github-starred/cs249r_book
2
0
Fork 0
mirror of https://github.com/harvard-edge/cs249r_book.git synced 2026-05-07 02:03:55 -05:00
Code Issues 85 Packages Projects Releases 24 Wiki Activity

#1305 [Bug] Labs lab00 cleanup

Browse Source
This commit is contained in:
Peter Koellner
2026-04-09 19:07:31 +02:00
committed by Vijay Janapa Reddi
parent ff5df70044
commit cbd418c5b9
1 changed files with 136 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files

210
labs/vol1/lab_00_introduction.py
View File

@@ -1,6 +1,8 @@
import marimo
__generated_with = "0.19.6"
from mlsysim import Hardware
__generated_with = "0.23.0"
app = marimo.App(width="full")
# ─────────────────────────────────────────────────────────────────────────────
@@ -59,12 +61,12 @@ async def _():
ledger = DesignLedger()
if getattr(ledger, "is_wasm", False):
await ledger.load_async()
return mo, ledger, COLORS, LAB_CSS, DecisionLog
return COLORS, DecisionLog, LAB_CSS, ledger, mo
# ─── CELL 1: HEADER ────────────────────────────────────────────────────────────
@app.cell
def _(mo, LAB_CSS):
def _(LAB_CSS, mo):
mo.vstack([
LAB_CSS,
mo.md("""
@@ -109,7 +111,7 @@ def _(mo, LAB_CSS):
# ─── CELL 2: BRIEFING ──────────────────────────────────────────────────────────
@app.cell(hide_code=True)
def _(mo, COLORS):
def _(COLORS, mo):
mo.Html(f"""
<div style="border-left: 4px solid {COLORS['BlueLine']};
background: white; border-radius: 0 12px 12px 0;
@@ -247,15 +249,15 @@ def _(mo):
"D) The dataset size — gathering more labeled training examples": "D",
},
label="""**Check your understanding.** A startup ships a model with 94% accuracy.
Six months later, accuracy has silently dropped to 81% in production — but no code
has changed. As an ML Systems engineer, which part of the system is your *primary*
domain for diagnosing and fixing this?""",
Six months later, accuracy has silently dropped to 81% in production — but no code
has changed. As an ML Systems engineer, which part of the system is your *primary*
domain for diagnosing and fixing this?""",
)
return (check1,)
@app.cell
def _(mo, check1):
def _(check1, mo):
mo.stop(
check1.value is None,
mo.vstack([
@@ -309,7 +311,7 @@ def _(mo, check1):
# ─── CONCEPT 2: PHYSICAL CONSTRAINTS PARTITION DEPLOYMENT ─────────────────────
@app.cell
def _(mo, check1):
def _(check1, mo):
mo.stop(check1.value is None)
mo.vstack([
@@ -384,47 +386,78 @@ def _(mo, check1):
# ─── CHECK 2 (multi-select) ────────────────────────────────────────────────────
@app.cell
def _(mo, check1):
def _(check1, mo):
mo.stop(check1.value is None)
check2 = mo.ui.multiselect(
options={
"Use a smaller model with fewer parameters": "model_size",
"Apply INT8 quantization to reduce precision": "quantization",
"Move the datacenter server physically closer": "move_server",
"Use a faster GPU with higher TFLOPS": "faster_gpu",
"Deploy the model directly on the vehicle": "edge_deploy",
},
label="""**Check your understanding.** An autonomous vehicle perception system
is routed to a cloud datacenter 2,000 km away. Round-trip latency is 40 ms.
The safety requirement is a 10 ms end-to-end decision loop.
Select **all approaches** that could actually solve the latency problem:""",
model_size = mo.ui.checkbox(
value=False,
label="Use a smaller model with fewer parameters"
)
return (check2,)
quantization = mo.ui.checkbox(
value=False,
label="Apply INT8 quantization to reduce precision"
)
move_server = mo.ui.checkbox(
value=False,
label="Move the datacenter server physically closer"
)
faster_gpu = mo.ui.checkbox(
value=False,
label="Use a faster GPU with higher TFLOPS"
)
edge_deploy = mo.ui.checkbox(
value=False,
label="Deploy the model directly on the vehicle"
)
mo.vstack([
mo.md("""**Check your understanding.** An autonomous vehicle perception system
is routed to a cloud datacenter 2,000 km away. Round-trip latency is 40 ms.
