Update app.py
Browse files
app.py
CHANGED
|
@@ -2,11 +2,10 @@ import gradio as gr
|
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
import os
|
| 4 |
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
FILENAME = "prompt-refiner-mistral7b-Q4_K_M.gguf"
|
| 10 |
|
| 11 |
SYSTEM_PROMPT = (
|
| 12 |
'You are an expert Prompt Engineer. '
|
|
@@ -28,65 +27,35 @@ SYSTEM_PROMPT = (
|
|
| 28 |
'Always respond in the same language as the input.'
|
| 29 |
)
|
| 30 |
|
| 31 |
-
# ============================================================
|
| 32 |
-
# CHARGEMENT DU MODELE (une seule fois au démarrage)
|
| 33 |
-
# ============================================================
|
| 34 |
-
print("Chargement du modèle...")
|
| 35 |
-
model_path = hf_hub_download(
|
| 36 |
-
repo_id = REPO_ID,
|
| 37 |
-
filename = FILENAME,
|
| 38 |
-
token = os.environ.get("HF_TOKEN"),
|
| 39 |
-
)
|
| 40 |
-
|
| 41 |
-
llm = Llama(
|
| 42 |
-
model_path = model_path,
|
| 43 |
-
n_ctx = 2048,
|
| 44 |
-
n_threads = 4,
|
| 45 |
-
n_gpu_layers = 0,
|
| 46 |
-
verbose = False,
|
| 47 |
-
)
|
| 48 |
-
print("Modèle chargé ✅")
|
| 49 |
-
|
| 50 |
-
# ============================================================
|
| 51 |
-
# FONCTION PRINCIPALE
|
| 52 |
-
# ============================================================
|
| 53 |
def refine_prompt(user_input: str) -> str:
|
| 54 |
if not user_input.strip():
|
| 55 |
return "❌ Veuillez entrer un prompt."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
|
| 57 |
-
prompt = f"<s>[INST] {SYSTEM_PROMPT}\n\nUser input to process: {user_input} [/INST]"
|
| 58 |
-
|
| 59 |
-
response = llm(
|
| 60 |
-
prompt,
|
| 61 |
-
max_tokens = 512,
|
| 62 |
-
temperature = 0.3,
|
| 63 |
-
top_p = 0.9,
|
| 64 |
-
repeat_penalty = 1.1,
|
| 65 |
-
stop = ["</s>", "[INST]"],
|
| 66 |
-
)
|
| 67 |
-
return response["choices"][0]["text"].strip()
|
| 68 |
-
|
| 69 |
-
# ============================================================
|
| 70 |
-
# INTERFACE GRADIO
|
| 71 |
-
# ============================================================
|
| 72 |
with gr.Blocks(title="Prompt Refiner") as demo:
|
| 73 |
gr.Markdown("# 🚀 Prompt Refiner\nEntre ton prompt et il sera structuré automatiquement.")
|
| 74 |
-
|
| 75 |
with gr.Row():
|
| 76 |
input_box = gr.Textbox(label="Ton prompt", placeholder="Ex: Crée un script YouTube sur l'IA...", lines=3)
|
| 77 |
output_box = gr.Textbox(label="Prompt raffiné", lines=10)
|
| 78 |
-
|
| 79 |
btn = gr.Button("Raffiner ✨", variant="primary")
|
| 80 |
btn.click(fn=refine_prompt, inputs=input_box, outputs=output_box)
|
| 81 |
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
["Create a YouTube script about AI for beginners"],
|
| 85 |
-
["Rédige un article de blog sur le marketing digital"],
|
| 86 |
-
["bonjour je mappelle wassim"],
|
| 87 |
-
["qsdqsdqsdqs"],
|
| 88 |
-
],
|
| 89 |
-
inputs=input_box,
|
| 90 |
-
)
|
| 91 |
|
| 92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
import os
|
| 4 |
|
| 5 |
+
HF_TOKEN = os.environ.get("HF_TOKEN")
|
| 6 |
+
MODEL_ID = "Wessym/prompt-refiner-mistral7b"
|
| 7 |
+
|
| 8 |
+
client = InferenceClient(model=MODEL_ID, token=HF_TOKEN)
|
|
|
|
| 9 |
|
| 10 |
SYSTEM_PROMPT = (
|
| 11 |
'You are an expert Prompt Engineer. '
|
|
|
|
| 27 |
'Always respond in the same language as the input.'
|
| 28 |
)
|
| 29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
def refine_prompt(user_input: str) -> str:
|
| 31 |
if not user_input.strip():
|
| 32 |
return "❌ Veuillez entrer un prompt."
|
| 33 |
+
try:
|
| 34 |
+
messages = [
|
| 35 |
+
{"role": "system", "content": SYSTEM_PROMPT},
|
| 36 |
+
{"role": "user", "content": f"User input to process: {user_input}"}
|
| 37 |
+
]
|
| 38 |
+
response = client.chat_completion(
|
| 39 |
+
messages = messages,
|
| 40 |
+
max_tokens = 512,
|
| 41 |
+
temperature = 0.3,
|
| 42 |
+
)
|
| 43 |
+
return response.choices[0].message.content.strip()
|
| 44 |
+
except Exception as e:
|
| 45 |
+
return f"❌ Erreur: {str(e)}"
|
| 46 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
with gr.Blocks(title="Prompt Refiner") as demo:
|
| 48 |
gr.Markdown("# 🚀 Prompt Refiner\nEntre ton prompt et il sera structuré automatiquement.")
|
|
|
|
| 49 |
with gr.Row():
|
| 50 |
input_box = gr.Textbox(label="Ton prompt", placeholder="Ex: Crée un script YouTube sur l'IA...", lines=3)
|
| 51 |
output_box = gr.Textbox(label="Prompt raffiné", lines=10)
|
|
|
|
| 52 |
btn = gr.Button("Raffiner ✨", variant="primary")
|
| 53 |
btn.click(fn=refine_prompt, inputs=input_box, outputs=output_box)
|
| 54 |
|
| 55 |
+
demo.launch()
|
| 56 |
+
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
+
Et le `requirements.txt` doit contenir seulement :
|
| 59 |
+
```
|
| 60 |
+
gradio
|
| 61 |
+
huggingface_hub
|