|
|
---
|
|
|
base_model: Qwen/Qwen2.5-Coder-1.5B-Instruct
|
|
|
language:
|
|
|
- en
|
|
|
library_name: transformers
|
|
|
license: apache-2.0
|
|
|
datasets:
|
|
|
- LSXPrime/ProseFlow-Actions-v1
|
|
|
tags:
|
|
|
- text-generation
|
|
|
- instruction
|
|
|
- proseflow
|
|
|
- unsloth
|
|
|
- qwen
|
|
|
- code-assistant
|
|
|
- writing-assistant
|
|
|
---
|
|
|
|
|
|
# ProseFlow-v1-1.5B-Instruct
|
|
|
|
|
|
**ProseFlow-v1-1.5B-Instruct** is a versatile instruction-tuned model designed to be the local AI engine for the [ProseFlow desktop application](https://github.com/LSXPrime/ProseFlow). This model excels at a wide variety of text-processing and code-related tasks, making it an ideal choice for users who want a high-performance, private, and offline-capable AI assistant integrated into their daily workflow.
|
|
|
|
|
|
This model was fine-tuned from [**Qwen/Qwen2.5-Coder-1.5B-Instruct**](https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct), inheriting its strong coding and logical reasoning capabilities, and has been further specialized to follow the specific, structured prompts used by ProseFlow "Actions".
|
|
|
|
|
|
The model was fine-tuned on the [**ProseFlow-Actions-v1**](https://huggingface.co/datasets/LSXPrime/ProseFlow-Actions-v1) dataset.
|
|
|
|
|
|
## Model Description
|
|
|
|
|
|
ProseFlow is a universal AI text processor that works via global hotkeys. Users create "Actions" – reusable instructions for the AI – to perform tasks like proofreading, summarizing, refactoring code, or changing the tone of a text. This model is the brain that executes those instructions.
|
|
|
|
|
|
**`ProseFlow-v1-1.5B-Instruct`** is the recommended and primary local model for the application. It strikes an excellent balance between performance, resource requirements, and capability.
|
|
|
|
|
|
### Key Strengths
|
|
|
|
|
|
Based on comprehensive evaluations against the `ProseFlow-Actions-v1` dataset, this model demonstrates:
|
|
|
|
|
|
* **Excellent Task Comprehension:** The model consistently understands the *intent* behind a wide variety of instructions, from simple text manipulation to complex business and logical tasks.
|
|
|
* **Superior Code Intelligence:** Thanks to its Qwen2.5-Coder base, the model is exceptionally proficient at code-related tasks like explaining code snippets, finding bugs, refactoring for efficiency, and adding comments.
|
|
|
* **High-Quality Text Generation:** It produces coherent, high-quality output for summarization, expansion, and creative writing prompts.
|
|
|
* **Strong Reasoning Capabilities:** The model can successfully solve multi-step word problems and perform logical deductions required by more advanced actions.
|
|
|
* **Versatility:** It performs reliably across dozens of distinct tasks, including sentiment analysis, data extraction (JSON conversion), and professional email drafting.
|
|
|
|
|
|
### Intended Use
|
|
|
|
|
|
This model is primarily intended to be used within the **ProseFlow desktop application**. Its prompt format and output style are specifically tuned to work seamlessly with the app's "Action" system.
|
|
|
|
|
|
When used in ProseFlow, it provides a powerful, private, and offline alternative to cloud-based AI services.
|
|
|
|
|
|
**Primary Use Cases:**
|
|
|
* Code refactoring, debugging, and documentation.
|
|
|
* Drafting and improving professional emails and documents.
|
|
|
* Summarizing long articles or meeting notes.
|
|
|
* Proofreading and enhancing creative or technical writing.
|
|
|
* Brainstorming ideas and generating structured content.
|
|
|
|
|
|
### Limitations and Considerations
|
|
|
|
|
|
While highly capable, this model has a few known behaviors:
|
|
|
|
|
|
* **Instruction Following vs. Helpfulness:** The model is so well-aligned to be a helpful assistant that it sometimes adds conversational headers or brief explanations to its output (e.g., "Here is the refactored code:"). This violates the strict "output only" constraint in some of the training prompts. In the context of the ProseFlow application, this is generally a minor issue, but it's a deviation from the prompt instructions.
|
|
|
* **Complex Logic:** While it can handle multi-step logic, it may fail on more abstract or tricky logical puzzles (e.g., identifying an anomaly in a nuanced list).
|
|
|
|
|
|
## How to Use in ProseFlow
|
|
|
|
|
|
1. [Download and install the ProseFlow application](https://github.com/LSXPrime/ProseFlow/releases).
|
|
|
2. Navigate to the **Providers -> Local Provider** tab.
|
|
|
3. Click "Manage Models..." and download `ProseFlow-v1-1.5B-Instruct` from the "Available for Download" list.
|
|
|
4. Once downloaded, select it from the "My Models" list.
|
|
|
5. Set your "Primary Service Type" in ProseFlow to **Local**.
|
|
|
6. You're all set! The application will now use this model for all AI actions.
|
|
|
|
|
|
## Training Details
|
|
|
|
|
|
* **Base Model:** [Qwen/Qwen2.5-Coder-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct)
|
|
|
* **Dataset:** [LSXPrime/ProseFlow-Actions-v1](https://huggingface.co/datasets/LSXPrime/ProseFlow-Actions-v1)
|
|
|
* **Fine-tuning Library:** [Unsloth](https://github.com/unslothai/unsloth)
|
|
|
* **Fine-tuning Method:** Supervised fine-tuning on a dataset of structured instruction-input-output triplets.
|
|
|
|
|
|
## License
|
|
|
|
|
|
This model is licensed under the [Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0). |
