onereason-0.8b-kuaishou-full-sft/README.md

library_name, license, tags, model-index

library_name	license	tags	model-index
transformers	other	llama-factory full generated_from_trainer	name results sft

sft

This model is a fine-tuned version of the local competition base model on the kuaishou_seed_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.7658

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• total_eval_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.03
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.1814	0.5185	500	1.2199
0.9387	1.0363	1000	1.0034
0.7721	1.5548	1500	0.8802
0.5989	2.0726	2000	0.8023
0.5697	2.5911	2500	0.7709
0.6001	3.0	2895	0.7658

Framework versions

• Transformers 5.6.0
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.22.2