| --- |
| license: apache-2.0 |
| language: |
| - en |
| base_model: |
| - drizzymedia/SynapseCoder-32B |
| pipeline_tag: text-generation |
| library_name: transformers |
| tags: |
| - code |
| - synapsecoder |
| - chat |
| - synapse |
| - ai-coding |
| - coding-agent |
| --- |
| |
| # SynapseCoder-32B |
|
|
| <a href="https://synapse.ai/" target="_blank" style="margin: 2px;"> |
| <img alt="Chat" src="https://img.shields.io/badge/%F0%9F%A7%A0%20Synapse%20Chat-6B4EFF" style="display: inline-block; vertical-align: middle;"/> |
| </a> |
| |
| ## Introduction |
|
|
| SynapseCoder is the advanced code-specialized language model series from Synapse AI. Built for modern software development, SynapseCoder provides powerful capabilities across code generation, code reasoning, debugging, refactoring, and AI-powered development workflows. |
|
|
| SynapseCoder brings significant improvements in: |
|
|
| - **Code generation**, **code understanding**, and **code fixing** |
| - **Software engineering reasoning** for real-world development tasks |
| - **AI coding agents** and autonomous developer workflows |
| - **Long-context programming support** for large codebases and complex projects |
|
|
| SynapseCoder-32B is the flagship coding model in the SynapseCoder family, designed to deliver professional-level programming assistance while maintaining strong general reasoning and mathematical capabilities. |
|
|
| **This repository contains the instruction-tuned 32B SynapseCoder model**, featuring: |
|
|
| - Type: Causal Language Model |
| - Training Stage: Pretraining & Post-training |
| - Architecture: Transformer with RoPE, SwiGLU, RMSNorm, and Attention QKV bias |
| - Number of Parameters: 32.5B |
| - Number of Non-Embedding Parameters: 31.0B |
| - Number of Layers: 64 |
| - Number of Attention Heads (GQA): 40 for Query and 8 for Key/Value |
| - Context Length: Up to 131,072 tokens |
|
|
| SynapseCoder is optimized for: |
|
|
| - Software development |
| - Code generation |
| - Code completion |
| - Debugging |
| - Refactoring |
| - Documentation generation |
| - AI coding assistants |
| - Autonomous coding agents |
|
|
| ## Requirements |
|
|
| SynapseCoder requires the latest version of Hugging Face `transformers`. |
|
|
| Older versions may cause compatibility issues during model loading. |
|
|
| ## Quickstart |
|
|
| ```python |
| from transformers import AutoModelForCausalLM, AutoTokenizer |
| |
| model_name = "drizzymedia/SynapseCoder-32B" |
| |
| model = AutoModelForCausalLM.from_pretrained( |
| model_name, |
| torch_dtype="auto", |
| device_map="auto" |
| ) |
| |
| tokenizer = AutoTokenizer.from_pretrained(model_name) |
| |
| prompt = "Write a Python quick sort algorithm." |
| |
| messages = [ |
| { |
| "role": "system", |
| "content": "You are SynapseCoder, an advanced AI coding assistant created by Synapse AI." |
| }, |
| { |
| "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) |
