848, |
| "timestamp": 1753206968.9056048 |
| } |
