Upload folder using huggingface_hub

.meta.json

@@ -1,4 +1,4 @@
 {
-  "torch": "2.11.0",
-  "transformers": "5.5.0"
+  "torch": "2.12.0",
+  "transformers": "5.9.0"
 }

README.md

@@ -193,16 +193,62 @@ model.save_pretrained(save_folder)
 ```text
 Gemma4ForConditionalGeneration(
   (model): Gemma4Model(
+    (vision_tower): Gemma4VisionModel(
+      (patch_embedder): Gemma4VisionPatchEmbedder(
+        (input_proj): Linear(in_features=768, out_features=8, bias=False)
+      )
+      (encoder): Gemma4VisionEncoder(
+        (rotary_emb): Gemma4VisionRotaryEmbedding()
+        (layers): ModuleList(
+          (0-1): 2 x Gemma4VisionEncoderLayer(
+            (self_attn): Gemma4VisionAttention(
+              (q_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=8, out_features=128, bias=False)
+              )
+              (k_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=8, out_features=128, bias=False)
+              )
+              (v_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=8, out_features=128, bias=False)
+              )
+              (o_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=128, out_features=8, bias=False)
+              )
+              (q_norm): Gemma4RMSNorm()
+              (k_norm): Gemma4RMSNorm()
+              (v_norm): Gemma4RMSNorm()
+            )
+            (mlp): Gemma4VisionMLP(
+              (gate_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=8, out_features=64, bias=False)
+              )
+              (up_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=8, out_features=64, bias=False)
+              )
+              (down_proj): Gemma4ClippableLinear(
+                (linear): Linear(in_features=64, out_features=8, bias=False)
+              )
+              (act_fn): GELUTanh()
+            )
+            (input_layernorm): Gemma4RMSNorm()
+            (post_attention_layernorm): Gemma4RMSNorm()
+            (pre_feedforward_layernorm): Gemma4RMSNorm()
+            (post_feedforward_layernorm): Gemma4RMSNorm()
+          )
+        )
+      )
+      (pooler): Gemma4VisionPooler()
+    )
     (language_model): Gemma4TextModel(
       (embed_tokens): Gemma4TextScaledWordEmbedding(262144, 8, padding_idx=0)
       (layers): ModuleList(
         (0): Gemma4TextDecoderLayer(
           (self_attn): Gemma4TextAttention(
+            (q_proj): Linear(in_features=8, out_features=256, bias=False)
             (q_norm): Gemma4RMSNorm()
             (k_norm): Gemma4RMSNorm()
             (v_norm): Gemma4RMSNorm()
             (k_proj): Linear(in_features=8, out_features=128, bias=False)
-            (q_proj): Linear(in_features=8, out_features=256, bias=False)
             (v_proj): Linear(in_features=8, out_features=128, bias=False)
             (o_proj): Linear(in_features=256, out_features=8, bias=False)
           )
@@ -223,11 +269,11 @@ Gemma4ForConditionalGeneration(
         )
         (1): Gemma4TextDecoderLayer(
           (self_attn): Gemma4TextAttention(
+            (q_proj): Linear(in_features=8, out_features=512, bias=False)
             (q_norm): Gemma4RMSNorm()
             (k_norm): Gemma4RMSNorm()
             (v_norm): Gemma4RMSNorm()
             (k_proj): Linear(in_features=8, out_features=256, bias=False)
-            (q_proj): Linear(in_features=8, out_features=512, bias=False)
             (v_proj): Linear(in_features=8, out_features=256, bias=False)
             (o_proj): Linear(in_features=512, out_features=8, bias=False)
           )
@@ -248,12 +294,8 @@ Gemma4ForConditionalGeneration(
         )
         (2): Gemma4TextDecoderLayer(
           (self_attn): Gemma4TextAttention(
-            (q_norm): Gemma4RMSNorm()
-            (k_norm): Gemma4RMSNorm()
-            (v_norm): Gemma4RMSNorm()
-            (k_proj): Linear(in_features=8, out_features=128, bias=False)
             (q_proj): Linear(in_features=8, out_features=256, bias=False)
-            (v_proj): Linear(in_features=8, out_features=128, bias=False)
+            (q_norm): Gemma4RMSNorm()
             (o_proj): Linear(in_features=256, out_features=8, bias=False)
           )
           (mlp): Gemma4TextMLP(
@@ -273,12 +315,8 @@ Gemma4ForConditionalGeneration(
         )
         (3): Gemma4TextDecoderLayer(
           (self_attn): Gemma4TextAttention(
-            (q_norm): Gemma4RMSNorm()
-            (k_norm): Gemma4RMSNorm()
-            (v_norm): Gemma4RMSNorm()
-            (k_proj): Linear(in_features=8, out_features=256, bias=False)
