laion/nemosci-tasrep-a1mfc-gfistaqc-scaff-maxeps-swes-r2eg-32b__Qwen3-32B
The laion/nemosci-tasrep-a1mfc-gfistaqc-scaff-maxeps-swes-r2eg-32b__Qwen3-32B model is a 32 billion parameter Qwen3-based language model fine-tuned by laion. It was trained on a diverse collection of scientific computing, repetition penalty, multifile composition, and code-related datasets. This model is specialized for tasks involving complex code generation, scientific problem-solving, and agentic reasoning within a 32768 token context window.
Loading preview...
Model Overview
This model, laion/nemosci-tasrep-a1mfc-gfistaqc-scaff-maxeps-swes-r2eg-32b__Qwen3-32B, is a fine-tuned variant of the Qwen3-32B architecture. It has been specifically adapted using a comprehensive set of datasets focused on advanced computational and reasoning tasks.
Key Fine-tuning Datasets
The model's training incorporated several specialized datasets, indicating an optimization for complex problem-solving and code-related applications:
- Scientific Computing: From
laion/nemotron-terminal-scientific_computing. - Repetition Penalty Traces: From
DCAgent/exp_tas_repetition_penalty_1.05_traces. - Multifile Composition: From
DCAgent/a1_multifile_composition. - GFI STAQC Embeddings: From
DCAgent/exp-gfi-staqc-embedding-mean-filtered-10K_glm_4.7_traces_jupiter. - Max Episodes Traces: From
DCAgent/exp_tas_max_episodes_512_traces. - Repo Scaffold: From
DCAgent/a1_repo_scaffold. - SWESmith Sandboxes: From
DCAgent/swesmith-sandboxes-with_tests-gpt-5-mini-passed_glm_4.7_traces. - Kimi-2.5-r2egym Sandboxes: From
penfever/Kimi-2.5-r2egym_sandboxes-maxeps-32k.
Training Configuration
The model was trained with a learning rate of 4e-05, using an AdamW optimizer, and a cosine learning rate scheduler with a 0.1 warmup ratio over 7 epochs. It utilized a distributed training setup across 96 devices, resulting in a total batch size of 96.
Potential Use Cases
Given its specialized training, this model is likely well-suited for applications requiring advanced reasoning, scientific computation, complex code generation, and handling multi-file programming contexts.