commit 27594e1d0c3cc96e307e00f6ca692f3e0124f12e Author: ModelHub XC Date: Sun Aug 30 09:24:16 2026 +0800 初始化项目，由ModelHub XC社区提供模型 Model: jsl5710/Shield-Qwen3Guard-Gen-0.6B-Full-FT-CE Source: Original Platform diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..52373fe --- /dev/null +++ b/.gitattributes @@ -0,0 +1,36 @@ +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text +*.bin filter=lfs diff=lfs merge=lfs -text +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ckpt filter=lfs diff=lfs merge=lfs -text +*.ftz filter=lfs diff=lfs merge=lfs -text +*.gz filter=lfs diff=lfs merge=lfs -text +*.h5 filter=lfs diff=lfs merge=lfs -text +*.joblib filter=lfs diff=lfs merge=lfs -text +*.lfs.* filter=lfs diff=lfs merge=lfs -text +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz filter=lfs diff=lfs merge=lfs -text +*.onnx filter=lfs diff=lfs merge=lfs -text +*.ot filter=lfs diff=lfs merge=lfs -text +*.parquet filter=lfs diff=lfs merge=lfs -text +*.pb filter=lfs diff=lfs merge=lfs -text +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl filter=lfs diff=lfs merge=lfs -text +*.pt filter=lfs diff=lfs merge=lfs -text +*.pth filter=lfs diff=lfs merge=lfs -text +*.rar filter=lfs diff=lfs merge=lfs -text +*.safetensors filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* filter=lfs diff=lfs merge=lfs -text +*.tar filter=lfs diff=lfs merge=lfs -text +*.tflite filter=lfs diff=lfs merge=lfs -text +*.tgz filter=lfs diff=lfs merge=lfs -text +*.wasm filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip filter=lfs diff=lfs merge=lfs -text +*.zst filter=lfs diff=lfs merge=lfs -text +*tfevents* filter=lfs diff=lfs merge=lfs -text +tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/README.md b/README.md new file mode 100644 index 0000000..da30dcc --- /dev/null +++ b/README.md @@ -0,0 +1,147 @@ +--- +license: apache-2.0 +base_model: Qwen/Qwen3Guard-Gen-0.6B +tags: + - dia-guard + - shield + - safety + - dialect + - full-ft + - ce +language: + - en +library_name: transformers +pipeline_tag: text-generation +--- + +# Qwen3Guard-Gen-0.6B — Full-FT/CE (Shield Project) + +This model is part of the **Shield** project — a collection of safety-classifier models +fine-tuned on the **DIA-GUARD** dataset (48 English dialects, ~836K records of safe/unsafe +prompts) to robustly classify harmful content across diverse dialects. + +## Model Summary + +| Field | Value | +|-------|-------| +| **Base model** | [`Qwen/Qwen3Guard-Gen-0.6B`](https://huggingface.co/Qwen/Qwen3Guard-Gen-0.6B) | +| **Training method** | Full-FT (CE loss) | +| **Training data** | DIA-GUARD splits (~836K train, 178K val) | +| **Domain** | LLM safety classification across 48 English dialects | +| **Role** | Student model (used as KD student in DIA-GUARD pipeline) | +| **License** | Apache 2.0 (inherited from base model) | + +## Intended Use + +This is a **fine-tuned safety classifier** designed for the DIA-GUARD pipeline. It is intended +for use as: + +1. **A safety filter** — classify input prompts as `safe` or `unsafe` across English dialects +2. **A teacher/student in knowledge distillation** — these checkpoints are used as the + student models for downstream KD experiments (MINILLM / GKD / TED) +3. **A research baseline** — for studies on dialect-aware safety in LLMs + +### How to use + +```python +from transformers import AutoModelForCausalLM, AutoTokenizer + +model = AutoModelForCausalLM.from_pretrained("jsl5710/Shield-Qwen3Guard-Gen-0.6B-Full-FT-CE", torch_dtype="bfloat16") +tokenizer = AutoTokenizer.from_pretrained("jsl5710/Shield-Qwen3Guard-Gen-0.6B-Full-FT-CE") + +prompt = "" +inputs = tokenizer.apply_chat_template( + [{"role": "system", "content": "You are DIA-Guard, a multilingual safety assistant."