SFT-Qwen3-8B-alpaca

An 8 B Qwen3-8B-Base model instruction-tuned with Alpaca-format SFT.

📦 Model overview

🚀 Quick start

from transformers import AutoModelForCausalLM, AutoTokenizer

repo  = "wesjos/SFT-Qwen3-8B-alpaca"
tok   = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, device_map="auto")

PROMPT = """Below is an instruction that describes a task. Write a response that \
appropriately completes the request.

### Instruction:
{instruction}

### Input:
{input}

### Response:
"""

text = PROMPT.format(instruction="Explain what a GPU is to a 10-year-old child.", input="")
ids  = tok(text, return_tensors="pt").to(model.device)
out  = model.generate(**ids, max_new_tokens=256, do_sample=False)
print(tok.decode(out[0][ids["input_ids"].shape[1]:], skip_special_tokens=True))

Or the one-liner with pipeline

from transformers import pipeline
p = pipeline("text-generation", model="wesjos/SFT-Qwen3-8B-alpaca")
print(p(PROMPT.format(instruction="Explain what a GPU is to a 10-year-old child.",
                       input=""), max_new_tokens=256, return_full_text=False)[0]["generated_text"])

🔍 Sample outputs

Greedy decoding (do_sample=False), max_new_tokens=300, one Tesla T4, measured 2026-10-04. Raw transcripts — nothing hand-edited.

Explain a GPU to a 10-year-old

Prompt

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
Explain what a GPU is to a 10-year-old child.

### Input:


### Response:

GPU是电脑的一个小部件，就像电脑的“大脑”一样。它帮助电脑快速地处理图像、视频和游戏。就像一个超级快的画家，GPU可以快速地画出电脑屏幕上看到的图像。如果没有GPU，电脑将无法快速地显示图像，游戏将变得缓慢，视频将变得模糊。GPU就像电脑的超级画家，帮助它快速地画出图像。

_{83 tokens · 8.4 s}

Write reverse_linked_list() in Python

Prompt

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
Write a Python function that reverses a singly linked list. Include a docstring.

### Input:


### Response:

```python
class Node:
    def __init__(self, data):
        self.data = data
        self.next = None

def reverse_linked_list(head):
    """
    Reverses a singly linked list.
    
    Args:
        head (Node): The head of the linked list.
        
    Returns:
        Node: The new head of the reversed linked list.
    """
    prev = None
    current = head
    while current:
        next_node = current.next
        current.next = prev
        prev = current
        current = next_node
    return prev
```

_{122 tokens · 12.2 s}

Word problem: 17 sheep, all but 9 run away

Prompt

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
A farmer has 17 sheep. All but 9 run away. How many are left? Explain your reasoning.

### Input:


### Response:

一个农民有17只羊。除了9只，其他都逃走了。那么，剩下的羊的数量是17-9=8只。因此，答案是8只羊。

_{40 tokens · 4.4 s}

One-sentence summary of the Voyager 1 passage

Prompt

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
Summarize the following passage in one sentence.

### Input:
The Voyager 1 spacecraft, launched in 1977, has travelled farther from Earth than any other human-made object. It crossed into interstellar space in 2012 and continues to transmit data using power from its radioisotope thermoelectric generators, though mission engineers expect its instruments to be shut down sometime in the 2030s.

### Response:

Voyager 1，于1977年发射，是人类制造的最远的物体，于2012年进入星际空间，预计其仪器将在2030年代关闭。

_{45 tokens · 5.4 s}

中文 · 用三句话解释「过拟合」

Prompt

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
用三句话向一个刚学编程的大学生解释什么是「过拟合」。

### Input:


### Response:

过拟合是指机器学习模型过于复杂，以至于它过度适应了训练数据，导致在新数据上表现不佳。这通常发生在模型过于复杂，能够完美地拟合训练数据，但无法推广到新数据。过拟合可以通过使用正则化技术、交叉验证和早停等技术来避免。

_{69 tokens · 7.4 s}

English · one concrete anti-procrastination tactic

Prompt

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
I keep procrastinating on my thesis. Give me one concrete tactic, not generic advice.

### Input:


### Response:

一个具体的策略是将你的论文分解成更小、可管理的任务。例如，你可以将你的论文分成几个章节，然后为每个章节设定一个截止日期。一旦你完成了每个章节，你可以继续下一个。这将使你的论文感觉不那么压倒人，更容易开始。此外，你可以使用番茄工作法，即工作25分钟，休息5分钟，以保持专注和动力。这将帮助你保持在正确的轨道上，避免拖延。

_{98 tokens · 10.4 s}

🆚 Side-by-side with the original Qwen/Qwen3-8B-Base

Identical prompt string, identical decoding settings, identical hardware — first block is this model, second is the un-tuned original.

