Modelpublic

felixwangg/Qwen2.5-Coder-7B-sft-plus-alpha-2-line-diff-ctx5-v2

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face

0likes14downloads

Model Card

<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/> <details><summary>See axolotl config</summary>

axolotl version: 0.13.2

yaml

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_line_diff_mask_skip_indent_ctx5_chat_v2
    type: pretokenized
    split: train
test_datasets:
  - path: felixwangg/prime_vul_plus_splitted_line_diff_mask_skip_indent_ctx5_chat_v2
    type: pretokenized
    split: validation
dataset_prepared_path: /u901/t577wang/SecSteer/axolotl-datasets/lora/Qwen2.5-Coder-7B/prime_vul_plus_splitted_line_diff_mask_skip_indent_ctx5_chat_v2_alpha2
val_set_size: 0
output_dir: /u901/t577wang/SecSteer/axolotl-outputs/lora/Qwen2.5-Coder-7B-sft-plus-alpha-2-line-diff-ctx5-v2
sequence_len: 4096
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
merge_lora: true

wandb_project: diff-mask-sft-primevul-ctx-5
wandb_entity: wtkuan
wandb_watch: "false"
wandb_name: Qwen2.5-Coder-7B-sft-plus-alpha-2-line-diff-ctx5-v2
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: 1
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: 2.0

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

</details><br>

u901/t577wang/SecSteer/axolotl-outputs/lora/Qwen2.5-Coder-7B-sft-plus-alpha-2-line-diff-ctx5-v2

This model is a fine-tuned version of Qwen/Qwen2.5-Coder-7B-Instruct on the felixwangg/primevulplussplittedlinediffmaskskipindentctx5chat_v2 dataset. It achieves the following results on the evaluation set:

—Loss: 0.7680
—Ppl: 2.1555
—Memory/max Active (gib): 42.7
—Memory/max Allocated (gib): 42.7
—Memory/device Reserved (gib): 60.61

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
—trainbatchsize: 4
—evalbatchsize: 4
—seed: 42
—distributed_type: multi-GPU
—num_devices: 2
—gradientaccumulationsteps: 8
—totaltrainbatch_size: 64
—totalevalbatch_size: 8
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 5
—training_steps: 57

Training results

Training Loss	Epoch	Step	Validation Loss	Ppl	Active (gib)	Allocated (gib)	Reserved (gib)
No log	0	0	0.8686	2.3836	42.36	42.36	51.15
6.8507	0.2632	15	0.8235	2.2784	42.7	42.7	60.61
6.0853	0.5263	30	0.7768	2.1745	42.7	42.7	60.61
5.7471	0.7895	45	0.7680	2.1555	42.7	42.7	60.61

Framework versions

—PEFT 0.18.1
—Transformers 4.57.6
—Pytorch 2.10.0+cu128
—Datasets 4.5.0
—Tokenizers 0.22.2