laion/nemosci-tasrep-a1mfc-dev1-maxeps-32b__Qwen3-32B
The laion/nemosci-tasrep-a1mfc-dev1-maxeps-32b__Qwen3-32B model is a 32 billion parameter language model fine-tuned from Qwen/Qwen3-32B. It was trained on a diverse set of scientific computing and agent-based trace datasets, including nemotron-terminal-scientific_computing and various DCAgent traces. This model is specialized for tasks related to scientific computing and agent behavior analysis, leveraging its 32768 token context length for complex sequences.
Loading preview...
Model Overview
This model, nemosci-tasrep-a1mfc-dev1-maxeps-32b__Qwen3-32B, is a 32 billion parameter language model derived from the Qwen/Qwen3-32B architecture. It has been specifically fine-tuned on a collection of specialized datasets, indicating an optimization for particular domains.
Key Training Data
The fine-tuning process utilized several distinct datasets, suggesting a focus on scientific and agent-based tasks:
/e/data1/datasets/playground/ot-baf/hf_hub/datasets--laion--nemotron-terminal-scientific_computing/e/data1/datasets/playground/ot-baf/hf_hub/datasets--DCAgent--exp_tas_repetition_penalty_1.05_traces/e/data1/datasets/playground/ot-baf/hf_hub/datasets--DCAgent--a1_multifile_composition/e/data1/datasets/playground/ot-baf/hf_hub/datasets--DCAgent--exp_tas_max_episodes_512_traces/e/data1/datasets/playground/ot-baf/hf_hub/datasets--DCAgent--dev_set_part1_10k_glm_4.7_traces_jupiter
Training Configuration
The model was trained with a learning rate of 4e-05 over 7 epochs, using a total batch size of 96 across 96 GPUs. An AdamW optimizer with cosine learning rate scheduler and 0.1 warmup ratio was employed. The training utilized Transformers 4.57.6, Pytorch 2.9.1+cu130, Datasets 4.7.0, and Tokenizers 0.22.2.