Module 5 · Language as probability · train0.py · ~25 min

5.2 Temperature

Always pick the most likely next letter and you get the name “a”, forever. Pick letters completely at random and you get “xqzv”. Every language model, ChatGPT included, has a dial between those two extremes. It’s called temperature.

1Watch

Your counting model from 5.1 rolls a weighted die built from real counts. Temperature reshapes that die before you roll it: raise every count to the power 1 / T, then divide by the new total so the chances still add up to 1.

T = 1: nothing changes. The die is exactly the counts.
T below 1 (cold): big counts get bigger relative to small ones. At T = 0.25, “a” starts 64% of names instead of 13.8%. Safe, repetitive.
T above 1 (hot): everything gets flattened toward equal. At T = 5 no first letter is above 6%. Creative, then chaotic.

Why does a power sharpen? Take a row with counts 4 and 1, so the favourite has 4 / 5 = 80%. At T = 0.5 the power is 1 / 0.5 = 2: the counts become 4² = 16 and 1² = 1, and the favourite now has 16 / 17 ≈ 94%. A power above 1 stretches the gap between big and small; a power below 1 (hot) squashes it.

T can’t be exactly 0 (you can’t divide by 0), but as T shrinks toward 0 the favourite takes everything: that’s the “always pick the most likely letter” end of the dial.

Nothing about the model changes: same counts, same knowledge. Only the way you choose from its predictions changes.

2Explore

Loading 32,033 names…

3Build

Fill the two TODOs in chances():

Turn each count into a weight by raising it to the power 1 / T. In Python ** is the power, and Python does it before any division, so watch the brackets.
Show me the line
weights = [counts[(current, b)] ** (1 / T) for b in options]
Set total to the sum of the weights, so the chances add up to 1.

Run it and compare the names at T = 0.3, 1 and 3.

import random

docs = [line.strip() for line in open('names.txt') if line.strip()]

def count_pairs(docs):  # from lesson 5.1
    counts = {}
    for name in docs:
        chars = '.' + name + '.'
        for a, b in zip(chars, chars[1:]):
            counts[(a, b)] = counts.get((a, b), 0) + 1
    return counts

counts = count_pairs(docs)


# Turn one row of counts into chances, reshaped by the temperature dial.
# Each count is raised to the power 1/T, then everything is divided by the
# total so the chances add up to 1.
def chances(counts, current, T):
    options = [b for (a, b) in counts if a == current]
    weights = [counts[(current, b)] for b in options]  # TODO: raise each count to the power (1 / T), brackets included
    total = 1  # TODO: the sum of the weights
    # {key: value for ...} builds a dict the way [... for ...] builds a list.
    return {b: w / total for b, w in zip(options, weights)}


def sample_name(counts, T, rng=random):
    name, current = '', '.'
    while True:
        row = chances(counts, current, T)
        nxt = rng.choices(list(row), weights=list(row.values()))[0]
        if nxt == '.':
            return name
        name += nxt
        current = nxt


for T in [0.3, 1, 3]:
    rng = random.Random(5)
    print(f'T={T}:', [sample_name(counts, T, rng) for _ in range(6)])

4Check yourself

1. You want a chatbot to give almost the same careful answer every time. Which temperature?

2. Predict: at T = 1, what is the chance a name starts with “a”?

3. Does changing the temperature change what the model has learned?

5Unlocked in microgpt

Line 187 sets temperature = 0.5: microgpt samples on the cool, careful side. Line 195 divides every score by the temperature before turning scores into chances (you’ll meet scores, called logits, and build that “softmax” step in lesson 6.4). Dividing scores by T does exactly what raising counts to 1/T does, if the scores are the logs of the counts: exp(log c / T) = c^1/T.

Notice the comment on line 187: it says the temperature should be “in (0, 1]”. Values above 1 still work mathematically, as the dial showed, but for names they mostly produce gibberish, so microgpt stays at or below 1.

Progress is saved in this browser. Sign in to keep it across devices.

microgpt.py60 / 175 lines learned

1"""

2The most atomic way to train and run inference for a GPT in pure, dependency-free Python.

3This file is the complete algorithm.

4Everything else is just efficiency.