             (q_proj): Linear(in_features=8, out_features=512, bias=False)
-            (v_proj): Linear(in_features=8, out_features=256, bias=False)
+            (q_norm): Gemma4RMSNorm()
             (o_proj): Linear(in_features=512, out_features=8, bias=False)
           )
           (mlp): Gemma4TextMLP(
@@ -303,66 +341,16 @@ Gemma4ForConditionalGeneration(
       (per_layer_model_projection): Linear(in_features=8, out_features=8, bias=False)
       (per_layer_projection_norm): Gemma4RMSNorm()
     )
-    (vision_tower): Gemma4VisionModel(
-      (patch_embedder): Gemma4VisionPatchEmbedder(
-        (input_proj): Linear(in_features=768, out_features=8, bias=False)
-      )
-      (encoder): Gemma4VisionEncoder(
-        (rotary_emb): Gemma4VisionRotaryEmbedding()
-        (layers): ModuleList(
-          (0-1): 2 x Gemma4VisionEncoderLayer(
-            (self_attn): Gemma4VisionAttention(
-              (q_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=8, out_features=128, bias=False)
-              )
-              (k_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=8, out_features=128, bias=False)
-              )
-              (v_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=8, out_features=128, bias=False)
-              )
-              (o_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=128, out_features=8, bias=False)
-              )
-              (q_norm): Gemma4RMSNorm()
-              (k_norm): Gemma4RMSNorm()
-              (v_norm): Gemma4RMSNorm()
-            )
-            (mlp): Gemma4VisionMLP(
-              (gate_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=8, out_features=64, bias=False)
-              )
-              (up_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=8, out_features=64, bias=False)
-              )
-              (down_proj): Gemma4ClippableLinear(
-                (linear): Linear(in_features=64, out_features=8, bias=False)
-              )
-              (act_fn): GELUTanh()
-            )
-            (input_layernorm): Gemma4RMSNorm()
-            (post_attention_layernorm): Gemma4RMSNorm()
-            (pre_feedforward_layernorm): Gemma4RMSNorm()
-            (post_feedforward_layernorm): Gemma4RMSNorm()
-          )
-        )
-      )
-      (pooler): Gemma4VisionPooler()
-    )
-    (embed_vision): Gemma4MultimodalEmbedder(
-      (embedding_projection): Linear(in_features=8, out_features=8, bias=False)
-      (embedding_pre_projection_norm): Gemma4RMSNorm()
-    )
     (audio_tower): Gemma4AudioModel(
       (subsample_conv_projection): Gemma4AudioSubSampleConvProjection(
         (layer0): Gemma4AudioSubSampleConvProjectionLayer(
           (conv): Conv2d(1, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
-          (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
+          (norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True, bias=False)
           (act): ReLU()
         )
         (layer1): Gemma4AudioSubSampleConvProjectionLayer(
           (conv): Conv2d(128, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
-          (norm): LayerNorm((32,), eps=1e-06, elementwise_affine=True)
+          (norm): LayerNorm((32,), eps=1e-06, elementwise_affine=True, bias=False)
           (act): ReLU()
         )
         (input_proj_linear): Linear(in_features=1024, out_features=64, bias=False)
@@ -426,6 +414,10 @@ Gemma4ForConditionalGeneration(
       )
       (output_proj): Linear(in_features=64, out_features=32, bias=True)
     )
+    (embed_vision): Gemma4MultimodalEmbedder(
+      (embedding_projection): Linear(in_features=8, out_features=8, bias=False)
+      (embedding_pre_projection_norm): Gemma4RMSNorm()
+    )
     (embed_audio): Gemma4MultimodalEmbedder(
       (embedding_projection): Linear(in_features=32, out_features=8, bias=False)
       (embedding_pre_projection_norm): Gemma4RMSNorm()
@@ -439,5 +431,5 @@ Gemma4ForConditionalGeneration(
 
 ### Test environment:
 
-- torch: 2.11.0
-- transformers: 5.5.0
+- torch: 2.12.0
+- transformers: 5.9.0

chat_template.jinja

@@ -1,9 +1,9 @@
-{%- macro format_parameters(properties, required) -%}
+{%- macro format_parameters(properties, required, filter_keys=false) -%}
     {%- set standard_keys = ['description', 'type', 'properties', 'required', 'nullable'] -%}
     {%- set ns = namespace(found_first=false) -%}
     {%- for key, value in properties | dictsort -%}
         {%- set add_comma = false -%}
-        {%- if key not in standard_keys -%}
+        {%- if not filter_keys or key not in standard_keys -%}
