I have a challenge and I am trying to solve this in order to move forward, it is the final piece of the puzzle for my model operations.
What I am trying to do?:*
is verify the videos that are being used in the Xval_test variable via the split operations here as per the example via here In Python sklearn, how do I retrieve the names of samples/variables in test/training data? :
X_train, Xval_test, Y_train, Yval_test = train_test_split(
X, Y, train_size=0.8, test_size=0.2, random_state=1, shuffle=True)
1.
What I tried?:
is calling the name from the actual tag via file_path name, however that is not working. (every time the code runs the names from the file path are taken and not from the actual split operations Xval_test variable. This causes an issue during the model.fit() procedures as it changes the 1D flattened tensor to (a number of rows, 1 column)
file_paths = []
for file_name in os.listdir(root):
file_path = os.path.join(root, file_name)
if os.path.isfile(file_path):
file_paths.append(file_path)
print('**********************************************************')
print('ALL Directory File Paths Completed', file_paths)
I am not sure if the files are being shuffled properly with my weak attempt as per the guidelines from the split() forum. (based on my knowledge, every time I run the code, those files would be shuffled to a new Xval_test set relative to it's specified split parameter 80:20.
2.
I tried calling the model.predict(), that presents no labels for which I was hoping that it did (maybe I am using it the wrong way for calling the indices, I don't know).
my_pred = model.predict(Xval_test).argmax(axis=1)
I tried calling np.argsmax():( I KNOW THE TOTAL AMT OF FILES IN Xval_test is 16 based on the split())
Y_valpred = np.argmax(model.predict(Xval_test), axis=1) # model
This returns just the class label and not it's contents e.g. the classes in the datastore are folders containing (walking and fencing) rather than the actual video labels such as (walking0.avi....100/n and fencing0.avi.....100n/) !!!???!
I am not certain of the operation to get the folder content's tags, the actual file itself. It is this that I am trying to get from the X_test variable.
(or maybe its the wrong variable or functioning I using, again I am lacking the knowledge to understand this, please assist so that I can move to the next stage).
3.
I tried printing all of the variable's from the previous operations to see where that name tag would be stored and it is stored in the name variable below as per my operations: (but how do I call these folder content's file tags forward to the X_test variable or as per my choice the model.predict() outputs in a column together with the other metrics. So far, this causes issues with the model.fit() function???)
for files3 in files2:
name = os.path.join(namelist, files3)
name1 = name.strip("./dataset/")
name2 = name1.strip("Fencing/")
name3 = name2.strip("Stabing/")
name3 = name3.replace('.av', '')
name4 = name3.split()
# print("This is name1 ", name1)
# name5 = pd.DataFrame({"vid_names": name4}).to_csv("results.csv")
# name1 = name1.replace('[]', '')
with open('vid_names.csv', 'a',newline='') as f:
writer = csv.writer(f)
writer = writer.writerow(name4)
# print("My Video Names => ", name3)
3A.
Thank you in advance, I am grateful for any guidance provided, Please assist!
QUESTIONS: ############################################
Ques: 1.
Is it possible to see what video label tags are segmented within the X_Test Variable?
Ques: 1A.
If yes, may I request your guidance here, please, on how this can be done?:
I have been researching for weeks and cannot seem to get this sorted, your efforts would be greatly appreciated.
Ques: 2. MY Expected OUTCOME:
I am trying to access the prediction. So, In the end I would get an output relative to the actual video tag that insinuates the actual video that was used in the prediction operation along with its class tag (see below):
Initially, the model.predict() operations outputs numerical data relative to the class label.
