merge_report.json

{
  "input": "/workspace/sn80/training_checkpoints/sn80_pi05_lora_teacher/e17_objspatial_replay_expert_s80_lr1e6/79",
  "output": "/workspace/sn80/candidates/e17_objspatial_replay_expert_dense",
  "lora_keys_before": 20,
  "lora_keys_after": 0,
  "merged_groups": 5,
  "dropped_groups": 4,
  "einsum_scale": 1.0,
  "merge_strength": 1.0,
  "merge_target": "expert",
  "groups": [
    {
      "path": "PaliGemma/llm/layers/attn/attn_vec_einsum",
      "kind": "dropped_frozen_lora",
      "keys": [
        "lora_a",
        "lora_b"
      ]
    },
    {
      "path": "PaliGemma/llm/layers/attn/attn_vec_einsum_1",
      "kind": "einsum",
      "shape": [
        18,
        8,
        256,
        1024
      ]
    },
    {
      "path": "PaliGemma/llm/layers/attn/kv_einsum",
      "kind": "dropped_frozen_lora",
      "keys": [
        "lora_a",
        "lora_b"
      ]
    },
    {
      "path": "PaliGemma/llm/layers/attn/kv_einsum_1",
      "kind": "einsum",
      "shape": [
        18,
        2,
        1,
        1024,
        256
      ]
    },
    {
      "path": "PaliGemma/llm/layers/attn/q_einsum",
      "kind": "dropped_frozen_lora",
      "keys": [
        "lora_a",
        "lora_b"
      ]
    },
    {
      "path": "PaliGemma/llm/layers/attn/q_einsum_1",
      "kind": "einsum",
      "shape": [
        18,
        8,
        1024,
        256
      ]
    },
    {
      "path": "PaliGemma/llm/layers/mlp",
      "kind": "dropped_frozen_lora",
      "keys": [
        "gating_einsum_lora_a",
        "gating_einsum_lora_b",
        "linear_lora_a",
        "linear_lora_b"
      ]
    },
    {
      "path": "PaliGemma/llm/layers/mlp_1",
      "kind": "ffn_gating",
      "shape": [
        18,
        2,
        1024,
        4096
      ]
    },
    {
      "path": "PaliGemma/llm/layers/mlp_1",
      "kind": "ffn_linear",
      "shape": [
        18,
        4096,
        1024
      ]
    }
  ]
}