sudormrfbin/qwen2.5-1.5b-lora-sensitive

See axolotl config

axolotl version: 0.13.0.dev0

base_model: Qwen/Qwen2.5-1.5B-Instruct

# Dataset configuration - training only, no validation
datasets:
  - path: /workspace/data/train.jsonl
    ds_type: json
    type: alpaca

val_set_size: 0
output_dir: ./outputs/qwen-sensitive-classifier

# LoRA configuration
adapter: lora
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - v_proj
  - k_proj
  - o_proj
  - gate_proj
  - down_proj
  - up_proj

# Training hyperparameters
sequence_len: 512
sample_packing: true
micro_batch_size: 16
gradient_accumulation_steps: 1
num_epochs: 4
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 0.0002

# Performance settings (optimized for H100)
bf16: auto
tf32: true
flash_attention: true
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false

# Logging - console only, no wandb
wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
logging_steps: 10
saves_per_epoch: 1

# Misc
warmup_ratio: 0.1
weight_decay: 0.0
train_on_inputs: false

outputs/qwen-sensitive-classifier

This model is a fine-tuned version of Qwen/Qwen2.5-1.5B-Instruct on the /workspace/data/train.jsonl dataset.

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 8
• training_steps: 84

Training results

Framework versions

• PEFT 0.17.1
• Transformers 4.57.0
• Pytorch 2.7.1+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1