Remove unused layer mapping

Author: jonatanklosko

lib/bumblebee/text/gemma.ex

@@ -449,8 +449,6 @@ defmodule Bumblebee.Text.Gemma do
         "decoder.blocks.{n}.self_attention.value" => "model.layers.{n}.self_attn.v_proj",
         "decoder.blocks.{n}.self_attention.output" => "model.layers.{n}.self_attn.o_proj",
         "decoder.blocks.{n}.self_attention_norm" => "model.layers.{n}.input_layernorm",
-        "decoder.blocks.{n}.self_attention.rotary_embedding" =>
-          "model.layers.{n}.self_attn.rotary_emb",
         "decoder.blocks.{n}.ffn.gate" => "model.layers.{n}.mlp.gate_proj",
         "decoder.blocks.{n}.ffn.intermediate" => "model.layers.{n}.mlp.up_proj",
         "decoder.blocks.{n}.ffn.output" => "model.layers.{n}.mlp.down_proj",

lib/bumblebee/text/gpt_neo_x.ex

@@ -426,8 +426,6 @@ defmodule Bumblebee.Text.GptNeoX do
           ),
         "decoder.blocks.{n}.self_attention.output" => "gpt_neox.layers.{n}.attention.dense",
         "decoder.blocks.{n}.self_attention_norm" => "gpt_neox.layers.{n}.input_layernorm",
-        "decoder.blocks.{n}.self_attention.rotary_embedding" =>
-          "gpt_neox.layers.{n}.self_attn.rotary_emb",
         "decoder.blocks.{n}.ffn.intermediate" => "gpt_neox.layers.{n}.mlp.dense_h_to_4h",
         "decoder.blocks.{n}.ffn.output" => "gpt_neox.layers.{n}.mlp.dense_4h_to_h",
         "decoder.blocks.{n}.output_norm" => "gpt_neox.layers.{n}.post_attention_layernorm",

lib/bumblebee/text/llama.ex

@@ -469,8 +469,6 @@ defmodule Bumblebee.Text.Llama do
         "decoder.blocks.{n}.self_attention.value" => "model.layers.{n}.self_attn.v_proj",
         "decoder.blocks.{n}.self_attention.output" => "model.layers.{n}.self_attn.o_proj",
         "decoder.blocks.{n}.self_attention_norm" => "model.layers.{n}.input_layernorm",
-        "decoder.blocks.{n}.self_attention.rotary_embedding" =>
-          "model.layers.{n}.self_attn.rotary_emb",
         "decoder.blocks.{n}.ffn.gate" => "model.layers.{n}.mlp.gate_proj",
         "decoder.blocks.{n}.ffn.intermediate" => "model.layers.{n}.mlp.up_proj",
         "decoder.blocks.{n}.ffn.output" => "model.layers.{n}.mlp.down_proj",

lib/bumblebee/text/mistral.ex

@@ -411,8 +411,6 @@ defmodule Bumblebee.Text.Mistral do
         "decoder.blocks.{n}.self_attention.value" => "model.layers.{n}.self_attn.v_proj",
         "decoder.blocks.{n}.self_attention.output" => "model.layers.{n}.self_attn.o_proj",
         "decoder.blocks.{n}.self_attention_norm" => "model.layers.{n}.input_layernorm",
-        "decoder.blocks.{n}.self_attention.rotary_embedding" =>
-          "model.layers.{n}.self_attn.rotary_emb",
         "decoder.blocks.{n}.ffn.gate" => "model.layers.{n}.mlp.gate_proj",
         "decoder.blocks.{n}.ffn.intermediate" => "model.layers.{n}.mlp.up_proj",
         "decoder.blocks.{n}.ffn.output" => "model.layers.{n}.mlp.down_proj",

lib/bumblebee/text/phi.ex

@@ -442,8 +442,6 @@ defmodule Bumblebee.Text.Phi do
         "decoder.blocks.{n}.self_attention.value" => "model.layers.{n}.self_attn.v_proj",
         "decoder.blocks.{n}.self_attention.output" => "model.layers.{n}.self_attn.dense",
         "decoder.blocks.{n}.self_attention_norm" => "model.layers.{n}.input_layernorm",
-        "decoder.blocks.{n}.self_attention.rotary_embedding" =>
-          "model.layers.{n}.self_attn.rotary_emb",
         "decoder.blocks.{n}.ffn.intermediate" => "model.layers.{n}.mlp.fc1",
         "decoder.blocks.{n}.ffn.output" => "model.layers.{n}.mlp.fc2",
         "output_norm" => "model.final_layernorm",

lib/bumblebee/text/phi3.ex

@@ -497,8 +497,6 @@ defmodule Bumblebee.Text.Phi3 do
           ),
         "decoder.blocks.{n}.self_attention.output" => "model.layers.{n}.self_attn.o_proj",
         "decoder.blocks.{n}.self_attention_norm" => "model.layers.{n}.input_layernorm",
-        "decoder.blocks.{n}.self_attention.rotary_embedding" =>
-          "model.layers.{n}.self_attn.rotary_emb",
         "decoder.blocks.{n}.ffn.gate" =>
           Shared.sliced_dense_params_source(
             "model.layers.{n}.mlp.gate_up_proj",