Temperature and sampling in NLP - Model Pipeline Trace

Model Pipeline - Temperature and sampling

This pipeline shows how temperature and sampling affect text generation in language models. Temperature controls randomness, and sampling picks the next word based on probabilities.

Data Flow - 6 Stages

1Input Text

1 sentence (variable length)→User provides a starting sentence or prompt→1 sentence (variable length)

"The weather today is"

↓

2Tokenization

1 sentence (variable length)→Split sentence into tokens (words or subwords)→1 sequence x 4 tokens

["The", "weather", "today", "is"]

↓

3Model Prediction

1 sequence x 4 tokens→Model predicts next word probabilities→1 sequence x vocabulary size (e.g., 50,000)

{"sunny": 0.3, "rainy": 0.2, "cloudy": 0.1, ...}

↓

4Apply Temperature

1 sequence x vocabulary size→Adjust probabilities by temperature to control randomness→1 sequence x vocabulary size

Temperature=0.5 makes distribution sharper; Temperature=1.5 makes it flatter

↓

5Sampling

1 sequence x vocabulary size→Randomly pick next word based on adjusted probabilities→1 token

"sunny"

↓

6Output Text

1 token→Add chosen token to sentence→1 sentence (variable length + 1 token)

"The weather today is sunny"

Training Trace - Epoch by Epoch

Loss
2.5 |****
2.0 |***
1.5 |**
1.0 |*
0.5 |
    +------------
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	2.5	0.30	Model starts learning word patterns with high loss and low accuracy
2	1.8	0.45	Loss decreases and accuracy improves as model learns better predictions
3	1.3	0.60	Model shows steady improvement in predicting next words
4	1.0	0.70	Loss continues to decrease; model becomes more confident
5	0.8	0.78	Training converges with good accuracy and low loss

Prediction Trace - 5 Layers

Layer 1: Tokenization

Layer 2: Model Prediction

Layer 3: Apply Temperature (T=0.5)

Layer 4: Sampling

Layer 5: Output Text

Model Quiz - 3 Questions

What does lowering the temperature value do to the word probabilities?

A Makes the distribution sharper, favoring high-probability words

B Makes the distribution flatter, increasing randomness

C Removes low-probability words completely

D Does not affect the probabilities

Key Insight

Temperature controls how creative or predictable the model's text is by adjusting word choice randomness. Sampling uses these adjusted probabilities to generate varied and interesting text outputs.

Practice

1. What does increasing the temperature parameter in text generation usually do?

easy

A. Makes the output more predictable and repetitive

B. Stops the model from generating any text

C. Makes the output more random and creative

D. Always selects the most probable next word

Solution

1. Step 1: Understand temperature effect on randomness

   Temperature controls how much randomness is added to the word selection process in text generation.

2. Step 2: Relate temperature to creativity

   Higher temperature increases randomness, making the output more creative and less predictable.

3. Final Answer:

   Makes the output more random and creative -> Option C

4. Quick Check:

   Higher temperature = more randomness [OK]

Hint: Higher temperature means more randomness in output [OK]

Common Mistakes:

• Thinking higher temperature makes output more predictable
• Confusing temperature with model size
• Assuming temperature stops generation

2. Which of the following code snippets correctly applies temperature scaling to logits before sampling in Python?

easy

A. probs = softmax(logits / temperature)

B. probs = softmax(logits * temperature)

C. probs = softmax(logits + temperature)

D. probs = softmax(logits - temperature)

Solution

1. Step 1: Recall temperature scaling formula

   Temperature is applied by dividing logits by temperature before softmax to adjust randomness.

2. Step 2: Identify correct operation

   Dividing logits by temperature scales the logits correctly; multiplying or adding is incorrect.

3. Final Answer:

   probs = softmax(logits / temperature) -> Option A

4. Quick Check:

   Divide logits by temperature before softmax [OK]

Hint: Divide logits by temperature before softmax [OK]

Common Mistakes:

• Multiplying logits by temperature instead of dividing
• Adding temperature to logits
• Subtracting temperature from logits

3. Given logits = [2.0, 1.0, 0.1] and temperature = 0.5, what is the approximate probability of the first token after applying softmax with temperature scaling?

medium

A. About 0.30

B. About 0.60

C. About 0.50

D. About 0.84

Solution

1. Step 1: Scale logits by dividing by temperature

   Divide each logit by 0.5: [2.0/0.5=4.0, 1.0/0.5=2.0, 0.1/0.5=0.2]

2. Step 2: Calculate softmax probabilities

   Compute exp values: exp(4.0)=54.6, exp(2.0)=7.39, exp(0.2)=1.22; sum=63.21; probability first token = 54.6/63.21 ≈ 0.86 (approx 0.86 considering rounding)

3. Final Answer:

   About 0.86 -> Option D

4. Quick Check:

   Lower temperature sharpens distribution, first token ~0.86 [OK]

Hint: Divide logits by temperature, then softmax to find probabilities [OK]

Common Mistakes:

• Multiplying logits by temperature instead of dividing
• Skipping exponentiation step
• Using temperature incorrectly in softmax

4. A developer writes this code to sample a token with temperature 1.5 but always gets the same token. What is the likely bug?

scaled_logits = logits * temperature
probs = softmax(scaled_logits)
sampled_token = sample_from(probs)

medium

A. They should divide logits by temperature, not multiply

B. They forgot to apply softmax

C. Temperature should be zero to get randomness

D. Sampling function is incorrect

Solution

1. Step 1: Identify temperature scaling mistake

   The code multiplies logits by temperature, which is incorrect; it should divide logits by temperature.

2. Step 2: Explain effect of wrong scaling

   Multiplying by temperature >1 increases logits, making softmax peakier and less random, causing same token output.

3. Final Answer:

   They should divide logits by temperature, not multiply -> Option A

4. Quick Check:

   Divide logits by temperature for correct scaling [OK]

Hint: Divide, don't multiply logits by temperature [OK]

Common Mistakes:

• Multiplying instead of dividing logits
• Setting temperature to zero
• Ignoring softmax step

5. You want to generate text that balances creativity and coherence. Which temperature value and sampling strategy combination is best?

hard

A. Temperature 0.1 with greedy sampling

B. Temperature around 0.7 with top-k sampling

C. Temperature 2.0 with random sampling

D. Temperature 1.5 with no sampling (always pick max)

Solution

1. Step 1: Understand temperature impact on creativity

   Temperature ~0.7 balances randomness and predictability, avoiding too repetitive or too random output.

2. Step 2: Choose sampling method for balance

   Top-k sampling limits choices to top probable tokens, improving coherence while allowing creativity.

3. Final Answer:

   Temperature around 0.7 with top-k sampling -> Option B

4. Quick Check:

   Moderate temperature + top-k = balanced creativity [OK]

Hint: Use moderate temperature and top-k for balanced text [OK]

Common Mistakes:

• Using very low temperature causing boring text
• Using very high temperature causing nonsense
• Ignoring sampling method effects