diff --git a/requirements.txt b/requirements.txt
index 89987bc0..ec9e41d1 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -11,7 +11,6 @@ jsonschema>=2.6.0
 numpy>=1.10
 pathlib2>=2.3.0
 Pillow>=4.1.1
-pigar==0.9.2
 plotly>=3.9.0
 psutil>=3.4.2
 pyparsing>=2.0.3
diff --git a/trains/backend_interface/task/repo/scriptinfo.py b/trains/backend_interface/task/repo/scriptinfo.py
index b2e12ed2..42817a01 100644
--- a/trains/backend_interface/task/repo/scriptinfo.py
+++ b/trains/backend_interface/task/repo/scriptinfo.py
@@ -26,103 +26,18 @@ class ScriptRequirements(object):
     def __init__(self, root_folder):
         self._root_folder = root_folder
 
-    @staticmethod
-    def get_installed_pkgs_detail(reqs):
-        """
-        HACK: bugfix of the original pigar get_installed_pkgs_detail
-
-        Get mapping for import top level name
-        and install package name with version.
-        """
-        mapping = dict()
-
-        for path in sys.path:
-            if os.path.isdir(path) and path.rstrip('/').endswith(
-                    ('site-packages', 'dist-packages')):
-                new_mapping = reqs._search_path(path)
-                # BUGFIX:
-                # override with previous, just like python resolves imports, the first match is the one used.
-                # unlike the original implementation, where the last one is used.
-                new_mapping.update(mapping)
-                mapping = new_mapping
-
-        # HACK: prefer tensorflow_gpu over tensorflow
-        if 'tensorflow_gpu' in new_mapping:
-            new_mapping['tensorflow'] = new_mapping['tensorflow_gpu']
-
-        return mapping
-
     def get_requirements(self):
         try:
-            from pigar import reqs
-            reqs.project_import_modules = ScriptRequirements._patched_project_import_modules
-            from pigar.__main__ import GenerateReqs
-            from pigar.log import logger
-            logger.setLevel(logging.WARNING)
-            try:
-                # first try our version, if we fail revert to the internal implantation
-                installed_pkgs = self.get_installed_pkgs_detail(reqs)
-            except Exception:
-                installed_pkgs = reqs.get_installed_pkgs_detail()
+            from ....utilities.pigar.reqs import get_installed_pkgs_detail
+            from ....utilities.pigar.__main__ import GenerateReqs
+            installed_pkgs = get_installed_pkgs_detail()
             gr = GenerateReqs(save_path='', project_path=self._root_folder, installed_pkgs=installed_pkgs,
                               ignores=['.git', '.hg', '.idea', '__pycache__', '.ipynb_checkpoints'])
-            reqs, try_imports, guess = gr.extract_reqs()
-            return self.create_requirements_txt(reqs)
+            reqs, try_imports, guess, local_pks = gr.extract_reqs(module_callback=ScriptRequirements.add_trains_used_packages)
+            return self.create_requirements_txt(reqs, local_pks)
         except Exception:
             return '', ''
 
-    @staticmethod
-    def _patched_project_import_modules(project_path, ignores):
-        """
-        copied form pigar req.project_import_modules
-        patching, os.getcwd() is incorrectly used
-        """
-        from pigar.modules import ImportedModules
-        from pigar.reqs import file_import_modules
-        modules = ImportedModules()
-        try_imports = set()
-        local_mods = list()
-        ignore_paths = collections.defaultdict(set)
-        if not ignores:
-            ignore_paths[project_path].add('.git')
-        else:
-            for path in ignores:
-                parent_dir = os.path.dirname(path)
-                ignore_paths[parent_dir].add(os.path.basename(path))
-
-        if os.path.isfile(project_path):
-            fake_path = Path(project_path).name
-            with open(project_path, 'rb') as f:
-                fmodules, try_ipts = file_import_modules(fake_path, f.read())
-                modules |= fmodules
-                try_imports |= try_ipts
-        else:
-            cur_dir = project_path  # os.getcwd()
-            for dirpath, dirnames, files in os.walk(project_path, followlinks=True):
-                if dirpath in ignore_paths:
-                    dirnames[:] = [d for d in dirnames
-                                   if d not in ignore_paths[dirpath]]
-                py_files = list()
-                for fn in files:
-                    # C extension.
-                    if fn.endswith('.so'):
-                        local_mods.append(fn[:-3])
-                    # Normal Python file.
