Upload 2 files
Browse files- UI.py +20 -15
- interface.py +192 -197
UI.py
CHANGED
@@ -3,7 +3,7 @@ import gradio as gr
|
|
3 |
import numpy as np # Importar numpy para np.inf
|
4 |
|
5 |
def create_interface(process_function_for_button):
|
6 |
-
with gr.Blocks(theme='gradio/soft') as demo:
|
7 |
gr.Markdown("# Modelado de Bioprocesos con Ecuaciones Personalizadas y Análisis por IA")
|
8 |
|
9 |
with gr.Row():
|
@@ -16,11 +16,10 @@ def create_interface(process_function_for_button):
|
|
16 |
legend_position_ui = gr.Dropdown(
|
17 |
label="Posición de la leyenda",
|
18 |
choices=['best', 'upper right', 'upper left', 'lower right', 'lower left', 'center left', 'center right', 'lower center', 'upper center', 'center'],
|
19 |
-
value='best'
|
20 |
)
|
21 |
with gr.Column(scale=1):
|
22 |
gr.Markdown("### Conteo de Ecuaciones a Probar")
|
23 |
-
# Asegurar que los valores son numéricos y dentro del rango
|
24 |
biomass_eq_count_ui = gr.Number(label="Biomasa (1-3)", value=1, minimum=1, maximum=3, step=1, precision=0)
|
25 |
substrate_eq_count_ui = gr.Number(label="Sustrato (1-3)", value=1, minimum=1, maximum=3, step=1, precision=0)
|
26 |
product_eq_count_ui = gr.Number(label="Producto (1-3)", value=1, minimum=1, maximum=3, step=1, precision=0)
|
@@ -30,10 +29,9 @@ def create_interface(process_function_for_button):
|
|
30 |
with gr.Row():
|
31 |
with gr.Column(): # Columna 1 siempre visible
|
32 |
biomass_eq1_ui = gr.Textbox(label="Ecuación de Biomasa 1", value="Xm * (1 - exp(-um * (t - t_lag)))", lines=2, placeholder="Ej: Xm * (1 - exp(-um * (t - t_lag)))")
|
33 |
-
biomass_param1_ui = gr.Textbox(label="Parámetros Biomasa 1", value="Xm, um, t_lag", info="Nombres, coma sep. 't' para tiempo. 'X_val' para X(t) en S/P.")
|
34 |
-
biomass_bound1_ui = gr.Textbox(label="Límites Biomasa 1", value="(0, np.inf), (0, np.inf), (0, np.inf)", info="Formato: (low,high). Use np.inf
|
35 |
|
36 |
-
# Definir Columnas 2 y 3 fuera del `with` si se manipula su visibilidad programáticamente
|
37 |
biomass_col2_container = gr.Column(visible=False)
|
38 |
with biomass_col2_container:
|
39 |
biomass_eq2_ui = gr.Textbox(label="Ecuación de Biomasa 2", value="X0 * exp(um * t)", lines=2)
|
@@ -42,7 +40,7 @@ def create_interface(process_function_for_button):
|
|
42 |
|
43 |
biomass_col3_container = gr.Column(visible=False)
|
44 |
with biomass_col3_container:
|
45 |
-
biomass_eq3_ui = gr.Textbox(label="Ecuación de Biomasa 3", lines=2, value="")
|
46 |
biomass_param3_ui = gr.Textbox(label="Parámetros Biomasa 3", value="")
|
47 |
biomass_bound3_ui = gr.Textbox(label="Límites Biomasa 3", value="")
|
48 |
|
@@ -89,9 +87,10 @@ def create_interface(process_function_for_button):
|
|
89 |
|
90 |
# Lógica para mostrar/ocultar campos de ecuación dinámicamente
|
91 |
def update_eq_visibility(count_value):
|
92 |
-
#
|
93 |
-
count = int(count_value)
|
94 |
-
|
|
|
95 |
|
96 |
biomass_eq_count_ui.change(fn=update_eq_visibility, inputs=biomass_eq_count_ui, outputs=[biomass_col2_container, biomass_col3_container])
|
97 |
substrate_eq_count_ui.change(fn=update_eq_visibility, inputs=substrate_eq_count_ui, outputs=[substrate_col2_container, substrate_col3_container])
|
@@ -101,10 +100,11 @@ def create_interface(process_function_for_button):
|
|
101 |
|
102 |
gr.Markdown("## Resultados del Análisis")
|
103 |
with gr.Row():
|
104 |
-
image_output = gr.Image(label="Gráfico Generado", type="pil", scale=2, show_download_button=True, height=600)
|
105 |
with gr.Column(scale=3):
|
106 |
-
analysis_output = gr.Markdown(label="Análisis del Modelo por IA")
|
107 |
|
|
|
108 |
all_inputs_for_button = [
|
109 |
file_input,
|
110 |
biomass_eq1_ui, biomass_eq2_ui, biomass_eq3_ui,
|
@@ -132,12 +132,17 @@ def create_interface(process_function_for_button):
|
|
132 |
)
|
133 |
|
134 |
# Inicializar visibilidad usando demo.load para que se aplique al cargar la UI
|
135 |
-
|
136 |
-
|