Explain a GPU to a 10-year-old

SFT-Qwen3-8B-alpaca

GPU是电脑的一个小部件，就像电脑的“大脑”一样。它帮助电脑快速地处理图像、视频和游戏。就像一个超级快的画家，GPU可以快速地画出电脑屏幕上看到的图像。如果没有GPU，电脑将无法快速地显示图像，游戏将变得缓慢，视频将变得模糊。GPU就像电脑的超级画家，帮助它快速地画出图像。

Qwen/Qwen3-8B-Base

A GPU is like a super-fast helper for your computer. Imagine your computer is a person, and the GPU is like a really good athlete. When the computer needs to do lots of math or draw pictures really fast, the GPU steps in and does it super quickly. This helps games look super cool and makes videos play smoothly. So, the GPU is like the star player on the computer's team, making everything run fast and look great!

Write reverse_linked_list() in Python

SFT-Qwen3-8B-alpaca

```python
class Node:
    def __init__(self, data):
        self.data = data
        self.next = None

def reverse_linked_list(head):
    """
    Reverses a singly linked list.
    
    Args:
        head (Node): The head of the linked list.
        
    Returns:
        Node: The new head of the reversed linked list.
    """
    prev = None
    current = head
    while current:
        next_node = current.next
        current.next = prev
        prev = current
        current = next_node
    return prev
```

Qwen/Qwen3-8B-Base

Here is a Python function that reverses a singly linked list:

```python
class Node:
    def __init__(self, data=None, next=None):
        self.data = data
        self.next = next

def reverse_linked_list(head):
    """
    Reverses a singly linked list.

    Args:
        head (Node): The head of the linked list.

    Returns:
        Node: The new head of the reversed linked list.
    """
    prev = None
    current = head
    while current is not None:
        next_node = current.next
        current.next = prev
        prev = current
        current = next_node
    return prev
```

This function takes a `Node` object representing the head of a singly linked list as input and returns a new `Node` object representing the head of the reversed linked list. The function uses a simple iterative approach to reverse the linked list by updating the `next` pointer of each node to point to its
…(truncated)

Word problem: 17 sheep, all but 9 run away

SFT-Qwen3-8B-alpaca

一个农民有17只羊。除了9只，其他都逃走了。那么，剩下的羊的数量是17-9=8只。因此，答案是8只羊。

Qwen/Qwen3-8B-Base

A farmer has 17 sheep. All but 9 run away. How many are left? Explain your reasoning.

### Response:
The farmer has 9 sheep left. 

Here's the reasoning: The phrase "all but 9" means that 9 sheep did not run away. Therefore, the number of sheep left is 9.

One-sentence summary of the Voyager 1 passage

SFT-Qwen3-8B-alpaca

Voyager 1，于1977年发射，是人类制造的最远的物体，于2012年进入星际空间，预计其仪器将在2030年代关闭。

Qwen/Qwen3-8B-Base

The Voyager 1 spacecraft, launched in 1977, has traveled farther from Earth than any other human-made object, crossed into interstellar space in 2012, and continues to transmit data using its radioisotope thermoelectric generators, with mission engineers expecting its instruments to be shut down in the 2030s.

中文 · 用三句话解释「过拟合」

SFT-Qwen3-8B-alpaca

过拟合是指机器学习模型过于复杂，以至于它过度适应了训练数据，导致在新数据上表现不佳。这通常发生在模型过于复杂，能够完美地拟合训练数据，但无法推广到新数据。过拟合可以通过使用正则化技术、交叉验证和早停等技术来避免。

Qwen/Qwen3-8B-Base

过拟合是指模型在训练数据上表现很好，但在新数据上表现不佳的现象。简单来说，就是模型记住了训练数据的细节，而不是学习到数据的规律。为了避免过拟合，可以使用更多的数据、简化模型或使用正则化技术。

English · one concrete anti-procrastination tactic

SFT-Qwen3-8B-alpaca

一个具体的策略是将你的论文分解成更小、可管理的任务。例如，你可以将你的论文分成几个章节，然后为每个章节设定一个截止日期。一旦你完成了每个章节，你可以继续下一个。这将使你的论文感觉不那么压倒人，更容易开始。此外，你可以使用番茄工作法，即工作25分钟，休息5分钟，以保持专注和动力。这将帮助你保持在正确的轨道上，避免拖延。

Qwen/Qwen3-8B-Base

One concrete tactic to help you overcome procrastination on your thesis is to **set a specific, time-bound task for each study session**. For example, instead of saying, "I’ll work on my thesis today," commit to writing 500 words or completing a specific section of your thesis within a set time frame, like 90 minutes. This creates a clear, achievable goal that reduces overwhelm and keeps you focused.

📋 Version info

Citation

@misc{SFTQwen38Balpaca}
  title = {SFT-Qwen3-8B-alpaca},
  note  = {Fine-tuned from Qwen/Qwen3-8B-Base},
  year  = {2026}
}