license, base_model, pipeline_tag, library_name

license	base_model	pipeline_tag	library_name
apache-2.0	deepseek-ai/DeepSeek-R1-Distill-Qwen-14B Qwen/Qwen2.5-14B-Instruct-1M	text-generation	transformers

Qwen2.5-14B-DeepSeek-R1-1M

A merged model combines the reasoning model's strengths (Qwen2.5-14B-DeepSeek-R1) and the long-context model capabilities (Qwen2.5-14B-Instruct-1M) for versatile performance.

Merge config

models:
  - model: "deepseek-ai/DeepSeek-R1-Distill-Qwen-14B"
    parameters:
      weight: 1
      density: 1

merge_method: ties
base_model: "Qwen/Qwen2.5-14B-Instruct-1M"
parameters:
  density: 1
  normalize: true
  int8_mask: true
dtype: bfloat16

and I needed to make some minor adjustments to the tokenizer configuration.

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "mkurman/Qwen2.5-14B-DeepSeek-R1-1M"
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Write a Python script to merge two CSV files."
messages = [
    {"role": "system", "content": "You are an expert programmer."},
    {"role": "user", "content": prompt}
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

You can use it on LM Studio or Ollama by utilizing the provided GGUF files.

License

Apache 2.0 for open-source contribution and collaboration.