Spaces:
Running
Running
Descarga CSV y gráfica pro (estilo limpio, año corto, anotación)
Browse files
app.py
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
import pandas as pd
|
| 3 |
import numpy as np
|
|
@@ -6,15 +7,39 @@ import matplotlib.pyplot as plt
|
|
| 6 |
import matplotlib.dates as mdates
|
| 7 |
from chronos import ChronosPipeline
|
| 8 |
|
| 9 |
-
#
|
| 10 |
-
|
|
|
|
|
|
|
| 11 |
|
| 12 |
PIPELINE = ChronosPipeline.from_pretrained(
|
| 13 |
MODEL_ID,
|
| 14 |
device_map="auto",
|
| 15 |
-
dtype=torch.float32,
|
| 16 |
)
|
| 17 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
def _prepare_series(df: pd.DataFrame, freq: str | None):
|
| 19 |
"""
|
| 20 |
Espera columnas: date,value
|
|
@@ -24,6 +49,7 @@ def _prepare_series(df: pd.DataFrame, freq: str | None):
|
|
| 24 |
"""
|
| 25 |
if "date" not in df.columns or "value" not in df.columns:
|
| 26 |
raise gr.Error("El CSV debe tener columnas: date,value")
|
|
|
|
| 27 |
df = df.copy()
|
| 28 |
df["date"] = pd.to_datetime(df["date"])
|
| 29 |
df = df.sort_values("date")
|
|
@@ -53,15 +79,54 @@ def _filter_by_sku(df: pd.DataFrame, sku: str | None):
|
|
| 53 |
sdf = df[df["sku"].astype(str) == str(sku)].copy()
|
| 54 |
if sdf.empty:
|
| 55 |
raise gr.Error(f"No hay datos para el SKU: {sku}")
|
| 56 |
-
|
| 57 |
-
sdf = sdf[["date", "value"]]
|
| 58 |
-
return sdf
|
| 59 |
return df[["date", "value"]].copy()
|
| 60 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
def forecast_fn(file, sku: str, horizon: int = 12, freq: str = "MS"):
|
| 62 |
"""
|
| 63 |
-
|
| 64 |
-
ejecuta pronóstico y devuelve tabla + gráfica.
|
| 65 |
"""
|
| 66 |
if file is None:
|
| 67 |
raise gr.Error("Sube un CSV con columnas: (sku,) date, value")
|
|
@@ -93,28 +158,17 @@ def forecast_fn(file, sku: str, horizon: int = 12, freq: str = "MS"):
|
|
| 93 |
"p90": np.round(p90, 4),
|
| 94 |
})
|
| 95 |
|
| 96 |
-
# Gráfica
|
| 97 |
-
fig =
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
ax.fill_between(out["date"], out["p10"], out["p90"], alpha=0.3, label="Banda P10–P90")
|
| 102 |
-
ax.set_title("Pronóstico con Chronos-T5 (P10 / P50 / P90)")
|
| 103 |
-
ax.set_xlabel("Fecha")
|
| 104 |
-
ax.set_ylabel("Valor")
|
| 105 |
-
# Formato de fechas: año corto + mes (yy-mm), p. ej. 24-07
|
| 106 |
-
ax.xaxis.set_major_locator(mdates.AutoDateLocator())
|
| 107 |
-
ax.xaxis.set_major_formatter(mdates.DateFormatter("%y-%m"))
|
| 108 |
-
plt.xticks(rotation=45)
|
| 109 |
-
ax.legend()
|
| 110 |
|
| 111 |
-
|
|
|
|
| 112 |
|
| 113 |
def list_skus(file):
|
| 114 |
-
"""
|
| 115 |
-
Al cargar un archivo, detecta los SKUs disponibles (si la columna existe)
|
| 116 |
-
y llena el dropdown dinámicamente.
