From 39fe636ff2deb5f10e0ab2f2a227b606f7151092 Mon Sep 17 00:00:00 2001
From: lroyia <lirh@chinaweal.com.cn>
Date: Wed, 25 Feb 2026 10:33:28 +0800
Subject: [PATCH] =?UTF-8?q?=E8=B0=83=E6=95=B4=E6=8A=A5=E9=94=99?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../com/example/glmocr/model/GlmOcrModel.java     | 15 ++++++++-------
 .../example/glmocr/model/GlmOcrTranslator.java    |  2 +-
 .../glmocr/tokenizer/TokenizerService.java        |  8 ++++----
 src/main/resources/application.yml                |  2 +-
 4 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/src/main/java/com/example/glmocr/model/GlmOcrModel.java b/src/main/java/com/example/glmocr/model/GlmOcrModel.java
index 28abe3c..abd47e6 100644
--- a/src/main/java/com/example/glmocr/model/GlmOcrModel.java
+++ b/src/main/java/com/example/glmocr/model/GlmOcrModel.java
@@ -11,6 +11,7 @@ import ai.djl.pytorch.engine.PtModel;
 import ai.djl.pytorch.engine.PtNDArray;
 import ai.djl.repository.zoo.Criteria;
 import ai.djl.repository.zoo.ZooModel;
+import ai.djl.training.util.ProgressBar;
 import ai.djl.translate.TranslateException;
 import ai.djl.translate.Translator;
 import ai.djl.translate.TranslatorContext;
@@ -76,7 +77,7 @@ public class GlmOcrModel {
                 .optModelPath(modelPath)
                 .optEngine("PyTorch")
                 .optTranslator(new GlmOcrTranslator(config))
-                .optProgress(new ProgressListener())
+                .optProgress(new ProgressBar())
                 .build();
 
         model = criteria.loadModel();
@@ -100,10 +101,10 @@ public class GlmOcrModel {
         init();
     }
 
-    private static class ProgressListener implements ai.djl.training.listener.ProgressListener {
-        @Override
-        public void progressUpdate(int progress, String message) {
-            log.debug("Model loading progress: {}% - {}", progress, message);
-        }
-    }
+//    private static class ProgressListener implements ai.djl.training.listener.ProgressListener {
+//        @Override
+//        public void progressUpdate(int progress, String message) {
+//            log.debug("Model loading progress: {}% - {}", progress, message);
+//        }
+//    }
 }
diff --git a/src/main/java/com/example/glmocr/model/GlmOcrTranslator.java b/src/main/java/com/example/glmocr/model/GlmOcrTranslator.java
index 72a2012..5150760 100644
--- a/src/main/java/com/example/glmocr/model/GlmOcrTranslator.java
+++ b/src/main/java/com/example/glmocr/model/GlmOcrTranslator.java
@@ -49,7 +49,7 @@ public class GlmOcrTranslator implements Translator<BufferedImage, String> {
         int width = processedImage.getWidth();
         int height = processedImage.getHeight();
         
-        float[] pixels = new float[3 * height * width];
+        double[] pixels = new double[3 * height * width];
         int idx = 0;
         
         for (int c = 0; c < 3; c++) {
diff --git a/src/main/java/com/example/glmocr/tokenizer/TokenizerService.java b/src/main/java/com/example/glmocr/tokenizer/TokenizerService.java
index c068631..214041a 100644
--- a/src/main/java/com/example/glmocr/tokenizer/TokenizerService.java
+++ b/src/main/java/com/example/glmocr/tokenizer/TokenizerService.java
@@ -39,7 +39,7 @@ public class TokenizerService {
             return new long[0];
         }
         Encoding encoding = tokenizer.encode(text);
-        return Arrays.stream(encoding.getIds()).asLongStream().toArray();
+        return Arrays.stream(encoding.getIds()).toArray();
     }
 
     public String decode(long[] ids) {
@@ -51,8 +51,7 @@ public class TokenizerService {
                 .filter(id -> id > 0 && id < 151936) // GLM token范围
                 .toArray();
         
-        String[] tokens = tokenizer.decode(filteredIds);
-        return String.join("", tokens);
+        return tokenizer.decode(filteredIds);
     }
 
     public String decodeWithPrompt(long[] ids, String prompt) {
@@ -65,7 +64,8 @@ public class TokenizerService {
     }
 
     public int getVocabSize() {
-        return tokenizer != null ? tokenizer.getVocabularySize() : 0;
+//        return tokenizer != null ? tokenizer.getVocabularySize() : 0;
+        return tokenizer != null ? tokenizer.getMaxLength() : 0;
     }
 
     public boolean isAvailable() {
diff --git a/src/main/resources/application.yml b/src/main/resources/application.yml
index ccaf1b9..5a63a30 100644
--- a/src/main/resources/application.yml
+++ b/src/main/resources/application.yml
@@ -12,7 +12,7 @@ spring:
 # GLM-OCR配置（纯Java本地部署）
 glm-ocr:
   # 模型本地路径（支持相对路径或绝对路径）
-  model-path: ./models/GLM-OCR
+  model-path: 'D:/development/community/GLM-OCR'
   # 推理设备: cpu, gpu(0), gpu(1)
   device: cpu
   # 精度: fp32, fp16, bf16, int8