(Trained with Unsloth)

config.json

@@ -0,0 +1,123 @@
+{
+  "architectures": [
+    "CsmForConditionalGeneration"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "audio_eos_token_id": 128003,
+  "audio_token_id": 128002,
+  "bos_token_id": 128000,
+  "codebook_eos_token_id": 0,
+  "codebook_pad_token_id": 2050,
+  "codec_config": {
+    "_frame_rate": 12.5,
+    "_name_or_path": "kyutai/mimi",
+    "architectures": [
+      "MimiModel"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "audio_channels": 1,
+    "codebook_dim": 256,
+    "codebook_size": 2048,
+    "compress": 2,
+    "dilation_growth_rate": 2,
+    "torch_dtype": "float16",
+    "head_dim": 64,
+    "hidden_act": "gelu",
+    "hidden_size": 512,
+    "initializer_range": 0.02,
+    "intermediate_size": 2048,
+    "kernel_size": 7,
+    "last_kernel_size": 3,
+    "layer_scale_initial_scale": 0.01,
+    "max_position_embeddings": 8000,
+    "model_type": "mimi",
+    "norm_eps": 1e-05,
+    "normalize": false,
+    "num_attention_heads": 8,
+    "num_filters": 64,
+    "num_hidden_layers": 8,
+    "num_key_value_heads": 8,
+    "num_quantizers": 32,
+    "num_residual_layers": 1,
+    "num_semantic_quantizers": 1,
+    "pad_mode": "constant",
+    "residual_kernel_size": 3,
+    "rope_theta": 10000.0,
+    "sampling_rate": 24000,
+    "sliding_window": 250,
+    "trim_right_ratio": 1.0,
+    "upsample_groups": 512,
+    "upsampling_ratios": [
+      8,
+      6,
+      5,
+      4
+    ],
+    "use_cache": false,
+    "use_causal_conv": true,
+    "use_conv_shortcut": false,
+    "use_streaming": false,
+    "vector_quantization_hidden_dimension": 256
+  },
+  "depth_decoder_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "backbone_hidden_size": 2048,
+    "torch_dtype": "float16",
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "initializer_range": 0.02,
+    "intermediate_size": 8192,
+    "max_position_embeddings": 33,
+    "mlp_bias": false,
+    "model_type": "csm_depth_decoder_model",
+    "num_attention_heads": 8,
+    "num_codebooks": 32,
+    "num_hidden_layers": 4,
+    "num_key_value_heads": 2,
+    "rms_norm_eps": 1e-05,
+    "rope_scaling": {
+      "factor": 32.0,
+      "high_freq_factor": 0.0078125,
+      "low_freq_factor": 0.001953125,
+      "original_max_position_embeddings": 16,
+      "rope_type": "llama3"
+    },
+    "rope_theta": 500000,
+    "use_cache": true,
+    "vocab_size": 2051
+  },
+  "torch_dtype": "float16",
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 8192,
+  "max_position_embeddings": 2048,
+  "mlp_bias": false,
+  "model_type": "csm",
+  "num_attention_heads": 32,
+  "num_codebooks": 32,
+  "num_hidden_layers": 16,
+  "num_key_value_heads": 8,
+  "pad_token_id": 128256,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 32.0,
+    "high_freq_factor": 0.5,
+    "low_freq_factor": 0.125,
+    "original_max_position_embeddings": 1024,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000,
+  "text_vocab_size": 128256,
+  "tie_codebooks_embeddings": true,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.56.2",
+  "unsloth_version": "2025.10.3",
+  "use_cache": true,
+  "vocab_size": 2051
+}

tokenizer_config.json

@@ -2060,5 +2060,6 @@
   "model_max_length": 131072,
   "pad_token": "<|end_of_text|>",
   "processor_class": "CsmProcessor",
-  "tokenizer_class": "PreTrainedTokenizerFast"
-}
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "chat_template": "\n{%- for message in messages %}\n    {#-- Validate role is a stringified integer --#}\n    {%- if not message['role'] is string or not message['role'].isdigit() %}\n        {{- raise_exception(\"The role must be an integer or a stringified integer (e.g. '0') designating the speaker id\") }}\n    {%- endif %}\n\n    {#-- Validate content is a list --#}\n    {%- set content = message['content'] %}\n    {%- if content is not iterable or content is string %}\n        {{- raise_exception(\"The content must be a list\") }}\n    {%- endif %}\n\n    {#-- Collect content types --#}\n    {%- set content_types = content | map(attribute='type') | list %}\n    {%- set is_last = loop.last %}\n\n    {#-- Last message validation --#}\n    {%- if is_last %}\n        {%- if 'text' not in content_types %}\n            {{- raise_exception(\"The last message must include one item of type 'text'\") }}\n        {%- elif (content_types | select('equalto', 'text') | list | length > 1) or (content_types | select('equalto', 'audio') | list | length > 1) %}\n            {{- raise_exception(\"At most two items are allowed in the last message: one 'text' and one 'audio'\") }}\n        {%- endif %}\n\n    {#-- All other messages validation --#}\n    {%- else %}\n        {%- if content_types | select('equalto', 'text') | list | length != 1\n              or content_types | select('equalto', 'audio') | list | length != 1 %}\n            {{- raise_exception(\"Each message (except the last) must contain exactly one 'text' and one 'audio' item\") }}\n        {%- elif content_types | reject('in', ['text', 'audio']) | list | length > 0 %}\n            {{- raise_exception(\"Only 'text' and 'audio' types are allowed in content\") }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n\n{%- for message in messages %}\n    {{- bos_token }}\n    {{- '[' + message['role'] + ']' }}\n    {{- message['content'][0]['text'] }}\n    {{- eos_token }}\n    {%- if message['content']|length > 1 %}\n        {{- '<|AUDIO|><|audio_eos|>' }}\n    {%- endif %}\n{%- endfor %}\n"
+}