slices:
  - sources:
      - model: Writer/palmyra-mini-thinking-a
        layer_range: [0, 12]  # Layers 0 to 11 (first half)
  - sources:
      - model: Writer/palmyra-mini-thinking-b
        layer_range: [12, 24] # Layers 12 to 23 (second half)

merge_method: passthrough
dtype: bfloat16  # Recommended for Qwen models
base_model: Qwen/Qwen2.5-1.5B  # CRITICAL: Specify the base model
output_dir: ./qwen-thinking-passthrough

💻 Usage

python

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "amylynn/palmyra-thinking-merge"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

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"])