mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-06-03 12:12:48 -05:00
Remove unnecessary matplotlib import from losses module
Issue: xor_crisis.py was failing with ImportError on matplotlib architecture mismatch Root cause: losses_dev.py imported matplotlib.pyplot but never used it Fix: - ✅ Removed unused imports: matplotlib.pyplot, time - ✅ Re-exported module 04_losses to update tinytorch package - ✅ Verified both milestone 02 scripts now run successfully The matplotlib import was causing failures on M2 Macs where matplotlib was installed for wrong architecture (x86_64 vs arm64). Since it was never used, removing it eliminates the dependency entirely. Tested: - ✅ milestones/02_xor_crisis_1969/xor_crisis.py (49% accuracy - expected failure) - ✅ milestones/02_xor_crisis_1969/xor_solved.py (100% accuracy - perfect!)
This commit is contained in:
Vijay Janapa Reddi
@@ -2,7 +2,7 @@
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "9d798b1c",
|
||||
"id": "dc4a8074",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -35,7 +35,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "91804987",
|
||||
"id": "08ab6b0b",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -59,7 +59,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "c09dc686",
|
||||
"id": "848eaef7",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -80,7 +80,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "51189bc1",
|
||||
"id": "90d6651a",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": false,
|
||||
@@ -94,8 +94,6 @@
|
||||
"#| export\n",
|
||||
"\n",
|
||||
"import numpy as np\n",
|
||||
"import matplotlib.pyplot as plt\n",
|
||||
"import time\n",
|
||||
"from typing import Optional\n",
|
||||
"\n",
|
||||
"def import_previous_module(module_name: str, component_name: str):\n",
|
||||
@@ -113,7 +111,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "cc227c2d",
|
||||
"id": "529a8e8a",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -189,7 +187,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "49e5039b",
|
||||
"id": "4e69ba6d",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -235,7 +233,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "b1e1cbd0",
|
||||
"id": "b9a1fa2c",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -247,7 +245,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "820e9937",
|
||||
"id": "bf3b7915",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -297,7 +295,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "854758b3",
|
||||
"id": "085562d6",
|
||||
"metadata": {
|
||||
"lines_to_next_cell": 1,
|
||||
"nbgrader": {
|
||||
@@ -348,7 +346,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "6b57e650",
|
||||
"id": "d274d1e1",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": true,
|
||||
@@ -389,7 +387,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "b8be9f2c",
|
||||
"id": "d51980c3",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -459,7 +457,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "aca5154a",
|
||||
"id": "1107bf9d",
|
||||
"metadata": {
|
||||
"lines_to_next_cell": 1,
|
||||
"nbgrader": {
|
||||
@@ -531,7 +529,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "7391538b",
|
||||
"id": "80f97626",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": true,
|
||||
@@ -577,7 +575,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "0b9b254c",
|
||||
"id": "14b2d795",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -670,7 +668,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "eb59fb50",
|
||||
"id": "c0a10af0",
|
||||
"metadata": {
|
||||
"lines_to_next_cell": 1,
|
||||
"nbgrader": {
|
||||
@@ -746,7 +744,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "c59fbbfd",
|
||||
"id": "24685fb9",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": true,
|
||||
@@ -797,7 +795,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "599727d1",
|
||||
"id": "68a261f3",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -906,7 +904,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "54a20f3f",
|
||||
"id": "b02977aa",
|
||||
"metadata": {
|
||||
"lines_to_next_cell": 1,
|
||||
"nbgrader": {
|
||||
@@ -982,7 +980,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "1bab9d23",
|
||||
"id": "722d5c07",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": true,
|
||||
@@ -1033,7 +1031,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "ca40b581",
|
||||
"id": "88bad600",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -1090,7 +1088,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "76b4eb81",
|
||||
"id": "b5a701fe",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": false,
|
||||
@@ -1146,7 +1144,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b90c91f0",
|
||||
"id": "35891a55",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": false,
|
||||
@@ -1211,7 +1209,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "e2fc1aa7",
|
||||
"id": "95e3f483",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -1286,7 +1284,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "573fa75d",
|
||||
"id": "c46f9468",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": false,
|
||||
@@ -1336,7 +1334,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b7f12c78",
|
||||
"id": "d95c49f1",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": false,
|
||||
@@ -1393,7 +1391,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "4c6ebac9",
|
||||
"id": "88afe536",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\"",
|
||||
"lines_to_next_cell": 1
|
||||
@@ -1457,7 +1455,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "d0b635c1",
|
||||
"id": "232f8764",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": false,
|
||||
@@ -1513,7 +1511,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "d770e887",
|
||||
"id": "f00b5616",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
@@ -1526,7 +1524,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "55fd411d",
|
||||
"id": "76ec7947",
|
||||
"metadata": {
|
||||
"nbgrader": {
|
||||
"grade": true,
|
||||
@@ -1606,7 +1604,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b66f2370",
|
||||
"id": "e635a52d",
|
||||
"metadata": {
|
||||
"lines_to_next_cell": 2
|
||||
},
|
||||
@@ -1619,7 +1617,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "ce0d9c33",
|
||||
"id": "da28d331",
|
||||
"metadata": {
|
||||
"cell_marker": "\"\"\""
|
||||
},
|
||||
|
||||
@@ -79,8 +79,6 @@ The `import_previous_module()` function below helps us cleanly import components
|
||||
#| export
|
||||
|
||||
import numpy as np
|
||||
import matplotlib.pyplot as plt
|
||||
import time
|
||||
from typing import Optional
|
||||
|
||||
def import_previous_module(module_name: str, component_name: str):
|
||||
|
||||
95
test_xor_original_1986.py
Normal file
95
test_xor_original_1986.py
Normal file
@@ -0,0 +1,95 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
Original 1986 XOR Solution - Rumelhart, Hinton, Williams
|
||||
Testing the MINIMAL architecture that solved the XOR crisis.
