LLMs-from-scratch/ch05/11_qwen3

Qwen3 From Scratch

This standalone-qwen3.ipynb Jupyter notebook in this folder contains a from-scratch implementation of Qwen3 0.6B.

 

Using Qwen3 0.6B via the llms-from-scratch package

For an easy way to use the Qwen3 from-scratch implementation, you can also use the llms-from-scratch PyPI package based on the source code in this repository at pkg/llms_from_scratch.

 

1) Installation

pip install llms_from_scratch tokenizers

 

2) Model and text generation settings

Specify which model to use:

USE_REASONING_MODEL = True   # The "thinking" model
USE_REASONING_MODEL = False  # The base model

Basic text generation settings that can be defined by the user. With 150 tokens, the model requires approximately 1.5 GB memory.

MAX_NEW_TOKENS = 150
TEMPERATURE = 0.
TOP_K = 1

 

3) Weight download and loading

This automatically downloads the weight file based on the model choice above:

from llms_from_scratch.qwen3 import download_from_huggingface

repo_id = "rasbt/qwen3-from-scratch"

if USE_REASONING_MODEL:
    filename = "qwen3-0.6B.pth"
    local_dir = "Qwen3-0.6B"    
else:
    filename = "qwen3-0.6B-base.pth"   
    local_dir = "Qwen3-0.6B-Base"

download_from_huggingface(
    repo_id=repo_id,
    filename=filename,
    local_dir=local_dir
)

The model weights are then loaded as follows:

from pathlib import Path
import torch

from llms_from_scratch.qwen3 import Qwen3Model, QWEN_CONFIG_06_B

model_file = Path(local_dir) / filename

model = Qwen3Model(QWEN_CONFIG_06_B)
model.load_state_dict(torch.load(model_file, weights_only=True, map_location="cpu"))

device = (
    torch.device("cuda") if torch.cuda.is_available() else
    torch.device("mps") if torch.backends.mps.is_available() else
    torch.device("cpu")
)
model.to(device)

 

4) Initialize tokenizer

The following code downloads and initializes the tokenizer:

from llms_from_scratch.qwen3 import Qwen3Tokenizer

if USE_REASONING_MODEL:
    tok_filename = "tokenizer.json"    
else:
    tok_filename = "tokenizer-base.json"   

tokenizer = Qwen3Tokenizer(
    tokenizer_file_path=tok_filename,
    repo_id=repo_id,
    add_generation_prompt=USE_REASONING_MODEL,
    add_thinking=USE_REASONING_MODEL
)

 

5) Generating text

Lastly, we can generate text via the following code:

prompt = "Give me a short introduction to large language models."
input_token_ids = tokenizer.encode(prompt)

from llms_from_scratch.ch05 import generate
import time

torch.manual_seed(123)

start = time.time()

output_token_ids = generate(
    model=model,
    idx=torch.tensor(input_token_ids, device=device).unsqueeze(0),
    max_new_tokens=150,
    context_size=QWEN_CONFIG_06_B["context_length"],
    top_k=1,
    temperature=0.
)

total_time = time.time() - start
print(f"Time: {total_time:.2f} sec")
print(f"{int(len(output_token_ids[0])/total_time)} tokens/sec")

if torch.cuda.is_available():
    max_mem_bytes = torch.cuda.max_memory_allocated()
    max_mem_gb = max_mem_bytes / (1024 ** 3)
    print(f"Max memory allocated: {max_mem_gb:.2f} GB")

output_text = tokenizer.decode(output_token_ids.squeeze(0).tolist())

print("\n\nOutput text:\n\n", output_text + "...")

When using the Qwen3 0.6B reasoning model, the output should look similar to the one shown below (this was run on an A100):

Time: 6.35 sec
25 tokens/sec
Max memory allocated: 1.49 GB


Output text:

 <|im_start|>user
Give me a short introduction to large language models.<|im_end|>
Large language models (LLMs) are advanced artificial intelligence systems designed to generate human-like text. They are trained on vast amounts of text data, allowing them to understand and generate coherent, contextually relevant responses. LLMs are used in a variety of applications, including chatbots, virtual assistants, content generation, and more. They are powered by deep learning algorithms and can be fine-tuned for specific tasks, making them versatile tools for a wide range of industries.<|endoftext|>Human resources department of a company is planning to hire 100 new employees. The company has a budget of $100,000 for the recruitment process. The company has a minimum wage of $10 per hour. The company has a total of...

 

Pro tip 1: speed up inference with compilation

For up to a 4× speed-up, replace

model.to(device)

with

model = torch.compile(model)
model.to(device)

Note: There is a significant multi-minute upfront cost when compiling, and the speed-up takes effect after the first generate call.

The following table shows a performance comparison on an A100 for consequent generate calls:

	Tokens/sec	Memory
Qwen3Model	25	1.49 GB
Qwen3Model compiled	107	1.99 GB

 

Pro tip 2: speed up inference with compilation

You can significantly boost inference performance using the KV cache Qwen3Model drop-in replacement when running the model on a CPU. (See my Understanding and Coding the KV Cache in LLMs from Scratch article to learn more about KV caches.)

from llms_from_scratch.kv_cache.qwen3 import Qwen3Model
from llms_from_scratch.kv_cache.generate import generate_text_simple

model = Qwen3Model(QWEN_CONFIG_06_B)
# ...
token_ids = generate_text_simple(
    model=model,
    idx=text_to_token_ids(PROMPT, tokenizer).to(device),
    max_new_tokens=MAX_NEW_TOKENS,
    context_size=QWEN_CONFIG_06_B["context_length"],
)

Note that the peak memory usage is only listed for Nvidia CUDA devices, as it is easier to calculate. However, the memory usage on other devices is likely similar as it uses a similar precision format, and the KV cache storage dominates here for the generated 150-token text (however, different devices may implement matrix multiplication differently and may result in different peak memory requirements).

Model	Mode	Hardware	Tokens/sec	GPU Memory (VRAM)
Qwen3Model	Regular	Mac Mini M4 CPU	1	-
Qwen3Model	Regular compiled	Mac Mini M4 CPU	-	-
Qwen3Model	KV cache	Mac Mini M4 CPU	80	-
Qwen3Model	KV cache compiled	Mac Mini M4 CPU	-	-
Qwen3Model	Regular	Mac Mini M4 GPU	21	-
Qwen3Model	Regular compiled	Mac Mini M4 GPU	-	-
Qwen3Model	KV cache	Mac Mini M4 GPU	32	-
Qwen3Model	KV cache compiled	Mac Mini M4 GPU	-	-
Qwen3Model	Regular	Nvidia A100 GPU	25	1.49 GB
Qwen3Model	Regular compiled	Nvidia A100 GPU	107	1.99 GB
Qwen3Model	KV cache	Nvidia A100 GPU	25	10.20 GB
Qwen3Model	KV cache compiled	Nvidia A100 GPU	24	10.61 GB

Note that all settings above have been tested to produce the same text outputs.