The safety requirement is a 10 ms end-to-end decision loop."""),
mo.md("""Select **all approaches** that could actually solve the latency problem:"""),
model_size,
quantization,
move_server,
faster_gpu,
edge_deploy
])
return edge_deploy, faster_gpu, model_size, move_server, quantization
@app.cell
def _(mo, check1, check2):
mo.stop(check1.value is None or len(check2.value) == 0)
def _(
check1,
check2empty,
edge_deploy,
faster_gpu,
mo,
model_size,
move_server,
quantization,
):
mo.stop(check1.value is None or check2empty())
_correct_set = {"move_server", "edge_deploy"}
_selected = set(check2.value)
_exactly_right = _selected == _correct_set
_has_wrong = bool(_selected - _correct_set)
_missing_right = bool(_correct_set - _selected)
_correct_set = {'move_server', 'edge_deploy'}
_has_wrong = model_size.value or quantization.value or faster_gpu.value
_missing_right = not (move_server.value and edge_deploy.value)
_exactly_right = move_server.value and edge_deploy.value and not _has_wrong
_option_labels = {
"model_size": "Use a smaller model",
"quantization": "Apply INT8 quantization",
"move_server": "Move the server physically closer",
"faster_gpu": "Use a faster GPU",
"edge_deploy": "Deploy on the vehicle",
'model_size': "Use a smaller model",
'quantization': "Apply INT8 quantization",
'move_server': "Move the server physically closer",
'faster_gpu': "Use a faster GPU",
'edge_deploy': "Deploy on the vehicle",
}
_rows = ""
for _key, _label in _option_labels.items():
_is_selected = _key in _selected
_is_selected = locals()[_key].value
_is_correct = _key in _correct_set
if _is_selected and _is_correct:
_icon, _bg, _col = "✅", "#f0fdf4", "#15803d"
@@ -468,7 +501,6 @@ def _(mo, check1, check2):
_bg_outer = "#f0fdf4" if _exactly_right else ("#fffbeb" if not _has_wrong else "#fef2f2")
mo.vstack([
check2,
mo.Html(f"""
<div style="background:{_bg_outer}; border:1.5px solid {_border};
border-radius:10px; padding:18px 20px; margin-top:8px;">
@@ -487,17 +519,17 @@ def _(mo, check1, check2):
), kind="danger"),
mo.accordion({
"Math Peek: Propagation Delay": mo.md("""
**Formula:**
$$
t_{\\text{round-trip}} = \\frac{2d}{c \\cdot n}
$$
**Formula:**
$$
t_{\\text{round-trip}} = \\frac{2d}{c \\cdot n}
$$
**Variables:**
- **d**: distance between client and server (km)
- **c**: speed of light in vacuum (299,792 km/s)
- **n**: fiber refractive index factor (~0.67)
- At d = 2,000 km: t = 2 × 2,000 / (299,792 × 0.67) ≈ 20 ms — exceeds 10 ms SLA by 2x
""")
**Variables:**
- **d**: distance between client and server (km)
- **c**: speed of light in vacuum (299,792 km/s)
- **n**: fiber refractive index factor (~0.67)
- At d = 2,000 km: t = 2 × 2,000 / (299,792 × 0.67) ≈ 20 ms — exceeds 10 ms SLA by 2x
""")
}),
])
return
@@ -506,8 +538,8 @@ $$
# ─── CONCEPT 3: THE DEPLOYMENT REGIMES ────────────────────────────────────────
@app.cell
def _(mo, check1, check2):
mo.stop(check1.value is None or len(check2.value) == 0)
def _(check1, check2empty, mo):
mo.stop(check1.value is None or check2empty())
mo.vstack([
mo.md("---"),
@@ -628,8 +660,8 @@ def _(mo, check1, check2):
# ─── CHECK 3 (constraint reasoning) ───────────────────────────────────────────
@app.cell
def _(mo, check1, check2):
mo.stop(check1.value is None or len(check2.value) == 0)
def _(check1, check2empty, mo):
mo.stop(check1.value is None or check2empty())
check3 = mo.ui.radio(
options={
@@ -639,18 +671,22 @@ def _(mo, check1, check2):
"D) TinyML — lowest power, can run for months on a battery": "D",
},
label="""**Check your understanding.** A hospital wants to deploy an AI system
that detects sepsis from ICU sensor readings. Requirements: results within 2 ms of
each sensor reading, no patient data can leave the hospital network, and the sensor
node must run for 6 months on a small battery without replacement.
that detects sepsis from ICU sensor readings. Requirements: results within 2 ms of
each sensor reading, no patient data can leave the hospital network, and the sensor
node must run for 6 months on a small battery without replacement.
