Upload folder using huggingface_hub

README.md

@@ -3,208 +3,251 @@ base_model: unsloth/Qwen3-1.7B-unsloth-bnb-4bit
 library_name: peft
 pipeline_tag: text-generation
 tags:
-- base_model:adapter:unsloth/Qwen3-1.7B-unsloth-bnb-4bit
+- medical
+- bilingual
+- french
+- english
 - dpo
 - lora
-- transformers
+- peft
 - trl
 - unsloth
+- qwen3
+- base_model:adapter:unsloth/Qwen3-1.7B-unsloth-bnb-4bit
+language:
+- fr
+- en
+datasets:
+- Maphe/medical-sft-5k
+- Maphe/medical-dpo-5k
 ---
 
-# Model Card for Model ID
-
-<!-- Provide a quick summary of what the model is/does. -->
-
-
-
-## Model Details
-
-### Model Description
-
-<!-- Provide a longer summary of what this model is. -->
-
-
-
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-
-### Model Sources [optional]
-
-<!-- Provide the basic links for the model. -->
-
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+# Qwen3 1.7B Medical Finetuned
 
-## Uses
+This repository contains a bilingual French/English medical LoRA adapter built on top of `unsloth/Qwen3-1.7B-unsloth-bnb-4bit`.
 
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+The training workflow used:
 
-### Direct Use
+1. supervised fine-tuning (SFT) on a curated medical instruction dataset;
+2. preference alignment with DPO on medical chosen/rejected pairs.
 
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+The adapter is intended for experimentation, evaluation, and educational use around medical-domain instruction tuning. It is not a medical device and must not be used as a substitute for a qualified health professional.
 
-[More Information Needed]
+## Model Details
 
-### Downstream Use [optional]
+- Base model: `unsloth/Qwen3-1.7B-unsloth-bnb-4bit`
+- Adapter type: PEFT LoRA
+- Task: causal language modeling / chat-style instruction following
+- Languages: French and English
+- Final artifact in this folder: DPO-aligned LoRA adapter
+- Upstream SFT dataset: `Maphe/medical-sft-5k`
+- Upstream DPO dataset: `Maphe/medical-dpo-5k`
 
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+### Training setup
 
-[More Information Needed]
+The project uses Unsloth, TRL, PEFT, and bitsandbytes with 4-bit loading.
 
-### Out-of-Scope Use
+LoRA configuration:
 
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+- `r = 16`
+- `lora_alpha = 16`
+- `lora_dropout = 0`
+- `bias = none`
+- Target modules: `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
 
-[More Information Needed]
+SFT configuration:
 
-## Bias, Risks, and Limitations
+- Epochs: `2`
+- Per-device batch size: `32`
+- Gradient accumulation: `16`
+- Learning rate: `2e-4`
+- Scheduler: `cosine`
+- Max sequence length: `1024`
+- Optimizer: `adamw_8bit`
+- Seed: `42`
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+DPO configuration:
 
-[More Information Needed]
+- Epochs: `1`
+- Per-device batch size: `4`
+- Gradient accumulation: `8`
+- Learning rate: `5e-5`
+- Beta: `0.1`
+- Scheduler: `cosine`
+- Max sequence length: `1024`
+- Optimizer: `adamw_8bit`
+- Seed: `42`
 
-### Recommendations
+## Training Data
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+Two project datasets were prepared and used in the workflow:
 
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+- `Maphe/medical-sft-5k` for supervised fine-tuning
+- `Maphe/medical-dpo-5k` for preference optimization
 
-## How to Get Started with the Model
+The SFT dataset aggregates bilingual medical QA and MCQ-style examples derived from these Hugging Face sources:
 
-Use the code below to get started with the model.
+- `ANR-MALADES/MediQAl`
+- `nthngdy/frenchmedmcqa`
+- `keivalya/MedQuad-MedicalQnADataset`
 
-[More Information Needed]
+The DPO dataset is built primarily from:
 
-## Training Details
+- `TsinghuaC3I/UltraMedical-Preference`
 
-### Training Data
+Project-side preprocessing includes:
 
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+- schema normalization across heterogeneous sources;
+- prompt/response formatting for chat training;
+- deduplication on textual pairs;
+- source quota sampling;
+- deterministic train/validation/test splitting for SFT;
+- heuristic PII anonymization with Presidio and regex-based detectors.
 
-[More Information Needed]
+The resulting model is optimized for:
 
-### Training Procedure
+- French and English medical questions;
+- short factual answers;
+- multiple-choice style medical questions;
+- structured, direct responses.
 
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+## Prompting Format
 
-#### Preprocessing [optional]
+The training prompt uses a fixed system instruction:
 
-[More Information Needed]
+`Tu es un assistant medical expert. Reponds de maniere claire, factuelle et structuree. Si la question est en anglais, reponds en anglais.`
 
+During training, assistant outputs were formatted in direct-answer mode with an empty Qwen thinking block. This adapter therefore works best with standard chat prompting and concise medical questions.
 
-#### Training Hyperparameters
+## Intended Uses
 
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+Appropriate uses:
 
-#### Speeds, Sizes, Times [optional]
+- research prototypes in domain adaptation;
+- comparison between base and finetuned medical assistants;
+- educational work on SFT + DPO pipelines;
+- internal experimentation on bilingual medical QA.
 
