Update README.md

README.md

@@ -6,6 +6,69 @@ base_model:
 pipeline_tag: summarization
 datasets:
 - CATIE-AQ/frenchSUM
+license_name: lfm1.0
 ---
 
-TODO: Write the README once the code for calculating the ROUGE score has finished running
+## Description
+Liquid.AI's [LFM2](https://huggingface.co/LiquidAI/LFM2-1.2B) model finetuned on [frenchSUM]([hf.co/CATIE-AQ/frenchSUM](https://huggingface.co/datasets/CATIE-AQ/frenchSUM)) dataset to summarize French texts in ~32 words.
+
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_id = "CATIE-AQ//LMF2_french_summary"
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto",
+    torch_dtype="bfloat16",
+    trust_remote_code=True,
+#    attn_implementation="flash_attention_2" <- uncomment on compatible GPU
+)
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+
+prompt = """Résume l'article suivant :\n""" + "text_to_summarize"
+tokenizer.padding_side = "left"
+tokenizer.truncation_side = "right"
+
+# Apply the chat template to prepare the input
+input_ids = tokenizer.apply_chat_template(
+    [{"role": "user", "content": prompt}],
+    add_generation_prompt=True,
+    return_tensors="pt",
+    tokenize=True,
+).to(model.device)
+
+# Generate the output from the model
+output = model.generate(
+    input_ids,
+    do_sample=True,
+    temperature=0.3,
+    min_p=0.15,
+    repetition_penalty=1.05,
+    max_new_tokens=2048,
+)
+
+summary_text = tokenizer.decode(
+    output[0][input_ids.shape[1]:], # Slice the output tensor
+    skip_special_tokens=True # Skip special tokens for a cleaner output
+)
+
+print(summary_text)
+```
+
+## Results
+
+#### orange_sum test split
+
+| Model                                | Parameters | ROUGE-1 | ROUGE-2 | ROUGE-L |
+|--------------------------------------|-------------|----------|----------|----------|
+| LMF2_french_summary (this model)     | 1.2B        |     |    |     |
+| [Barthez](https://huggingface.co/moussaKam/barthez) | 165M | 31.44 | 12.77 | 22.23 |
+| [mBarthez](https://huggingface.co/moussaKam/mbarthez) | 458M | 32.67 | 13.73 | 23.18 |
+
+
+## License
+[LMF1.0](https://huggingface.co/LiquidAI/LFM2-1.2B/blob/main/LICENSE)