6@karpathy

7"""

9import os # os.path.exists

10import math # math.log, math.exp

11import random # random.seed, random.choices, random.gauss, random.shuffle

12random.seed(42) # Let there be order among chaos

14# Let there be a Dataset `docs`: list[str] of documents (e.g. a list of names)

15if not os.path.exists('input.txt'):

16 import urllib.request

17 names_url = 'https://raw.githubusercontent.com/karpathy/makemore/988aa59/names.txt'

18 urllib.request.urlretrieve(names_url, 'input.txt')

19docs = [line.strip() for line in open('input.txt') if line.strip()]

20random.shuffle(docs)

21print(f"num docs: {len(docs)}")

23# Let there be a Tokenizer to translate strings to sequences of integers ("tokens") and back

24uchars = sorted(set(''.join(docs))) # unique characters in the dataset become token ids 0..n-1

25BOS = len(uchars) # token id for a special Beginning of Sequence (BOS) token

26vocab_size = len(uchars) + 1 # total number of unique tokens, +1 is for BOS

27print(f"vocab size: {vocab_size}")

29# Let there be Autograd to recursively apply the chain rule through a computation graph

30class Value:

31 __slots__ = ('data', 'grad', '_children', '_local_grads') # Python optimization for memory usage

33 def __init__(self, data, children=(), local_grads=()):

34 self.data = data # scalar value of this node calculated during forward pass

35 self.grad = 0 # derivative of the loss w.r.t. this node, calculated in backward pass

36 self._children = children # children of this node in the computation graph

37 self._local_grads = local_grads # local derivative of this node w.r.t. its children

39 def __add__(self, other):

40 other = other if isinstance(other, Value) else Value(other)

41 return Value(self.data + other.data, (self, other), (1, 1))

43 def __mul__(self, other):

44 other = other if isinstance(other, Value) else Value(other)

45 return Value(self.data * other.data, (self, other), (other.data, self.data))

47 def __pow__(self, other): return Value(self.data**other, (self,), (other * self.data**(other-1),))

48 def log(self): return Value(math.log(self.data), (self,), (1/self.data,))

49 def exp(self): return Value(math.exp(self.data), (self,), (math.exp(self.data),))

50 def relu(self): return Value(max(0, self.data), (self,), (float(self.data > 0),))

51 def __neg__(self): return self * -1

52 def __radd__(self, other): return self + other

53 def __sub__(self, other): return self + (-other)

54 def __rsub__(self, other): return other + (-self)

55 def __rmul__(self, other): return self * other

56 def __truediv__(self, other): return self * other**-1

57 def __rtruediv__(self, other): return other * self**-1

59 def backward(self):

60 topo = []

61 visited = set()

62 def build_topo(v):

63 if v not in visited:

64 visited.add(v)

65 for child in v._children:

66 build_topo(child)

67 topo.append(v)

68 build_topo(self)

69 self.grad = 1

70 for v in reversed(topo):

71 for child, local_grad in zip(v._children, v._local_grads):

72 child.grad += local_grad * v.grad

74# Initialize the parameters, to store the knowledge of the model

75n_layer = 1 # depth of the transformer neural network (number of layers)

76n_embd = 16 # width of the network (embedding dimension)

77block_size = 16 # maximum context length of the attention window (note: the longest name is 15 characters)

78n_head = 4 # number of attention heads

79head_dim = n_embd // n_head # derived dimension of each head

80matrix = lambda nout, nin, std=0.08: [[Value(random.gauss(0, std)) for _ in range(nin)] for _ in range(nout)]

81state_dict = {'wte': matrix(vocab_size, n_embd), 'wpe': matrix(block_size, n_embd), 'lm_head': matrix(vocab_size, n_embd)}

82for i in range(n_layer):

83 state_dict[f'layer{i}.attn_wq'] = matrix(n_embd, n_embd)

84 state_dict[f'layer{i}.attn_wk'] = matrix(n_embd, n_embd)

85 state_dict[f'layer{i}.attn_wv'] = matrix(n_embd, n_embd)

86 state_dict[f'layer{i}.attn_wo'] = matrix(n_embd, n_embd)

87 state_dict[f'layer{i}.mlp_fc1'] = matrix(4 * n_embd, n_embd)

88 state_dict[f'layer{i}.mlp_fc2'] = matrix(n_embd, 4 * n_embd)

89params = [p for mat in state_dict.values() for row in mat for p in row] # flatten params into a single list[Value]

90print(f"num params: {len(params)}")

92# Define the model architecture: a function mapping tokens and parameters to logits over what comes next

93# Follow GPT-2, blessed among the GPTs, with minor differences: layernorm -> rmsnorm, no biases, GeLU -> ReLU

94def linear(x, w):

95 return [sum(wi * xi for wi, xi in zip(wo, x)) for wo in w]

97def softmax(logits):

98 max_val = max(val.data for val in logits)

99 exps = [(val - max_val).exp() for val in logits]

100 total = sum(exps)

101 return [e / total for e in exps]

102

103def rmsnorm(x):

104 ms = sum(xi * xi for xi in x) / len(x)

105 scale = (ms + 1e-5) ** -0.5

106 return [xi * scale for xi in x]

107

108def gpt(token_id, pos_id, keys, values):

109 tok_emb = state_dict['wte'][token_id] # token embedding

110 pos_emb = state_dict['wpe'][pos_id] # position embedding

111 x = [t + p for t, p in zip(tok_emb, pos_emb)] # joint token and position embedding