137 |
-
|
|
|
|
|
|
|
|
|
138 |
b_vis2_upd, b_vis3_upd = update_eq_visibility(b_c_int)
|
139 |
s_vis2_upd, s_vis3_upd = update_eq_visibility(s_c_int)
|
140 |
p_vis2_upd, p_vis3_upd = update_eq_visibility(p_c_int)
|
|
|
141 |
return b_vis2_upd, b_vis3_upd, s_vis2_upd, s_vis3_upd, p_vis2_upd, p_vis3_upd
|
142 |
|
143 |
demo.load(
|
|
|
3 |
import numpy as np # Importar numpy para np.inf
|
4 |
|
5 |
def create_interface(process_function_for_button):
|
6 |
+
with gr.Blocks(theme='gradio/soft') as demo:
|
7 |
gr.Markdown("# Modelado de Bioprocesos con Ecuaciones Personalizadas y Análisis por IA")
|
8 |
|
9 |
with gr.Row():
|
|
|
16 |
legend_position_ui = gr.Dropdown(
|
17 |
label="Posición de la leyenda",
|
18 |
choices=['best', 'upper right', 'upper left', 'lower right', 'lower left', 'center left', 'center right', 'lower center', 'upper center', 'center'],
|
19 |
+
value='best'
|
20 |
)
|
21 |
with gr.Column(scale=1):
|
22 |
gr.Markdown("### Conteo de Ecuaciones a Probar")
|
|
|
23 |
biomass_eq_count_ui = gr.Number(label="Biomasa (1-3)", value=1, minimum=1, maximum=3, step=1, precision=0)
|
24 |
substrate_eq_count_ui = gr.Number(label="Sustrato (1-3)", value=1, minimum=1, maximum=3, step=1, precision=0)
|
25 |
product_eq_count_ui = gr.Number(label="Producto (1-3)", value=1, minimum=1, maximum=3, step=1, precision=0)
|
|
|
29 |
with gr.Row():
|
30 |
with gr.Column(): # Columna 1 siempre visible
|
31 |
biomass_eq1_ui = gr.Textbox(label="Ecuación de Biomasa 1", value="Xm * (1 - exp(-um * (t - t_lag)))", lines=2, placeholder="Ej: Xm * (1 - exp(-um * (t - t_lag)))")
|
32 |
+
biomass_param1_ui = gr.Textbox(label="Parámetros Biomasa 1", value="Xm, um, t_lag", info="Nombres, coma sep. Use 't' para tiempo. 'X_val' para X(t) en S/P.")
|
33 |
+
biomass_bound1_ui = gr.Textbox(label="Límites Biomasa 1", value="(0, np.inf), (0, np.inf), (0, np.inf)", info="Formato: (low,high). Use np.inf.")
|
34 |
|
|
|
35 |
biomass_col2_container = gr.Column(visible=False)
|
36 |
with biomass_col2_container:
|
37 |
biomass_eq2_ui = gr.Textbox(label="Ecuación de Biomasa 2", value="X0 * exp(um * t)", lines=2)
|
|
|
40 |
|
41 |
biomass_col3_container = gr.Column(visible=False)
|
42 |
with biomass_col3_container:
|
43 |
+
biomass_eq3_ui = gr.Textbox(label="Ecuación de Biomasa 3", lines=2, value="")
|
44 |
biomass_param3_ui = gr.Textbox(label="Parámetros Biomasa 3", value="")
|
45 |
biomass_bound3_ui = gr.Textbox(label="Límites Biomasa 3", value="")
|
46 |
|
|
|
87 |
|
88 |
# Lógica para mostrar/ocultar campos de ecuación dinámicamente
|
89 |
def update_eq_visibility(count_value):
|
90 |
+
# Asegurar que el valor es entero antes de la comparación
|
91 |
+
count = int(count_value)
|
92 |
+
# Retorna diccionarios de actualización para `gr.update`
|
93 |
+
return { "visible": count >= 2 }, { "visible": count >= 3 }
|
94 |
|
95 |
biomass_eq_count_ui.change(fn=update_eq_visibility, inputs=biomass_eq_count_ui, outputs=[biomass_col2_container, biomass_col3_container])
|
96 |
substrate_eq_count_ui.change(fn=update_eq_visibility, inputs=substrate_eq_count_ui, outputs=[substrate_col2_container, substrate_col3_container])
|
|
|
100 |
|
101 |
gr.Markdown("## Resultados del Análisis")
|
102 |
with gr.Row():
|
103 |
+
image_output = gr.Image(label="Gráfico Generado", type="pil", scale=2, show_download_button=True, height=600)
|
104 |
with gr.Column(scale=3):
|
105 |
+
analysis_output = gr.Markdown(label="Análisis del Modelo por IA")
|
106 |
|
107 |
+
# Lista de todos los inputs para el botón de submit
|
108 |
all_inputs_for_button = [
|
109 |
file_input,
|
110 |
biomass_eq1_ui, biomass_eq2_ui, biomass_eq3_ui,
|
|
|
132 |
)
|
133 |
|
134 |
# Inicializar visibilidad usando demo.load para que se aplique al cargar la UI
|
135 |
+
# Esto asegura que el estado inicial de visibilidad es correcto
|
136 |
+
def set_initial_visibility_on_load_wrapper(b_c_val, s_c_val, p_c_val):