-                    if fn.endswith('.py'):
-                        local_mods.append(fn[:-3])
-                        py_files.append(fn)
-                if '__init__.py' in files:
-                    local_mods.append(os.path.basename(dirpath))
-                for file in py_files:
-                    fpath = os.path.join(dirpath, file)
-                    fake_path = fpath.split(cur_dir)[1][1:]
-                    with open(fpath, 'rb') as f:
-                        fmodules, try_ipts = file_import_modules(fake_path, f.read())
-                        modules |= fmodules
-                        try_imports |= try_ipts
-
-        return ScriptRequirements.add_trains_used_packages(modules), try_imports, local_mods
-
     @staticmethod
     def add_trains_used_packages(modules):
         # hack: forcefully insert storage modules if we have them
@@ -159,7 +74,7 @@ class ScriptRequirements(object):
         return modules
 
     @staticmethod
-    def create_requirements_txt(reqs):
+    def create_requirements_txt(reqs, local_pks=None):
         # write requirements.txt
         try:
             conda_requirements = ''
@@ -188,6 +103,11 @@ class ScriptRequirements(object):
         # python version header
         requirements_txt = '# Python ' + sys.version.replace('\n', ' ').replace('\r', ' ') + '\n'
 
+        if local_pks:
+            requirements_txt += '\n# Local modules found - skipping:\n'
+            for k, v in local_pks.sorted_items():
+                requirements_txt += '# {0} == {1}\n'.format(k, v.version)
+
         # requirement summary
         requirements_txt += '\n'
         for k, v in reqs.sorted_items():
@@ -203,6 +123,13 @@ class ScriptRequirements(object):
         requirements_txt += '\n' + \
                             '# Detailed import analysis\n' \
                             '# **************************\n'
+
+        if local_pks:
+            for k, v in local_pks.sorted_items():
+                requirements_txt += '\n'
+                requirements_txt += '# IMPORT LOCAL PACKAGE {0}\n'.format(k)
+                requirements_txt += ''.join(['# {0}\n'.format(c) for c in v.comments.sorted_items()])
+
         for k, v in reqs.sorted_items():
             requirements_txt += '\n'
             if k == '-e':
@@ -262,9 +189,9 @@ class _JupyterObserver(object):
         # load pigar
         # noinspection PyBroadException
         try:
-            from pigar.reqs import get_installed_pkgs_detail, file_import_modules
-            from pigar.modules import ReqsModules
-            from pigar.log import logger
+            from ....utilities.pigar.reqs import get_installed_pkgs_detail, file_import_modules
+            from ....utilities.pigar.modules import ReqsModules
+            from ....utilities.pigar.log import logger
             logger.setLevel(logging.WARNING)
         except Exception:
             file_import_modules = None
diff --git a/trains/utilities/pigar/__init__.py b/trains/utilities/pigar/__init__.py
new file mode 100644
index 00000000..40a96afc
--- /dev/null
+++ b/trains/utilities/pigar/__init__.py
@@ -0,0 +1 @@
+# -*- coding: utf-8 -*-
diff --git a/trains/utilities/pigar/__main__.py b/trains/utilities/pigar/__main__.py
new file mode 100644
index 00000000..7508b799
--- /dev/null
+++ b/trains/utilities/pigar/__main__.py
@@ -0,0 +1,133 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import os
+import codecs
+
+from .reqs import project_import_modules, is_std_or_local_lib
+from .utils import lines_diff
+from .log import logger
+from .modules import ReqsModules
+
+
+class GenerateReqs(object):
+
+    def __init__(self, save_path, project_path, ignores,
+                 installed_pkgs, comparison_operator='=='):
+        self._save_path = save_path
+        self._project_path = project_path
+        self._ignores = ignores
+        self._installed_pkgs = installed_pkgs
+        self._maybe_local_mods = set()
+        self._local_mods = dict()
+        self._comparison_operator = comparison_operator
+
+    def extract_reqs(self, module_callback=None):
+        """Extract requirements from project."""
+
+        reqs = ReqsModules()
+        guess = ReqsModules()
+        local = ReqsModules()
+        modules, try_imports, local_mods = project_import_modules(
+            self._project_path, self._ignores)
+        if module_callback:
+            modules = module_callback(modules)
+        app_name = os.path.basename(self._project_path)
+        if app_name in local_mods:
+            local_mods.remove(app_name)
+
+        # Filtering modules
+        candidates = self._filter_modules(modules, local_mods)
+
+        logger.info('Check module in local environment.')
+        for name in candidates:
+            logger.info('Checking module: %s', name)
+            if name in self._installed_pkgs:
+                pkg_name, version = self._installed_pkgs[name]
+                reqs.add(pkg_name, version, modules[name])
+            else:
+                guess.add(name, 0, modules[name])
+
+        # add local modules, so we know what is used but not installed.
+        for name in self._local_mods:
+            if name in modules:
+                relpath = os.path.relpath(self._local_mods[name], self._project_path)
+                if not relpath.startswith('.'):
+                    relpath = '.' + os.path.sep + relpath
+                local.add(name, relpath, modules[name])
+
+        return reqs, try_imports, guess, local
+
+    def _write_reqs(self, reqs):
+        print('Writing requirements to "{0}"'.format(
+            self._save_path))
+        with open(self._save_path, 'w+') as f:
+            f.write('# Requirements automatically generated by pigar.\n'
+                    '# https://github.com/damnever/pigar\n')
+            for k, v in reqs.sorted_items():
+                f.write('\n')
+                f.write(''.join(['# {0}\n'.format(c)
+                                 for c in v.comments.sorted_items()]))
+                if k == '-e':
+                    f.write('{0} {1}\n'.format(k, v.version))
+                elif v:
+                    f.write('{0} {1} {2}\n'.format(
+                        k, self._comparison_operator, v.version))
+                else:
+                    f.write('{0}\n'.format(k))
+
+    def _best_matchs(self, name, pkgs):
+        # If imported name equals to package name.
