I am following this tutorial with my custom data and my custom S3 buckets where train and validation data are. I am getting the following error:
Customer Error: imread read blank (None) image for file: /opt/ml/input/data/train/s3://image-classification/image_classification_model_data/train/img-001.png
I have all my training data are in one folder named 'train' I have set up my lst file like this suggested by doc,
22 1 s3://image-classification/image_classification_model_data/train/img-001.png
86 0 s3://image-classification/image_classification_model_data/train/img-002.png
...
My other configurations:
s3_bucket = 'image-classification'
prefix = 'image_classification_model_data'
s3train = 's3://{}/{}/train/'.format(s3_bucket, prefix)
s3validation = 's3://{}/{}/validation/'.format(s3_bucket, prefix)
s3train_lst = 's3://{}/{}/train_lst/'.format(s3_bucket, prefix)
s3validation_lst = 's3://{}/{}/validation_lst/'.format(s3_bucket, prefix)
train_data = sagemaker.inputs.TrainingInput(s3train, distribution='FullyReplicated',
content_type='application/x-image', s3_data_type='S3Prefix')
validation_data = sagemaker.inputs.TrainingInput(s3validation, distribution='FullyReplicated',
content_type='application/x-image', s3_data_type='S3Prefix')
train_data_lst = sagemaker.inputs.TrainingInput(s3train_lst, distribution='FullyReplicated',
content_type='application/x-image', s3_data_type='S3Prefix')
validation_data_lst = sagemaker.inputs.TrainingInput(s3validation_lst, distribution='FullyReplicated',
content_type='application/x-image', s3_data_type='S3Prefix')
data_channels = {'train': train_data, 'validation': validation_data, 'train_lst': train_data_lst,
'validation_lst': validation_data_lst}
I checked the images downloaded and checked physically, I see the image. Now sure what this error gets thrown out as blank. Any suggestion would be great.
Sagemaker copies the input data you specify in s3train into the instance in /opt/ml/input/data/train/ and that's why you have an error, because as you can see from the error message is trying to concatenate the filename in the lst file with the path where it expect the image to be. So just put only the filenames in your lstand should be fine (remove the s3 path).
Related
I have a managed jupyter notebook in Vertex AI that I want to schedule. The notebook works just fine as long as I start it manually, but as soon as it is scheduled, it fails. There are in fact many things that go wrong when scheduled, some of them are fixable. Before explaining what my trouble is, let me first give some details of the context.
The notebook gathers information from an API for several stores and saves the data in different folders before processing it, saving csv-files to store-specific folders and to bigquery. So, in the location of the notebook, I have:
The notebook
Functions needed for the handling of data (as *.py files)
A series of folders, some of which have subfolders which also have subfolders
When I execute this manually, no problem. Everything works well and all files end up exactly where they should, as well as in different bigQuery tables.
However, when scheduling the execution of the notebook, everything goes wrong. First, the files *.py cannot be read (as import). No problem, I added the functions in the notebook.
Now, the following error is where I am at a loss, because I have no idea why it does work or how to fix it. The code that leads to the error is the following:
internal = "https://api.************************"
df_descriptions = []
storess = internal
response_stores = requests.get(storess,auth = HTTPBasicAuth(userInternal, keyInternal))
pathlib.Path("stores/request_1.json").write_bytes(response_stores.content)
filepath = "stores"
files = os.listdir(filepath)
for file in files:
with open(filepath + "/"+file) as json_string:
jsonstr = json.load(json_string)
information = pd.json_normalize(jsonstr)
df_descriptions.append(information)
StoreINFO = pd.concat(df_descriptions)
StoreINFO = StoreINFO.dropna()
StoreINFO = StoreINFO[StoreINFO['storeIdMappings'].map(lambda d: len(d)) > 0]
cloud_store_ids = list(set(StoreINFO.cloudStoreId))
LastWeek = datetime.date.today()- timedelta(days=2)
LastWeek =np.datetime64(LastWeek)
and the error reported is:
FileNotFoundError Traceback (most recent call last)
/tmp/ipykernel_165/2970402631.py in <module>
5 storess = internal
6 response_stores = requests.get(storess,auth = HTTPBasicAuth(userInternal, keyInternal))
----> 7 pathlib.Path("stores/request_1.json").write_bytes(response_stores.content)
8
9 filepath = "stores"
/opt/conda/lib/python3.7/pathlib.py in write_bytes(self, data)
1228 # type-check for the buffer interface before truncating the file
1229 view = memoryview(data)
-> 1230 with self.open(mode='wb') as f:
1231 return f.write(view)
1232
/opt/conda/lib/python3.7/pathlib.py in open(self, mode, buffering, encoding, errors, newline)
1206 self._raise_closed()
1207 return io.open(self, mode, buffering, encoding, errors, newline,
-> 1208 opener=self._opener)
1209
1210 def read_bytes(self):
/opt/conda/lib/python3.7/pathlib.py in _opener(self, name, flags, mode)
1061 def _opener(self, name, flags, mode=0o666):
1062 # A stub for the opener argument to built-in open()
-> 1063 return self._accessor.open(self, flags, mode)
1064
1065 def _raw_open(self, flags, mode=0o777):
FileNotFoundError: [Errno 2] No such file or directory: 'stores/request_1.json'
I would gladly find another way to do this, for instance by using GCS buckets, but my issue is the existence of sub-folders. There are many stores and I do not wish to do this operation manually because some retailers for which I am doing this have over 1000 stores. My python code generates all these folders and as I understand it, this is not feasible in GCS.
