Upload v22 DPO (length-normalized, MOSHI_MODE=speech_only) on kalbin/moshi-on-policy-dpo-v20-kyutai

config.json

@@ -0,0 +1,144 @@
+{
+  "architectures": [
+    "MoshiForConditionalGeneration"
+  ],
+  "attention_dropout": 0.0,
+  "audio_encoder_config": {
+    "_frame_rate": 12.5,
+    "_name_or_path": "kyutai/mimi",
+    "add_cross_attention": false,
+    "architectures": [
+      "MimiModel"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "audio_channels": 1,
+    "bos_token_id": null,
+    "codebook_dim": 256,
+    "codebook_size": 2048,
+    "compress": 2,
+    "cross_attention_hidden_size": null,
+    "decoder_start_token_id": null,
+    "dilation_growth_rate": 2,
+    "dtype": "bfloat16",
+    "eos_token_id": null,
+    "finetuning_task": null,
+    "head_dim": 64,
+    "hidden_act": "gelu",
+    "hidden_size": 512,
+    "initializer_range": 0.02,
+    "intermediate_size": 2048,
+    "is_decoder": false,
+    "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",
+    "pad_token_id": null,
+    "prefix": null,
+    "pruned_heads": {},
+    "residual_kernel_size": 3,
+    "rope_parameters": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    },
+    "sampling_rate": 24000,
+    "sep_token_id": null,
+    "sliding_window": 250,
+    "task_specific_params": null,
+    "tf_legacy_loss": false,
+    "tie_encoder_decoder": false,
+    "tie_word_embeddings": true,
+    "tokenizer_class": null,
+    "torchscript": false,
+    "trim_right_ratio": 1.0,
+    "upsample_groups": 512,
+    "upsampling_ratios": [
+      8,
+      6,
+      5,
+      4
+    ],
+    "use_bfloat16": false,
+    "use_cache": false,
+    "use_causal_conv": true,
+    "use_conv_shortcut": false,
+    "use_streaming": false,
+    "vector_quantization_hidden_dimension": 256
+  },
+  "audio_vocab_size": 2048,
+  "bos_token_id": null,
+  "depth_decoder_config": {
+    "add_cross_attention": false,
+    "attention_dropout": 0.0,
+    "audio_vocab_size": 2048,
+    "bos_token_id": null,
+    "cross_attention_hidden_size": null,
+    "decoder_start_token_id": null,
+    "dtype": "bfloat16",
+    "eos_token_id": null,
+    "ffn_dim": 5632,
+    "finetuning_task": null,
+    "head_dim": 64,
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "initializer_range": 0.02,
+    "input_size": 4096,
+    "is_decoder": false,
+    "max_position_embeddings": 9,
+    "model_type": "moshi_depth",
+    "num_attention_heads": 16,
+    "num_codebooks": 8,
+    "num_hidden_layers": 6,
+    "num_key_value_heads": 16,
+    "pad_token_id": null,
+    "prefix": null,
+    "pruned_heads": {},
+    "rms_norm_eps": 1e-08,
+    "sep_token_id": null,
+    "sliding_window": 8,
+    "task_specific_params": null,
+    "tf_legacy_loss": false,
+    "tie_encoder_decoder": false,
+    "tie_word_embeddings": false,
+    "tokenizer_class": null,
+    "torchscript": false,
+    "use_bfloat16": false,
+    "use_cache": true,
+    "vocab_size": 32000
+  },
+  "dtype": "bfloat16",
+  "eos_token_id": null,
+  "ffn_dim": 22528,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "max_position_embeddings": 3000,
+  "model_type": "moshi",
+  "num_attention_heads": 32,
+  "num_codebooks": 8,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": null,
+  "rms_norm_eps": 1e-08,
+  "rope_parameters": {
+    "rope_theta": 10000.0,
+    "rope_type": "default"
+  },
+  "sliding_window": 3000,
+  "tie_word_embeddings": false,
+  "transformers_version": "5.8.0",
+  "use_cache": false,
+  "vocab_size": 32000
+}

generation_config.json

@@ -0,0 +1,16 @@
+{
+  "cache_implementation": "sliding_window",
+  "depth_decoder_config": {
+    "cache_implementation": "sliding_window",
+    "do_sample": true,
+    "max_length": 9,
+    "min_length": 9,
+    "temperature": 0.8,
+    "top_k": 250,
+    "transformers_version": "4.46.0.dev0"
+  },
+  "do_sample": true,
+  "temp": 0.7,
+  "top_k": 25,
+  "transformers_version": "5.8.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d33bfc246a0d9044c2b686a1f04fb12c5d03550dadd4f3f4a3d17c3965a1e72
+size 15567851962

tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "backend": "tokenizers",
+  "bos_token_id": null,
+  "chat_template": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token_id": null,
+  "is_local": false,
+  "local_files_only": false,
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token_id": null,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<unk>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:412577bfcaf949cd8a0f4cb3c9d1594ece3be9e86354d96a4dddcbb9d4956356
+size 5329