Spaces:
Running on Zero
Running on Zero
File size: 7,402 Bytes
a0270e2 2cb1107 a0270e2 2cb1107 a0270e2 2cb1107 a0270e2 2cb1107 a0270e2 | 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 | """
Hugging Face Spaces Interactive Demo for Parallel Constrained Decoding.
Optimized for Nvidia ZeroGPU (A10G) and PyTorch.
"""
import os
import json
import time
from typing import Dict, Any, Generator
import gradio as gr
from core.schema import StructuredSchema
from core.engine import run_parallel_generation, run_naive_generation
# ZeroGPU decorator support
try:
import spaces
gpu_decorator = spaces.GPU(duration=60)
except Exception:
def gpu_decorator(fn):
return fn
# Load presets from presets/ directory
PRESETS = {}
presets_dir = os.path.join(os.path.dirname(__file__), "presets")
if os.path.exists(presets_dir):
for fname in sorted(os.listdir(presets_dir)):
if fname.endswith(".json"):
try:
with open(os.path.join(presets_dir, fname), "r") as f:
data = json.load(f)
title = data.get("title", fname)
PRESETS[title] = {
"context": data.get("context", ""),
"schema": json.dumps(data.get("schema", {}), indent=2)
}
except Exception as e:
print(f"Error loading {fname}: {e}")
preset_titles = list(PRESETS.keys())
default_title = preset_titles[0] if preset_titles else None
default_context = PRESETS[default_title]["context"] if default_title else ""
default_schema = PRESETS[default_title]["schema"] if default_title else "{}"
@gpu_decorator
def run_comparison(context_str: str, schema_json_str: str):
if not context_str or not context_str.strip():
yield (
"<div style='color: #dc2626; font-weight: 600; padding: 6px 12px;'>Please provide a context prompt.</div>",
"{}",
"0.0 ms",
"{}",
"0.0 ms"
)
return
try:
schema_dict = json.loads(schema_json_str)
schema = StructuredSchema(schema_dict)
except Exception as e:
yield (
f"<div style='color: #dc2626; font-weight: 600; padding: 6px 12px;'>Invalid Schema JSON: {e}</div>",
"{}",
"0.0 ms",
"{}",
"0.0 ms"
)
return
try:
# 1. Run Parallel Constrained Decoding first
parallel_res = run_parallel_generation(context_str, schema)
parallel_ms = parallel_res["elapsed_ms"]
parallel_json_str = json.dumps(parallel_res["parsed_json"], indent=2)
parallel_time_badge = f"{parallel_ms:.1f} ms"
summary_intermediate = f"""
<div style="background: #f0fdf4; border: 1px solid #bbf7d0; border-radius: 9999px; padding: 6px 16px; display: inline-flex; align-items: center; gap: 8px; font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; font-size: 14px;">
<span style="color: #16a34a; font-weight: 700;">Parallel Done: {parallel_time_badge}</span>
<span style="color: #94a3b8;">·</span>
<span style="color: #64748b;">Evaluating normal autoregressive baseline...</span>
</div>
"""
yield (
summary_intermediate,
parallel_json_str,
parallel_time_badge,
"// Running sequential autoregressive baseline forward passes...",
"Evaluating..."
)
# 2. Run Naive generation baseline
naive_res = run_naive_generation(context_str, schema)
naive_ms = naive_res["elapsed_ms"]
naive_json_str = json.dumps(naive_res["parsed_json"], indent=2) if naive_res.get("parsed_json") else naive_res.get("raw_text", "")
naive_time_badge = f"{naive_ms:.1f} ms"
speedup = round(naive_ms / max(parallel_ms, 1.0), 1)
final_summary_html = f"""
<div style="background: #f0fdf4; border: 1px solid #bbf7d0; border-radius: 9999px; padding: 8px 20px; display: inline-flex; align-items: center; gap: 10px; font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; font-size: 15px; box-shadow: 0 1px 3px rgba(0,0,0,0.05);">
<span style="color: #16a34a; font-weight: 800; font-size: 16px; letter-spacing: 0.5px;">{speedup}x FASTER</span>
<span style="color: #cbd5e1; font-weight: 600;">·</span>
<span style="color: #334155; font-weight: 600; font-family: monospace;">{parallel_time_badge} vs {naive_time_badge}</span>
</div>
"""
yield (
final_summary_html,
parallel_json_str,
parallel_time_badge,
naive_json_str,
naive_time_badge
)
except Exception as err:
import traceback
err_msg = f"{err}\n{traceback.format_exc()}"
yield (
f"<div style='color: #dc2626; background: #fef2f2; border: 1px solid #fecaca; border-radius: 8px; padding: 10px 14px; font-family: monospace; font-size: 13px;'>Error: {err}</div>",
"{}",
"0.0 ms",
f"Error details:\n{err_msg}",
"0.0 ms"
)
with gr.Blocks(title="Parallel Constrained Decision Engine") as demo:
gr.Markdown("# Parallel Constrained vs Normal Inference (Qwen2.5 1.5B)")
gr.Markdown("Parallel Constrained Decoding evaluates all schema fields simultaneously against broadcast prefix KV-cache states, delivering substantial latency reductions with 100% schema adherence.")
with gr.Row():
preset_dropdown = gr.Dropdown(
choices=preset_titles,
value=default_title,
label="Select Preset Scenario",
scale=4
)
btn_run = gr.Button("⚡ Run Comparison", variant="primary", scale=1)
summary_banner = gr.HTML(value="")
with gr.Row():
with gr.Column(scale=1):
gr.Markdown("### Parallel Constrained (Qwen2.5 1.5B)")
timer_parallel = gr.Textbox(label="Elapsed Time", value="0.0 ms", interactive=False, max_lines=1)
output_parallel = gr.Code(label="Parallel JSON (Values + Calibrated Probabilities)", language="json", interactive=False, lines=18)
with gr.Column(scale=1):
gr.Markdown("### Normal Inference (Qwen2.5 1.5B)")
timer_naive = gr.Textbox(label="Elapsed Time", value="0.0 ms", interactive=False, max_lines=1)
output_naive = gr.Code(label="Autoregressive JSON Output", language="json", interactive=False, lines=18)
with gr.Accordion("Inspect Context Document & Schema Definition", open=False):
context_input = gr.Textbox(
label="Context Document",
value=default_context,
lines=6
)
schema_input = gr.Code(
label="Schema Definition (JSON)",
value=default_schema,
language="json",
lines=10
)
def on_preset_change(title):
if title in PRESETS:
return PRESETS[title]["context"], PRESETS[title]["schema"]
return "", "{}"
preset_dropdown.change(
fn=on_preset_change,
inputs=[preset_dropdown],
outputs=[context_input, schema_input]
)
btn_run.click(
fn=run_comparison,
inputs=[context_input, schema_input],
outputs=[summary_banner, output_parallel, timer_parallel, output_naive, timer_naive]
)
if __name__ == "__main__":
demo.queue().launch(server_name="0.0.0.0", server_port=int(os.environ.get("PORT", 7860)))
|