|
137 |
+
# Obtener los valores iniciales de los gr.Number components
|
138 |
+
# y aplicar la lógica de visibilidad.
|
139 |
+
# Los valores de los Number inputs pueden ser float, convertirlos a int
|
140 |
+
b_c_int, s_c_int, p_c_int = int(b_c_val), int(s_c_val), int(p_c_val)
|
141 |
+
|
142 |
b_vis2_upd, b_vis3_upd = update_eq_visibility(b_c_int)
|
143 |
s_vis2_upd, s_vis3_upd = update_eq_visibility(s_c_int)
|
144 |
p_vis2_upd, p_vis3_upd = update_eq_visibility(p_c_int)
|
145 |
+
|
146 |
return b_vis2_upd, b_vis3_upd, s_vis2_upd, s_vis3_upd, p_vis2_upd, p_vis3_upd
|
147 |
|
148 |
demo.load(
|
interface.py
CHANGED
@@ -2,44 +2,33 @@
|
|
2 |
import numpy as np
|
3 |
import pandas as pd
|
4 |
import matplotlib
|
5 |
-
matplotlib.use('Agg')
|
6 |
import matplotlib.pyplot as plt
|
7 |
from PIL import Image
|
8 |
import io
|
9 |
import json
|
10 |
-
import traceback
|
11 |
|
12 |
-
|
13 |
-
from
|
14 |
-
# from decorators import gpu_decorator # El decorador @gpu es de HF Spaces, Modal lo maneja diferente
|
15 |
|
16 |
-
# Variables globales
|
17 |
USE_MODAL_FOR_LLM_ANALYSIS = False
|
18 |
-
generate_analysis_from_modal = None
|
19 |
|
20 |
def parse_bounds_str(bounds_str_input, num_params):
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
# Manejar el caso de cadena vacía o solo espacios en blanco
|
26 |
-
if not bounds_str.strip():
|
27 |
-
print(f"Cadena de límites vacía para {num_params} params. Usando límites (-inf, inf).")
|
28 |
return [-np.inf] * num_params, [np.inf] * num_params
|
29 |
|
30 |
try:
|
31 |
-
|
32 |
-
#
|
33 |
-
bounds_str = bounds_str.lower().replace('inf', 'np.inf')
|
34 |
-
|
35 |
-
# Evaluar la cadena para convertirla en una lista de tuplas o listas
|
36 |
-
# Ejemplo de entrada esperada: " (0, np.inf), (0,10), (np.nan, np.nan) "
|
37 |
-
# Asegurar que esté encerrado en corchetes para que eval produzca una lista
|
38 |
-
if not bounds_str.startswith('['):
|
39 |
bounds_str = f"[{bounds_str}]"
|
40 |
|
41 |
-
parsed_bounds_list = eval(bounds_str)
|
42 |
-
|
43 |
if not isinstance(parsed_bounds_list, list):
|
44 |
raise ValueError("La cadena de límites no evaluó a una lista.")
|
45 |
|
@@ -51,9 +40,11 @@ def parse_bounds_str(bounds_str_input, num_params):
|
|
51 |
for item in parsed_bounds_list:
|
52 |
if not (isinstance(item, (tuple, list)) and len(item) == 2):
|
53 |
raise ValueError(f"Cada límite debe ser una tupla/lista de dos elementos (low, high). Se encontró: {item}")
|
54 |
-
|
|
|
55 |
low = -np.inf if (item[0] is None or (isinstance(item[0], float) and np.isnan(item[0]))) else float(item[0])
|
56 |
high = np.inf if (item[1] is None or (isinstance(item[1], float) and np.isnan(item[1]))) else float(item[1])
|
|
|
57 |
lower_bounds.append(low)
|
58 |
upper_bounds.append(high)
|
59 |
|
@@ -64,30 +55,30 @@ def parse_bounds_str(bounds_str_input, num_params):
|
|
64 |
|
65 |
|
66 |
def call_llm_analysis_service(prompt: str) -> str:
|
67 |
-
"""Llama al servicio LLM (ya sea localmente o a través de Modal)."""
