Model: modrill/math_think_11_qwen3_4b_base_task_arithmetic_scaling_0_3 Source: Original Platform
license, tags, library_name, base_model
| license | tags | library_name | base_model | |||
|---|---|---|---|---|---|---|
| cc-by-nc-4.0 |
|
transformers | LlamaFactory/math_think_11_qwen3_4b_base_sft |
math_think_11_qwen3_4b_base_task_arithmetic_scaling_0_3
Task Arithmetic merge between math_think_11 Qwen3-4B-Base SFT and Qwen/Qwen3-4B-Base, with scaling coefficient 0.3.
- Method: task arithmetic (
theta = theta_base + scaling * (theta_sft - theta_base)) - Base model:
math_think_11_qwen3_4b_base_sft(Qwen3-4B-Base) - Other model:
Qwen/Qwen3-4B-Base - Scaling:
0.3 - Renamed from:
scaling_0_3
Description
Model synced from source: modrill/math_think_11_qwen3_4b_base_task_arithmetic_scaling_0_3
Languages
Jinja
100%