DCAgent/a1-swesmith
DCAgent/a1-swesmith is a fine-tuned Qwen3-8B model developed by DCAgent, optimized for specific tasks through training on the swesmith-sandboxes-with_tests-gpt-5-mini-passed_glm_4.7_traces dataset. This 8 billion parameter model leverages the Qwen3 architecture to enhance performance on tasks related to its specialized training data. Its primary application is within the domain defined by the fine-tuning dataset, suggesting a focus on code-related or agentic reasoning tasks.
Loading preview...
Overview
DCAgent/a1-swesmith is a specialized language model derived from the Qwen3-8B architecture. It has undergone fine-tuning by DCAgent on a unique dataset, specifically /e/scratch/jureap59/raoof1/sft_data/hf_hub/datasets--DCAgent--swesmith-sandboxes-with_tests-gpt-5-mini-passed_glm_4.7_traces/snapshots/b9b0e0d113e9c37dd035f03644315478acc04487_thinking_preprocessed. This fine-tuning process aims to adapt the base Qwen3-8B model for particular applications or performance characteristics related to the training data's content.
Key Training Details
The model was trained with a learning rate of 4e-05 over 7 epochs, utilizing a multi-GPU setup with 16 devices and a total batch size of 16. The optimizer used was ADAMW_TORCH_FUSED with specific beta and epsilon parameters, and a cosine learning rate scheduler with a 0.1 warmup ratio. The training leveraged Transformers 4.57.6 and Pytorch 2.9.1+cu130.
Intended Use & Limitations
While specific details on intended uses and limitations are not provided in the model card, its fine-tuning on a dataset named swesmith-sandboxes-with_tests-gpt-5-mini-passed_glm_4.7_traces suggests a potential focus on tasks involving code, testing, or agentic reasoning within sandboxed environments. Users should refer to the dataset's characteristics to infer the model's specialized capabilities and potential limitations.