텍스트 비교

두 텍스트 파일의 차이점을 찾아보세요

실시간 편집

공백 변경 숨기기

변경 없는 행 숨기기

줄바꿈 비활성화

레이아웃

비교 단위

텍스트 스타일

구문 강조

Diffchecker Desktop 가장 안전하게 Diffchecker를 사용하는 방법. 데스크톱 앱을 사용하면 비교 데이터가 외부로 전송되지 않습니다!데스크톱 앱 받기

opt_state_diff

생성일 3년 전비교 결과 만료 없음

4 삭제

행
총
삭제

글자
총
삭제

이 기능을 계속 사용하려면 업그레이드해 주세요 Diffchecker Pro 요금제 보기

171 행

4 추가

행
총
추가

글자
총
추가

이 기능을 계속 사용하려면 업그레이드해 주세요 Diffchecker Pro 요금제 보기

171 행

(EmptyState(), (ScaleByAdamState(count=Array(0, dtype=int32), mu=React(

(EmptyState(), (ScaleByAdamState(count=Array(130, dtype=int32), mu=React(

max_iters=10,

bottleneck=128,

SEQLEN=16,

embed_dim=128,

input_proj=LinearProj(

bias=f32[128],

weight=f32[128,128],

input_dim=128,

output_dim=128,

use_bias=True

input_act=NewGELU(),

out_head=LinearProj(

bias=f32[2],

weight=f32[128,2],

input_dim=128,

output_dim=2,

use_bias=True

embed_layer=Embedding(num_embeddings=8, embedding_size=128, weight=f32[8,128]),

main_block=RecurrentModule(

gelu=NewGELU(),

reshape_layer=LinearProj(

bias=f32[128],

weight=f32[256,128],

input_dim=256,

output_dim=128,

use_bias=True

key=u32[2],

key=None,

attention_blocks=[

AttentionBlock(

activation=NewGELU(),

attn_gate=LiteAttention(

input_dim=256,

weight=LinearProj(

bias=f32[256],

weight=f32[256,256],

input_dim=256,

output_dim=256,

use_bias=False

)

ln1=LayerNorm(

shape=256,

eps=1e-05,

use_weight=True,

use_bias=True,

weight=f32[256],

bias=f32[256]

ln2=LayerNorm(

shape=256,

eps=1e-05,

use_weight=True,

use_bias=True,

weight=f32[256],

bias=f32[256]

mlp=MLP(

layers=[

LinearProj(

bias=f32[256],

weight=f32[256,256],

input_dim=256,

output_dim=256,

use_bias=True

Lambda(fn=NewGELU()),

LinearProj(

bias=f32[256],

weight=f32[256,256],

input_dim=256,

output_dim=256,

use_bias=True

)

dropout=Dropout(p=None, inference=None)

)

]

id=Identity(),

pos_enc=f32[16,128]

), nu=React(

max_iters=10,

bottleneck=128,

SEQLEN=16,

embed_dim=128,

input_proj=LinearProj(

bias=f32[128],

weight=f32[128,128],

input_dim=128,

output_dim=128,

use_bias=True

input_act=NewGELU(),

out_head=LinearProj(

bias=f32[2],

weight=f32[128,2],

input_dim=128,

output_dim=2,

use_bias=True

embed_layer=Embedding(num_embeddings=8, embedding_size=128, weight=f32[8,128]),

main_block=RecurrentModule(

gelu=NewGELU(),

reshape_layer=LinearProj(

bias=f32[128],

weight=f32[256,128],

input_dim=256,

output_dim=128,

use_bias=True

key=u32[2],

key=None,

attention_blocks=[

AttentionBlock(

activation=NewGELU(),

attn_gate=LiteAttention(

input_dim=256,

weight=LinearProj(

bias=f32[256],

weight=f32[256,256],

input_dim=256,

output_dim=256,

use_bias=False

)

ln1=LayerNorm(

shape=256,

eps=1e-05,

use_weight=True,

use_bias=True,

weight=f32[256],

bias=f32[256]

ln2=LayerNorm(

shape=256,

eps=1e-05,

use_weight=True,

use_bias=True,

weight=f32[256],

bias=f32[256]

mlp=MLP(

layers=[

LinearProj(

bias=f32[256],

weight=f32[256,256],

input_dim=256,

output_dim=256,

use_bias=True

Lambda(fn=NewGELU()),

LinearProj(

bias=f32[256],

weight=f32[256,256],

input_dim=256,

output_dim=256,

use_bias=True

)

dropout=Dropout(p=None, inference=None)

)

]

id=Identity(),

pos_enc=f32[16,128]

)), EmptyState(), ScaleByScheduleState(count=Array(0, dtype=int32))))

)), EmptyState(), ScaleByScheduleState(count=Array(130, dtype=int32))))

