Reporting on first findings

Solid performance that we'll seek to improve by running some sweeps
Created on November 15|Last edited on November 15
Comment
﻿
Section 1﻿
This set of panels contains runs from a private project, which cannot be shown in this report
﻿
model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input_1 (InputLayer)        [(None, 32, 32, 3)]       0         
                                                                 
 conv2d (Conv2D)             (None, 32, 32, 16)        448       
                                                                 
 tf.nn.relu (TFOpLambda)     (None, 32, 32, 16)        0         
                                                                 
 max_pooling2d (MaxPooling2D  (None, 16, 16, 16)       0         
 )                                                               
                                                                 
 conv2d_1 (Conv2D)           (None, 16, 16, 32)        4640      
                                                                 
 tf.nn.relu_1 (TFOpLambda)   (None, 16, 16, 32)        0         
                                                                 
 max_pooling2d_1 (MaxPooling  (None, 8, 8, 32)         0         
 2D)                                                             
                                                                 
 conv2d_2 (Conv2D)           (None, 8, 8, 32)          9248      
                                                                 
 tf.nn.relu_2 (TFOpLambda)   (None, 8, 8, 32)          0         
                                                                 
 max_pooling2d_2 (MaxPooling  (None, 4, 4, 32)         0         
 2D)                                                             
                                                                 
 global_average_pooling2d (G  (None, 32)               0         
 lobalAveragePooling2D)                                          
                                                                 
 dense (Dense)               (None, 32)                1056      
                                                                 
 dropout (Dropout)           (None, 32)                0         
                                                                 
 last (Dense)                (None, 2)                 66        
                                                                 
=================================================================
Total params: 15,458
Trainable params: 15,458
Non-trainable params: 0
_________________________________________________________________
﻿
As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5%
﻿
﻿
Add a comment