|
68 |
if USE_MODAL_FOR_LLM_ANALYSIS and generate_analysis_from_modal:
|
69 |
print("interface.py: Usando la función de análisis LLM de Modal...")
|
70 |
try:
|
71 |
-
# La función wrapper en modal_app.py obtiene MODEL_PATH y MAX_LENGTH de config.py
|
72 |
return generate_analysis_from_modal(prompt)
|
73 |
except Exception as e_modal_call:
|
74 |
print(f"Error llamando a la función Modal LLM: {e_modal_call}")
|
|
|
75 |
return f"Error al contactar el servicio de análisis IA (Modal): {e_modal_call}"
|
76 |
else:
|
77 |
-
# --- Implementación de Fallback (o si no se usa Modal) ---
|
78 |
print("interface.py: Usando la función de análisis LLM local (fallback)...")
|
79 |
try:
|
80 |
-
# Esta parte necesitaría que cargues el modelo localmente
|
81 |
-
# como lo hacías en tu versión original de interface.py
|
82 |
from config import MODEL_PATH, MAX_LENGTH, DEVICE # Importar configuración local
|
83 |
from transformers import AutoTokenizer, AutoModelForCausalLM # Importaciones locales
|
84 |
|
85 |
print(f"Fallback: Cargando modelo {MODEL_PATH} localmente en {DEVICE}...")
|
86 |
tokenizer_local = AutoTokenizer.from_pretrained(MODEL_PATH)
|
87 |
-
model_local = AutoModelForCausalLM.from_pretrained(MODEL_PATH).to(DEVICE)
|
88 |
|
89 |
-
|
90 |
-
|
|
|
|
|
|
|
|
|
91 |
outputs = model_local.generate(
|
92 |
**inputs,
|
93 |
max_new_tokens=MAX_LENGTH,
|
@@ -100,12 +91,12 @@ def call_llm_analysis_service(prompt: str) -> str:
|
|
100 |
return analysis.strip()
|
101 |
except Exception as e_local_llm:
|
102 |
print(f"Error en el fallback LLM local: {e_local_llm}")
|
|
|
103 |
return f"Análisis (fallback local): Error al cargar/ejecutar modelo LLM local: {e_local_llm}."
|
104 |
|
105 |
-
|
106 |
def process_and_plot(
|
107 |
-
file_obj,
|
108 |
-
# Entradas de la UI (desempaquetadas)
|
109 |
biomass_eq1_ui, biomass_eq2_ui, biomass_eq3_ui,
|
110 |
biomass_param1_ui, biomass_param2_ui, biomass_param3_ui,
|
111 |
biomass_bound1_ui, biomass_bound2_ui, biomass_bound3_ui,
|
@@ -122,165 +113,163 @@ def process_and_plot(
|
|
122 |
substrate_eq_count_ui,
|
123 |
product_eq_count_ui
|
124 |
):
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
for i in range(len(eq_list)):
|
183 |
-
eq_str = eq_list[i]
|
184 |
-
param_s = param_str_list[i]
|
185 |
-
bound_s = bound_str_list[i]
|
186 |
-
|
187 |
-
if not eq_str or not param_s:
|
188 |
-
print(f"Ecuación o parámetros vacíos para {model_type} #{i+1}, saltando.")
|
189 |
continue
|
190 |
-
|
191 |
-
print(f"Procesando {model_type} #{i+1}: Eq='{eq_str}', Params='{param_s}'")
|
192 |
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
y_pred, popt = model_handler.fit_model(model_type, time_data, exp_data, bounds=(l_b, u_b), biomass_params_fitted=current_biomass_params)
|
202 |
|
203 |
-
#
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
'
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
ax
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
param_display_texts = []
|
255 |
for idx, res_detail in enumerate(plot_results_list):
|
256 |
-
|
257 |
-
|
258 |
-
|
|
|
|
|
|
|
|
|
259 |
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
-
|
266 |
-
|
267 |
-
|
268 |
-
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
|
277 |
-
|
278 |
-
|
279 |
-
|
280 |
-
|
281 |
-
|
282 |
-
|
283 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
284 |
1. **Resumen General:** Una breve descripción del experimento y qué se intentó modelar.