I am trying to access the actual file label as well:
For example, what I want the predictions to look like is as follows:
X_test_labs Pred_labs Actual_File Pred_Score
0 Fencing Fencing fencing0.avi 0.99650866
1 Walking Fencing walking6.avi 0.9948837
2 Walking Walking walking21.avi 0.9967557
3 Fencing Fencing fencing32.avi 0.9930409
4 Walking Fencing walking43.avi 0.9961387
5 Walking Walking walking48.avi 0.6467387
6 Walking Walking walking50.avi 0.5465369
7 Walking Walking walking9.avi 0.3478027
8 Fencing Fencing fencing22.avi 0.1247543
9 Fencing Fencing fencing46.avi 0.7477777
10 Walking Walking walking37.avi 0.8499399
11 Fencing Fencing fencing19.avi 0.8887722
12 Walking Walking walking12.avi 0.7775351
13 Fencing Fencing fencing33.avi 0.4323323
14 Fencing Fencing fencing51.avi 0.7812434
15 Fencing Fencing fencing8.avi 0.8723476
I am not sure how to achieve this task, this one is a little more tricky for me than anticipated
This is my code*
'''*******Load Dependencies********'''
from keras.regularizers import l2
from keras.layers import Dense
from keras_tqdm import TQDMNotebookCallback
from tqdm.keras import TqdmCallback
from tensorflow import keras
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import math
import tensorflow as tf
from tqdm import tqdm
import videoto3d
import seaborn as sns
import scikitplot as skplt
from sklearn import preprocessing
from sklearn.metrics import classification_report, confusion_matrix
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score, classification_report, confusion_matrix
from sklearn.metrics import confusion_matrix, accuracy_score, precision_score, recall_score, f1_score
from keras.utils.vis_utils import plot_model
from keras.utils import np_utils
from tensorflow.keras.optimizers import Adam
from keras.models import Sequential
from keras.losses import categorical_crossentropy
from keras.layers import (Activation, Conv3D, Dense, Dropout, Flatten,MaxPooling3D)
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import os
import argparse
import time
import sys
import openpyxl
import os
import re
import csv
from keras import models
import cv2
import pickle
import glob
from numpy import load
np.seterr(divide='ignore', invalid='ignore')
print('**********************************************************')
print('Graphical Representation Of Accuracy & Validation Results Completed')
def plot_history(history, result_dir):
plt.plot(history.history['val_accuracy'], marker='.')
plt.plot(history.history['accuracy'], marker='.')
plt.title('model accuracy')
plt.xlabel('epoch')
plt.ylabel('accuracy')
plt.grid()
plt.legend(['Val_acc', 'Test_acc'], loc='lower right')
plt.savefig(os.path.join(result_dir, 'model_accuracy.png'))
plt.close()
plt.plot(history.history['val_loss'], marker='.')
plt.plot(history.history['loss'], marker='.')
plt.title('model Loss')
plt.xlabel('epoch')
plt.ylabel('loss')
plt.grid()
plt.legend(['Val_loss', 'Test_loss'], loc='upper right')
plt.savefig(os.path.join(result_dir, 'model_loss.png'))
plt.close()
# Saving History Accuracy & Validation Acuuracy Results To Directory
print('**********************************************************')
print('Generating History Acuuracy Results Completed')
def save_history(history, result_dir):
loss = history.history['loss']
acc = history.history['accuracy']
val_loss = history.history['val_loss']
val_acc = history.history['val_accuracy']
nb_epoch = len(acc)
# Creating The Results File To Directory = Store Results
print('**********************************************************')
print('Saving History Acuuracy Results To Directory Completed')
with open(os.path.join(result_dir, 'result.txt'), 'w') as fp:
fp.write('epoch\tloss\tacc\tval_loss\tval_acc\n')
# print(fp)
for i in range(nb_epoch):
fp.write('{}\t{}\t{}\t{}\t{}\n'.format(
i, loss[i], acc[i], val_loss[i], val_acc[i]))
print('**********************************************************')
print('Loading All Specified Video Data Samples From Directory Completed')
def loaddata(video_dir, vid3d, nclass, result_dir, color=False, skip=True):
files = os.listdir(video_dir)
with open('files.csv', 'w') as f:
writer = csv.writer(f)
writer.writerow(files)
root = '/Users/symbadian/3DCNN_latest_Version/3DCNNtesting/dataset/'
dirlist = [item for item in os.listdir(
root) if os.path.isdir(os.path.join(root, item))]
