Update tiny models for HiggsAudioV2ForConditionalGeneration

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

chat_template.jinja

@@ -0,0 +1,74 @@
+{{- bos_token }}
+{#- This block extracts the system message, so we can slot it into the right place. #}
+{%- if messages[0]['role'] == 'system' %}
+    {%- if messages[0]['content'] is string %}
+        {%- set system_message = messages[0]['content']|trim %}
+    {%- elif messages[0]['content'] is iterable and messages[0]['content'][0]['type'] == 'text' %}
+        {%- set system_message = messages[0]['content'][0]['text']|trim %}
+    {%- else %}
+        {{- raise_exception("System message content must be a string or contain text type!") }}
+    {%- endif %}
+    {%- set messages = messages[1:] %}
+{%- else %}
+    {{- raise_exception("A system message is required but not provided!") }}
+{%- endif %}
+
+{#- System message #}
+{{- "<|start_header_id|>system<|end_header_id|>\n\n" }}
+{{- system_message }}
+
+{#- Check for scene message and handle it specially #}
+{%- if messages and messages[0]['role'] == 'scene' %}
+    {{- "\n\n<|scene_desc_start|>\n" }}
+    {%- if messages[0]['content'] is string %}
+        {{- messages[0]['content'] | trim }}
+    {%- elif messages[0]['content'] is iterable %}
+        {%- for content_item in messages[0]['content'] %}
+            {%- if content_item['type'] == 'text' %}
+                {%- set text_content = content_item['text'] | trim %}
+                {{- text_content }}
+                {%- if loop.first and not loop.last %}
+                    {{- "\n\n" }}
+                {%- endif %}
+                {%- if not loop.first and not loop.last and messages[0]['content'][loop.index]['type'] != 'audio' %}
+                    {{- "\n" }}
+                {%- endif %}
+            {%- elif content_item['type'] == 'audio' %}
+                {{- ' <|audio_out_bos|><|AUDIO_OUT|><|audio_eos|>' }}
+                {%- if not loop.last %}
+                    {{- "\n" }}
+                {%- endif %}
+            {%- endif %}
+        {%- endfor %}
+    {%- endif %}
+    {{- "\n<|scene_desc_end|>" }}
+    {%- set messages = messages[1:] %}
+{%- endif %}
+
+{{- "<|eot_id|>" }}
+
+{#- Loop through all messages #}
+{%- for message in messages %}
+    {{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n' }}
+    {%- if message['role'] == 'assistant' %}
+        {%- if message['content'] is not iterable or message['content'][0]['type'] != 'audio' %}
+            {{- raise_exception("Assistant messages must contain audio content only!") }}
+        {%- endif %}
+        {{- '<|audio_out_bos|><|AUDIO_OUT|><|audio_eos|>' }}
+    {%- else %}
+        {%- if message['content'] is string %}
+            {{- message['content'] | trim }}
+        {%- elif message['content'] is iterable %}
+            {%- for content_item in message['content'] %}
+                {%- if content_item['type'] == 'text' %}
+                    {{- content_item['text'] | trim }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|eot_id|>' }}
+{%- endfor %}
+
+{%- if add_generation_prompt %}
+    {{- '<|start_header_id|>assistant<|end_header_id|>\n\n<|audio_out_bos|>' }}
+{%- endif %}

config.json

@@ -0,0 +1,43 @@
+{
+  "architectures": [
+    "HiggsAudioV2ForConditionalGeneration"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "audio_bos_token_id": 128254,
+  "audio_delay_token_id": 128253,
+  "audio_stream_bos_id": 15,
+  "audio_stream_eos_id": 14,
+  "audio_token_id": 128255,
+  "bos_token_id": 128000,
+  "codebook_size": 16,
+  "dtype": "float32",
+  "eos_token_id": 128001,
+  "head_dim": 16,
+  "hidden_act": "silu",
+  "hidden_size": 32,
+  "initializer_range": 0.02,
+  "intermediate_size": 37,
+  "max_position_embeddings": 512,
+  "mlp_bias": false,
+  "model_type": "higgs_audio_v2",
+  "num_attention_heads": 2,
+  "num_codebooks": 2,
+  "num_hidden_layers": 2,
+  "num_key_value_heads": 2,
+  "pad_token_id": 128001,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_parameters": {
+    "factor": 32.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_theta": 500000.0,
+    "rope_type": "llama3"
+  },
+  "tie_word_embeddings": false,
+  "transformers_version": "5.16.0.dev0",
+  "use_cache": true,
+  "vocab_size": 128256
+}

generation_config.json

@@ -0,0 +1,10 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 128000,
+  "eos_token_id": 128001,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 128001,
+  "transformers_version": "5.16.0.dev0",
+  "use_cache": true
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1ce2743f666a5c3c5eeff2e87a1ee82de3e2818bf96e65cde80e7be3c347d05
+size 16519168

preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+  "feature_extractor_type": "DacFeatureExtractor",
+  "feature_size": 1,
+  "hop_length": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 24000
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a222563314bf6ffe3471622bff017ff5bb0630f2924faf44216195ebfef2af3
+size 17209675

tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<|begin_of_text|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|end_of_text|>",
+  "is_local": true,
+  "local_files_only": false,
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 512,
+  "pad_token": "<|end_of_text|>",
+  "processor_class": "HiggsAudioV2Processor",
+  "tokenizer_class": "TokenizersBackend",
+  "trust_remote": true
+}