How can I solve this issue?
GCS uses a flat namespace, so folders don't actually exist, but can be simulated as given in this documentation.For your requirement, you can either use absolute path (starting with "/" -- not relative) or create the "stores" directory (with "mkdir"). For more information you can check this blog.
using detect_fn = tf.keras.models.load_model(PATH_TO_SAVED_MODEL) I get the error above. My filepath is generated when downloading the model:
def download_model(model_name, model_date):
base_url = 'http://download.tensorflow.org/models/object_detection/tf2/'
model_file = model_name + '.tar.gz'
model_dir = tf.keras.utils.get_file(fname=model_name,
origin=base_url + model_date + '/' + model_file,
untar=True)
return str(model_dir)
and is ~/.keras/datasets/centernet_hg104_1024x1024_coco17/saved_model with a saved_model.pb file within the directory. It seems to be asking for a saveed_model.pb file but does not accept the directory I provide which contains the file. Any idea what the issue is? is it an error with the tf.keras.utils.get_file() downloading or is it an error with my loading?
# ***** Reading the Data********
if processed_first:
#Reading all the features and labels for this chunk
shared_list = []
p = threading.Thread(target=read_lab_fea, args=(cfg_file, is_production, shared_list, output_folder))
p.start()
p.join()
data_name = shared_list[0]
data_end_index = shared_list[1]
fea_dict = shared_list[2]
lab_dict = shared_list[3]
arch_dict = shared_list[4]
data_set = shared_list[5]
enter image description here
First I did run kaldi's run.sh file
When I did that, I corrected cmd.sh's contents.
Original --> call queue.pl
to --> call run.pl
Because I met bug when i run original source
Reference : https://www.google.com/url?q=https://groups.google.com/g/kaldi-help/c/tokwXTLdGFY?pli%3D1&sa=D&source=editors&ust=1631002151871000&usg=AOvVaw1FYQHJEmI-kkAAeAB2tcKt
enter image description here
I found that fea_dict and lab_dict in data_io.py has no shared element. How can I progress the TIMIT tutorial experiments?
I'm doing experiment using cfg/TIMIT_baselines/TIMIT_MLP_mfcc_basic.cfg file. Just correcting absolute directory of linux.
I refered https://github.com/mravanelli/pytorch-kaldi/issues/185
run copy-feats.
I saw--> kaldierror::KaldiFatalError
Issue
Following the official instruction, I'm trying to add an extra training dataset and train a model on local cpu environment.
But I don't change the content of base_config.cfg and config.cfg files.
How can I fix these errors to build a model and evaluate it?
Error
I'm not sure about the first one is an issue or not, and I have no idea to fill the config.cfg file.
The config.cfg file was an empty even after executing the code on the below procedure so far section.
The error message was shown when executing train command.
ℹ Using CPU
✘ Error parsing config overrides
paths -> train not a section value that can be overwritten
Code
$ python3 -m spacy train config.cfg --output ./output --paths.train train.spacy --paths.dev train.spacy
Procedure so far
$ python3 -m spacy init fill-config base_config.cfg config.cfg
#former output
configparser.DuplicateSectionError: While reading from '<string>' [line 90]: section 'paths' already exists
#current output
File "/usr/local/lib/python3.9/site-packages/spacy/util.py", line 137, in get
raise RegistryError(
catalogue.RegistryError: [E893] Could not find function 'spacy.MultiHashEmbed.v2' in function registry 'architectures'. If you're using a custom function, make sure the code is available. If the function is provided by a third-party package, e.g. spacy-transformers, make sure the package is installed in your environment.