+        if name in pkgs:
+            return [pkgs[pkgs.index(name)]]
+        # If not, return all possible packages.
+        return pkgs
+
+    def _filter_modules(self, modules, local_mods):
+        candidates = set()
+
+        logger.info('Filtering modules ...')
+        for module in modules:
+            logger.info('Checking module: %s', module)
+            if not module or module.startswith('.'):
+                continue
+            if module in local_mods:
+                self._maybe_local_mods.add(module)
+            module_std_local = is_std_or_local_lib(module)
+            if module_std_local is True:
+                continue
+            if isinstance(module_std_local, str):
+                self._local_mods[module] = module_std_local
+                continue
+            candidates.add(module)
+
+        return candidates
+
+    def _invalid_reqs(self, reqs):
+        for name, detail in reqs.sorted_items():
+            print(
+                '  {0} referenced from:\n    {1}'.format(
+                    name,
+                    '\n    '.join(detail.comments.sorted_items())
+                )
+            )
+
+    def _save_old_reqs(self):
+        if os.path.isfile(self._save_path):
+            with codecs.open(self._save_path, 'rb', 'utf-8') as f:
+                self._old_reqs = f.readlines()
+
+    def _reqs_diff(self):
+        if not hasattr(self, '_old_reqs'):
+            return
+        with codecs.open(self._save_path, 'rb', 'utf-8') as f:
+            new_reqs = f.readlines()
+        is_diff, diffs = lines_diff(self._old_reqs, new_reqs)
+        msg = 'Requirements file has been covered, '
+        if is_diff:
+            msg += 'there is the difference:'
+            print('{0}\n{1}'.format(msg, ''.join(diffs)), end='')
+        else:
+            msg += 'no difference.'
+            print(msg)
diff --git a/trains/utilities/pigar/extractor/__init__.py b/trains/utilities/pigar/extractor/__init__.py
new file mode 100644
index 00000000..b9b0ac03
--- /dev/null
+++ b/trains/utilities/pigar/extractor/__init__.py
@@ -0,0 +1,12 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+
+from ..utils import PY32
+
+
+if PY32:
+    from .thread_extractor import ThreadExtractor as Extractor
+else:
+    from .gevent_extractor import GeventExtractor as Extractor
diff --git a/trains/utilities/pigar/extractor/extractor.py b/trains/utilities/pigar/extractor/extractor.py
new file mode 100644
index 00000000..08464f6c
--- /dev/null
+++ b/trains/utilities/pigar/extractor/extractor.py
@@ -0,0 +1,36 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import multiprocessing
+
+
+class BaseExtractor(object):
+
+    def __init__(self, names, max_workers=None):
+        self._names = names
+        self._max_workers = max_workers or (multiprocessing.cpu_count() * 4)
+
+    def run(self, job):
+        try:
+            self.extract(job)
+            self.wait_complete()
+        except KeyboardInterrupt:
+            print('** Shutting down ...')
+            self.shutdown()
+        else:
+            print('^.^ Extracting all packages done!')
+        finally:
+            self.final()
+
+    def extract(self, job):
+        raise NotImplemented
+
+    def wait_complete(self):
+        raise NotImplemented
+
+    def shutdown(self):
+        raise NotImplemented
+
+    def final(self):
+        pass
diff --git a/trains/utilities/pigar/extractor/gevent_extractor.py b/trains/utilities/pigar/extractor/gevent_extractor.py
new file mode 100644
index 00000000..ba7363c5
--- /dev/null
+++ b/trains/utilities/pigar/extractor/gevent_extractor.py
@@ -0,0 +1,50 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import sys
+
+import greenlet
+from gevent.pool import Pool
+
+from .extractor import BaseExtractor
+from ..log import logger
+
+
+class GeventExtractor(BaseExtractor):
+
+    def __init__(self, names, max_workers=222):
+        super(self.__class__, self).__init__(names, max_workers)
+        self._pool = Pool(self._max_workers)
+        self._exited_greenlets = 0
+
+    def extract(self, job):
+        job = self._job_wrapper(job)
+        for name in self._names:
+            if self._pool.full():
+                self._pool.wait_available()
+            self._pool.spawn(job, name)
+
+    def _job_wrapper(self, job):
+        def _job(name):
+            result = None
+            try:
+                result = job(name)
+            except greenlet.GreenletExit:
+                self._exited_greenlets += 1
+            except Exception:
+                e = sys.exc_info()[1]
+                logger.error('Extracting "{0}", got: {1}'.format(name, e))
+            return result
+        return _job
+
+    def wait_complete(self):
+        self._pool.join()
+
+    def shutdown(self):
+        self._pool.kill(block=True)
+
+    def final(self):
+        count = self._exited_greenlets
+        if count != 0:
+            print('** {0} running job exited.'.format(count))
diff --git a/trains/utilities/pigar/extractor/thread_extractor.py b/trains/utilities/pigar/extractor/thread_extractor.py
new file mode 100644
index 00000000..e0a4ce73
--- /dev/null
+++ b/trains/utilities/pigar/extractor/thread_extractor.py
@@ -0,0 +1,45 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import concurrent.futures
+
+from .extractor import BaseExtractor
+
+from ..log import logger
+
+
+class ThreadExtractor(BaseExtractor):
+    """Extractor use thread pool execute tasks.
