|
--- |
|
license: llama3.1 |
|
language: |
|
- en |
|
base_model: |
|
- prithivMLmods/Triangulum-10B |
|
pipeline_tag: text-generation |
|
library_name: transformers |
|
tags: |
|
- LlamaWithQuestions |
|
- CoT |
|
- Reasoner |
|
- LWQ |
|
--- |
|
|
|
 |
|
|
|
# **LwQ-Reasoner-10B** |
|
|
|
LwQ-Reasoner-10B (Llama with Questions), based on the Llama 3.1 collection of multilingual large language models (LLMs), is a set of pre-trained and instruction-tuned generative models optimized for multilingual dialogue use cases. These models outperform many available open-source alternatives. Model Architecture: Llama 3.1 is an auto-regressive language model utilizing an optimized transformer architecture. The tuned versions undergo supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to better align with human preferences for helpfulness and safety. LwQ-Reasoner-10B is trained on synthetic reasoning datasets for mathematical reasoning and context-based problem-solving, with a focus on following instructions or keywords embedded in the input. |
|
|
|
# **Use with transformers** |
|
|
|
Starting with `transformers >= 4.43.0` onward, you can run conversational inference using the Transformers `pipeline` abstraction or by leveraging the Auto classes with the `generate()` function. |
|
|
|
Make sure to update your transformers installation via `pip install --upgrade transformers`. |
|
|
|
```python |
|
import transformers |
|
import torch |
|
|
|
model_id = "prithivMLmods/LwQ-Reasoner-10B" |
|
|
|
pipeline = transformers.pipeline( |
|
"text-generation", |
|
model=model_id, |
|
model_kwargs={"torch_dtype": torch.bfloat16}, |
|
device_map="auto", |
|
) |
|
|
|
messages = [ |
|
{"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"}, |
|
{"role": "user", "content": "Who are you?"}, |
|
] |
|
|
|
outputs = pipeline( |
|
messages, |
|
max_new_tokens=256, |
|
) |
|
print(outputs[0]["generated_text"][-1]) |
|
``` |
|
# **Config and Base** |
|
|
|
```json |
|
{ |
|
"_name_or_path": "prithivMLmods/Triangulum-10B", |
|
"architectures": [ |
|
"LlamaForCausalLM" |
|
] |
|
} |
|
``` |
|
# **Intended Use** |
|
|
|
1. **Multilingual Dialogue Systems**: LwQ-Reasoner-10B is designed for creating conversational agents capable of engaging in dialogues across multiple languages, making it suitable for global customer support and multilingual chatbots. |
|
|
|
2. **Instruction-Following Tasks**: The model excels at tasks requiring adherence to specific instructions or keywords embedded in the input, such as form completion, task automation, and guided workflows. |
|
|
|
3. **Mathematical Reasoning**: With specialized training on synthetic reasoning datasets, LwQ-Reasoner-10B can perform complex mathematical reasoning and problem-solving, making it useful for educational platforms, tutoring systems, and research assistance. |
|
|
|
4. **Context-Based Problem Solving**: The model is optimized to handle contextually rich problems, allowing it to generate context-aware responses for applications such as summarization, question answering, and decision support. |
|
|
|
5. **Content Generation**: It can generate high-quality content, including articles, reports, summaries, and creative writing, across various domains and languages. |
|
|
|
6. **Knowledge Retrieval**: LwQ-Reasoner-10B can retrieve and synthesize information from its trained data to answer factual questions, assist in research, and support knowledge-intensive tasks. |
|
|
|
# **Limitations** |
|
|
|
1. **Performance Variability Across Languages**: While the model supports multiple languages, its performance may vary depending on the language, with better results for languages more prevalent in its training data. |
|
|
|
2. **Handling of Niche Topics**: The model may struggle to provide accurate information or generate high-quality content for highly specialized or niche topics not covered extensively in its training data. |
|
|
|
3. **Complex Multi-Step Reasoning**: Although trained on reasoning datasets, the model may still occasionally produce incorrect or incomplete results for multi-step or highly complex reasoning tasks. |
|
|
|
4. **Bias and Ethical Concerns**: Since LwQ-Reasoner-10B is trained on large, publicly available datasets, it may inherit biases present in the data, leading to potential ethical concerns or inappropriate outputs in certain contexts. |
|
|
|
5. **Context Limitations**: The model has a finite context window, which may lead to incomplete understanding or response generation for tasks requiring extensive context or very long input texts. |
|
|
|
6. **Resource Intensive**: As a large-scale model with 10 billion parameters, it requires substantial computational resources for both inference and deployment, limiting its use in resource-constrained environments. |
|
|
|
7. **Instruction Ambiguity**: The model’s performance can degrade when instructions are ambiguous, vague, or conflicting, potentially leading to outputs that do not align with user expectations. |
