PCC drop in ttnn.softmax PCC=0.5832070037132617

[sonalibaskaran2499]

Describe the bug

ttnn. softmax having pcc drop = 0.5832070037132617
The affected models are listed below:

test_all_models_torch[gemma/pytorch-google/gemma-1.1-7b-it-single_device-inference]                                                         language    red         p150         INCORRECT_RESULT           PASSING       0.9724180883623303      0.99       PCC_EN   single_device    131.686   N/A
test_all_models_torch[llama/causal_lm/pytorch-llama_3_1_8b_instruct-single_device-inference]                                                language    red         p150         INCORRECT_RESULT           PASSING       0.9702271827420873      0.99       PCC_EN   single_device    171.237   N/A
test_all_models_torch[mistral/pytorch-mistral_nemo_instruct_2407-single_device-inference]                                                   language    red         p150         INCORRECT_RESULT           PASSING       0.9821535219381753      0.99       PCC_EN   single_device    187.107   N/A
test_all_models_torch[phi2/causal_lm/pytorch-microsoft/phi-2-single_device-inference]                                                       language    red         n150         INCORRECT_RESULT           PASSING       0.9344665590421978      0.98       PCC_EN   single_device    142.625   N/A
test_all_models_torch[phi2/causal_lm/pytorch-microsoft/phi-2-single_device-inference]                                                       language    red         p150         INCORRECT_RESULT           PASSING       0.9310092454805221      0.98       PCC_EN   single_device    109.987   N/A

Steps to Reproduce

git checkout sonalib/softmax
git submodule update --init --recursive
git lfs pull --include ‘tensors/*.tensorbin'
pytest /tt-metal/tests/ttnn/unit_tests/operations/softmax/main.py -svv

Logs

softmax_ttnn.log