base_model: Qwen/Qwen2.5-Coder-7B-Instruct
model_type: Qwen2ForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false

# Pre-tokenized datasets produced by scripts/preprocess_diff_mask_chat.py
# (from felixwangg/prime_vul_plus_splitted which has a fixed validation split).
# Columns: input_ids, attention_mask, labels, diff_mask.
# Labels are already -100 for non-assistant tokens; axolotl keeps them as-is.
datasets:
  - path: felixwangg/prime_vul_plus_splitted_token_diff_mask_skip_indent_ctx0_chat_v2
    type: pretokenized
    split: train
test_datasets:
  - path: felixwangg/prime_vul_plus_splitted_token_diff_mask_skip_indent_ctx0_chat_v2
    type: pretokenized
    split: validation
dataset_prepared_path: /u901/t577wang/SecSteer-v2/axolotl-datasets/lora/Qwen2.5-Coder-7B/prime_vul_plus_splitted_token_diff_mask_skip_indent_ctx0_chat_v2
val_set_size: 0
output_dir: /u901/t577wang/SecSteer-v2/axolotl-outputs/lora/Qwen2.5-Coder-7B-evol-33k-stage1-sec-stage2-token-diff-ctx0
sequence_len: 4096
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir: /u901/t577wang/SecSteer-v2/axolotl-outputs/lora/Qwen2.5-Coder-7B-evol-33k-stage1/checkpoint-424
lora_r: 16
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
merge_lora: false

wandb_project: diff-mask-evol-33k-primevul-ctx-0
wandb_entity: wtkuan
wandb_watch: "false"
wandb_name: Qwen2.5-Coder-7B-evol-33k-stage1-sec-stage2-token-diff-ctx0
wandb_log_model: "false"

gradient_accumulation_steps: 8
micro_batch_size: 4
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 4e-05

bf16: true
tf32: false

gradient_checkpointing: true
resume_from_checkpoint:
logging_steps: 1
flash_attention: true

num_epochs: 2
warmup_ratio: 0.1
early_stopping_patience: 1000
eval_steps: 15
save_steps: 15
save_total_limit: 1000
load_best_model_at_end: true

weight_decay: 0.02
special_tokens:

# Diff-mask weighted loss: CE(logit_t, label_t) * (1 + alpha * diff_mask_{t+1})
# Security-sensitive tokens (diff_mask=1) get weight (1 + diff_mask_alpha).
# Requires PYTHONPATH to include the repo root so diff_mask_trainer is importable.
diff_mask_alpha: 0.5

plugins:
  - diff_mask_trainer.plugin.DiffMaskPlugin
  # - sec_bench_callback.SecBenchPlugin

u901/t577wang/SecSteer-v2/axolotl-outputs/lora/Qwen2.5-Coder-7B-evol-33k-stage1-sec-stage2-token-diff-ctx0

This model is a fine-tuned version of Qwen/Qwen2.5-Coder-7B-Instruct on the felixwangg/prime_vul_plus_splitted_token_diff_mask_skip_indent_ctx0_chat_v2 dataset. It achieves the following results on the evaluation set:

Loss: 0.7336
Ppl: 2.0827
Memory/max Active (gib): 42.65
Memory/max Allocated (gib): 42.65
Memory/device Reserved (gib): 61.17

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: 4e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 8
total_train_batch_size: 64
total_eval_batch_size: 8
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: 11
training_steps: 114

Training results

Training Loss	Epoch	Step	Validation Loss	Ppl	Active (gib)	Allocated (gib)	Reserved (gib)
No log	0	0	0.8212	2.2732	42.34	42.34	51.09
6.4789	0.2632	15	0.7739	2.1683	42.65	42.65	60.59
5.8481	0.5263	30	0.7503	2.1177	42.65	42.65	61.17
5.8538	0.7895	45	0.7405	2.0970	42.65	42.65	61.17
6.0125	1.0526	60	0.7365	2.0887	42.65	42.65	61.17
6.1222	1.3158	75	0.7347	2.0848	42.65	42.65	61.17
5.9562	1.5789	90	0.7339	2.0831	42.65	42.65	61.17
5.9736	1.8421	105	0.7336	2.0825	42.65	42.65	61.17
6.6086	2.0	114	0.7336	2.0827	42.65	42.65	61.17

Framework versions

PEFT 0.19.1
Transformers 5.5.4
Pytorch 2.11.0+cu130
Datasets 4.5.0
Tokenizers 0.22.2

felixwangg/Qwen2.5-Coder-7B-evol-33k-stage1-sec-stage2-token-diff-ctx0

u901/t577wang/SecSteer-v2/axolotl-outputs/lora/Qwen2.5-Coder-7B-evol-33k-stage1-sec-stage2-token-diff-ctx0

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

felixwangg/
Qwen2.5-Coder-7B-evol-33k-stage1-sec-stage2-token-diff-ctx0