Which deployment paradigm is the *only* one that satisfies all three requirements simultaneously?""",
Which deployment paradigm is the *only* one that satisfies all three requirements simultaneously?""",
)
mo.vstack([
check3,
])
return (check3,)
@app.cell
def _(mo, check1, check2, check3):
mo.stop(check1.value is None or len(check2.value) == 0 or check3.value is None)
def _(check1, check2empty, check3, mo):
mo.stop(check1.value is None or check2empty() or check3.value is None)
_correct = check3.value == "D"
_feedback = {
@@ -683,7 +719,6 @@ def _(mo, check1, check2, check3):
}
mo.vstack([
check3,
mo.callout(
mo.md(_feedback[check3.value]),
kind="success" if _correct else "warn",
@@ -701,8 +736,8 @@ def _(mo, check1, check2, check3):
# ─── INTERFACE ORIENTATION INTRO ───────────────────────────────────────────────
@app.cell
def _(mo, check1, check2, check3):
mo.stop(check1.value is None or len(check2.value) == 0 or check3.value is None)
def _(check1, check2empty, check3, mo):
mo.stop(check1.value is None or check2empty() or check3.value is None)
mo.vstack([
mo.md("---"),
@@ -721,8 +756,8 @@ def _(mo, check1, check2, check3):
@app.cell
def _(mo, check1, check2, check3, COLORS):
mo.stop(check1.value is None or len(check2.value) == 0 or check3.value is None)
def _(COLORS, check1, check2empty, check3, mo):
mo.stop(check1.value is None or check2empty() or check3.value is None)
# ── ZONE ANATOMY DIAGRAM ─────────────────────────────────────────
_zone_html = """
@@ -953,9 +988,9 @@ def _(mo, check1, check2, check3, COLORS):
# ─── DEPLOYMENT CONTEXT SELECTION ─────────────────────────────────────────────
@app.cell
def _(mo, check1, check2, check3):
def _(check1, check2empty, check3, mo):
mo.stop(
check1.value is None or len(check2.value) == 0 or check3.value is None,
check1.value is None or check2empty() or check3.value is None,
mo.md("_Complete all three checks above to unlock your deployment context selection._")
)
@@ -978,8 +1013,8 @@ def _(mo, check1, check2, check3):
@app.cell
def _(mo, check1, check2, check3):
mo.stop(check1.value is None or len(check2.value) == 0 or check3.value is None)
def _(check1, check2empty, check3, mo):
mo.stop(check1.value is None or check2empty() or check3.value is None)
context_selector = mo.ui.radio(
options={
@@ -996,16 +1031,26 @@ def _(mo, check1, check2, check3):
# ─── CONTEXT REVEAL + STAKEHOLDER MESSAGE + LEDGER INIT ───────────────────────
@app.cell(hide_code=True)
def _(mo, DecisionLog):
def _(DecisionLog):
decision_input, decision_ui = DecisionLog()
return decision_input, decision_ui
return (decision_ui,)
@app.cell
def _(mo, check1, check2, check3, context_selector, ledger, COLORS, decision_input, decision_ui):
def _(
COLORS,
check1,
check2empty,
check2value_list,
check3,
context_selector,
decision_ui,
ledger,
mo,
):
mo.stop(
check1.value is None
or len(check2.value) == 0
or check2empty()
or check3.value is None
or context_selector.value is None,
mo.vstack([
@@ -1121,7 +1166,7 @@ def _(mo, check1, check2, check3, context_selector, ledger, COLORS, decision_inp
"deployment_context": _key,
"check1_answer": check1.value,
"check1_correct": check1.value == "C",
"check2_selections": list(check2.value),
"check2_selections": check2value_list(),
"check3_answer": check3.value,
"check3_correct": check3.value == "D",
})
@@ -1200,7 +1245,7 @@ def _(mo, check1, check2, check3, context_selector, ledger, COLORS, decision_inp
# ─── CELL 20: SYNTHESIS ────────────────────────────────────────────────────────
@app.cell(hide_code=True)
def _(mo, COLORS):
def _(COLORS, mo):
mo.vstack([
mo.md("---"),
@@ -1291,7 +1336,7 @@ def _(mo, COLORS):
# ─── CELL 21: LEDGER_HUD ───────────────────────────────────────────────────────
@app.cell
def _(mo, ledger, COLORS):
def _(COLORS, ledger, mo):
_track = ledger.get_track() or "NONE"
_color_map = {
"cloud": COLORS["BlueLine"],
@@ -1325,6 +1370,23 @@ def _(mo, ledger, COLORS):
""")
return
# --- Auxiliary methods ---------------------------------------------------------
@app.cell
def _(edge_deploy, faster_gpu, model_size, move_server, quantization):
def check2empty():
return not (model_size.value or quantization.value or move_server.value or faster_gpu.value or edge_deploy.value)
def check2value_list():
check2values = []
# have to use items directly here for dependency evaluation
if (model_size.value or quantization.value or move_server.value or faster_gpu.value or edge_deploy.value):
for item in ['model_size', 'quantization', 'move_server' ,'faster_gpu', 'edge_deploy']:
if globals()[item].value:
check2values.append(item)
return check2values
return check2empty, check2value_list
if __name__ == "__main__":
app.run()