+
+    Can be used to extract /simple/<pkg_name> or /pypi/<pkg_name>/json.
+
+    FIXME: can not deliver SIG_INT to threads in Python 2.
+    """
+
+    def __init__(self, names, max_workers=None):
+        super(self.__class__, self).__init__(names, max_workers)
+        self._futures = dict()
+
+    def extract(self, job):
+        """Extract url by package name."""
+        with concurrent.futures.ThreadPoolExecutor(
+                max_workers=self._max_workers) as executor:
+            for name in self._names:
+                self._futures[executor.submit(job, name)] = name
+
+    def wait_complete(self):
+        """Wait for futures complete done."""
+        for future in concurrent.futures.as_completed(self._futures.keys()):
+            try:
+                error = future.exception()
+            except concurrent.futures.CancelledError:
+                break
+            name = self._futures[future]
+            if error is not None:
+                err_msg = 'Extracting "{0}", got: {1}'.format(name, error)
+                logger.error(err_msg)
+
+    def shutdown(self):
+        for future in self._futures:
+            future.cancel()
diff --git a/trains/utilities/pigar/log.py b/trains/utilities/pigar/log.py
new file mode 100644
index 00000000..57d3ee15
--- /dev/null
+++ b/trains/utilities/pigar/log.py
@@ -0,0 +1,9 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import logging.handlers
+
+
+logger = logging.getLogger('pigar')
+logger.setLevel(logging.WARNING)
diff --git a/trains/utilities/pigar/modules.py b/trains/utilities/pigar/modules.py
new file mode 100644
index 00000000..da71529a
--- /dev/null
+++ b/trains/utilities/pigar/modules.py
@@ -0,0 +1,111 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import collections
+
+
+# FIXME: Just a workaround, not a radical cure..
+_special_cases = {
+    "dogpile.cache": "dogpile.cache",
+    "dogpile.core": "dogpile.core",
+    "ruamel.yaml": "ruamel.yaml",
+    "ruamel.ordereddict": "ruamel.ordereddict",
+}
+
+
+class Modules(dict):
+    """Modules object will be used to store modules information."""
+
+    def __init__(self):
+        super(Modules, self).__init__()
+
+
+class ImportedModules(Modules):
+
+    def __init__(self):
+        super(ImportedModules, self).__init__()
+
+    def add(self, name, file, lineno):
+        if name is None:
+            return
+
+        names = list()
+        special_name = '.'.join(name.split('.')[:2])
+        # Flask extension.
+        if name.startswith('flask.ext.'):
+            names.append('flask')
+            names.append('flask_' + name.split('.')[2])
+        # Special cases..
+        elif special_name in _special_cases:
+            names.append(_special_cases[special_name])
+        # Other.
+        elif '.' in name and not name.startswith('.'):
+            names.append(name.split('.')[0])
+        else:
+            names.append(name)
+
+        for nm in names:
+            if nm not in self:
+                self[nm] = _Locations()
+            self[nm].add(file, lineno)
+
+    def __or__(self, obj):
+        for name, locations in obj.items():
+            for file, linenos in locations.items():
+                for lineno in linenos:
+                    self.add(name, file, lineno)
+        return self
+
+
+class ReqsModules(Modules):
+
+    _Detail = collections.namedtuple('Detail', ['version', 'comments'])
+
+    def __init__(self):
+        super(ReqsModules, self).__init__()
+        self._sorted = None
+
+    def add(self, package, version, locations):
+        if package in self:
+            self[package].comments.extend(locations)
+        else:
+            self[package] = self._Detail(version, locations)
+
+    def sorted_items(self):
+        if self._sorted is None:
+            self._sorted = sorted(self.items())
+        return self._sorted
+
+    def remove(self, *names):
+        for name in names:
+            if name in self:
+                self.pop(name)
+        self._sorted = None
+
+
+class _Locations(dict):
+    """_Locations store code locations(file, linenos)."""