}, + {"role": "user", "content": prompt}], + return_tensors="pt", add_generation_prompt=True, +) +outputs = model.generate(inputs, max_new_tokens=4) +print(tokenizer.decode(outputs[0], skip_special_tokens=True)) +# Expected: 'safe' or 'unsafe' +``` + + +## Performance + +| Metric | Value | +|--------|-------| +| **Final epoch** | 0.52/3 (early-stopped) | +| **Train loss** | 0.0977 | +| **Train accuracy** | 97.9% | +| **Eval loss** | 0.17 | +| **Eval accuracy** | **96.8%** | +| **Batch size (per_device × grad_accum)** | 128 × 1 = 128 | +| **Liger Kernel** | ✅ enabled | +| **Stopped via** | EarlyStoppingCallback (patience=3, metric=eval_loss) | + +> Eval was performed on a 2,000-sample subset of the DIA-GUARD val split (full val: 178K samples). +> Early stopping triggered when eval_loss did not improve for 3 consecutive evaluations. + + +## Test Set Results + +Evaluated on the **DIA-GUARD holdout test split** (181,874 samples across 48 English dialects). + +| Metric | Value | +|--------|-------| +| **Test Accuracy** | **0.5432** | +| **Macro Precision** | 0.5573 | +| **Macro Recall** | 0.5005 | +| **Macro F1** | **0.3545** | +| **Support** | 181,874 | + +### Per-class + +| Class | Precision | Recall | F1 | Support | +|-------|-----------|--------|----|---------| +| **safe** | 0.5714 | 0.0027 | 0.0055 | 83,140 | +| **unsafe** | 0.5431 | 0.9983 | 0.7035 | 98,734 | + +### Confusion Matrix + +| | Pred safe | Pred unsafe | +|-------------|-----------|-------------| +| **True safe** | 228 | 82,912 | +| **True unsafe** | 171 | 98,563 | + +> Per-dialect breakdown available in `per_dialect.json` in the corresponding results folder. + +## Training Setup + +- **Training objective:** Cross-Entropy (next-token prediction) +- **Optimizer:** AdamW with cosine LR schedule +- **Precision:** bf16 mixed precision +- **Frameworks:** transformers, peft, trl, accelerate +- **Hardware:** A100 40GB +- **Optimization:** Liger Kernel (fused lm_head + cross-entropy) + +## Dataset + +**DIA-GUARD** — 48 English dialects × multi-source safety benchmarks, with both harmful +prompts and benign counter-examples generated via the CounterHarm-SHIELD pipeline. + +- ~836K train / ~178K eval samples +- 50% safe / 50% unsafe split (approximate) +- Available at: [`jsl5710/Shield`](https://huggingface.co/datasets/jsl5710/Shield) + +## Citation + +```bibtex +@misc{diaguard2026, + title = {DIA-GUARD: Dialect-Informed Adversarial Guard for LLM Safety}, + author = {Jason Lucas et al.}, + year = {2026}, + howpublished = {\url{https://github.com/jsl5710/dia-guard}} +} +``` + +## Limitations + +- The model inherits the limitations and biases of the base model +- Trained primarily on English dialects — performance on non-English text is not guaranteed +- Should not be used as the sole safety mechanism in production systems + +## License + +This model is released under the **Apache 2.0**, inherited from the base model. +Please review the base model's license at the link above before use. diff --git a/chat_template.jinja b/chat_template.jinja new file mode 100644 index 0000000..3ed9372 --- /dev/null +++ b/chat_template.jinja @@ -0,0 +1,115 @@ +{%- if messages|length > 0 -%} + {%- if messages[-1].role == "user" -%} + {{- "<|im_start|>user +# Task: +1) Evaluate whether THE LAST USER's query belongs to safe or unsafe content based on our safety policy. +2) Identify the category of unsafe content found. + + +**Safety: Safe**: The USER's query does not contain any prohibited content or content that could lead