TFL CIT

[ymodak]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
• TensorFlow installed from (source or binary):
• TensorFlow version (or github SHA if from source):

Provide colab notebooks as per your model

-- The sample colab notebooks given below demonstrate:-
   Build model in TF Keras --> Convert to TF Lite (performing quantization techniques) --> Invoke the TF Lite interpreter
-- You may use these notebooks as a reference point to generate your model behavior and attach to this issue.

• Train a TensorFlow Model Colab
• TensorFlow Lite Debugger Colab

Command used to run the converter or code if you’re using the Python API

# Copy and paste here the exact command

The output from the converter invocation

# Copy and paste the output here.

Also, please include a link to the saved model or GraphDef

# Put link here or attach to the issue.

Failure details If the conversion is successful, but the generated model is wrong, state what is wrong:

• Producing wrong results and/or decrease in accuracy
• Producing correct results, but the model is slower than expected (model generated from old converter)

RNN conversion support If converting TF RNN to TFLite fused RNN ops, please prefix [RNN] in the title.

Any other info / logs

Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.

[ymodak]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
• TensorFlow installed from (source or binary):
• TensorFlow version (or github SHA if from source):

Command used to run the converter or code if you’re using the Python API If possible, please share a link to Colab/Jupyter/any notebook.

# Copy and paste here the exact command

The output from the converter invocation

# Copy and paste the output here.

Also, please include a link to the saved model or GraphDef

# Put link here or attach to the issue.

Failure details If the conversion is successful, but the generated model is wrong, state what is wrong:

• Producing wrong results and/or decrease in accuracy
• Producing correct results, but the model is slower than expected (model generated from old converter)

RNN conversion support If converting TF RNN to TFLite fused RNN ops, please prefix [RNN] in the title.

Any other info / logs

Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.

[MeghnaNatraj]

(suggested update)

1. System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
• TensorFlow installed from (source or binary):
• TensorFlow version (or github SHA if from source):

2. Code (that demonstrates how to reproduce your issue)

Option A: Using reference colab notebooks

The reference colab notebooks given below demonstrate: (TensorFlow Model Colab) Build model in TF Keras --> (TensorFlow Lite Model Colab) Convert to TF Lite (performing quantization techniques) and run TFLite Inference. You may use these colab notebooks as a reference point to generate your model behavior and attach links.

1) Reference TensorFlow Model Colab 2) Reference TensorFlow Lite Model Colab

# Put links here or attach to the issue

Option B: Paste your code here or provide a link to your custom end-to-end colab

# Provide your code here (or put links here or attach to the issue)
# - Include code to invoke the TFLite Converter Python API
# - Include the output (with errors) from the converter invocation 

3. Models

Provide links to your TensorFlow model and (optionally) TensorFlow Lite Model

# Put links here or attach to the issue

4. Failure details

If the conversion is successful, but the generated model is wrong, then state what is wrong:

• Producing wrong results and/or decrease in accuracy
• Producing correct results, but the model is slower than expected (model generated from old converter)

5. (optional) RNN conversion support

If converting TF RNN to TFLite fused RNN ops, please prefix [RNN] in the title.

6. (optional) Any other info / logs

Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.