Available names: spacy-legacy.MaxoutWindowEncoder.v1, spacy-legacy.MishWindowEncoder.v1, spacy-legacy.TextCatEnsemble.v1, spacy-legacy.Tok2Vec.v1, spacy-legacy.WandbLogger.v1, spacy.CharacterEmbed.v1, spacy.EntityLinker.v1, spacy.HashEmbedCNN.v1, spacy.MaxoutWindowEncoder.v2, spacy.MishWindowEncoder.v2, spacy.MultiHashEmbed.v1, spacy.PretrainCharacters.v1, spacy.PretrainVectors.v1, spacy.Tagger.v1, spacy.TextCatBOW.v1, spacy.TextCatCNN.v1, spacy.TextCatEnsemble.v2, spacy.TextCatLowData.v1, spacy.Tok2Vec.v2, spacy.Tok2VecListener.v1, spacy.TorchBiLSTMEncoder.v1, spacy.TransitionBasedParser.v1, spacy.TransitionBasedParser.v2
The config.cfg file was empty even after executing the above code.
base_config.cfg downloaded from the form on the official instruction
# This is an auto-generated partial config. To use it with 'spacy train'
# you can run spacy init fill-config to auto-fill all default settings:
# python -m spacy init fill-config ./base_config.cfg ./config.cfg
[paths]
train = null
dev = null
[system]
gpu_allocator = null
[nlp]
lang = "en"
pipeline = ["tok2vec","ner"]
batch_size = 1000
[components]
[components.tok2vec]
factory = "tok2vec"
[components.tok2vec.model]
#architectures = "spacy.Tok2Vec.v2"
[components.tok2vec.model.embed]
#architectures = "spacy.MultiHashEmbed.v2"
width = ${components.tok2vec.model.encode.width}
attrs = ["ORTH", "SHAPE"]
rows = [5000, 2500]
include_static_vectors = false
[components.tok2vec.model.encode]
#architectures = "spacy.MaxoutWindowEncoder.v2"
width = 96
depth = 4
window_size = 1
maxout_pieces = 3
[components.ner]
factory = "ner"
[components.ner.model]
#architectures = "spacy.TransitionBasedParser.v2"
state_type = "ner"
extra_state_tokens = false
hidden_width = 64
maxout_pieces = 2
use_upper = true
nO = null
[components.ner.model.tok2vec]
#architectures = "spacy.Tok2VecListener.v1"
width = ${components.tok2vec.model.encode.width}
[corpora]
[corpora.train]
#readers = "spacy.Corpus.v1"
path = ${paths.train}
max_length = 0
[corpora.dev]
#readers = "spacy.Corpus.v1"
path = ${paths.dev}
max_length = 0
[training]
dev_corpus = "corpora.dev"
train_corpus = "corpora.train"
[training.optimizer]
#optimizers = "Adam.v1"
[training.batcher]
#batchers = "spacy.batch_by_words.v1"
discard_oversize = false
tolerance = 0.2
[training.batcher.size]
#schedules = "compounding.v1"
start = 100
stop = 1000
compound = 1.001
[initialize]
vectors = ${paths.vectors}
It looks like you double-pasted the config or something? From the errors you'll note that it says you have two [paths] sections. About halfway through your file there's a comment like this:
# This is an auto-generated partial config. To use it with 'spacy train'
Try deleting everything from there and down and then doing it again.
Pretraining custom weights in Spacy pipeline gives error.
When I run this using a jupyter notebook:
!python -m spacy pretrain config.cfg ./output_pretrain --paths.raw_text ./data.jsonl
The following error occurs:
[i] Using CPU
[i] Loading config from: config.cfg
[x] Error parsing config overrides
paths -> raw_text not a section value that can be overwritten
My Environment
Operating System: Windows 10.0.19041
Python Version Used: 3.8.8
spaCy Version Used:3.0.5
The way the spacy pretrain command works is like this: you provide it a config file, and an output directory. Optionally, you can override certain settings from the config file on the command line. For instance, if your config file would contain this block:
[paths]
train = null
dev = null
raw_text = "path/to/toy_data"
Then on the command line you can do --raw_text="path/to/production_data" to change that one specific value.
Now, it looks like you're getting this error message because you don't have an entry raw_text in the paths block, so you can't overwrite it.
In fact, I wonder whether you have defined a pretraining block in your config file?
If not - have a look at spacy init config. You can provide it with the option -pt which will fill in default values for the pretraining step, specifically it will create something like this (among the other 'normal' settings):
[paths]
raw_text = null
...
[pretraining]
max_epochs = 1000
dropout = 0.2
n_save_every = null
component = "tok2vec"
layer = ""
corpus = "corpora.pretrain"
[pretraining.batcher]
#batchers = "spacy.batch_by_words.v1"
size = 3000
discard_oversize = false
tolerance = 0.2
get_length = null
[pretraining.objective]
#architectures = "spacy.PretrainCharacters.v1"
maxout_pieces = 3
hidden_size = 300
n_characters = 4
[pretraining.optimizer]
#optimizers = "Adam.v1"
beta1 = 0.9
beta2 = 0.999
L2_is_weight_decay = true
L2 = 0.01
grad_clip = 1.0
use_averages = true
eps = 0.00000001
learn_rate = 0.001
...
[corpora.pretrain]
#readers = "spacy.JsonlCorpus.v1"
path = ${paths.raw_text}
min_length = 5
max_length = 500
limit = 0
Notice how the raw_text variable is used by a corpus reader, that will stream the raw text for your pretraining step. So you need to make sure all of this is wired up correctly before you can run the pretraining.