|
285 |
2. **Análisis por Componente (Biomasa, Sustrato, Producto):**
|
286 |
a. Para cada ecuación probada:
|
@@ -294,13 +283,19 @@ def process_and_plot(
|
|
294 |
4. **Sugerencias y Próximos Pasos:**
|
295 |
a. ¿Cómo se podría mejorar el modelado (ej. probar otras ecuaciones, transformar datos, revisar calidad de datos experimentales)?
|
296 |
b. ¿Qué experimentos adicionales podrían realizarse para validar o refinar los modelos?
|
297 |
-
5. **Conclusión Final:** Un veredicto general sobre el éxito del modelado y la utilidad de los resultados obtenidos.
|
298 |
|
299 |
Utiliza un lenguaje claro y accesible, pero manteniendo el rigor técnico. El análisis debe ser útil para alguien que busca entender la cinética de su bioproceso."""
|
300 |
-
|
301 |
-
|
302 |
-
|
303 |
-
|
304 |
-
|
|
|
305 |
|
306 |
-
|
|
|
|
|
|
|
|
|
|
|
|
2 |
import numpy as np
|
3 |
import pandas as pd
|
4 |
import matplotlib
|
5 |
+
matplotlib.use('Agg')
|
6 |
import matplotlib.pyplot as plt
|
7 |
from PIL import Image
|
8 |
import io
|
9 |
import json
|
10 |
+
import traceback
|
11 |
|
12 |
+
from models import BioprocessModel # Asegúrate que esto apunta a tu models.py
|
13 |
+
# from decorators import gpu_decorator # Mantener comentado si usas Modal
|
|
|
14 |
|
15 |
+
# Variables globales inyectadas por modal_app.py o app.py
|
16 |
USE_MODAL_FOR_LLM_ANALYSIS = False
|
17 |
+
generate_analysis_from_modal = None
|
18 |
|
19 |
def parse_bounds_str(bounds_str_input, num_params):
|
20 |
+
bounds_str = str(bounds_str_input).strip()
|
21 |
+
if not bounds_str:
|
22 |
+
print(f"Cadena de límites vacía para {num_params} params. Usando (-inf, inf).")
|
|
|
|
|
|
|
|
|
23 |
return [-np.inf] * num_params, [np.inf] * num_params
|
24 |
|
25 |
try:
|
26 |
+
bounds_str = bounds_str.lower().replace('inf', 'np.inf').replace('none', 'None')
|
27 |
+
if not (bounds_str.startswith('[') and bounds_str.endswith(']')): # Asegurar que sea una lista
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
bounds_str = f"[{bounds_str}]"
|
29 |
|
30 |
+
parsed_bounds_list = eval(bounds_str, {'np': np, 'inf': np.inf, 'None': None}) # Evaluar con np
|
31 |
+
|
32 |
if not isinstance(parsed_bounds_list, list):
|
33 |
raise ValueError("La cadena de límites no evaluó a una lista.")
|
34 |
|
|
|
40 |
for item in parsed_bounds_list:
|
41 |
if not (isinstance(item, (tuple, list)) and len(item) == 2):
|
42 |
raise ValueError(f"Cada límite debe ser una tupla/lista de dos elementos (low, high). Se encontró: {item}")
|
43 |
+
|
44 |
+
# Convertir a float y manejar None/np.nan
|
45 |
low = -np.inf if (item[0] is None or (isinstance(item[0], float) and np.isnan(item[0]))) else float(item[0])
|
46 |
high = np.inf if (item[1] is None or (isinstance(item[1], float) and np.isnan(item[1]))) else float(item[1])
|
47 |
+
|
48 |
lower_bounds.append(low)
|
49 |
upper_bounds.append(high)
|
50 |
|
|
|
55 |
|
56 |
|
57 |
def call_llm_analysis_service(prompt: str) -> str:
|
|
|
58 |
if USE_MODAL_FOR_LLM_ANALYSIS and generate_analysis_from_modal:
|
59 |
print("interface.py: Usando la función de análisis LLM de Modal...")
|
60 |
try:
|
|
|
61 |
return generate_analysis_from_modal(prompt)
|
62 |
except Exception as e_modal_call:
|
63 |
print(f"Error llamando a la función Modal LLM: {e_modal_call}")
|
64 |
+
traceback.print_exc() # Imprimir el traceback de la llamada a Modal
|
65 |
return f"Error al contactar el servicio de análisis IA (Modal): {e_modal_call}"
|
66 |
else:
|
|
|
67 |
print("interface.py: Usando la función de análisis LLM local (fallback)...")