print('Get the filesname and path')
print('DIRLIST Directory Completed', dirlist)
file_paths = []
for file_name in os.listdir(root):
file_path = os.path.join(root, file_name)
if os.path.isfile(file_path):
file_paths.append(file_path)
print('**********************************************************')
print('ALL Directory File Paths Completed', file_paths)
roots, dirsy, fitte = next(os.walk(root), ([], [], []))
print('**********************************************************')
print('ALL Directory ROOTED', roots, fitte, dirsy)
X = []
print('X labels==>', X) # This stores all variable data in an object format
labellist = []
pbar = tqdm(total=len(files)) # generate progress bar for file processing
print('**********************************************************')
print('Generating/Join Class Labels For Video Dataset For Input Completed')
# Accessing files and labels from dataset directory
for filename in files:
pbar.update(1)
if filename == '.DS_Store':#.DS_Store
continue
namelist = os.path.join(video_dir, filename)
files2 = os.listdir(namelist)
###############################################################################
######### NEEDS TO FIX THIS Data Adding to CSV Rather Than REWRITTING #########
for files3 in files2:
name = os.path.join(namelist, files3)
#Call a function that extract the frames details of all file names
label = vid3d.get_UCF_classname(filename)
if label not in labellist:
if len(labellist) >= nclass:
continue
labellist.append(label)
# This X variable is the point where the lables are store (I think??!?!)
X.append(vid3d.video3d(name, color=color, skip=skip))
pbar.close()
# generating labellist/ writing to directory
print('******************************************************')
print('Saving All Class Labels For Referencing To Directory Completed')
with open(os.path.join(result_dir, 'classes.txt'), 'w') as fp:
for i in range(len(labellist)):
# print('These are labellist i classes',i) #Not This
fp.write('{}\n'.format(labellist[i]))
# print('These are my labels: ==>',mylabel)
for num, label in enumerate(labellist):
for i in range(len(labels)):
if label == labels[i]:
labels[i] = num
# print('This is labels i',labels[i]) #Not this
if color: # conforming image channels of image for input sequence
return np.array(X).transpose((0, 2, 3, 4, 1)), labels
else:
return np.array(X).transpose((0, 2, 3, 1)), labels
print('**********************************************************')
print('Generating Args Informative Messages/ Tuning Parameters Options Completed')
def main():
parser = argparse.ArgumentParser(description='A 3D Convolution Model For Action Recognition')
parser.add_argument('--batch', type=int, default=130)
parser.add_argument('--epoch', type=int, default=100)
parser.add_argument('--videos', type=str, default='dataset',help='Directory Where Videos Are Stored')# UCF101
parser.add_argument('--nclass', type=int, default= 2)
parser.add_argument('--output', type=str, required=True)
parser.add_argument('--color', type=bool, default=False)
parser.add_argument('--skip', type=bool, default=True)
parser.add_argument('--depth', type=int, default=10)
args = parser.parse_args()
# print('This is the Option Arguments ==>',args)
print('**********************************************************')
print('Specifying Input Size and Channels Completed')
img_rows, img_cols, frames = 32, 32, args.depth
channel = 3 if args.color else 1
print('**********************************************************')
print('Saving Dataset As NPZ To Directory Completed')
fname_npz = 'dataset_{}_{}_{}.npz'.format(args.nclass, args.depth, args.skip)
vid3d = videoto3d.Videoto3D(img_rows, img_cols, frames)
nb_classes = args.nclass
# loading the data
if os.path.exists(fname_npz):
loadeddata = np.load(fname_npz)
X, Y = loadeddata["X"], loadeddata["Y"]
else:
x, y = loaddata(args.videos, vid3d, args.nclass,args.output, args.color, args.skip)
X = x.reshape((x.shape[0], img_rows, img_cols, frames, channel))
Y = np_utils.to_categorical(y, nb_classes)
X = X.astype('float32')
#save npzdata to file
np.savez(fname_npz, X=X, Y=Y)
print('Saved Dataset To dataset.npz. Completed')
print('X_shape:{}\nY_shape:{}'.format(X.shape, Y.shape))
print('**********************************************************')
print('Initialise Model Layers & Layer Parameters Completed')