|
| 117 |
-
"""
|
| 118 |
if file is None:
|
| 119 |
return gr.update(choices=[], value=None), None
|
| 120 |
df = pd.read_csv(file.name)
|
|
@@ -123,28 +177,40 @@ def list_skus(file):
|
|
| 123 |
if not skus:
|
| 124 |
return gr.update(choices=[], value=None), df.head(10)
|
| 125 |
return gr.update(choices=skus, value=skus[0]), df.head(10)
|
| 126 |
-
# Sin columna 'sku'
|
| 127 |
return gr.update(choices=[], value=None), df.head(10)
|
| 128 |
|
| 129 |
with gr.Blocks(title="Pronóstico por SKU (Chronos-T5)") as demo:
|
| 130 |
-
gr.Markdown(
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
with gr.Row():
|
| 132 |
file = gr.File(label="CSV (sku,date,value o date,value)", file_types=[".csv"])
|
| 133 |
sku_dd = gr.Dropdown(choices=[], value=None, label="SKU (si el CSV tiene columna 'sku')")
|
| 134 |
horizon = gr.Slider(1, 36, value=12, step=1, label="Horizonte (pasos)")
|
| 135 |
freq = gr.Dropdown(choices=["", "D", "W", "MS", "M"], value="MS",
|
| 136 |
label="Frecuencia. ''=inferir, MS=mensual")
|
| 137 |
-
# Vista previa de datos
|
| 138 |
preview = gr.Dataframe(label="Vista previa (primeras filas)")
|
| 139 |
-
|
| 140 |
-
# Cuando cambia el archivo, llenar dropdown de SKUs y mostrar preview
|
| 141 |
file.change(list_skus, inputs=file, outputs=[sku_dd, preview])
|
| 142 |
|
| 143 |
-
btn = gr.Button("Generar pronóstico")
|
| 144 |
out_table = gr.Dataframe(label="Tabla de pronóstico")
|
| 145 |
out_plot = gr.Plot(label="Gráfica")
|
| 146 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
|
| 148 |
if __name__ == "__main__":
|
| 149 |
demo.queue().launch(server_name="0.0.0.0", server_port=7860, show_error=True)
|
| 150 |
-
|
|
|
|
| 1 |
+
import io
|
| 2 |
import gradio as gr
|
| 3 |
import pandas as pd
|
| 4 |
import numpy as np
|
|
|
|
| 7 |
import matplotlib.dates as mdates
|
| 8 |
from chronos import ChronosPipeline
|
| 9 |
|
| 10 |
+
# ----------------------------
|
| 11 |
+
# Modelo (ligero para Space free)
|
| 12 |
+
# ----------------------------
|
| 13 |
+
MODEL_ID = "amazon/chronos-t5-small" # o "amazon/chronos-t5-mini"
|
| 14 |
|
| 15 |
PIPELINE = ChronosPipeline.from_pretrained(
|
| 16 |
MODEL_ID,
|
| 17 |
device_map="auto",
|
| 18 |
+
dtype=torch.float32, # usar dtype (no torch_dtype)
|
| 19 |
)
|
| 20 |
|
| 21 |
+
# ----------------------------
|
| 22 |
+
# Estilo "más pro" para las gráficas
|
| 23 |
+
# ----------------------------
|
| 24 |
+
plt.rcParams.update({
|
| 25 |
+
"figure.figsize": (9, 4.8),
|
| 26 |
+
"figure.facecolor": "#ffffff",
|
| 27 |
+
"axes.facecolor": "#ffffff",
|
| 28 |
+
"axes.grid": True,
|
| 29 |
+
"grid.color": "#e6e6e6",
|
| 30 |
+
"grid.linestyle": "-",
|
| 31 |
+
"grid.linewidth": 0.6,
|
| 32 |
+
"axes.spines.top": False,
|
| 33 |
+
"axes.spines.right": False,
|
| 34 |
+
"axes.spines.left": False,
|
| 35 |
+
"axes.spines.bottom": False,
|
| 36 |
+
"axes.titlesize": 16,
|
| 37 |
+
"axes.titleweight": "semibold",
|
| 38 |
+
"axes.labelsize": 12,
|
| 39 |
+
"legend.frameon": True,
|
| 40 |
+
"legend.framealpha": 0.9,
|
| 41 |
+
})
|
| 42 |
+
|
| 43 |
def _prepare_series(df: pd.DataFrame, freq: str | None):
|
| 44 |
"""
|
| 45 |
Espera columnas: date,value
|
|
|
|
| 49 |
"""
|
| 50 |
if "date" not in df.columns or "value" not in df.columns:
|
| 51 |
raise gr.Error("El CSV debe tener columnas: date,value")
|
| 52 |
+
|
| 53 |
df = df.copy()
|
| 54 |
df["date"] = pd.to_datetime(df["date"])
|
| 55 |
df = df.sort_values("date")
|
|
|
|
| 79 |
sdf = df[df["sku"].astype(str) == str(sku)].copy()
|
| 80 |
if sdf.empty:
|
| 81 |
raise gr.Error(f"No hay datos para el SKU: {sku}")
|
| 82 |
+
return sdf[["date", "value"]]
|
|
|
|
|
|
|
| 83 |
return df[["date", "value"]].copy()
|
| 84 |
|
| 85 |
+
def _nice_plot(df_hist: pd.DataFrame, df_fc: pd.DataFrame) -> plt.Figure:
|
| 86 |
+
"""Grafica con estilo pro + año corto en eje X y anotación del último P50."""
