Dump activation shardings

[charlesli640]

Description

To dump activation shardings to golden file for further comparison. It can include in unit test in case further code change touches activation shardings. This PR is the initial submission for draft review. The change is based on PR 3034

Output

The output format is readable and comparable by both human and machine. For example llama3.1-70b/v6e-16/slice_1 activation dump as below

"Activation Sharding Dump": [
    {
      "llama2/inputs: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "llama2/lnx: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "attention/inputs_q: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_attn_length_no_exp', 'activation_attn_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "attention/input_kv: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_attn_length_no_exp', 'activation_attn_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "attention/query: bfloat16[192,2048,64,128]": {
        "logic_axes": "('activation_kv_batch', 'activation_attn_length_no_exp', 'activation_kv_heads', 'activation_kv_head_dim')",
        "PartitionSpec": "P('fsdp', None, None, None)"
      }
    },
    {
      "attention/key: bfloat16[192,2048,8,128]": {
        "logic_axes": "('activation_kv_batch', 'activation_attn_length_no_exp', 'activation_kv_heads', 'activation_kv_head_dim')",
        "PartitionSpec": "P('fsdp', None, None, None)"
      }
    },
    {
      "attention/value: bfloat16[192,2048,8,128]": {
        "logic_axes": "('activation_kv_batch', 'activation_attn_length_no_exp', 'activation_kv_heads', 'activation_kv_head_dim')",
        "PartitionSpec": "P('fsdp', None, None, None)"
      }
    },
    {
      "Unknown: bfloat16[192,64,2048,128]": {
        "logic_axes": "Unknown",
        "PartitionSpec": "P('fsdp', None, None, None)"
      }
    },
    {
      "Unknown: bfloat16[192,8,2048,128]": {
        "logic_axes": "Unknown",
        "PartitionSpec": "P('fsdp', None, None, None)"
      }
    },
    {
      "attention/out: bfloat16[192,2048,64,128]": {
        "logic_axes": "('activation_batch', 'activation_attn_length_no_exp', 'activation_heads', 'activation_kv')",
        "PartitionSpec": "P('fsdp', None, None, None)"
      }
    },
    {
      "llama2/attention_lnx: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "llama2/hidden_states: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "linears/x: bfloat16[192,2048,28672]": {
        "logic_axes": "('activation_batch', 'activation_length_no_exp', 'activation_mlp')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "llama2/mlp_lnx: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "llama2/mlp_lnx: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    },
    {
      "llama2/layer_output: bfloat16[192,2048,8192]": {
        "logic_axes": "('activation_batch', 'activation_norm_length', 'activation_embed')",
        "PartitionSpec": "P('fsdp', None, None)"
      }
    }
]

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

❌ Patch coverage is 54.79452% with 33 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/MaxText/sharding.py	33.33%	10 Missing ⚠️
src/MaxText/layers/pipeline.py	0.00%	6 Missing ⚠️
src/MaxText/layers/attention_mla.py	66.66%	5 Missing ⚠️
src/MaxText/layers/attentions.py	77.77%	4 Missing ⚠️
src/MaxText/layers/decoders.py	50.00%	4 Missing ⚠️
src/MaxText/train_compile.py	0.00%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

[gobbleturk]

I think this LGTM although there are a lot of names to review! How did you generate these names?