Replace unicode escaped characters in ipynb files

PiperOrigin-RevId: 854213846

Author: AleksMat

colabs/finetuning.ipynb

@@ -106,7 +106,7 @@
         {
           "data": {
             "text/plain": [
-              "[\u003c_Gemma2SpecialTokens.BOS: 2\u003e, 1596, 603, 671, 3287, 13060]"
+              "[<_Gemma2SpecialTokens.BOS: 2>, 1596, 603, 671, 3287, 13060]"
             ]
           },
           "execution_count": 22,
@@ -213,7 +213,7 @@
         "id": "3ny2J07G2X7i"
       },
       "source": [
-        "We can decode an example from the batch to inspect the model input. We see that the `\u003cstart_of_turn\u003e` / `\u003cend_of_turn\u003e` where correctly added to follow Gemma dialog format."
+        "We can decode an example from the batch to inspect the model input. We see that the `<start_of_turn>` / `<end_of_turn>` where correctly added to follow Gemma dialog format."
       ]
     },
     {
@@ -238,9 +238,9 @@
           "name": "stdout",
           "output_type": "stream",
           "text": [
-            "\u003cstart_of_turn\u003euser\n",
-            "Would love any other tips from anyone, but specially from someone who’s been where I’m at.\u003cend_of_turn\u003e\n",
-            "\u003cstart_of_turn\u003emodel\n",
+            "<start_of_turn>user\n",
+            "Would love any other tips from anyone, but specially from someone who’s been where I’m at.<end_of_turn>\n",
+            "<start_of_turn>model\n",
             "J'apprécierais vraiment d'autres astuces, mais particulièrement par quelqu'un qui était était déjà là où je me trouve.\n"
           ]
         }
@@ -398,7 +398,7 @@
               "version_minor": 0
             },
             "text/plain": [
-              "train:   0%|          | 0/301 [00:00\u003c?, ?it/s]"
+              "train:   0%|          | 0/301 [00:00<?, ?it/s]"
             ]
           },
           "metadata": {},
@@ -641,7 +641,7 @@
             "layout": "IPY_MODEL_e0685ee78f8c499390ac0e1024ab26a9",
             "placeholder": "​",
             "style": "IPY_MODEL_a022879a38e849cfac7178cea469f36f",
-            "value": " 301/301 [04:24\u0026lt;00:00,  6.86s/it]"
+            "value": " 301/301 [04:24&lt;00:00,  6.86s/it]"
           }
         },
         "9cc2c80d152248c4bade0955c25554ad": {