|
68 |
try:
|
|
|
|
|
69 |
from config import MODEL_PATH, MAX_LENGTH, DEVICE # Importar configuración local
|
70 |
from transformers import AutoTokenizer, AutoModelForCausalLM # Importaciones locales
|
71 |
|
72 |
print(f"Fallback: Cargando modelo {MODEL_PATH} localmente en {DEVICE}...")
|
73 |
tokenizer_local = AutoTokenizer.from_pretrained(MODEL_PATH)
|
74 |
+
model_local = AutoModelForCausalLM.from_pretrained(MODEL_PATH).to(DEVICE)
|
75 |
|
76 |
+
model_context_window = getattr(model_local.config, 'max_position_embeddings', getattr(model_local.config, 'sliding_window', 4096))
|
77 |
+
max_prompt_len = model_context_window - MAX_LENGTH - 50
|
78 |
+
if max_prompt_len <= 0 : max_prompt_len = model_context_window // 2
|
79 |
+
|
80 |
+
inputs = tokenizer_local(prompt, return_tensors="pt", truncation=True, max_length=max_prompt_len).to(DEVICE)
|
81 |
+
with torch.no_grad():
|
82 |
outputs = model_local.generate(
|
83 |
**inputs,
|
84 |
max_new_tokens=MAX_LENGTH,
|
|
|
91 |
return analysis.strip()
|
92 |
except Exception as e_local_llm:
|
93 |
print(f"Error en el fallback LLM local: {e_local_llm}")
|
94 |
+
traceback.print_exc()
|
95 |
return f"Análisis (fallback local): Error al cargar/ejecutar modelo LLM local: {e_local_llm}."
|
96 |
|
97 |
+
|
98 |
def process_and_plot(
|
99 |
+
file_obj,
|
|
|
100 |
biomass_eq1_ui, biomass_eq2_ui, biomass_eq3_ui,
|
101 |
biomass_param1_ui, biomass_param2_ui, biomass_param3_ui,
|
102 |
biomass_bound1_ui, biomass_bound2_ui, biomass_bound3_ui,
|
|
|
113 |
substrate_eq_count_ui,
|
114 |
product_eq_count_ui
|
115 |
):
|
116 |
+
try: # Bloque try-except general para capturar cualquier error y retornar consistentemente
|
117 |
+
analysis_text = "Iniciando análisis..."
|
118 |
+
default_image = Image.new('RGB', (600, 400), color = 'white') # Imagen placeholder
|
119 |
+
|
120 |
+
if file_obj is None:
|
121 |
+
return default_image, "Error: Por favor, sube un archivo Excel."
|
122 |
+
|
123 |
+
try:
|
124 |
+
df = pd.read_excel(file_obj.name)
|
125 |
+
except Exception as e:
|
126 |
+
return default_image, f"Error al leer el archivo Excel: {e}\n{traceback.format_exc()}"
|
127 |
+
|
128 |
+
expected_cols = ['Tiempo', 'Biomasa', 'Sustrato', 'Producto']
|
129 |
+
for col in expected_cols:
|
130 |
+
if col not in df.columns:
|
131 |
+
return default_image, f"Error: La columna '{col}' no se encuentra en el archivo Excel."
|
132 |
+
|
133 |
+
time_data = df['Tiempo'].values
|
134 |
+
biomass_data_exp = df['Biomasa'].values
|
135 |
+
substrate_data_exp = df['Sustrato'].values
|
136 |
+
product_data_exp = df['Producto'].values
|
137 |
+
|
138 |
+
active_biomass_eqs = int(biomass_eq_count_ui)
|
139 |
+
active_substrate_eqs = int(substrate_eq_count_ui)
|
140 |
+
active_product_eqs = int(product_eq_count_ui)
|
141 |
+
|
142 |
+
all_eq_inputs = {
|
143 |
+
'biomass': (
|
144 |
+
[biomass_eq1_ui, biomass_eq2_ui, biomass_eq3_ui][:active_biomass_eqs],
|
145 |
+
[biomass_param1_ui, biomass_param2_ui, biomass_param3_ui][:active_biomass_eqs],
|
146 |
+
[biomass_bound1_ui, biomass_bound2_ui, biomass_bound3_ui][:active_biomass_eqs],
|
147 |
+
biomass_data_exp
|
148 |
+
),
|
149 |
+
'substrate': (
|
150 |
+
[substrate_eq1_ui, substrate_eq2_ui, substrate_eq3_ui][:active_substrate_eqs],
|
151 |
+
[substrate_param1_ui, substrate_param2_ui, substrate_param3_ui][:active_substrate_eqs],
|
152 |
+
[substrate_bound1_ui, substrate_bound2_ui, substrate_bound3_ui][:active_substrate_eqs],
|
153 |
+
substrate_data_exp
|
154 |
+
),
|
155 |
+
'product': (
|
156 |
+