|
| 87 |
+
fig, ax = plt.subplots()
|
| 88 |
+
|
| 89 |
+
# Líneas
|
| 90 |
+
ax.plot(df_hist["date"], df_hist["value"], label="Histórico", linewidth=2.2, color="C0")
|
| 91 |
+
ax.plot(df_fc["date"], df_fc["p50"], label="Pronóstico (P50)", linewidth=2.4, color="C1")
|
| 92 |
+
|
| 93 |
+
# Banda de incertidumbre
|
| 94 |
+
ax.fill_between(df_fc["date"], df_fc["p10"], df_fc["p90"],
|
| 95 |
+
alpha=0.18, label="Banda P10–P90", color="C1", edgecolor="none")
|
| 96 |
+
|
| 97 |
+
# Formato de fechas: año corto (yy-mm)
|
| 98 |
+
ax.xaxis.set_major_locator(mdates.AutoDateLocator())
|
| 99 |
+
ax.xaxis.set_major_formatter(mdates.DateFormatter("%y-%m"))
|
| 100 |
+
plt.setp(ax.get_xticklabels(), rotation=0, ha="center")
|
| 101 |
+
|
| 102 |
+
# Etiquetas y título
|
| 103 |
+
ax.set_title("Pronóstico con Chronos-T5 (P10 / P50 / P90)")
|
| 104 |
+
ax.set_xlabel("Fecha")
|
| 105 |
+
ax.set_ylabel("Valor")
|
| 106 |
+
|
| 107 |
+
# Anotar último P50
|
| 108 |
+
x_last = df_fc["date"].iloc[-1]
|
| 109 |
+
y_last = df_fc["p50"].iloc[-1]
|
| 110 |
+
ax.scatter([x_last], [y_last], color="C1", s=30, zorder=3)
|
| 111 |
+
ax.annotate(f"P50={y_last:.1f}", xy=(x_last, y_last),
|
| 112 |
+
xytext=(10, 10), textcoords="offset points",
|
| 113 |
+
fontsize=10, bbox=dict(boxstyle="round,pad=0.25", fc="#f5f5f5", ec="#cccccc"))
|
| 114 |
+
|
| 115 |
+
# Leyenda fuera de la serie para no tapar
|
| 116 |
+
leg = ax.legend(loc="upper left")
|
| 117 |
+
for lh in leg.legendHandles:
|
| 118 |
+
try:
|
| 119 |
+
lh.set_alpha(1.0)
|
| 120 |
+
except Exception:
|
| 121 |
+
pass
|
| 122 |
+
|
| 123 |
+
fig.tight_layout(pad=1.2)
|
| 124 |
+
return fig
|
| 125 |
+
|
| 126 |
def forecast_fn(file, sku: str, horizon: int = 12, freq: str = "MS"):
|
| 127 |
"""
|
| 128 |
+
Lee CSV, filtra por SKU (si hay), prepara serie,
|
| 129 |
+
ejecuta pronóstico y devuelve tabla + gráfica + archivo descargable.