저장된 비교 결과

원본

파일 열기

(EmptyState(), (ScaleByAdamState(count=Array(0, dtype=int32), mu=React(
  max_iters=10,
  bottleneck=128,
  SEQLEN=16,
  embed_dim=128,
  input_proj=LinearProj(
    bias=f32[128],
    weight=f32[128,128],
    input_dim=128,
    output_dim=128,
    use_bias=True
  ),
  input_act=NewGELU(),
  out_head=LinearProj(
    bias=f32[2],
    weight=f32[128,2],
    input_dim=128,
    output_dim=2,
    use_bias=True
  ),
  embed_layer=Embedding(num_embeddings=8, embedding_size=128, weight=f32[8,128]),
  main_block=RecurrentModule(
    gelu=NewGELU(),
    reshape_layer=LinearProj(
      bias=f32[128],
      weight=f32[256,128],
      input_dim=256,
      output_dim=128,
      use_bias=True
    ),
    key=u32[2],
    attention_blocks=[
      AttentionBlock(
        activation=NewGELU(),
        attn_gate=LiteAttention(
          input_dim=256,
          weight=LinearProj(
            bias=f32[256],
            weight=f32[256,256],
            input_dim=256,
            output_dim=256,
            use_bias=False
          )
        ),
        ln1=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        ln2=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        mlp=MLP(
          layers=[
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            ),
            Lambda(fn=NewGELU()),
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            )
          ],
          dropout=Dropout(p=None, inference=None)
        )
      )
    ]
  ),
  id=Identity(),
  pos_enc=f32[16,128]
), nu=React(
  max_iters=10,
  bottleneck=128,
  SEQLEN=16,
  embed_dim=128,
  input_proj=LinearProj(
    bias=f32[128],
    weight=f32[128,128],
    input_dim=128,
    output_dim=128,
    use_bias=True
  ),
  input_act=NewGELU(),
  out_head=LinearProj(
    bias=f32[2],
    weight=f32[128,2],
    input_dim=128,
    output_dim=2,
    use_bias=True
  ),
  embed_layer=Embedding(num_embeddings=8, embedding_size=128, weight=f32[8,128]),
  main_block=RecurrentModule(
    gelu=NewGELU(),
    reshape_layer=LinearProj(
      bias=f32[128],
      weight=f32[256,128],
      input_dim=256,
      output_dim=128,
      use_bias=True
    ),
    key=u32[2],
    attention_blocks=[
      AttentionBlock(
        activation=NewGELU(),
        attn_gate=LiteAttention(
          input_dim=256,
          weight=LinearProj(
            bias=f32[256],
            weight=f32[256,256],
            input_dim=256,
            output_dim=256,
            use_bias=False
          )
        ),
        ln1=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        ln2=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        mlp=MLP(
          layers=[
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            ),
            Lambda(fn=NewGELU()),
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            )
          ],
          dropout=Dropout(p=None, inference=None)
        )
      )
    ]
  ),
  id=Identity(),
  pos_enc=f32[16,128]
)), EmptyState(), ScaleByScheduleState(count=Array(0, dtype=int32))))

수정본

파일 열기

(EmptyState(), (ScaleByAdamState(count=Array(130, dtype=int32), mu=React(
  max_iters=10,
  bottleneck=128,
  SEQLEN=16,
  embed_dim=128,
  input_proj=LinearProj(
    bias=f32[128],
    weight=f32[128,128],
    input_dim=128,
    output_dim=128,
    use_bias=True
  ),
  input_act=NewGELU(),
  out_head=LinearProj(
    bias=f32[2],
    weight=f32[128,2],
    input_dim=128,
    output_dim=2,
    use_bias=True
  ),
  embed_layer=Embedding(num_embeddings=8, embedding_size=128, weight=f32[8,128]),
  main_block=RecurrentModule(
    gelu=NewGELU(),
    reshape_layer=LinearProj(
      bias=f32[128],
      weight=f32[256,128],
      input_dim=256,
      output_dim=128,
      use_bias=True
    ),
    key=None,
    attention_blocks=[
      AttentionBlock(
        activation=NewGELU(),
        attn_gate=LiteAttention(
          input_dim=256,
          weight=LinearProj(
            bias=f32[256],
            weight=f32[256,256],
            input_dim=256,
            output_dim=256,
            use_bias=False
          )
        ),
        ln1=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        ln2=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        mlp=MLP(
          layers=[
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            ),
            Lambda(fn=NewGELU()),
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            )
          ],
          dropout=Dropout(p=None, inference=None)
        )
      )
    ]
  ),
  id=Identity(),
  pos_enc=f32[16,128]
), nu=React(
  max_iters=10,
  bottleneck=128,
  SEQLEN=16,
  embed_dim=128,
  input_proj=LinearProj(
    bias=f32[128],
    weight=f32[128,128],
    input_dim=128,
    output_dim=128,
    use_bias=True
  ),
  input_act=NewGELU(),
  out_head=LinearProj(
    bias=f32[2],
    weight=f32[128,2],
    input_dim=128,
    output_dim=2,
    use_bias=True
  ),
  embed_layer=Embedding(num_embeddings=8, embedding_size=128, weight=f32[8,128]),
  main_block=RecurrentModule(
    gelu=NewGELU(),
    reshape_layer=LinearProj(
      bias=f32[128],
      weight=f32[256,128],
      input_dim=256,
      output_dim=128,
      use_bias=True
    ),
    key=None,
    attention_blocks=[
      AttentionBlock(
        activation=NewGELU(),
        attn_gate=LiteAttention(
          input_dim=256,
          weight=LinearProj(
            bias=f32[256],
            weight=f32[256,256],
            input_dim=256,
            output_dim=256,
            use_bias=False
          )
        ),
        ln1=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        ln2=LayerNorm(
          shape=256,
          eps=1e-05,
          use_weight=True,
          use_bias=True,
          weight=f32[256],
          bias=f32[256]
        ),
        mlp=MLP(
          layers=[
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            ),
            Lambda(fn=NewGELU()),
            LinearProj(
              bias=f32[256],
              weight=f32[256,256],
              input_dim=256,
              output_dim=256,
              use_bias=True
            )
          ],
          dropout=Dropout(p=None, inference=None)
        )
      )
    ]
  ),
  id=Identity(),
  pos_enc=f32[16,128]
)), EmptyState(), ScaleByScheduleState(count=Array(130, dtype=int32))))