Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsThe error usually comes from one of two issues: Python code uses the wrong capitalization (session instead of Session), or TensorFlow 1-style session code is running with TensorFlow 2. For a legacy graph-based program, use tf.compat.v1.Session; for a native TensorFlow 2 program, remove session calls and use eager execution.
First, identify which error you have
Check the exact line named in the traceback and the module your code imports. TensorFlow documents the class as Session, with a capital S—not session. In TensorFlow 2, the legacy class is exposed at tf.compat.v1.Session, rather than the old root-level tf.Session path. The TensorFlow v2.16.1 API reference, last updated April 26, 2024, documents this compatibility API at tf.compat.v1.Session.
- If your code says
tf.session(), correct the capitalization and use the compatibility path if you need a session. - If it says
tf.Session(), the code likely follows TensorFlow 1-era examples while running TensorFlow 2.
Before changing TensorFlow APIs, also verify locally that the active Python environment is the one where you installed TensorFlow and that the import resolves to the intended package. A file or directory in your project named tensorflow can also shadow the installed package. These checks help distinguish an API mismatch from an import or environment problem; the traceback and active environment determine which applies.
Choose between compatibility and migration
| Approach | Best fit | What it means |
|---|---|---|
| Keep TF1-style sessions | Existing code depends on graph execution, Session, or other TF1 assumptions that you are not ready to replace. |
Use the tf.compat.v1 compatibility API. This preserves legacy behavior, not a native TensorFlow 2 design. |
| Move to native TF2 | You can update the program to run with eager execution and modern TensorFlow patterns. | Remove explicit session creation and sess.run(...); operations execute eagerly unless you use tf.function for graph compilation. |
TensorFlow describes Session as a TensorFlow 1 API and states that it does not work with eager execution or tf.function. See the Session API reference and TensorFlow migration guide.
#1 Best Overall
Fix A: keep the TF1-style session code
Replace the old root-level session call with the compatibility API:
import tensorflow as tf
with tf.compat.v1.Session() as sess:
result = sess.run(some_tensor)Use this route when the surrounding code genuinely relies on TF1 graph/session execution. Other TF1-era APIs may also need compatibility paths, so changing only the constructor may reveal additional errors.
TensorFlow’s migration overview also documents a broader compatibility option:
import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()This keeps TensorFlow 1 behavior on a TensorFlow 2 installation; it is not a migration to native TF2. Choose compatibility behavior deliberately for a codebase whose graph and session assumptions you understand, rather than treating it as a universal fix. See the migration overview.
Fix B: migrate to native TensorFlow 2
In TensorFlow 2, eager execution is enabled by default: operations run immediately and produce concrete values. Remove explicit session creation and calls to sess.run(...), then use tensors and variables directly. For example:
import tensorflow as tf
x = tf.constant(6)
y = tf.constant(7)
result = tf.multiply(x, y)
print(result.numpy())If a function benefits from graph compilation, define it with tf.function instead of wrapping the program in a session. For new models, TensorFlow’s migration overview points to object-based tracking with tf.keras.layers.Layer, tf.keras.Model, or tf.Module, rather than TF1 graph collections.
A complete migration can involve more than replacing one API name. TensorFlow’s migration guide covers updating API symbols, removing obsolete APIs, making forward passes work with eager execution, and changing training and save/load flows. The edits depend on the surrounding program and TensorFlow version.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why toggling eager execution late is not a reliable fix
Correcting the attribute may expose a second problem: a session is incompatible with eager execution. TensorFlow documents that eager execution cannot be enabled after APIs have created or executed graphs, and execution-mode changes are program-level compatibility decisions. Do not mix TF1 session assumptions with TF2 eager execution casually. Decide at program startup whether to preserve TF1 compatibility or migrate to native TF2; a late eager-mode toggle is not a general repair. See the Session API reference and migration guide.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