+
+    def __init__(self):
+        super(_Locations, self).__init__()
+        self._sorted = None
+
+    def add(self, file, lineno):
+        if file in self and lineno not in self[file]:
+            self[file].append(lineno)
+        else:
+            self[file] = [lineno]
+
+    def extend(self, obj):
+        for file, linenos in obj.items():
+            for lineno in linenos:
+                self.add(file, lineno)
+
+    def sorted_items(self):
+        if self._sorted is None:
+            self._sorted = [
+                '{0}: {1}'.format(f, ','.join([str(n) for n in sorted(ls)]))
+                for f, ls in sorted(self.items())
+            ]
+        return self._sorted
diff --git a/trains/utilities/pigar/reqs.py b/trains/utilities/pigar/reqs.py
new file mode 100644
index 00000000..3de530ca
--- /dev/null
+++ b/trains/utilities/pigar/reqs.py
@@ -0,0 +1,398 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import os
+import sys
+import fnmatch
+import importlib
+import imp
+import ast
+import doctest
+import collections
+import functools
+from pathlib2 import Path
+try:
+    from types import FileType  # py2
+except ImportError:
+    from io import IOBase as FileType  # py3
+
+from .log import logger
+from .utils import parse_git_config
+from .modules import ImportedModules
+
+
+def project_import_modules(project_path, ignores):
+    """
+    copied form pigar req.project_import_modules patching, os.getcwd() is incorrectly used
+    """
+    modules = ImportedModules()
+    try_imports = set()
+    local_mods = list()
+    ignore_paths = collections.defaultdict(set)
+    if not ignores:
+        ignore_paths[project_path].add('.git')
+    else:
+        for path in ignores:
+            parent_dir = os.path.dirname(path)
+            ignore_paths[parent_dir].add(os.path.basename(path))
+
+    if os.path.isfile(project_path):
+        fake_path = Path(project_path).name
+        with open(project_path, 'rb') as f:
+            fmodules, try_ipts = file_import_modules(fake_path, f.read())
+            modules |= fmodules
+            try_imports |= try_ipts
+    else:
+        cur_dir = project_path  # os.getcwd()
+        for dirpath, dirnames, files in os.walk(project_path, followlinks=True):
+            if dirpath in ignore_paths:
+                dirnames[:] = [d for d in dirnames
+                               if d not in ignore_paths[dirpath]]
+            py_files = list()
+            for fn in files:
+                # C extension.
+                if fn.endswith('.so'):
+                    local_mods.append(fn[:-3])
+                # Normal Python file.
+                if fn.endswith('.py'):
+                    local_mods.append(fn[:-3])
+                    py_files.append(fn)
+            if '__init__.py' in files:
+                local_mods.append(os.path.basename(dirpath))
+            for file in py_files:
+                fpath = os.path.join(dirpath, file)
+                fake_path = fpath.split(cur_dir)[1][1:]
+                with open(fpath, 'rb') as f:
+                    fmodules, try_ipts = file_import_modules(fake_path, f.read())
+                    modules |= fmodules
+                    try_imports |= try_ipts
+
+    return modules, try_imports, local_mods
+
+
+def file_import_modules(fpath, fdata):
+    """Get single file all imported modules."""
+    modules = ImportedModules()
+    str_codes = collections.deque([(fdata, 1)])
+    try_imports = set()
+
+    while str_codes:
+        str_code, lineno = str_codes.popleft()
+        ic = ImportChecker(fpath, lineno)
+        try:
+            parsed = ast.parse(str_code)
+            ic.visit(parsed)
+        # Ignore SyntaxError in Python code.
+        except SyntaxError:
+            pass
+        modules |= ic.modules
+        str_codes.extend(ic.str_codes)
+        try_imports |= ic.try_imports
+        del ic
+
+    return modules, try_imports
+
+
+class ImportChecker(object):
+
+    def __init__(self, fpath, lineno):
+        self._fpath = fpath
+        self._lineno = lineno - 1
+        self._modules = ImportedModules()
+        self._str_codes = collections.deque()
+        self._try_imports = set()
+
+    def visit_Import(self, node, try_=False):
+        """As we know: `import a [as b]`."""
+        lineno = node.lineno + self._lineno
+        for alias in node.names:
+            self._modules.add(alias.name, self._fpath, lineno)
+            if try_:
+                self._try_imports.add(alias.name)
+
+    def visit_ImportFrom(self, node, try_=False):
+        """
+        As we know: `from a import b [as c]`. If node.level is not 0,
+        import statement like this `from .a import b`.
+        """
+        mod_name = node.module
+        level = node.level
+        if mod_name is None:
+            level -= 1
+            mod_name = ''
+        for alias in node.names:
+            name = level*'.' + mod_name + '.' + alias.name
+            self._modules.add(name, self._fpath, node.lineno + self._lineno)
+            if try_:
+                self._try_imports.add(name)
+
+    def visit_TryExcept(self, node):
+        """
+        If modules which imported by `try except` and not found,
+        maybe them come from other Python version.
+        """
+        for ipt in node.body:
+            if ipt.__class__.__name__.startswith('Import'):
+                method = 'visit_' + ipt.__class__.__name__
+                getattr(self, method)(ipt, True)
+        for handler in node.handlers:
+            for ipt in handler.body:
+                if ipt.__class__.__name__.startswith('Import'):
+                    method = 'visit_' + ipt.__class__.__name__
+                    getattr(self, method)(ipt, True)
+
+    # For Python 3.3+
+    visit_Try = visit_TryExcept
+
+    def visit_Exec(self, node):
+        """
+        Check `expression` of `exec(expression[, globals[, locals]])`.
+        **Just available in python 2.**
+        """
+        if hasattr(node.body, 's'):
+            self._str_codes.append((node.body.s, node.lineno + self._lineno))
+        # PR#13: https://github.com/damnever/pigar/pull/13
+        # Sometimes exec statement may be called with tuple in Py2.7.6
+        elif hasattr(node.body, 'elts') and len(node.body.elts) >= 1:
+            self._str_codes.append(
+                (node.body.elts[0].s, node.lineno + self._lineno))
+
+    def visit_Expr(self, node):
+        """
+        Check `expression` of `eval(expression[, globals[, locals]])`.