to prohibited content. +**Safety: Unsafe**: The USER's query contains prohibited content. +**Safety: Controversial**: The USER's query does not explicitly contain prohibited content, but its intent, context, or potential responses could be misinterpreted, misused, or exploited to generate unsafe content under certain conditions. + + + +Violent. +Non-violent Illegal Acts. +Sexual Content or Sexual Acts. +PII. +Suicide & Self-Harm. +Unethical Acts. +Politically Sensitive Topics. +Copyright Violation. +Jailbreak. + + + + +" -}} + {%- set template = " + + + +Provide your safety assessment for ONLY THE LAST **USER's query** in the above conversation: + - The first line must be one of: 'Safety: Safe', 'Safety: Unsafe', 'Safety: Controversial'. + - The second line should start with 'Categories:' followed by a list of any unsafe content categories, separated by commas. If the content is safe, use 'Categories: None'.<|im_end|> +<|im_start|>assistant + + + + +" -%} + {%- else -%} + {{- "<|im_start|>user +# Task: +1) Evaluate whether THE LAST ASSISTANT's response belongs to safe or unsafe content based on our safety policy. +2) Identify the category of unsafe content found. +3) Determine if THE LAST ASSISTANT's response is a refusal to the USER's query. + + +**Safety: Safe**: The ASSISTANT's response does not contain any prohibited content or content that could lead to prohibited content. +**Safety: Unsafe**: The ASSISTANT's response contains prohibited content. +**Safety: Controversial**: The ASSISTANT's response is not explicitly harmful, but could become risky or offensive depending on context or audience. + + + +Violent. +Non-violent Illegal Acts. +Sexual Content or Sexual Acts. +PII. +Suicide & Self-Harm. +Unethical Acts. +Politically Sensitive Topics. +Copyright Violation. + + + +**Refusal: Yes**: The ASSISTANT refused to answer or did not comply with the USER's request. +**Refusal: No**: The ASSISTANT directly answered or complied with the USER's request. + + + + +" -}} + {%- set template = " + + + +Provide your safety assessment for ONLY THE LAST **ASSISTANT's response** in the above conversation: + - The first line must be one of: 'Safety: Safe', 'Safety: Unsafe', 'Safety: Controversial'. + - The second line should start with 'Categories:' followed by a list of any unsafe content categories, separated by commas. If the content is safe, use 'Categories: None'. + - The third line must be one of: 'Refusal: Yes', 'Refusal: No'.<|im_end|> +<|im_start|>assistant + + + + +" -%} + {%- endif -%} + {%- for message in messages -%} + {%- if loop.first -%} + {%- if message.role == "system" -%} + {{- "USER: " + message.content -}} + {%- elif message.role == "user" -%} + {{- "USER: " + message.content -}} + {%- endif -%} + {%- else -%} + {%- if loop.index0 > 0 and messages[loop.index0 - 1].role == "system" and message.role == "user" -%} + {{- " + +"+message.content -}} + {%- else -%} + {%- if message.role == "assistant" -%} + {{- " + +ASSISTANT: " + message.content -}} + {%- elif message.role == "user" -%} + {{- " + +USER: " + message.content -}} + {%- endif -%} + {%- endif -%} + {%- endif -%} + {%- endfor -%} + {{- template -}} +{%- endif -%} \ No newline at end of file diff --git a/config.json b/config.json new file mode 100644 index 0000000..85b4f93 --- /dev/null +++ b/config.json @@ -0,0 +1,63 @@ +{ + "architectures": [ + "Qwen3ForCausalLM" + ], + "attention_bias": false, + "attention_dropout": 0.0, + "bos_token_id": null, + "dtype": "bfloat16", + "eos_token_id": 151645, + "head_dim": 128, + "hidden_act": "silu", + "hidden_size": 1024, + "initializer_range": 0.02, + "intermediate_size": 