|
||||
"""
|
||||
import sys
|
||||
sys.path.insert(0, '.')
|
||||
|
||||
import numpy as np
|
||||
from tinytorch import Tensor, Linear, Sigmoid, BinaryCrossEntropyLoss, SGD
|
||||
|
||||
print("=" * 70)
|
||||
print("🏛️ ORIGINAL 1986 XOR SOLUTION")
|
||||
print("Rumelhart, Hinton, Williams - 'Learning representations by back-propagating errors'")
|
||||
print("=" * 70)
|
||||
|
||||
# Pure XOR
|
||||
X_data = np.array([[0.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 1.0]], dtype=np.float32)
|
||||
y_data = np.array([[0.0], [1.0], [1.0], [0.0]], dtype=np.float32)
|
||||
|
||||
X = Tensor(X_data)
|
||||
y = Tensor(y_data)
|
||||
|
||||
print("\n🏗️ Architecture (1986 style):")
|
||||
print(" Input: 2 neurons")
|
||||
print(" Hidden: 2 neurons (MINIMAL!)")
|
||||
print(" Output: 1 neuron")
|
||||
print(" Activation: Sigmoid (ReLU didn't exist yet!)")
|
||||
print(" Total params: 9 (2×2 weights + 2 bias + 2×1 weights + 1 bias)")
|
||||
|
||||
# Original architecture: 2-2-1 with Sigmoid
|
||||
hidden = Linear(2, 2) # Only 2 hidden neurons!
|
||||
sigmoid_hidden = Sigmoid()
|
||||
output = Linear(2, 1)
|
||||
sigmoid_output = Sigmoid()
|
||||
|
||||
loss_fn = BinaryCrossEntropyLoss()
|
||||
optimizer = SGD([p for p in hidden.parameters()] + [p for p in output.parameters()], lr=1.0)
|
||||
|
||||
print("\n🔥 Training with original 1986 architecture...")
|
||||
epochs = 2000 # May need more epochs with only 2 hidden units
|
||||
|
||||
for epoch in range(epochs):
|
||||
# Forward (all sigmoid, like 1986!)
|
||||
h = hidden(X)
|
||||
h_act = sigmoid_hidden(h) # Sigmoid in hidden layer
|
||||
out = output(h_act)
|
||||
pred = sigmoid_output(out) # Sigmoid in output layer
|
||||
loss = loss_fn(pred, y)
|
||||
|
||||
# Backward
|
||||
loss.backward()
|
||||
|
||||
# Update
|
||||
optimizer.step()
|
||||
optimizer.zero_grad()
|
||||
|
||||
if (epoch + 1) % 400 == 0:
|
||||
accuracy = ((pred.data > 0.5).astype(float) == y.data).mean()
|
||||
print(f"Epoch {epoch+1:4d}/{epochs} Loss: {loss.data:.4f} Accuracy: {accuracy:.1%}")
|
||||
|
||||
# Final evaluation
|
||||
print("\n✅ Final Results:")
|
||||
final_accuracy = ((pred.data > 0.5).astype(float) == y.data).mean()
|
||||
|
||||
for i in range(4):
|
||||
x_in = X_data[i]
|
||||
y_true = int(y_data[i, 0])
|
||||
y_pred_prob = pred.data[i, 0]
|
||||
y_pred = int(y_pred_prob > 0.5)
|
||||
status = "✅" if y_pred == y_true else "❌"
|
||||
print(f" Input: {x_in} → Pred: {y_pred} (prob: {y_pred_prob:.3f}) True: {y_true} {status}")
|
||||
|
||||
print(f"\n📊 Final Accuracy: {final_accuracy:.1%}")
|
||||
print(f"📊 Final Loss: {loss.data:.4f}")
|
||||
|
||||
if final_accuracy == 1.0:
|
||||
print("\n🎉 SUCCESS! XOR solved with MINIMAL 1986 architecture!")
|
||||
print(" This is exactly what ended the AI Winter!")
|
||||
else:
|
||||
print(f"\n⚠️ Accuracy: {final_accuracy:.1%} - may need more training")
|
||||
|
||||
# Show what the hidden units learned
|
||||
print("\n🧠 What the 2 hidden neurons learned:")
|
||||
print(" (Examining activation patterns)")
|
||||
h_activations = sigmoid_hidden(hidden(X)).data
|
||||
print(f"\n Hidden unit activations for each input:")
|
||||
for i, x_in in enumerate(X_data):
|
||||
print(f" {x_in}: h1={h_activations[i,0]:.3f}, h2={h_activations[i,1]:.3f}")
|
||||
|
||||
print("\n" + "=" * 70)
|
||||
print("💡 Historical Note:")
|
||||
print(" This 2-2-1 architecture ended the 17-year AI Winter!")
|
||||
print(" Proved that backprop + hidden layers solve 'impossible' problems")
|
||||
print("=" * 70)
|
||||
2
tinytorch/core/losses.py
generated
2
tinytorch/core/losses.py
generated
@@ -19,8 +19,6 @@ __all__ = ['import_previous_module', 'MSELoss', 'CrossEntropyLoss', 'BinaryCross
|
||||
|
||||
# %% ../../modules/source/04_losses/losses_dev.ipynb 3
|
||||
import numpy as np
|
||||
import matplotlib.pyplot as plt
|
||||
import time
|
||||
from typing import Optional
|
||||
|
||||
def import_previous_module(module_name: str, component_name: str):
|
||||
|
||||
Reference in New Issue
Block a user