# Sequential groups a linear stack of layers into a tf.keras.Model.
# Sequential provides training and inference features on this model
model = Sequential()
model.add(Conv3D(32, kernel_size=(3, 3, 3),input_shape=(X.shape[1:]), padding='same'))
model.add(Activation('relu'))
model.add(Conv3D(32, kernel_size=(3, 3, 3), padding='same'))
model.add(MaxPooling3D(pool_size=(3, 3, 3), padding='same'))
model.add(Conv3D(64, kernel_size=(3, 3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv3D(64, kernel_size=(3, 3, 3), padding='same'))
model.add(MaxPooling3D(pool_size=(3, 3, 3), padding='same'))
model.add(Conv3D(128, kernel_size=(3, 3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv3D(128, kernel_size=(3, 3, 3), padding='same'))
model.add(MaxPooling3D(pool_size=(3, 3, 3), padding='same'))
model.add(Dropout(0.5))
model.add(Conv3D(256, kernel_size=(3, 3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv3D(256, kernel_size=(3, 3, 3), padding='same'))
model.add(MaxPooling3D(pool_size=(3, 3, 3), padding='same'))
model.add(Dropout(0.5))
model.add(Flatten())
# Dense function to convert FCL to 512 values
model.add(Dense(512, activation='sigmoid'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes, activation='softmax'))
model.compile(loss=categorical_crossentropy,optimizer=Adam(), metrics=['accuracy'])
model.summary()
print('this is the model shape')
model.output_shape
plot_model(model, show_shapes=True,to_file=os.path.join(args.output, 'model.png'))
print('**********************************************************')
print("Train Test Method HoldOut Performance")
X_train, Xval_test, Y_train, Yval_test = train_test_split(
X, Y, train_size=0.8, test_size=0.2, random_state=1, stratify=Y, shuffle=True)
print('**********************************************************')
print('Deploying Data Fitting/ Performance Accuracy Guidance Completed')
#Stop operations when experiencing no learning
rlronp = tf.keras.callbacks.ReduceLROnPlateau(monitor="val_loss", factor=0.5, patience=1, mode='auto', min_delta=0.0001, cooldown=1, min_lr=0.0001)
# Fit the training data
history = model.fit(X_train, Y_train, validation_split=0.20, batch_size=args.batch,epochs=args.epoch, verbose=1, callbacks=[rlronp], shuffle=True)
# Predict X_Test (Xval_test) data and Labels
predict_labels = model.predict(Xval_test, batch_size=args.batch,verbose=1,use_multiprocessing=True)
classes = np.argmax(predict_labels, axis = 1)
label = np.argmax(Yval_test,axis = 1)
print('This the BATCH size', args.batch)
print('This the DEPTH size', args.depth)
print('This the EPOCH size', args.epoch)
print('This the TRAIN SPLIT size', len(X_train))
print('This the TEST SPLIT size', len(Xval_test))
# https://stackoverflow.com/questions/52261597/keras-model-fit-verbose-formatting
# A json file enhances the model performance by a simple to save/load model
model_json = model.to_json()
if not os.path.isdir(args.output):
os.makedirs(args.output)
with open(os.path.join(args.output, 'ucf101_3dcnnmodel.json'), 'w') as json_file:
json_file.write(model_json)
# hd5 contains multidimensional arrays of scientific data
model.save_weights(os.path.join(args.output, 'ucf101_3dcnnmodel.hd5'))
''' Evaluation is a process
'''
print('**********************************************************')
print('Displying Test Loss & Test Accuracy Completed')
loss, acc = model.evaluate(Xval_test, Yval_test, verbose=2, batch_size=args.batch, use_multiprocessing=True) # verbose 0
print('this is args output', args.output)
plot_history(history, args.output)
save_history(history, args.output)
print('**********************************************************')
# Generating Picture Of Confusion matrix
print('**********************************************************')
print('Generating CM InputData/Classification Report Completed')
#Ground truth (correct) target values.
