2results about How to "Realize hierarchical control" patented technology

The embodiment of the invention provides a training method and platform for a diffusion language model, equipment, a medium and a product. According to the scheme, the method comprises the steps of obtaining a plurality of intermediate prediction results generated by an initial diffusion language model for sample cue words; for each of the plurality of intermediate prediction results, determining a reward value of the prediction result; and adjusting parameters of the initial diffusion language model based on the reward value to obtain a trained diffusion language model. According to the scheme, the disadvantage that the contribution degrees corresponding to the rewards and the intermediate prediction results are not matched can be effectively avoided, refined hierarchical regulation and control of the model optimization process can be realized, the overall training effect of the model can be improved, and the reasoning performance of the diffusion language model after training can be improved.

The application provides a kind of based on multi-scale spectrum constraint's generated synthetic seismic data optimization method, comprising: actual seismic data and original synthetic seismic data are normalized and preprocessed;Actual seismic data is executed two-dimensional discrete wavelet transform, and low-frequency approximation sub-band and multi-stage high-frequency detail sub-band are obtained, and the condition input feature is constructed by channel splicing;Condition generative adversarial network is constructed, generator fuses multi-scale wavelet condition feature and generates optimization data, and discriminator carries out adversarial training;The same wavelet transform is executed to generated data, and multi-scale spectrum constraint loss is constructed, respectively matches low-frequency sub-band and each stage high-frequency detail sub-band, and establishes hierarchical spectrum constraint mechanism;Low frequency and high frequency constraint loss are weighted and fused, and are jointly optimized with adversarial loss to form total loss function, and generator parameter is iteratively updated;KL divergence is calculated to the optimized synthetic seismic data and actual data, and is filtered and output according to distribution similarity.