colabs/lora_finetuning.ipynb

@@ -402,9 +402,9 @@
       },
       "cell_type": "markdown",
       "source": [
-        "### `\u003cbos\u003e` / `\u003ceos\u003e`\n",
+        "### `<bos>` / `<eos>`\n",
         "\n",
-        "In Gemma models, the begin of sentence token (`\u003cbos\u003e`) should appear only once at the begining of the input. You can add it either explicitly or with `add_bos=True`:"
+        "In Gemma models, the begin of sentence token (`<bos>`) should appear only once at the begining of the input. You can add it either explicitly or with `add_bos=True`:"
       ]
     },
     {
@@ -432,7 +432,7 @@
         {
           "data": {
             "text/plain": [
-              "[\u003c_Gemma2SpecialTokens.BOS: 2\u003e, 4521, 2134, 235341]"
+              "[<_Gemma2SpecialTokens.BOS: 2>, 4521, 2134, 235341]"
             ]
           },
           "execution_count": 55,
@@ -465,7 +465,7 @@
         {
           "data": {
             "text/plain": [
-              "[\u003c_Gemma2SpecialTokens.BOS: 2\u003e, 4521, 2134, 235341]"
+              "[<_Gemma2SpecialTokens.BOS: 2>, 4521, 2134, 235341]"
             ]
           },
           "execution_count": 59,
@@ -481,9 +481,9 @@
       },
       "cell_type": "markdown",
       "source": [
-        "Similarly, the model can output a `\u003ceos\u003e` token to indicate the prediction is complete.\n",
+        "Similarly, the model can output a `<eos>` token to indicate the prediction is complete.\n",
         "\n",
-        "When fine-tuning Gemma, you can train the model to predict `\u003ceos\u003e` tokens."
+        "When fine-tuning Gemma, you can train the model to predict `<eos>` tokens."
       ]
     },
     {
@@ -509,7 +509,7 @@
         {
           "data": {
             "text/plain": [
-              "[4521, 2134, 235341, \u003c_Gemma2SpecialTokens.EOS: 1\u003e]"
+              "[4521, 2134, 235341, <_Gemma2SpecialTokens.EOS: 1>]"
             ]
           },
           "execution_count": 60,
@@ -525,11 +525,11 @@
       },
       "cell_type": "markdown",
       "source": [
-        "### `\u003cstart_of_turn\u003e` / `\u003cend_of_turn\u003e`\n",
+        "### `<start_of_turn>` / `<end_of_turn>`\n",
         "\n",
-        "When using the instruction-tuned version of Gemma, the `\u003cstart_of_turn\u003e` / `\u003cend_of_turn\u003e` tokens allow to specify who from the user or the model is talking.\n",
+        "When using the instruction-tuned version of Gemma, the `<start_of_turn>` / `<end_of_turn>` tokens allow to specify who from the user or the model is talking.\n",
         "\n",
-        "The `\u003cstart_of_turn\u003e` should be followed by either:\n",
+        "The `<start_of_turn>` should be followed by either:\n",
         "\n",
         "* `user`\n",
         "* `model`\n",
@@ -543,14 +543,14 @@
       },
       "cell_type": "code",
       "source": [
-        "token_ids = tokenizer.encode(\"\"\"\u003cstart_of_turn\u003euser\n",
-        "Knock knock.\u003cend_of_turn\u003e\n",
-        "\u003cstart_of_turn\u003emodel\n",
-        "Who's there ?\u003cend_of_turn\u003e\n",
-        "\u003cstart_of_turn\u003euser\n",
-        "Gemma.\u003cend_of_turn\u003e\n",
-        "\u003cstart_of_turn\u003emodel\n",
-        "Gemma who?\u003cend_of_turn\u003e\"\"\")"
+        "token_ids = tokenizer.encode(\"\"\"<start_of_turn>user\n",
+        "Knock knock.<end_of_turn>\n",
+        "<start_of_turn>model\n",
+        "Who's there ?<end_of_turn>\n",
+        "<start_of_turn>user\n",
+        "Gemma.<end_of_turn>\n",
+        "<start_of_turn>model\n",
+        "Gemma who?<end_of_turn>\"\"\")"
       ],
       "outputs": [],
       "execution_count": null
@@ -584,7 +584,7 @@
               "type": "string"
             },
             "text/plain": [
-              "'\u003cstart_of_turn\u003e'"
+              "'<start_of_turn>'"
             ]
           },
           "execution_count": 77,
@@ -600,11 +600,11 @@
       },
       "cell_type": "markdown",
       "source": [
-        "### `\u003cstart_of_image\u003e`\n",
+        "### `<start_of_image>`\n",
         "\n",
-        "In Gemma 3, to indicate the position of an image in the text, the prompt should contain the special `\u003cstart_of_image\u003e` token. Internally, Gemma model will automatically expand the token to insert the soft images tokens.\n",
+        "In Gemma 3, to indicate the position of an image in the text, the prompt should contain the special `<start_of_image>` token. Internally, Gemma model will automatically expand the token to insert the soft images tokens.\n",
         "\n",
-        "(Note: There's also a `\u003cend_of_image\u003e` token, but is handled internally by the model)"
+        "(Note: There's also a `<end_of_image>` token, but is handled internally by the model)"
       ]
     },
     {
@@ -624,7 +624,7 @@
       "source": [
         "In all Gemma versions, a few tokens (`99`) are unused. This allow custom applications to define and fine-tune their own custom tokens for their application. Those tokens are available through `tokenizer.special_tokens.CUSTOM + xx`, with `xx` being a number between `0` and `98`\n",
         "\n",
-        "\u003c!-- TODO(epot): Add option to customize the special tokens --\u003e"
+        "<!-- TODO(epot): Add option to customize the special tokens -->"
       ]
     },
     {
@@ -656,7 +656,7 @@
               "type": "string"
             },
             "text/plain": [
-              "'\u003cunused17\u003e'"
+              "'<unused17>'"
             ]
           },
           "execution_count": 28,
@@ -694,12 +694,12 @@
       "source": [
         "tokenizer = gm.text.Gemma3Tokenizer(\n",
         "    custom_tokens={\n",
-        "        0: '\u003cmy_custom_tag\u003e',\n",
-        "        17: '\u003cmy_other_tag\u003e',\n",
+        "        0: '<my_custom_tag>',\n",
+        "        17: '<my_other_tag>',\n",
         "    },\n",
         ")\n",
         "\n",
-        "tokenizer.encode('\u003cmy_other_tag\u003e')"
+        "tokenizer.encode('<my_other_tag>')"
       ],
       "outputs": [
         {
@@ -786,7 +786,7 @@
               "type": "string"
             },
             "text/plain": [
-              "'\u003cmy_other_tag\u003e'"
+              "'<my_other_tag>'"
             ]
           },
           "execution_count": 36,

