Lambda-Equulei-1.5B-xLingual

Lambda-Equulei-1.5B-xLingual is a multilingual conversational model fine-tuned from Qwen2-1.5B, specifically designed for cross-lingual chat and experimental conversations across 30+ languages. It brings advanced multilingual understanding and natural dialogue capabilities in a compact size, ideal for international communication tools, language learning platforms, and global conversational assistants.

Key Features

1. Multilingual Conversational Excellence
   Trained to engage in natural, flowing conversations across 30+ languages, Lambda-Equulei-1.5B-xLingual enables seamless cross-cultural communication and supports diverse linguistic contexts for global applications.

2. Extensive Language Support (30+ Languages)
   Capable of understanding, responding, and maintaining context fluently in over 30 languages including English, Chinese, Spanish, French, German, Japanese, Korean, Arabic, Hindi, Portuguese, Russian, Italian, Dutch, and many more regional languages.

3. Compact yet Conversationally Rich
   While only 1.5B parameters, this model delivers strong performance for natural dialogue, context retention, cultural awareness, and nuanced conversations with minimal resource demands.

4. Experimental Conversational AI
   Provides dynamic, context-aware responses that adapt to different conversational styles, cultural nuances, and communication patterns across languages.

Quickstart with Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "prithivMLmods/Lambda-Equulei-1.5B-xLingual"
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Hello! Can you help me practice Spanish conversation?"
messages = [
    {"role": "system", "content": "You are a helpful multilingual assistant capable of conversing naturally in over 30 languages."},
    {"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]

Intended Use

• Multilingual Chat Applications: Natural conversation support across 30+ languages for global platforms.
• Language Learning Tools: Interactive practice partners for students learning new languages.
• International Customer Support: Cross-cultural communication for global businesses and services.
• Cultural Exchange Platforms: Facilitating meaningful conversations between speakers of different languages.
• Lightweight Multilingual Bots: Embedded use cases in mobile apps, web platforms, or resource-constrained environments.

Limitations

1. Experimental Nature:
   As an experimental conversational model, responses may vary in quality and consistency across different languages and contexts.

2. Language Proficiency Variation:
   While supporting 30+ languages, proficiency levels may differ between major languages (English, Chinese, Spanish) and less common regional languages.

3. Parameter Scale Constraints:
   Though efficient, the 1.5B parameter size may limit performance on highly complex multilingual tasks compared to larger models.

4. Bias from Base Model:
   Inherits any biases from Qwen2-1.5B's pretraining. Cultural sensitivity and output validation recommended for sensitive applications.

5. Context Length Limitations:
   May struggle with very long conversations or complex multi-turn dialogues requiring extensive context retention.