Overview

DCAgent/a1-inferredbugs is an 8 billion parameter language model, fine-tuned from the Qwen/Qwen3-8B architecture. This model was developed by DCAgent and trained using a specific dataset: /e/scratch/jureap59/raoof1/sft_data/hf_hub/datasets--DCAgent--inferredbugs-sandboxes_glm_4.7_traces_jupiter/snapshots/fa94a8d5ac1d6f3953e1d93f260ce88f5da381ff_thinking_preprocessed. The fine-tuning process involved 7 epochs with a learning rate of 4e-05 and a total training batch size of 16 across 16 GPUs.

Key Characteristics

• Base Model: Qwen3-8B
• Parameter Count: 8 billion
• Context Length: 32768 tokens
• Training Data: Specialized dataset focused on "inferredbugs-sandboxes_glm_4.7_traces_jupiter", indicating a potential specialization in bug detection, analysis, or related tasks within sandboxed or traced environments.

Training Details

The model was trained using the following key hyperparameters:

• Optimizer: AdamW_Torch_Fused with betas=(0.9, 0.98) and epsilon=1e-08.
• LR Scheduler: Cosine type with a warmup ratio of 0.1.
• Epochs: 7.0

Intended Use Cases

While specific intended uses and limitations are not detailed in the provided README, the training data suggests that this model is likely optimized for tasks involving the analysis of traces or sandboxed environments to infer or identify bugs. Its large context window of 32768 tokens would be beneficial for processing extensive logs, code snippets, or execution traces.