             {%- if ns.found_first %},{% endif -%}
             {%- set ns.found_first = true -%}
             {{ key }}:{
@@ -65,7 +65,7 @@
                 {%- elif value is mapping -%}
                     {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
                     properties:{
-                    {{- format_parameters(value, value['required'] | default([])) -}}
+                    {{- format_parameters(value, value['required'] | default([]), filter_keys=true) -}}
                     }
                 {%- endif -%}
                 {%- if value['required'] -%}
@@ -178,18 +178,21 @@
 {#- Handle System/Tool Definitions Block -#}
 {%- if (enable_thinking is defined and enable_thinking) or tools or messages[0]['role'] in ['system', 'developer'] -%}
     {{- '<|turn>system\n' -}}
-
     {#- Inject Thinking token at the very top of the FIRST system turn -#}
     {%- if enable_thinking is defined and enable_thinking -%}
         {{- '<|think|>\n' -}}
         {%- set ns.prev_message_type = 'think' -%}
     {%- endif -%}
-
     {%- if messages[0]['role'] in ['system', 'developer'] -%}
-        {{- messages[0]['content'] | trim -}}
+        {%- if messages[0]['content'] is string -%}
+            {{- messages[0]['content'] | trim -}}
+        {%- elif messages[0]['content'] is sequence -%}
+            {%- for item in messages[0]['content'] -%}
+                {{- item['text'] | trim + ' '-}}
+            {%- endfor -%}
+        {%- endif -%}
         {%- set loop_messages = messages[1:] -%}
     {%- endif -%}
-
     {%- if tools -%}
         {%- for tool in tools %}
             {{- '<|tool>' -}}
@@ -198,7 +201,6 @@
         {%- endfor %}
         {%- set ns.prev_message_type = 'tool' -%}
     {%- endif -%}
-
     {{- '<turn|>\n' -}}
 {%- endif %}
 
@@ -293,6 +295,15 @@
                                 {%- endif -%}
                             {%- endfor -%}
                             {{- format_tool_response_block(ns_tname.name, ns_txt.s) -}}
+                            {%- for part in tool_body -%}
+                                {%- if part.get('type') == 'image' -%}
+                                    {{- '<|image|>' -}}
+                                {%- elif part.get('type') == 'audio' -%}
+                                    {{- '<|audio|>' -}}
+                                {%- elif part.get('type') == 'video' -%}
+                                    {{- '<|video|>' -}}
+                                {%- endif -%}
+                            {%- endfor -%}
                         {%- else -%}
                             {{- format_tool_response_block(ns_tname.name, tool_body) -}}
                         {%- endif -%}
@@ -302,6 +313,7 @@
                 {%- endfor -%}
             {%- endif -%}
 
+            {%- set captured_content -%}
             {%- if message['content'] is string -%}
                 {%- if role == 'model' -%}
                     {{- strip_thinking(message['content']) -}}
@@ -328,10 +340,14 @@
                     {%- endif -%}
                 {%- endfor -%}
             {%- endif -%}
+            {%- endset -%}
+
+            {{- captured_content -}}
+            {%- set has_content = captured_content | trim | length > 0 -%}
 
         {%- if ns.prev_message_type == 'tool_call' and not ns_tr_out.flag -%}
             {{- '<|tool_response>' -}}
-        {%- elif not (ns_tr_out.flag and not message.get('content')) -%}
+        {%- elif not (ns_tr_out.flag and not has_content) -%}
             {{- '<turn|>\n' -}}
         {%- endif -%}
     {%- endif -%}

config.json

@@ -111,7 +111,7 @@
     "vocab_size_per_layer_input": 262144
   },
   "tie_word_embeddings": true,
-  "transformers_version": "5.5.0",
+  "transformers_version": "5.9.0",
   "video_token_id": 258884,
   "vision_config": {
     "_name_or_path": "",

generation_config.json

@@ -10,6 +10,6 @@
   "temperature": 1.0,
   "top_k": 64,
   "top_p": 0.95,
-  "transformers_version": "5.5.0",
+  "transformers_version": "5.9.0",
   "trust_remote_code": true
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6c2045c91d5ee53e379d6f3c4947b914c36ab4308f708e25858c2eb5bd694bc3
-size 9483948
+oid sha256:8582ee51fce3dac0a80d9a90008785011e2e24d9d26234822df7361d2092877b
+size 9470740

tokenizer_config.json

@@ -18,6 +18,7 @@
   ],
   "image_token": "<|image|>",
   "is_local": false,
+  "local_files_only": false,
   "mask_token": "<mask>",
   "model_max_length": 1000000000000000019884624838656,
   "model_specific_special_tokens": {