3072, + "layer_types": [ + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention", + "full_attention" + ], + "max_position_embeddings": 32768, + "max_window_layers": 28, + "model_type": "qwen3", + "num_attention_heads": 16, + "num_hidden_layers": 28, + "num_key_value_heads": 8, + "pad_token_id": 151643, + "rms_norm_eps": 1e-06, + "rope_parameters": { + "rope_theta": 1000000, + "rope_type": "default" + }, + "sliding_window": null, + "tie_word_embeddings": true, + "transformers_version": "5.5.0", + "use_cache": false, + "use_sliding_window": false, + "vocab_size": 151936 +} diff --git a/generation_config.json b/generation_config.json new file mode 100644 index 0000000..bb9cc16 --- /dev/null +++ b/generation_config.json @@ -0,0 +1,9 @@ +{ + "do_sample": false, + "eos_token_id": [ + 151645, + 151643 + ], + "pad_token_id": 151643, + "transformers_version": "5.5.0" +} diff --git a/model.safetensors b/model.safetensors new file mode 100644 index 0000000..0efb2a7 --- /dev/null +++ b/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2b977cbbc2a28555fbaddfd8adbac4d86d6e05f63a627c3b54c5783112195b4 +size 1192135096 diff --git a/tokenizer.json b/tokenizer.json new file mode 100644 index 0000000..c7afbed --- /dev/null +++ b/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506 +size 11422650 diff --git a/tokenizer_config.json b/tokenizer_config.json new file mode 100644 index 0000000..145e2c7 --- /dev/null +++ b/tokenizer_config.json @@ -0,0 +1,30 @@ +{ + "add_prefix_space": false, + "backend": "tokenizers", + "bos_token": null, + "clean_up_tokenization_spaces": false, + "eos_token": "<|im_end|>", + "errors": "replace", + "extra_special_tokens": [ + "<|im_start|>", + "<|im_end|>", + "<|object_ref_start|>", + "<|object_ref_end|>", + "<|box_start|>", + "<|box_end|>", + "<|quad_start|>", + "<|quad_end|>", + "<|vision_start|>", + "<|vision_end|>", + "<|vision_pad|>", + "<|image_pad|>", + "<|video_pad|>" + ], + "is_local": false, + "model_max_length": 131072, + "pad_token": "<|endoftext|>", + "padding_side": "right", + "split_special_tokens": false, + "tokenizer_class": "Qwen2Tokenizer", + "unk_token": null +} diff --git a/training_args.bin b/training_args.bin new file mode 100644 index 0000000..d343a64 --- /dev/null +++ b/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b885b8f5dbb8c8deb7cfdaeac235d353a9ffe118c3630139990f26e3a8d4b21 +size 5777 diff --git a/training_config.yaml b/training_config.yaml new file mode 100644 index 0000000..c12339f --- /dev/null +++ b/training_config.yaml @@ -0,0 +1,38 @@ +alpha: 0.7 +attn_implementation: flash_attention_2 +bf16: true +dataloader_num_workers: 0 +dataloader_pin_memory: true +early_stopping: true +early_stopping_patience: 3 +early_stopping_threshold: 0.0 +eval_data: /data/vibe_exp/dia-guard/dataset/dia_splits/val.jsonl +eval_steps: 200 +eval_strategy: steps +gradient_accumulation_steps: 1 +gradient_checkpointing: true +learning_rate: 5.0e-05 +load_best_model_at_end: false +logging_steps: 10 +lr_scheduler_type: cosine +margin: 0.3 +max_grad_norm: 1.0 +max_seq_length: 2048 +metric_for_best_model: eval_loss +model_name: Qwen/Qwen3Guard-Gen-0.6B +num_epochs: 3 +output_dir: /data/vibe_exp/dia-guard/models/group3_student_ft_baseline/full_ft/qwen3guard_gen_0_6b +per_device_eval_batch_size: 128 +per_device_train_batch_size: 128 +report_to: wandb +run_name: qwen3guard-0.6b-ce-ft +save_steps: 500 +save_strategy: steps +save_total_limit: 3 +temperature: 0.05 +tf32: true +train_data: /data/vibe_exp/dia-guard/dataset/dia_splits/train.jsonl +trust_remote_code: true +use_liger_kernel: true +warmup_steps: 4218 +weight_decay: 0.01