Rakancorle1/qwen2.5-3b_Instruct_policy_traj_30k_full
Rakancorle1/qwen2.5-3b_Instruct_policy_traj_30k_full is a 3.1 billion parameter instruction-tuned causal language model, fine-tuned from Qwen/Qwen2.5-3B-Instruct. This model has been specialized through further training on the Policy_Traj_0826_30k_train dataset. It is designed for tasks requiring adherence to specific policy trajectories, leveraging its Qwen2.5 architecture and a 32K context length.
Loading preview...
Model Overview
This model, Rakancorle1/qwen2.5-3b_Instruct_policy_traj_30k_full, is a specialized instruction-tuned language model based on the Qwen2.5-3B-Instruct architecture. It features approximately 3.1 billion parameters and a context length of 32,768 tokens, making it suitable for processing moderately long sequences.
Key Specialization
The primary differentiator for this model is its fine-tuning on the Policy_Traj_0826_30k_train dataset. This targeted training suggests an optimization for tasks that involve understanding, generating, or following specific policy-related trajectories or sequences of actions.
Training Details
The fine-tuning process utilized the following hyperparameters:
- Learning Rate: 1e-05
- Batch Size: 2 (train), 8 (eval)
- Gradient Accumulation: 8 steps, leading to a total effective batch size of 64
- Optimizer: AdamW with cosine learning rate scheduler and 0.1 warmup ratio
- Epochs: 3.0
Potential Use Cases
Given its fine-tuning on a policy trajectory dataset, this model could be particularly useful for applications such as:
- Generating responses or actions that align with predefined policies.
- Simulating policy-driven behaviors.
- Analyzing sequences of events in the context of specific policies.
Limitations
As indicated in the original model card, more information is needed regarding its intended uses, limitations, and detailed training/evaluation data. Users should conduct thorough testing for their specific applications.