112 x = rmsnorm(x) # note: not redundant due to backward pass via the residual connection

113

114 for li in range(n_layer):

115 # 1) Multi-head Attention block

116 x_residual = x

117 x = rmsnorm(x)

118 q = linear(x, state_dict[f'layer{li}.attn_wq'])

119 k = linear(x, state_dict[f'layer{li}.attn_wk'])

120 v = linear(x, state_dict[f'layer{li}.attn_wv'])

121 keys[li].append(k)

122 values[li].append(v)

123 x_attn = []

124 for h in range(n_head):

125 hs = h * head_dim

126 q_h = q[hs:hs+head_dim]

127 k_h = [ki[hs:hs+head_dim] for ki in keys[li]]

128 v_h = [vi[hs:hs+head_dim] for vi in values[li]]

129 attn_logits = [sum(q_h[j] * k_h[t][j] for j in range(head_dim)) / head_dim**0.5 for t in range(len(k_h))]

130 attn_weights = softmax(attn_logits)

131 head_out = [sum(attn_weights[t] * v_h[t][j] for t in range(len(v_h))) for j in range(head_dim)]

132 x_attn.extend(head_out)

133 x = linear(x_attn, state_dict[f'layer{li}.attn_wo'])

134 x = [a + b for a, b in zip(x, x_residual)]

135 # 2) MLP block

136 x_residual = x

137 x = rmsnorm(x)

138 x = linear(x, state_dict[f'layer{li}.mlp_fc1'])

139 x = [xi.relu() for xi in x]

140 x = linear(x, state_dict[f'layer{li}.mlp_fc2'])

141 x = [a + b for a, b in zip(x, x_residual)]

142

143 logits = linear(x, state_dict['lm_head'])

144 return logits

145

146# Let there be Adam, the blessed optimizer and its buffers

147learning_rate, beta1, beta2, eps_adam = 0.01, 0.85, 0.99, 1e-8

148m = [0.0] * len(params) # first moment buffer

149v = [0.0] * len(params) # second moment buffer

150

151# Repeat in sequence

152num_steps = 1000 # number of training steps

153for step in range(num_steps):

154

155 # Take single document, tokenize it, surround it with BOS special token on both sides

156 doc = docs[step % len(docs)]

157 tokens = [BOS] + [uchars.index(ch) for ch in doc] + [BOS]

158 n = min(block_size, len(tokens) - 1)

159

160 # Forward the token sequence through the model, building up the computation graph all the way to the loss

161 keys, values = [[] for _ in range(n_layer)], [[] for _ in range(n_layer)]

162 losses = []

163 for pos_id in range(n):

164 token_id, target_id = tokens[pos_id], tokens[pos_id + 1]

165 logits = gpt(token_id, pos_id, keys, values)

166 probs = softmax(logits)

167 loss_t = -probs[target_id].log()

168 losses.append(loss_t)

169 loss = (1 / n) * sum(losses) # final average loss over the document sequence. May yours be low.

170

171 # Backward the loss, calculating the gradients with respect to all model parameters

172 loss.backward()

173

174 # Adam optimizer update: update the model parameters based on the corresponding gradients

175 lr_t = learning_rate * (1 - step / num_steps) # linear learning rate decay

176 for i, p in enumerate(params):

177 m[i] = beta1 * m[i] + (1 - beta1) * p.grad

178 v[i] = beta2 * v[i] + (1 - beta2) * p.grad ** 2

179 m_hat = m[i] / (1 - beta1 ** (step + 1))

180 v_hat = v[i] / (1 - beta2 ** (step + 1))

181 p.data -= lr_t * m_hat / (v_hat ** 0.5 + eps_adam)

182 p.grad = 0

183

184 print(f"step {step+1:4d} / {num_steps:4d} | loss {loss.data:.4f}", end='\r')

185

186# Inference: may the model babble back to us

187temperature = 0.5 # in (0, 1], control the "creativity" of generated text, low to high

188print("\n--- inference (new, hallucinated names) ---")

189for sample_idx in range(20):

190 keys, values = [[] for _ in range(n_layer)], [[] for _ in range(n_layer)]

191 token_id = BOS

192 sample = []

193 for pos_id in range(block_size):

194 logits = gpt(token_id, pos_id, keys, values)

195 probs = softmax([l / temperature for l in logits])

196 token_id = random.choices(range(vocab_size), weights=[p.data for p in probs])[0]

197 if token_id == BOS:

198 break

199 sample.append(uchars[token_id])

200 print(f"sample {sample_idx+1:2d}: {''.join(sample)}")

this lessonlearnedKarpathy’s original

Was this lesson clear?

Next available lesson → 5.3 Loss = surprise