122 lines
4.4 KiB
Markdown
122 lines
4.4 KiB
Markdown
---
|
|
license: apache-2.0
|
|
language:
|
|
- en
|
|
base_model:
|
|
- prithivMLmods/Qwen3-4B-ft-bf16
|
|
pipeline_tag: text-generation
|
|
library_name: transformers
|
|
tags:
|
|
- RL
|
|
- text-generation-inference
|
|
- blitzar
|
|
- coder
|
|
- trl
|
|
- code
|
|
datasets:
|
|
- livecodebench/code_generation_lite
|
|
- PrimeIntellect/verifiable-coding-problems
|
|
- likaixin/TACO-verified
|
|
- open-r1/codeforces-cots
|
|
---
|
|
|
|
|
|
|
|
# **Blitzar-Coder-4B-F.1**
|
|
|
|
> **Blitzar-Coder-4B-F.1** is a high-efficiency, multi-language coding model fine-tuned on **Qwen3-4B** using **larger coding traces datasets** spanning **10+ programming languages** including Python, Java, C#, C++, C, Go, JavaScript, TypeScript, Rust, and more. This model delivers exceptional code generation, debugging, and reasoning capabilities—making it an ideal tool for developers seeking advanced programming assistance under constrained compute.
|
|
|
|
> \[!note]
|
|
> GGUF: [https://huggingface.co/prithivMLmods/Blitzar-Coder-4B-F.1-GGUF](https://huggingface.co/prithivMLmods/Blitzar-Coder-4B-F.1-GGUF)
|
|
|
|
---
|
|
|
|
## **Key Features**
|
|
|
|
1. **Multi-Language Code Mastery**
|
|
Fine-tuned on **extensive coding traces datasets** covering **10+ programming languages** (Python, Java, C#, C++, C, Go, JavaScript, TypeScript, Rust, Swift, Kotlin, and more), enabling cross-language development and translation.
|
|
|
|
2. **Advanced Code Generation & Reasoning**
|
|
Supports complex algorithm synthesis, code optimization, debugging workflows, and architectural design patterns across multiple paradigms—from systems programming to web development.
|
|
|
|
3. **Cross-Language Development Support**
|
|
Seamlessly handles polyglot codebases, API integrations, and framework-specific implementations while maintaining language-specific best practices and idioms.
|
|
|
|
4. **Intelligent Code Analysis**
|
|
Performs code reviews, identifies performance bottlenecks, suggests refactoring opportunities, and provides detailed explanations for complex programming concepts.
|
|
|
|
5. **Structured Output for Development**
|
|
Generates clean code documentation, API specifications, configuration files, and technical documentation in various formats including **Markdown**, **JSON**, **YAML**, and inline comments.
|
|
|
|
6. **Optimized 4B Footprint for Developer Workflows**
|
|
Balanced for performance and efficiency, deployable on **developer workstations**, **CI/CD pipelines**, and **edge development environments** without compromising code quality.
|
|
|
|
---
|
|
|
|
## **Quickstart with Transformers**
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
model_name = "prithivMLmods/Blitzar-Coder-4B-F.1"
|
|
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|
model_name,
|
|
torch_dtype="auto",
|
|
device_map="auto"
|
|
)
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
|
prompt = "Create a REST API endpoint in Python using FastAPI that handles file uploads with validation and returns processing status."
|
|
|
|
messages = [
|
|
{"role": "system", "content": "You are an expert programming assistant skilled in multiple languages and development practices."},
|
|
{"role": "user", "content": prompt}
|
|
]
|
|
|
|
text = tokenizer.apply_chat_template(
|
|
messages,
|
|
tokenize=False,
|
|
add_generation_prompt=True
|
|
)
|
|
|
|
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
|
|
|
|
generated_ids = model.generate(
|
|
**model_inputs,
|
|
max_new_tokens=512
|
|
)
|
|
generated_ids = [
|
|
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
|
|
]
|
|
|
|
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
|
print(response)
|
|
```
|
|
|
|
---
|
|
|
|
## **Intended Use**
|
|
|
|
* Multi-language code generation and debugging assistance
|
|
* Cross-platform development and code translation between languages
|
|
* Code review, optimization, and refactoring suggestions
|
|
* Technical documentation and API specification generation
|
|
* Developer productivity tools and IDE integrations
|
|
* Educational coding tutorials and programming concept explanations
|
|
|
|
---
|
|
|
|
## **Limitations**
|
|
|
|
* Optimized for coding tasks—may underperform on general conversation
|
|
* Context limitations may affect analysis of very large codebases
|
|
* Focused on programming domains—creative writing capabilities are limited
|
|
* Best suited for technical development workflows rather than casual chat
|
|
|
|
---
|
|
|
|
## **References**
|
|
|
|
1. [Qwen2.5 Technical Report (2024)](https://arxiv.org/pdf/2412.15115)
|
|
2. [YaRN: Efficient Context Window Extension of Large Language Models](https://arxiv.org/pdf/2309.00071) |