y_valtest_arg = np.argmax(Yval_test, axis=1)
#Estimated targets as returned by a classifier
Y_valpred = np.argmax(model.predict(Xval_test), axis=1) # model
print('y_valtest_arg Shape is ==>', y_valtest_arg.shape)
print('Y_valpred Shape is ==>', Y_valpred.shape)
print('**********************************************************')
print('Classification_Report On Model Performance Completed==')
print(classification_report(y_valtest_arg.round(), Y_valpred.round(), target_names=filehandle, zero_division=1))
'''Intitate Confusion Matrix'''
# print('Model Confusion Matrix Per Test Data Completed===>')
cm = confusion_matrix(y_valtest_arg, Y_valpred, normalize=None)
print('Display Confusion Matrix ===>', cm)
print('**********************************************************')
print('Model Overall Accuracy')
print('Model Test loss:', loss)
print('**********************************************************')
print('Model Test accuracy:', acc)
print('**********************************************************')
if __name__ == '__main__':
main()
I Think the solution is around the prediction, train, test split and the evaluation arguments. However, I am lacking the knowledge to access the details required from the train, test, split(), if that's where the issue is.
I am really thankful for your guidance in advance, thanks a whole lot for clarifying this for me and closing the gaps in my understanding. really appreciate this!!!
Sorry, this might turn out to be a very basic problem but I'm stumped:
I'm trying to get the lengths of skeleton branches. I'm able to skeletonize my image and can see the skeleton with draw.overlay_skeleton_2d() but when I try to get any statistics on the skeleton back I get an error:
TypeError Traceback (most recent call last)
<ipython-input-3-5924e1b8ffc8> in <module>
4 # fig, ax = plt.subplots()
5 # draw.overlay_skeleton_2d(img, skeleton0, dilate=1, axes=ax)
----> 6 branch_data = summarize(Skeleton(skeleton0))
7 branch_data.head()
~\AppData\Roaming\Python\Python38\site-packages\skan\csr.py in __init__(self, skeleton_image, spacing, source_image, _buffer_size_offset, keep_images, junction_mode, unique_junctions)
383 self.nbgraph = csr_to_nbgraph(graph, pixel_values)
384 self.coordinates = coords
--> 385 self.paths = _build_skeleton_path_graph(
386 self.nbgraph, _buffer_size_offset=_buffer_size_offset
387 )
~\AppData\Roaming\Python\Python38\site-packages\skan\csr.py in _build_skeleton_path_graph(graph, _buffer_size_offset)
259 def _build_skeleton_path_graph(graph, *, _buffer_size_offset=None):
260 if _buffer_size_offset is None:
--> 261 max_num_cycles = graph.indices.size // 4
262 _buffer_size_offset = max_num_cycles
263 degrees = np.diff(graph.indptr)
TypeError: expected dtype object, got 'numpy.dtype[int32]'
Any tips would be great. I'm still a Python beginner so, like I said, the mistake might be a very basic one.