|
| 130 |
"""
|
| 131 |
if file is None:
|
| 132 |
raise gr.Error("Sube un CSV con columnas: (sku,) date, value")
|
|
|
|
| 158 |
"p90": np.round(p90, 4),
|
| 159 |
})
|
| 160 |
|
| 161 |
+
# Gráfica pulida
|
| 162 |
+
fig = _nice_plot(df, out)
|
| 163 |
+
|
| 164 |
+
# --- Descarga: generar CSV en memoria y devolver bytes ---
|
| 165 |
+
csv_bytes = out.to_csv(index=False).encode("utf-8")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 166 |
|
| 167 |
+
# Retornamos: tabla, gráfica, y contenido para DownloadButton
|
| 168 |
+
return out, fig, csv_bytes
|
| 169 |
|
| 170 |
def list_skus(file):
|
| 171 |
+
"""Detecta SKUs (si existe la columna) y llena el dropdown dinámicamente."""
|
|
|
|
|
|
|
|
|
|
| 172 |
if file is None:
|
| 173 |
return gr.update(choices=[], value=None), None
|
| 174 |
df = pd.read_csv(file.name)
|
|
|
|
| 177 |
if not skus:
|
| 178 |
return gr.update(choices=[], value=None), df.head(10)
|
| 179 |
return gr.update(choices=skus, value=skus[0]), df.head(10)
|
|
|
|
| 180 |
return gr.update(choices=[], value=None), df.head(10)
|
| 181 |
|
| 182 |
with gr.Blocks(title="Pronóstico por SKU (Chronos-T5)") as demo:
|
| 183 |
+
gr.Markdown(
|
| 184 |
+
"## Pronóstico de Demanda por **SKU**\n"
|
| 185 |
+
"CSV con columnas: **sku (opcional)**, **date**, **value**. "
|
| 186 |
+
"Selecciona el SKU y genera el forecast."
|
| 187 |
+
)
|
| 188 |
with gr.Row():
|
| 189 |
file = gr.File(label="CSV (sku,date,value o date,value)", file_types=[".csv"])
|
| 190 |
sku_dd = gr.Dropdown(choices=[], value=None, label="SKU (si el CSV tiene columna 'sku')")
|
| 191 |
horizon = gr.Slider(1, 36, value=12, step=1, label="Horizonte (pasos)")
|
| 192 |
freq = gr.Dropdown(choices=["", "D", "W", "MS", "M"], value="MS",
|
| 193 |
label="Frecuencia. ''=inferir, MS=mensual")
|
|
|
|
| 194 |
preview = gr.Dataframe(label="Vista previa (primeras filas)")
|
|
|
|
|
|
|
| 195 |
file.change(list_skus, inputs=file, outputs=[sku_dd, preview])
|
| 196 |
|
| 197 |
+
btn = gr.Button("Generar pronóstico", variant="primary")
|
| 198 |
out_table = gr.Dataframe(label="Tabla de pronóstico")
|
| 199 |
out_plot = gr.Plot(label="Gráfica")
|
| 200 |
+
|
| 201 |
+
# Botón de descarga (entrega bytes -> archivo "forecast.csv")
|
| 202 |
+
download_btn = gr.DownloadButton(
|
| 203 |
+
label="⬇️ Descargar pronóstico (CSV)",
|
| 204 |
+
value=None, # se asigna en tiempo de ejecución
|
| 205 |
+
file_name="forecast.csv" # nombre sugerido
|
| 206 |
+
)
|
| 207 |
+
|
| 208 |
+
btn.click(
|
| 209 |
+
forecast_fn,
|
| 210 |
+
inputs=[file, sku_dd, horizon, freq],
|
| 211 |
+
outputs=[out_table, out_plot, download_btn],
|
| 212 |
+
api_name="/forecast"
|
| 213 |
+
)
|
| 214 |
|
| 215 |
if __name__ == "__main__":
|
| 216 |
demo.queue().launch(server_name="0.0.0.0", server_port=7860, show_error=True)
|
|
|