colabs/multimodal.ipynb

@@ -118,7 +118,7 @@
           "output_type": "stream",
           "text": [
             "INFO:2025-06-06 02:43:16,896:jax._src.xla_bridge:749: Unable to initialize backend 'pathways': Could not initialize backend 'pathways'\n",
-            "INFO:2025-06-06 02:43:16,897:jax._src.xla_bridge:749: Unable to initialize backend 'proxy': INVALID_ARGUMENT: IFRT proxy server address must be '\u003ctransport-type\u003e://\u003cbackend-address\u003e' (e.g., 'grpc://localhost'), but got \n",
+            "INFO:2025-06-06 02:43:16,897:jax._src.xla_bridge:749: Unable to initialize backend 'proxy': INVALID_ARGUMENT: IFRT proxy server address must be '<transport-type>://<backend-address>' (e.g., 'grpc://localhost'), but got \n",
             "INFO:2025-06-06 02:43:16,900:jax._src.xla_bridge:749: Unable to initialize backend 'mlcr': Could not initialize backend 'mlcr'\n",
             "INFO:2025-06-06 02:43:16,901:jax._src.xla_bridge:749: Unable to initialize backend 'sliceme': Could not initialize backend 'sliceme'\n"
           ]
@@ -194,10 +194,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -213,10 +213,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -233,10 +233,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -252,10 +252,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -272,10 +272,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -291,10 +291,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -311,10 +311,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -330,10 +330,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -350,10 +350,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -369,10 +369,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -407,7 +407,7 @@
         "\n",
         "To create your own tool, you can inherit from the `gm.tools.Tool` class. You should provide:\n",
         "\n",
-        "* A description \u0026 example, so the model knows how to use your tool\n",
+        "* A description & example, so the model knows how to use your tool\n",
         "* Implement the `call` method. The `call` function can take arbitrary `**kwargs`, but the name of the args should match the ones defined in `tool_kwargs` and `tool_kwargs_doc`"
       ]
     },
@@ -435,12 +435,12 @@
         "      query='Which day of the week are we today ?',\n",
         "      thought='The `datetime.strptime` uses %a for day of the week',\n",
         "      tool_kwargs={'format': '%a'},\n",
-        "      tool_kwargs_doc={'format': '\u003cANY datetime.strptime expression\u003e'},\n",
+        "      tool_kwargs_doc={'format': '<ANY datetime.strptime expression>'},\n",
         "      result='Sat',\n",
         "      answer='Today is Saturday.',\n",
         "  )\n",
         "\n",
-        "  def call(self, format: str) -\u003e str:\n",
+        "  def call(self, format: str) -> str:\n",
         "    dt = datetime.datetime.now()\n",
         "    return dt.strftime(format)\n"
       ],
@@ -499,10 +499,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},
@@ -518,10 +518,10 @@
         {
           "data": {
             "text/html": [
-              "\u003chr\u003e"
+              "<hr>"
             ],
             "text/plain": [
-              "\u003cIPython.core.display.HTML object\u003e"
+              "<IPython.core.display.HTML object>"
             ]
           },
           "metadata": {},