Rgds,
Olie
*Edits included
This is my code:
import cv2
from skan import draw, skeleton_to_csgraph, Skeleton, summarize
from skimage import morphology
# import numpy as np
from matplotlib import pyplot as plt
%matplotlib notebook
# load image and convert to bool
img = cv2.imread(save_p + "\\" + img_name + ".png")
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img = img.astype(bool)
# skeletonize
skeleton0 = morphology.skeletonize(img)
### check skeleton
# fig, ax = plt.subplots()
# draw.overlay_skeleton_2d(img, skeleton0, dilate=1, axes=ax)
### from Documentation (linked below)
branch_data = summarize(Skeleton(skeleton0)) # ignoring scaling for now
branch_data.head()
# Skel_obj = Skeleton(skeleton0) # testing this but same problem
# Skel_obj.path_lengths(0)
I have attempted to follow a couple of posts (e.g. this but for that particular case the module wont install and anyway... it seems like skan.summarize is exactly what I need - documentation
Name: numpy
Version: 1.23.4
Name: scikit-image
Version: 0.19.3
Name: pandas
Version: 1.3.4
Name: skan
Version: 0.10.0
Name: numba
Version: 0.50.1
Windows 10 PC
This is same as this issue. Transformers version 4.19.4.
I have already seen this very similar question and the related issue, but these solutions don't work in my case.
This is the code I'm using:
args = TrainingArguments(
f"{model_name}-hyperp-{task}",
evaluation_strategy = "epoch",
learning_rate=2e-5,
per_device_train_batch_size=batch_size,
per_device_eval_batch_size=batch_size,
num_train_epochs=3,
weight_decay=0.01,
skip_memory_metrics=True, # https://github.com/huggingface/transformers/issues/12177 [picking error]
)
trainer = Trainer(
model_init=model_init, # function to initialize model (using 'from_pretrained')
args=args,
train_dataset = tokenized_datasets["train"],
eval_dataset = tokenized_datasets["validation"],
data_collator=data_collator,
tokenizer=tokenizer,
compute_metrics=compute_metrics,
)
trainer.hyperparameter_search(
hp_space=lambda _: tune_config,
backend="ray",
n_trials=10,
resources_per_trial={"cpu": 1, "gpu": 0},
scheduler=scheduler,
keep_checkpoints_num=1,
checkpoint_score_attr="training_iteration",
progress_reporter=reporter,
local_dir="/ray_results/",
name="tune_transformer_pbt",
log_to_file=True,
)
The error is:
TypeError Traceback (most recent call last)
[<ipython-input-50-3716493001d6>](https://localhost:8080/#) in <module>()
33 local_dir="/ray_results/",
34 name="tune_transformer_pbt",
---> 35 log_to_file=True,
36 )
37
...lots of lines...
[/usr/local/lib/python3.7/dist-packages/ray/cloudpickle/cloudpickle_fast.py](https://localhost:8080/#) in dump(self, obj)
618 def dump(self, obj):
619 try:
--> 620 return Pickler.dump(self, obj)
621 except RuntimeError as e:
622 if "recursion" in e.args[0]:
TypeError: can't pickle _thread.lock objects
The complete error trace can be found at the 1st link above.
I have already tried the suggested way to set skip_memory_metrics=True in TrainingArguments, but this doesn't work.
I was using Google Colab.
I am practicing with this life expectancy dataset from Kaggle (https://www.kaggle.com/datasets/kumarajarshi/life-expectancy-who?select=Life+Expectancy+Data.csv) and I want to train and visualize a classification and regression tree model. however, I keep getting an error that says "InvocationException: GraphViz's executables not found". I am wondering if this is because of the nature of the continuous numerical target dataset type? how can I visualize the model?
