Tiny-Llama-Llama-Dolphin-laser-1b-moe

Tiny-Llama-Llama-Dolphin-laser-1b-moe is a Mixure of Experts (MoE) made with the following models using LazyMergekit:

• TinyLlama/TinyLlama-1.1B-intermediate-step-715k-1.5T
• cognitivecomputations/TinyDolphin-2.8.2-1.1b-laser
• cognitivecomputations/TinyDolphin-2.8.1-1.1b
• TinyLlama/TinyLlama-1.1B-Chat-v1.0

🧩 Configuration

base_model: cognitivecomputations/TinyDolphin-2.8.2-1.1b-laser
experts:
  - source_model: TinyLlama/TinyLlama-1.1B-intermediate-step-715k-1.5T
    positive_prompts:
    - "Write a Python script that sorts a list of integers using the bubble sort algorithm."
    - "Write a JavaScript function that redirects a web page to another page after 5 seconds."
    negative_prompts:
    - "Discuss the latest world events."
    - "Narrate a fictional story about a knight's quest."
  - source_model: cognitivecomputations/TinyDolphin-2.8.2-1.1b-laser
    positive_prompts:
    - "Describe the steps to troubleshoot a fluid dynamics issue with a water fountain."
    - "If we have 3 marbles, and two roll under the counter, and one is found, how many marbles are there?"
    negative_prompts:
    - "Tell me about your favorite book."
    - "Write a Python script that sorts a list of integers."
  - source_model: cognitivecomputations/TinyDolphin-2.8.1-1.1b
    positive_prompts:
    - "Write a short story about a knight's quest to find a lost treasure, and then summarize it in one paragraph."
    - "Summarize the following article with details and clarity."
    negative_prompts:
    - "Give me a sample of code in Rust."
    - "Describe the steps to troubleshoot a fluid dynamics issue."
  - source_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
    positive_prompts:
    - "Tell me about your favorite book and why you like it."
    - "Chat with me about something I've been thinking of."
    negative_prompts:
    - "Write a Python script that sorts a list of integers."
    - "Summarize the following article with details and clarity."
gate_mode: hidden

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "jtatman/Tiny-Llama-Llama-Dolphin-laser-1b-moe"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])