README.md

---
license: apache-2.0
base_model: Ilides/coser-1-by-ilides
tags:
  - coser-1.1-code
  - ilides
  - coding-assistant
  - qwen3.5
  - lora
language:
  - en
pipeline_tag: text-generation
---

# Coser 1.1-code by ilides (GGUF)

**Coser 1.1-code** es la evolución de [Coser 1](https://huggingface.co/Ilides/coser-1-by-ilides), fine-tuned con **2,000 ejemplos de código** de datasets públicos (Code-Feedback, CodeAlpaca, Magicoder, CodeInstruct).

Publicado por **[ilides](https://huggingface.co/Ilides)**.

## Versiones disponibles

| Repositorio | Formato | Uso |
|-------------|---------|-----|
| [Ilides/coser-1.1-code](https://huggingface.co/Ilides/coser-1.1-code) | Safetensors (HF) | Transformers, fine-tuning |
| [Ilides/coser-1.1-code-GGUF](https://huggingface.co/Ilides/coser-1.1-code-GGUF) | GGUF Q8_0 / F16 | llama.cpp, LM Studio |

## Stats de entrenamiento

| Métrica | Valor |
|---------|-------|
| Base | Coser 1 by ilides (Qwen3.5-0.8B) |
| Dataset | 2,000 ejemplos código |
| Método | LoRA r=16 + QLoRA 4-bit |
| Steps | 181 |
| Loss | 0.8206850051879883 |
| Token accuracy | 79.2% |
| Tiempo | None min |
| GPU | NVIDIA GeForce RTX 3050 |

## Benchmark real (NVIDIA GeForce RTX 3050)

| Prompt | tok/s |
|--------|-------|
| Write a Python function that reverses a linked lis... | 17.0 |
| Write a JavaScript async function to fetch and par... | 22.0 |
| Explain what binary search is and write it in Pyth... | 21.5 |
| Write a SQL query to find duplicate emails in a us... | 19.4 |
| Fix this bug: my Python function returns None inst... | 21.8 |
| **Promedio** | **20.3** |

## Ejemplos reales

### Write a Python function that reverses a linked list.

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

def reverse_linked_list(head):
    prev = None
    curr = head
    while curr:
        next_node = curr.next
        prev.next = curr
        curr.next = prev
        prev = curr
        curr = next_node
    return prev
```
```

### Write a JavaScript async function to fetch and parse JSON from an API.

```
```javascript
async function fetchJsonAPI(url) {
  // Fetch the JSON data from the API
  const response = await fetch(url);
  const data = await response.json();

  // Handle any potential errors or exceptions
  if (!data) {
    console.error("Error fetching JSON from API");
    return null;
  }

  // Process the JSON data as needed
  // For example, convert to a dictionary, parse a JSON object, or store in an array
  // Process the data here
  // ...

  return data;
}
```
```

### Explain what binary search is and write it in Python.

```
Binary search is a search algorithm that takes an array of elements and returns the index of the target element in the array. It has the following properties:

1. It operates on the sorted elements of the array.
2. It has a time complexity of O(log n), which is very efficient.
3. It uses two pointers, one starting at the beginning of the array and the other at the end.

In Python, we can use the built-in `bisect` module to implement binary search. The `bisect.bisect_left` function returns the insertion point for a given value in a sorted list, and the `bisect.bisect_right` function returns the insertion point for a given value just after the insertion point of the previous value.

Here is a Python code snippet that demonstrates binary search using the `bisect` module:

```python
import bisect

# Given sorted array
arr = [1, 3, 5, 7, 9, 11, 13, 15, 17, 19]

# Search for a value in the arr
```


## Uso (Transformers)

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "Ilides/coser-1.1-code"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id, device_map="auto", trust_remote_code=True, torch_dtype=torch.bfloat16
)
messages = [
    {"role": "system", "content": "You are Coser 1.1-code by ilides, an expert AI coding assistant."},
    {"role": "user", "content": "Write a Python function to reverse a string."},
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=256, temperature=0.7)
print(tokenizer.decode(out[0], skip_special_tokens=True))
```

## Créditos

- Base: [Ilides/coser-1-by-ilides](https://huggingface.co/Ilides/coser-1-by-ilides)
- Datasets: Code-Feedback, python_code_instructions, CodeInstruct-20K, magicoder-python-5k
- Autor: ilides