code:
import warnings
warnings.filterwarnings('ignore')
import pandas as pd
import numpy as np
import seaborn as sn
from sklearn import datasets
from sklearn import metrics
from sklearn import tree
from sklearn.tree import DecisionTreeClassifier
from sklearn.tree import DecisionTreeRegressor
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from sklearn.preprocessing import LabelEncoder
from sklearn.tree import export_graphviz
import matplotlib.pyplot as plt,pydotplus
from IPython.display import Image,display
data = pd.read_csv('Life Expectancy Data.csv')
data = data.dropna(how = 'any')
#feature selection
data = data.drop(columns=['infant deaths', ' thinness 5-9 years', 'Alcohol', 'percentage expenditure', 'Hepatitis B', 'Total expenditure', 'Population', ' thinness 5-9 years', 'Year', 'Country'])
# Creating a instance of label Encoder.
le = LabelEncoder()
# Using .fit_transform function to fit label
# encoder and return encoded label
label = le.fit_transform(data['Status'])
# removing the column 'Status' from df
data.drop('Status', axis=1, inplace=True)
# Appending the array to our dataFrame
# with column name 'Status'
data['Status'] = label
#training model
model_data = data
X = data.drop(columns=['Life expectancy '])
y = data['Life expectancy ']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2)
model = DecisionTreeRegressor()
model.fit(X_train, y_train)
#visualizing tree
LEtree = tree.export_graphviz(model,
feature_names = ['Adult Mortality', 'Measles', ' BMI', 'under-five deaths', 'Polio', 'Diphtheria', ' HIV/AIDS', 'GDP', ' thinness 1-19 years', 'Income composition of resources', 'Schooling', 'Status'],
class_names = y,
label = 'all',
rounded = True,
filled = True)
graph=pydotplus.graph_from_dot_data(LEtree)
display(Image(graph.create_png()))
full error message:
InvocationException Traceback (most recent call last)
Input In [27], in <cell line: 2>()
1 graph=pydotplus.graph_from_dot_data(LEtree)
----> 2 display(Image(graph.create_png()))
File ~\Anaconda3\lib\site-packages\pydotplus\graphviz.py:1797, in Dot.__init__.<locals>.<lambda>(f, prog)
1792 # Automatically creates all the methods enabling the creation
1793 # of output in any of the supported formats.
1794 for frmt in self.formats:
1795 self.__setattr__(
1796 'create_' + frmt,
-> 1797 lambda f=frmt, prog=self.prog: self.create(format=f, prog=prog)
1798 )
1799 f = self.__dict__['create_' + frmt]
1800 f.__doc__ = (
1801 '''Refer to the docstring accompanying the'''
1802 ''''create' method for more information.'''
1803 )
File ~\Anaconda3\lib\site-packages\pydotplus\graphviz.py:1959, in Dot.create(self, prog, format)
1957 self.progs = find_graphviz()
1958 if self.progs is None:
-> 1959 raise InvocationException(
1960 'GraphViz\'s executables not found')
1962 if prog not in self.progs:
1963 raise InvocationException(
1964 'GraphViz\'s executable "%s" not found' % prog)
InvocationException: GraphViz's executables not found
Try Installing the Graphviz in a proper directory
you can install in Anaconda from conda-command-prompt using the below command -
conda install -c conda-forge python-graphviz
and replace the previously installed graphviz directory this might help you with the problem
I have no problems using the default model in the sentiment analysis pipeline.
# Allocate a pipeline for sentiment-analysis
nlp = pipeline('sentiment-analysis')
nlp('I am a black man.')
>>>[{'label': 'NEGATIVE', 'score': 0.5723695158958435}]
But, when I try to customise the pipeline a little by adding a specific model. It throws a KeyError.
nlp = pipeline('sentiment-analysis',
tokenizer = AutoTokenizer.from_pretrained("DeepPavlov/bert-base-cased-conversational"),
model = AutoModelWithLMHead.from_pretrained("DeepPavlov/bert-base-cased-conversational"))
nlp('I am a black man.')
>>>---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
<ipython-input-55-af7e46d6c6c9> in <module>
3 tokenizer = AutoTokenizer.from_pretrained("DeepPavlov/bert-base-cased-conversational"),
4 model = AutoModelWithLMHead.from_pretrained("DeepPavlov/bert-base-cased-conversational"))
----> 5 nlp('I am a black man.')
