Spaces:
Running
Running
File size: 11,256 Bytes
0ad7e2a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 |
"""
Train tab for Video Model Studio UI
"""
import gradio as gr
import logging
from typing import Dict, Any, List, Optional
from .base_tab import BaseTab
from ..config import TRAINING_PRESETS, MODEL_TYPES, ASK_USER_TO_DUPLICATE_SPACE
from ..utils import TrainingLogParser
logger = logging.getLogger(__name__)
class TrainTab(BaseTab):
"""Train tab for model training"""
def __init__(self, app_state):
super().__init__(app_state)
self.id = "train_tab"
self.title = "4️⃣ Train"
def handle_training_start(self, preset, model_type, *args):
"""Handle training start with proper log parser reset"""
# Safely reset log parser if it exists
if hasattr(self.app, 'log_parser') and self.app.log_parser is not None:
self.app.log_parser.reset()
else:
logger.warning("Log parser not initialized, creating a new one")
self.app.log_parser = TrainingLogParser()
# Start training
return self.app.trainer.start_training(
MODEL_TYPES[model_type],
*args,
preset_name=preset
)
def create(self, parent=None) -> gr.TabItem:
"""Create the Train tab UI components"""
with gr.TabItem(self.title, id=self.id) as tab:
with gr.Row():
with gr.Column():
with gr.Row():
self.components["train_title"] = gr.Markdown("## 0 files available for training (0 bytes)")
with gr.Row():
with gr.Column():
self.components["training_preset"] = gr.Dropdown(
choices=list(TRAINING_PRESETS.keys()),
label="Training Preset",
value=list(TRAINING_PRESETS.keys())[0]
)
self.components["preset_info"] = gr.Markdown()
with gr.Row():
with gr.Column():
self.components["model_type"] = gr.Dropdown(
choices=list(MODEL_TYPES.keys()),
label="Model Type",
value=list(MODEL_TYPES.keys())[0]
)
self.components["model_info"] = gr.Markdown(
value=self.app.get_model_info(list(MODEL_TYPES.keys())[0])
)
with gr.Row():
self.components["lora_rank"] = gr.Dropdown(
label="LoRA Rank",
choices=["16", "32", "64", "128", "256", "512", "1024"],
value="128",
type="value"
)
self.components["lora_alpha"] = gr.Dropdown(
label="LoRA Alpha",
choices=["16", "32", "64", "128", "256", "512", "1024"],
value="128",
type="value"
)
with gr.Row():
self.components["num_epochs"] = gr.Number(
label="Number of Epochs",
value=70,
minimum=1,
precision=0
)
self.components["batch_size"] = gr.Number(
label="Batch Size",
value=1,
minimum=1,
precision=0
)
with gr.Row():
self.components["learning_rate"] = gr.Number(
label="Learning Rate",
value=2e-5,
minimum=1e-7
)
self.components["save_iterations"] = gr.Number(
label="Save checkpoint every N iterations",
value=500,
minimum=50,
precision=0,
info="Model will be saved periodically after these many steps"
)
with gr.Column():
with gr.Row():
self.components["start_btn"] = gr.Button(
"Start Training",
variant="primary",
interactive=not ASK_USER_TO_DUPLICATE_SPACE
)
self.components["pause_resume_btn"] = gr.Button(
"Resume Training",
variant="secondary",
interactive=False
)
self.components["stop_btn"] = gr.Button(
"Stop Training",
variant="stop",
interactive=False
)
with gr.Row():
with gr.Column():
self.components["status_box"] = gr.Textbox(
label="Training Status",
interactive=False,
lines=4
)
with gr.Accordion("See training logs"):
self.components["log_box"] = gr.TextArea(
label="Finetrainers output (see HF Space logs for more details)",
interactive=False,
lines=40,
max_lines=200,
autoscroll=True
)
return tab
def connect_events(self) -> None:
"""Connect event handlers to UI components"""
# Model type change event
def update_model_info(model):
params = self.app.get_default_params(MODEL_TYPES[model])
info = self.app.get_model_info(MODEL_TYPES[model])
return {
self.components["model_info"]: info,
self.components["num_epochs"]: params["num_epochs"],
self.components["batch_size"]: params["batch_size"],
self.components["learning_rate"]: params["learning_rate"],
self.components["save_iterations"]: params["save_iterations"]
}
self.components["model_type"].change(
fn=lambda v: self.app.update_ui_state(model_type=v),
inputs=[self.components["model_type"]],
outputs=[]
).then(
fn=update_model_info,
inputs=[self.components["model_type"]],
outputs=[
self.components["model_info"],
self.components["num_epochs"],
self.components["batch_size"],
self.components["learning_rate"],
self.components["save_iterations"]
]
)
# Training parameters change events
self.components["lora_rank"].change(
fn=lambda v: self.app.update_ui_state(lora_rank=v),
inputs=[self.components["lora_rank"]],
outputs=[]
)
self.components["lora_alpha"].change(
fn=lambda v: self.app.update_ui_state(lora_alpha=v),
inputs=[self.components["lora_alpha"]],
outputs=[]
)
self.components["num_epochs"].change(
fn=lambda v: self.app.update_ui_state(num_epochs=v),
inputs=[self.components["num_epochs"]],
outputs=[]
)
self.components["batch_size"].change(
fn=lambda v: self.app.update_ui_state(batch_size=v),
inputs=[self.components["batch_size"]],
outputs=[]
)
self.components["learning_rate"].change(
fn=lambda v: self.app.update_ui_state(learning_rate=v),
inputs=[self.components["learning_rate"]],
outputs=[]
)
self.components["save_iterations"].change(
fn=lambda v: self.app.update_ui_state(save_iterations=v),
inputs=[self.components["save_iterations"]],
outputs=[]
)
# Training preset change event
self.components["training_preset"].change(
fn=lambda v: self.app.update_ui_state(training_preset=v),
inputs=[self.components["training_preset"]],
outputs=[]
).then(
fn=self.app.update_training_params,
inputs=[self.components["training_preset"]],
outputs=[
self.components["model_type"],
self.components["lora_rank"],
self.components["lora_alpha"],
self.components["num_epochs"],
self.components["batch_size"],
self.components["learning_rate"],
self.components["save_iterations"],
self.components["preset_info"]
]
)
# Training control events
self.components["start_btn"].click(
fn=self.handle_training_start, # Use safer method instead of lambda
inputs=[
self.components["training_preset"],
self.components["model_type"],
self.components["lora_rank"],
self.components["lora_alpha"],
self.components["num_epochs"],
self.components["batch_size"],
self.components["learning_rate"],
self.components["save_iterations"],
self.app.tabs["manage_tab"].components["repo_id"]
],
outputs=[
self.components["status_box"],
self.components["log_box"]
]
).success(
fn=self.app.get_latest_status_message_logs_and_button_labels,
outputs=[
self.components["status_box"],
self.components["log_box"],
self.components["start_btn"],
self.components["stop_btn"],
self.components["pause_resume_btn"]
]
)
self.components["pause_resume_btn"].click(
fn=self.app.handle_pause_resume,
outputs=[
self.components["status_box"],
self.components["log_box"],
self.components["start_btn"],
self.components["stop_btn"],
self.components["pause_resume_btn"]
]
)
self.components["stop_btn"].click(
fn=self.app.handle_stop,
outputs=[
self.components["status_box"],
self.components["log_box"],
self.components["start_btn"],
self.components["stop_btn"],
self.components["pause_resume_btn"]
]
) |