+        Check `expression` of `exec(expression[, globals[, locals]])`
+        in python 3.
+        Check `name` of `__import__(name[, globals[, locals[,
+        fromlist[, level]]]])`.
+        Check `name` or `package` of `importlib.import_module(name,
+        package=None)`.
+        """
+        # Built-in functions
+        value = node.value
+        if isinstance(value, ast.Call):
+            if hasattr(value.func, 'id'):
+                if (value.func.id == 'eval' and
+                        hasattr(node.value.args[0], 's')):
+                    self._str_codes.append(
+                        (node.value.args[0].s, node.lineno + self._lineno))
+                # **`exec` function in Python 3.**
+                elif (value.func.id == 'exec' and
+                        hasattr(node.value.args[0], 's')):
+                    self._str_codes.append(
+                        (node.value.args[0].s, node.lineno + self._lineno))
+                # `__import__` function.
+                elif (value.func.id == '__import__' and
+                        len(node.value.args) > 0 and
+                        hasattr(node.value.args[0], 's')):
+                    self._modules.add(node.value.args[0].s, self._fpath,
+                                      node.lineno + self._lineno)
+            # `import_module` function.
+            elif getattr(value.func, 'attr', '') == 'import_module':
+                module = getattr(value.func, 'value', None)
+                if (module is not None and
+                        getattr(module, 'id', '') == 'importlib'):
+                    args = node.value.args
+                    arg_len = len(args)
+                    if arg_len > 0 and hasattr(args[0], 's'):
+                        name = args[0].s
+                        if not name.startswith('.'):
+                            self._modules.add(name, self._fpath,
+                                              node.lineno + self._lineno)
+                        elif arg_len == 2 and hasattr(args[1], 's'):
+                            self._modules.add(args[1].s, self._fpath,
+                                              node.lineno + self._lineno)
+
+    def visit_FunctionDef(self, node):
+        """
+        Check docstring of function, if docstring is used for doctest.
+        """
+        docstring = self._parse_docstring(node)
+        if docstring:
+            self._str_codes.append((docstring, node.lineno + self._lineno + 2))
+
+    def visit_ClassDef(self, node):
+        """
+        Check docstring of class, if docstring is used for doctest.
+        """
+        docstring = self._parse_docstring(node)
+        if docstring:
+            self._str_codes.append((docstring, node.lineno + self._lineno + 2))
+
+    def visit(self, node):
+        """Visit a node, no recursively."""
+        for node in ast.walk(node):
+            method = 'visit_' + node.__class__.__name__
+            getattr(self, method, lambda x: x)(node)
+
+    @staticmethod
+    def _parse_docstring(node):
+        """Extract code from docstring."""
+        docstring = ast.get_docstring(node)
+        if docstring:
+            parser = doctest.DocTestParser()
+            try:
+                dt = parser.get_doctest(docstring, {}, None, None, None)
+            except ValueError:
+                # >>> 'abc'
+                pass
+            else:
+                examples = dt.examples
+                return '\n'.join([example.source for example in examples])
+        return None
+
+    @property
+    def modules(self):
+        return self._modules
+
+    @property
+    def str_codes(self):
+        return self._str_codes
+
+    @property
+    def try_imports(self):
+        return set((name.split('.')[0] if name and '.' in name else name)
+                   for name in self._try_imports)
+
+
+def _checked_cache(func):
+    checked = dict()
+
+    @functools.wraps(func)
+    def _wrapper(name):
+        if name not in checked:
+            checked[name] = func(name)
+        return checked[name]
+
+    return _wrapper
+
+
+@_checked_cache
+def is_std_or_local_lib(name):
+    """Check whether it is stdlib module.
+    True if std lib
+    False if installed package
+    str if local library
+    """
+    exist = True
+    module_info = ('', '', '')
+    try:
+        module_info = imp.find_module(name)
+    except ImportError:
+        try:
+            # __import__(name)
+            importlib.import_module(name)
+            module_info = imp.find_module(name)
+            sys.modules.pop(name)
+        except ImportError:
+            exist = False
+    # Testcase: ResourceWarning
+    if isinstance(module_info[0], FileType):
+        module_info[0].close()
+    mpath = module_info[1]
+    if exist and mpath is not None:
+        if ('site-packages' in mpath or
+                'dist-packages' in mpath or
+                'bin/' in mpath and mpath.endswith('.py')):
+            exist = False
+        elif ((sys.prefix not in mpath) and
+              (sys.base_exec_prefix not in mpath) and
+              (sys.base_prefix not in mpath)):
+            exist = mpath
+
+    return exist
+
+
+def get_installed_pkgs_detail():
+    """
+    HACK: bugfix of the original pigar get_installed_pkgs_detail
+
+    Get mapping for import top level name
+    and install package name with version.
+    """
+    mapping = dict()
+
+    for path in sys.path:
+        if os.path.isdir(path) and path.rstrip('/').endswith(
+                ('site-packages', 'dist-packages')):
+            new_mapping = _search_path(path)
+            # BUGFIX:
+            # override with previous, just like python resolves imports, the first match is the one used.