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+Out-of-scope uses:
 
-[More Information Needed]
+- diagnosis or treatment decisions without clinician oversight;
+- emergency triage;
+- autonomous clinical decision support;
+- legal, regulatory, or production-grade medical advice systems;
+- any workflow requiring guaranteed factuality or safety.
 
 ## Evaluation
 
-<!-- This section describes the evaluation protocols and provides the results. -->
-
-### Testing Data, Factors & Metrics
-
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
-
-#### Factors
-
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-
-[More Information Needed]
-
-#### Metrics
-
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-
-[More Information Needed]
-
-### Results
-
-[More Information Needed]
-
-#### Summary
+The repository contains a comparative evaluation between the base model and the SFT checkpoint on `500` examples.
 
+Important: the metrics below are for the SFT checkpoint, not for this final DPO adapter. At the time of writing, no dedicated post-DPO benchmark has been added to the repository.
 
+Available evaluation artifacts:
 
-## Model Examination [optional]
+- `notebooks/eval_results/qwen3_base_vs_sft_output_summary.json`
+- `notebooks/eval_results/qwen3_base_vs_sft_output.jsonl`
+- `notebooks/eval_results/qwen3_base_vs_sft_output.csv`
 
-<!-- Relevant interpretability work for the model goes here -->
+Summary of SFT-vs-base results:
 
-[More Information Needed]
+- Mean METEOR on free-text answers: `0.1361 -> 0.1653` (`+0.0292`)
+- MCQ first-letter score: `0.0515 -> 0.4378` (`+0.3863`)
+- MCQ correct answers: `12 -> 102`
 
-## Environmental Impact
+Interpretation:
 
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+- the finetuning substantially improved MCQ behavior in this project benchmark;
+- gains on open-ended generation were positive but more modest;
+- automatic metrics remain insufficient to validate clinical quality.
 
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+## Biases, Risks, and Limitations
 
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
+This model inherits limitations from both the base model and the medical datasets used during fine-tuning.
 
-## Technical Specifications [optional]
+Known risks:
 
-### Model Architecture and Objective
+- hallucinated or overconfident medical statements;
+- incomplete coverage of diseases, populations, and care settings;
+- source-data bias toward specific question styles;
+- imperfect anonymization in upstream preparation;
+- limited evaluation depth;
+- possible mismatch between benchmark gains and real clinical usefulness.
 
-[More Information Needed]
+This adapter should be used only with strong human review and explicit user-facing warnings.
 
-### Compute Infrastructure
+## How to Use
 
-[More Information Needed]
+Example with PEFT and Transformers:
 
-#### Hardware
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
 
-[More Information Needed]
+base_model_id = "unsloth/Qwen3-1.7B-unsloth-bnb-4bit"
+adapter_path = "Maphe/qwen3-1.7b-medical-finetuned"
 
-#### Software
+tokenizer = AutoTokenizer.from_pretrained(base_model_id)
+base_model = AutoModelForCausalLM.from_pretrained(base_model_id)
+model = PeftModel.from_pretrained(base_model, adapter_path)
 
-[More Information Needed]
+messages = [
+    {
+        "role": "system",
+        "content": (
+            "Tu es un assistant medical expert. "
+            "Reponds de maniere claire, factuelle et structuree. "
+            "Si la question est en anglais, reponds en anglais."
+        ),
+    },
+    {"role": "user", "content": "Quels sont les symptomes principaux du diabete de type 2 ?"},
+]
 
-## Citation [optional]
+prompt = tokenizer.apply_chat_template(
+    messages,
+    tokenize=False,
+    add_generation_prompt=True,
+)
+inputs = tokenizer(prompt, return_tensors="pt")
+outputs = model.generate(**inputs, max_new_tokens=256, do_sample=False)
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
 
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+If you use Unsloth in the same way as in the project notebook, load the base model first and then the LoRA adapter exported in this repository.
 
-**BibTeX:**
+## Repository Context
 
-[More Information Needed]
+This model card is derived from the accompanying project materials:
 
-**APA:**
+- root project documentation in `README.md`
+- training notebook: `notebooks/colab_qwen3_unsloth_finetune.ipynb`
+- evaluation notebook: `notebooks/colab_qwen3_unsloth_eval_compare.ipynb`
 
-[More Information Needed]
+The local training artifacts produced by the project include:
 
-## Glossary [optional]
+- SFT adapter: `notebooks/qwen3-medical-lora/`
+- DPO adapter: `notebooks/qwen3-medical-dpo-lora/`
+- SFT checkpoints: `notebooks/sft_output/checkpoint-*`
+- DPO checkpoint: `notebooks/dpo_output/checkpoint-157`
 
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+## License
 
-[More Information Needed]
+No final consolidated license statement has been added yet in the project for the combined derivative artifact. Before public release, verify:
 
-## More Information [optional]
+- the license of the base model;
+- the license terms of each source dataset;
+- whether redistribution of this adapter is compatible with those upstream terms.
 
-[More Information Needed]
+## Contact
 
-## Model Card Authors [optional]
+Project owner / publisher: `Maphe`
 
-[More Information Needed]
+If you publish this model publicly, it is worth adding:
 
-## Model Card Contact
+- the source repository URL;
+- exact dataset revisions;
+- a dedicated post-DPO evaluation section;
+- explicit medical safety disclaimers in the serving application.
 
-[More Information Needed]
 ### Framework versions
 
-- PEFT 0.19.1
+- PEFT 0.19.1