[product_eq1_ui, product_eq2_ui, product_eq3_ui][:active_product_eqs],
|
157 |
+
[product_param1_ui, product_param2_ui, product_param3_ui][:active_product_eqs],
|
158 |
+
[product_bound1_ui, product_bound2_ui, product_bound3_ui][:active_product_eqs],
|
159 |
+
product_data_exp
|
160 |
+
)
|
161 |
+
}
|
162 |
+
|
163 |
+
model_handler = BioprocessModel()
|
164 |
+
|
165 |
+
fitted_results_for_plot = {'biomass': [], 'substrate': [], 'product': []}
|
166 |
+
results_for_llm_prompt = {'biomass': [], 'substrate': [], 'product': []}
|
167 |
+
biomass_params_for_s_p = None
|
168 |
+
|
169 |
+
for model_type, (eq_list, param_str_list, bound_str_list, exp_data) in all_eq_inputs.items():
|
170 |
+
if not np.any(exp_data) and len(exp_data) > 0: # Check if all data points are zero or NaN
|
171 |
+
print(f"Datos experimentales para {model_type} son todos cero o NaN, saltando ajuste.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
continue
|
|
|
|
|
173 |
|
174 |
+
for i in range(len(eq_list)):
|
175 |
+
eq_str = eq_list[i]
|
176 |
+
param_s = param_str_list[i]
|
177 |
+
bound_s = bound_str_list[i]
|
178 |
+
|
179 |
+
if not eq_str or not param_s:
|
180 |
+
print(f"Ecuación o parámetros vacíos para {model_type} #{i+1}, saltando.")
|
181 |
+
continue
|
|
|
182 |
|
183 |
+
print(f"Procesando {model_type} #{i+1}: Eq='{eq_str}', Params='{param_s}'")
|
184 |
+
|
185 |
+
try:
|
186 |
+
model_handler.set_model(model_type, eq_str, param_s)
|
187 |
+
num_p = len(model_handler.models[model_type]['params'])
|
188 |
+
l_b, u_b = parse_bounds_str(bound_s, num_p)
|
189 |
+
|
190 |
+
current_biomass_params = biomass_params_for_s_p if model_type in ['substrate', 'product'] else None
|
191 |
+
|
192 |
+
y_pred, popt = model_handler.fit_model(model_type, time_data, exp_data, bounds=(l_b, u_b), biomass_params_fitted=current_biomass_params)
|
193 |
+
|
194 |
+
current_params = model_handler.params[model_type]
|
195 |
+
r2_val = model_handler.r2.get(model_type, float('nan'))
|
196 |
+
rmse_val = model_handler.rmse.get(model_type, float('nan'))
|
197 |
+
|
198 |
+
fitted_results_for_plot[model_type].append({
|
199 |
+
'equation': eq_str,
|
200 |
+
'y_pred': y_pred,
|
201 |
+
'params': current_params,
|
202 |
+
'R2': r2_val
|
203 |
+
})
|
204 |
+
results_for_llm_prompt[model_type].append({
|
205 |
+
'equation': eq_str,
|
206 |
+
'params_fitted': current_params,
|
207 |
+
'R2': r2_val,
|
208 |
+
'RMSE': rmse_val
|
209 |
+
})
|
210 |
+
|
211 |
+
if model_type == 'biomass' and biomass_params_for_s_p is None:
|
212 |
+
biomass_params_for_s_p = current_params
|
213 |
+
print(f"Parámetros de Biomasa (para S/P): {biomass_params_for_s_p}")
|
214 |
+
|
215 |
+
except Exception as e:
|
216 |
+
error_msg = f"Error ajustando {model_type} #{i+1} ('{eq_str}'): {e}\n{traceback.format_exc()}"
|
217 |
+
print(error_msg)
|
218 |
+
return default_image, error_msg
|
219 |
+
|
220 |
+
# Generar gráfico
|
221 |
+
fig, axs = plt.subplots(3, 1, figsize=(10, 18), sharex=True)
|
222 |
+
plot_config = {
|
223 |
+
axs[0]: (biomass_data_exp, 'Biomasa', fitted_results_for_plot['biomasa']),
|
224 |
+
axs[1]: (substrate_data_exp, 'Sustrato', fitted_results_for_plot['sustrato']),
|
225 |
+
axs[2]: (product_data_exp, 'Producto', fitted_results_for_plot['producto'])
|
226 |
+
}
|
227 |
+
|
228 |
+
for ax, data_actual, ylabel, plot_results_list in plot_config.items():
|
229 |
+
if np.any(data_actual): # Solo plotear si hay datos
|
230 |
+
ax.plot(time_data, data_actual, 'o', label=f'Datos {ylabel}', markersize=5, alpha=0.7)
|
231 |
+
else:
|
232 |
+