+            # unlike the original implementation, where the last one is used.
+            new_mapping.update(mapping)
+            mapping = new_mapping
+
+    # HACK: prefer tensorflow_gpu over tensorflow
+    if 'tensorflow_gpu' in new_mapping:
+        new_mapping['tensorflow'] = new_mapping['tensorflow_gpu']
+
+    return mapping
+
+
+def _search_path(path):
+    mapping = dict()
+
+    for file in os.listdir(path):
+        # Install from PYPI.
+        if fnmatch.fnmatch(file, '*-info'):
+            top_level = os.path.join(path, file, 'top_level.txt')
+            pkg_name, version = file.split('-')[:2]
+            if version.endswith('dist'):
+                version = version.rsplit('.', 1)[0]
+            # Issue for ubuntu: sudo pip install xxx
+            elif version.endswith('egg'):
+                version = version.rsplit('.', 1)[0]
+            mapping[pkg_name] = (pkg_name, version)
+            if not os.path.isfile(top_level):
+                continue
+            with open(top_level, 'r') as f:
+                for line in f:
+                    mapping[line.strip()] = (pkg_name, version)
+
+        # Install from local and available in GitHub.
+        elif fnmatch.fnmatch(file, '*-link'):
+            link = os.path.join(path, file)
+            if not os.path.isfile(link):
+                continue
+            # Link path.
+            with open(link, 'r') as f:
+                for line in f:
+                    line = line.strip()
+                    if line != '.':
+                        dev_dir = line
+            if not dev_dir:
+                continue
+            # Egg info path.
+            info_dir = [_file for _file in os.listdir(dev_dir)
+                        if _file.endswith('egg-info')]
+            if not info_dir:
+                continue
+            info_dir = info_dir[0]
+            top_level = os.path.join(dev_dir, info_dir, 'top_level.txt')
+            # Check whether it can be imported.
+            if not os.path.isfile(top_level):
+                continue
+
+            # Check .git dir.
+            git_path = os.path.join(dev_dir, '.git')
+            if os.path.isdir(git_path):
+                config = parse_git_config(git_path)
+                url = config.get('remote "origin"', {}).get('url')
+                if not url:
+                    continue
+                branch = 'branch "master"'
+                if branch not in config:
+                    for section in config:
+                        if 'branch' in section:
+                            branch = section
+                            break
+                if not branch:
+                    continue
+                branch = branch.split()[1][1:-1]
+
+                pkg_name = info_dir.split('.egg')[0]
+                git_url = 'git+{0}@{1}#egg={2}'.format(url, branch, pkg_name)
+                with open(top_level, 'r') as f:
+                    for line in f:
+                        mapping[line.strip()] = ('-e', git_url)
+    return mapping
diff --git a/trains/utilities/pigar/unpack.py b/trains/utilities/pigar/unpack.py
new file mode 100644
index 00000000..bfdafd31
--- /dev/null
+++ b/trains/utilities/pigar/unpack.py
@@ -0,0 +1,118 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import tarfile
+import zipfile
+import re
+import string
+import io
+
+
+class Archive(object):
+    """Archive provides a consistent interface for unpacking
+    compressed file.
+    """
+
+    def __init__(self, filename, fileobj):
+        self._filename = filename
+        self._fileobj = fileobj
+        self._file = None
+        self._names = None
+        self._read = None
+
+    @property
+    def filename(self):
+        return self._filename
+
+    @property
+    def names(self):
+        """If name list is not required, do not get it."""
+        if self._file is None:
+            self._prepare()
+        if not hasattr(self, '_namelist'):
+            self._namelist = self._names()
+        return self._namelist
+
+    def close(self):
+        """Close file object."""
+        if self._file is not None:
+            self._file.close()
+        if hasattr(self, '_namelist'):
+            del self._namelist
+        self._filename = self._fileobj = None
+        self._file = self._names = self._read = None
+
+    def read(self, filename):
+        """Read one file from archive."""
+        if self._file is None:
+            self._prepare()
+        return self._read(filename)
+
+    def unpack(self, to_path):
+        """Unpack compressed files to path."""
+        if self._file is None:
+            self._prepare()
+        self._safe_extractall(to_path)
+
+    def _prepare(self):
+        if self._filename.endswith(('.tar.gz', '.tar.bz2', '.tar.xz')):
+            self._prepare_tarball()
+        # An .egg file is actually just a .zip file
+        # with a different extension, .whl too.