6
7
~/opt/anaconda3/lib/python3.7/site-packages/transformers/pipelines.py in __call__(self, *args, **kwargs)
721 outputs = super().__call__(*args, **kwargs)
722 scores = np.exp(outputs) / np.exp(outputs).sum(-1, keepdims=True)
--> 723 return [{"label": self.model.config.id2label[item.argmax()], "score": item.max().item()} for item in scores]
724
725
~/opt/anaconda3/lib/python3.7/site-packages/transformers/pipelines.py in <listcomp>(.0)
721 outputs = super().__call__(*args, **kwargs)
722 scores = np.exp(outputs) / np.exp(outputs).sum(-1, keepdims=True)
--> 723 return [{"label": self.model.config.id2label[item.argmax()], "score": item.max().item()} for item in scores]
724
725
KeyError: 58129
I am facing the same problem. I am working with a model from XML-R fine-tuned with squadv2 data set ("a-ware/xlmroberta-squadv2"). In my case, the KeyError is 16.
Link
Looking for help on the issue I have found this information: link I hope you find it helpful.
Answer (from the link)
The pipeline throws an exception when the model predicts a token that is not part of the document (e.g. final special token [SEP])
My problem:
from transformers import XLMRobertaTokenizer, XLMRobertaForQuestionAnswering
from transformers import pipeline
nlp = pipeline('question-answering',
model = XLMRobertaForQuestionAnswering.from_pretrained('a-ware/xlmroberta-squadv2'),
tokenizer= XLMRobertaTokenizer.from_pretrained('a-ware/xlmroberta-squadv2'))
nlp(question = "Who was Jim Henson?", context ="Jim Henson was a nice puppet")
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
<ipython-input-15-b5a8ece5e525> in <module>()
1 context = "Jim Henson was a nice puppet"
2 # --------------- CON INTERROGACIONES
----> 3 nlp(question = "Who was Jim Henson?", context =context)
1 frames
/usr/local/lib/python3.6/dist-packages/transformers/pipelines.py in <listcomp>(.0)
1745 ),
1746 }
-> 1747 for s, e, score in zip(starts, ends, scores)
1748 ]
1749
KeyError: 16
Solution 1: Adding punctuation at the end of the context
In order to avoid the bug of trying to extract the final token (which may be an special one as [SEP]) I added an element (in this case a punctuation mark) at the end of the context:
nlp(question = "Who was Jim Henson?", context ="Jim Henson was a nice puppet.")
[OUT]
{'answer': 'nice puppet.', 'end': 28, 'score': 0.5742837190628052, 'start': 17}
Solution 2: Do not use pipeline()
The original model can handle itself to retrieve the correct token`s index.
from transformers import XLMRobertaTokenizer, XLMRobertaForQuestionAnswering
import torch
tokenizer = XLMRobertaTokenizer.from_pretrained('a-ware/xlmroberta-squadv2')
model = XLMRobertaForQuestionAnswering.from_pretrained('a-ware/xlmroberta-squadv2')
question, text = "Who was Jim Henson?", "Jim Henson was a nice puppet"
encoding = tokenizer(question, text, return_tensors='pt')
input_ids = encoding['input_ids']
attention_mask = encoding['attention_mask']
start_scores, end_scores = model(input_ids, attention_mask=attention_mask, output_attentions=False)[:2]
all_tokens = tokenizer.convert_ids_to_tokens(input_ids[0])
answer = ' '.join(all_tokens[torch.argmax(start_scores) : torch.argmax(end_scores)+1])
answer = tokenizer.convert_tokens_to_ids(answer.split())
answer = tokenizer.decode(answer)
Update
Looking in more detail your case, I found that the default model for Conversational task in the pipeline is distilbert-base-cased (source code).
The first solution I posted is not a good solution indeed. Trying other questions I got the same error. However, the model itself outside the pipeline works fine (as I showed in solution 2). Thus, I believe that not all models can be introduced in the pipeline. If anyone has more information about it please help us out. Thanks.