ax.text(0.5, 0.5, f"No hay datos para {ylabel}", transform=ax.transAxes, ha='center', va='center', fontsize=12, color='gray')
|
233 |
+
|
|
|
234 |
for idx, res_detail in enumerate(plot_results_list):
|
235 |
+
label = f'Modelo {idx+1} (R²:{res_detail["R2"]:.3f})'
|
236 |
+
ax.plot(time_data, res_detail['y_pred'], '-', label=label, linewidth=2)
|
237 |
+
ax.set_xlabel('Tiempo')
|
238 |
+
ax.set_ylabel(ylabel)
|
239 |
+
ax.grid(True, linestyle=':', alpha=0.7)
|
240 |
+
if show_legend_ui:
|
241 |
+
ax.legend(loc=legend_position_ui, fontsize='small')
|
242 |
|
243 |
+
if show_params_ui and plot_results_list:
|
244 |
+
param_display_texts = []
|
245 |
+
for idx, res_detail in enumerate(plot_results_list):
|
246 |
+
params_text = f"Modelo {idx+1}:\n" + "\n".join([f" {k}: {v:.4g}" for k,v in res_detail['params'].items()])
|
247 |
+
param_display_texts.append(params_text)
|
248 |
+
full_param_text = "\n---\n".join(param_display_texts)
|
249 |
+
|
250 |
+
text_x_pos = 0.02
|
251 |
+
text_y_pos = 0.98
|
252 |
+
v_align = 'top'
|
253 |
+
if legend_position_ui and 'upper' in legend_position_ui:
|
254 |
+
text_y_pos = 0.02
|
255 |
+
v_align = 'bottom'
|
256 |
+
|
257 |
+
ax.text(text_x_pos, text_y_pos, full_param_text, transform=ax.transAxes, fontsize=7,
|
258 |
+
verticalalignment=v_align, bbox=dict(boxstyle='round,pad=0.3', fc='lightyellow', alpha=0.8))
|
259 |
+
|
260 |
+
plt.tight_layout(rect=[0, 0, 1, 0.96])
|
261 |
+
fig.suptitle("Resultados del Ajuste de Modelos Cinéticos", fontsize=16)
|
262 |
+
|
263 |
+
buf = io.BytesIO()
|
264 |
+
plt.savefig(buf, format='png', dpi=150)
|
265 |
+
buf.seek(0)
|
266 |
+
image = Image.open(buf)
|
267 |
+
plt.close(fig)
|
268 |
+
|
269 |
+
# Construir prompt para LLM y llamar al servicio
|
270 |
+
prompt_intro = "Eres un experto en modelado cinético de bioprocesos. Analiza los siguientes resultados del ajuste de modelos a datos experimentales:\n\n"
|
271 |
+
prompt_details = json.dumps(results_for_llm_prompt, indent=2, ensure_ascii=False)
|
272 |
+
prompt_instructions = """\n\nPor favor, proporciona un análisis detallado y crítico en español, estructurado de la siguiente manera:
|
273 |
1. **Resumen General:** Una breve descripción del experimento y qué se intentó modelar.
|
274 |
2. **Análisis por Componente (Biomasa, Sustrato, Producto):**
|
275 |
a. Para cada ecuación probada:
|
|
|
283 |
4. **Sugerencias y Próximos Pasos:**
|
284 |
a. ¿Cómo se podría mejorar el modelado (ej. probar otras ecuaciones, transformar datos, revisar calidad de datos experimentales)?
|
285 |
b. ¿Qué experimentos adicionales podrían realizarse para validar o refinar los modelos?
|
286 |
+
5. **Conclusión Final:** Un veredicto general conciso sobre el éxito del modelado y la utilidad de los resultados obtenidos.
|
287 |
|
288 |
Utiliza un lenguaje claro y accesible, pero manteniendo el rigor técnico. El análisis debe ser útil para alguien que busca entender la cinética de su bioproceso."""
|
289 |
+
|
290 |
+
full_prompt = prompt_intro + prompt_details + prompt_instructions
|
291 |
+
|
292 |
+
analysis_text = call_llm_analysis_service(full_prompt)
|
293 |
+
|
294 |
+
return image, analysis_text
|
295 |
|
296 |
+
except Exception as general_e:
|
297 |
+
# Captura cualquier excepción no manejada y la muestra en la UI
|
298 |
+
error_trace = traceback.format_exc()
|
299 |
+
error_message_full = f"Error inesperado en process_and_plot: {general_e}\n{error_trace}"
|
300 |
+
print(error_message_full)
|
301 |
+
return Image.new('RGB', (600, 400), color = 'red'), error_message_full # Retorna imagen roja de error
|