+        elif self._filename.endswith(('.zip', '.egg', '.whl')):
+            self._prepare_zip()
+        else:
+            raise ValueError("unreadable: {0}".format(self._filename))
+
+    def _safe_extractall(self, to_path='.'):
+        unsafe = []
+        for name in self.names:
+            if not self.is_safe(name):
+                unsafe.append(name)
+        if unsafe:
+            raise ValueError("unsafe to unpack: {}".format(unsafe))
+        self._file.extractall(to_path)
+
+    def _prepare_zip(self):
+        self._file = zipfile.ZipFile(self._fileobj)
+        self._names = self._file.namelist
+        self._read = self._file.read
+
+    def _prepare_tarball(self):
+        # tarfile has no read method
+        def _read(filename):
+            f = self._file.extractfile(filename)
+            return f.read()
+
+        self._file = tarfile.open(mode='r:*', fileobj=self._fileobj)
+        self._names = self._file.getnames
+        self._read = _read
+
+    def is_safe(self, filename):
+        return not (filename.startswith(("/", "\\")) or
+                    (len(filename) > 1 and filename[1] == ":" and
+                    filename[0] in string.ascii_letter) or
+                    re.search(r"[.][.][/\\]", filename))
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, exc_type, exc_value, traceback):
+        self.close()
+
+
+def top_level(url, data):
+    """Read top level names from compressed file."""
+    sb = io.BytesIO(data)
+    txt = None
+    with Archive(url, sb) as archive:
+        file = None
+        for name in archive.names:
+            if name.lower().endswith('top_level.txt'):
+                file = name
+                break
+        if file:
+            txt = archive.read(file).decode('utf-8')
+    sb.close()
+    return [name.replace('/', '.') for name in txt.splitlines()] if txt else []
diff --git a/trains/utilities/pigar/utils.py b/trains/utilities/pigar/utils.py
new file mode 100644
index 00000000..6c8ebf34
--- /dev/null
+++ b/trains/utilities/pigar/utils.py
@@ -0,0 +1,144 @@
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function, division, absolute_import
+
+import os
+import re
+import sys
+import difflib
+
+
+PY32 = sys.version_info[:2] == (3, 2)
+
+if sys.version_info[0] == 3:
+    binary_type = bytes
+else:
+    binary_type = str
+
+
+class Dict(dict):
+    """Convert dict key object to attribute."""
+
+    def __init__(self, *args, **kwargs):
+        super(Dict, self).__init__(*args, **kwargs)
+
+    def __getattr__(self, name):
+        try:
+            return self[name]
+        except KeyError:
+            raise AttributeError('"{0}"'.format(name))
+
+    def __setattr__(self, name, value):
+        self[name] = value
+
+
+def parse_reqs(fpath):
+    pkg_v_re = re.compile(r'^(?P<pkg>[^><==]+)[><==]{,2}(?P<version>.*)$')
+    """Parse requirements file."""
+    reqs = dict()
+    with open(fpath, 'r') as f:
+        for line in f:
+            if line.startswith('#'):
+                continue
+            m = pkg_v_re.match(line.strip())
+            if m:
+                d = m.groupdict()
+                reqs[d['pkg'].strip()] = d['version'].strip()
+    return reqs
+
+
+def cmp_to_key(cmp_func):
+    """Convert a cmp=fcuntion into a key=function."""
+    class K(object):
+        def __init__(self, obj, *args):
+            self.obj = obj
+
+        def __lt__(self, other):
+            return cmp_func(self.obj, other.obj) < 0
+
+        def __gt__(self, other):
+            return cmp_func(self.obj, other.obj) > 0
+
+        def __eq__(self, other):
+            return cmp_func(self.obj, other.obj) == 0
+
+    return K
+
+
+def compare_version(version1, version2):
+    """Compare version number, such as 1.1.1 and 1.1b2.0."""
+    v1, v2 = list(), list()
+
+    for item in version1.split('.'):
+        if item.isdigit():
+            v1.append(int(item))
+        else:
+            v1.extend([i for i in _group_alnum(item)])
+    for item in version2.split('.'):
+        if item.isdigit():
+            v2.append(int(item))
+        else:
+            v2.extend([i for i in _group_alnum(item)])
+
+    while v1 and v2:
+        item1, item2 = v1.pop(0), v2.pop(0)
+        if item1 > item2:
+            return 1
+        elif item1 < item2:
+            return -1
+
+    if v1:
+        return 1
+    elif v2:
+        return -1
+    return 0
+
+
+def _group_alnum(s):
+    tmp = list()
+    flag = 1 if s[0].isdigit() else 0
+    for c in s:
+        if c.isdigit():
+            if flag == 0:
+                yield ''.join(tmp)
+                tmp = list()
+                flag = 1
+            tmp.append(c)
+        elif c.isalpha():
+            if flag == 1:
+                yield int(''.join(tmp))
+                tmp = list()
+                flag = 0
+            tmp.append(c)
+    last = ''.join(tmp)
+    yield (int(last) if flag else last)
+
+
+def parse_git_config(path):
+    """Parse git config file."""
+    config = dict()
+    section = None
+
+    with open(os.path.join(path, 'config'), 'r') as f:
+        for line in f:
+            line = line.strip()
+            if line.startswith('['):
+                section = line[1: -1].strip()
+                config[section] = dict()
+            elif section:
+                key, value = line.replace(' ', '').split('=')
+                config[section][key] = value
+    return config
+
+
+def lines_diff(lines1, lines2):
+    """Show difference between lines."""
+    is_diff = False
+    diffs = list()
+
+    for line in difflib.ndiff(lines1, lines2):
+        if not is_diff and line[0] in ('+', '-'):
+            is_diff = True
+        diffs.append(line)
+
+    return is_diff, diffs