Upload model

config.json

@@ -1,13 +1,18 @@
 {
+  "architectures": [
+    "SpeechLLMModel"
+  ],
   "audio_enc_dim": 1024,
   "audio_encoder_name": "microsoft/wavlm-large",
   "audio_processor_name": "microsoft/wavlm-base",
   "auto_map": {
-    "AutoConfig": "config.SpeechLLMModelConfig"
+    "AutoConfig": "config.SpeechLLMModelConfig",
+    "AutoModel": "model.SpeechLLMModel"
   },
   "llm_dim": 2048,
   "llm_model_checkpoint": "hf_repo/llm_model_checkpoint",
   "llm_model_name": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
   "model_type": "custom_model",
+  "torch_dtype": "float32",
   "transformers_version": "4.38.2"
 }

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca3b8867386b18c6aadb0499a682799dcbd76a573769d4bd28e3b3db0637d34c
+size 4972078744

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c276e96e5b2bf5bc6824b157179f3cb4304de0e674ffbb793cc2cb62b8dd1cd6
+size 836841304

model.py

@@ -0,0 +1,119 @@
+import torch
+from torch import nn
+import torchaudio
+from transformers import PreTrainedModel, AutoModelForCausalLM, AutoTokenizer, HubertModel, AutoFeatureExtractor, AutoModel
+from .config import SpeechLLMModelConfig
+from peft import LoraConfig, get_peft_model
+
+class TransformerAudioEnoder(nn.Module):
+    def __init__(self, model_name='microsoft/wavlm-large', finetune=False):
+        super().__init__()
+        self.encoder = AutoModel.from_pretrained(model_name)
+
+    def forward(self, x):
+        return self.encoder(x).last_hidden_state
+    
+    def return_device(self):
+        return next(self.parameters()).device
+
+
+class CNNConnector(nn.Module):
+    def __init__(self, in_channels, out_channels, k=2):
+        super().__init__()
+        self.layer = nn.Sequential(
+            nn.ReLU(),
+            nn.Conv1d(in_channels, out_channels//2, kernel_size=5,
+                      stride=1, padding=0),
+            nn.ReLU(),
+            nn.Conv1d(out_channels//2, out_channels, kernel_size=5,
+                      stride=k, padding=0),
+            nn.ReLU(),
+            nn.Conv1d(out_channels, out_channels, kernel_size=5,
+                      stride=1, padding=0),
+        )
+
+    def forward(self, x):
+        return self.layer(x.transpose(1,2)).transpose(1,2)
+    
+    
+class SpeechLLMModel(PreTrainedModel):
+    config_class = SpeechLLMModelConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.audio_processor = AutoFeatureExtractor.from_pretrained(config.audio_processor_name)
+        self.audio_encoder = TransformerAudioEnoder(config.audio_encoder_name)
+        self.connector = CNNConnector(config.audio_enc_dim, config.llm_dim)
+
+        self.llm_model = AutoModelForCausalLM.from_pretrained(config.llm_model_checkpoint)
+        self.llm_tokenizer = AutoTokenizer.from_pretrained(config.llm_model_name)
+
+        # self.llm_model = AutoModelForCausalLM.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
+        # self.llm_tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
+
+        # peft_config = LoraConfig(
+        #     r=8,
+        #     lora_alpha=16,
+        #     target_modules=['q_proj', 'k_proj', 'v_proj', 'o_proj', 'up_proj', 'down_proj', 'gate_proj'],
+        #     lora_dropout=0.05,
+        #     task_type="CAUSAL_LM",
+        # )
+        # self.llm_model = get_peft_model(self.llm_model, peft_config)
+
+    def encode(self, mel, pre_tokenized_ids, post_tokenized_ids, output_tokenized_ids):
+        batch_size = mel.shape[0]
+
+        with torch.no_grad():
+            speech_embeds = self.audio_encoder(mel)
+            speech_embeds = self.connector(speech_embeds)
+            
+        embedder = self.llm_model.model.embed_tokens
+        pre_prompt_embeds = embedder(pre_tokenized_ids)
+        post_prompt_embeds = embedder(post_tokenized_ids)
+        output_prompt_embeds = embedder(output_tokenized_ids)
+
+        combined_embeds = torch.cat([pre_prompt_embeds, speech_embeds, post_prompt_embeds, output_prompt_embeds], dim=1)
+        atts = torch.ones(combined_embeds.size()[:-1], dtype=torch.long).to(combined_embeds.device)
+
+        input_token_length = pre_tokenized_ids.shape[1] + speech_embeds.shape[1] + post_tokenized_ids.shape[1]
+        label_ids = torch.cat([
+            torch.ones([batch_size, input_token_length], device=combined_embeds.device) * -100,
+            output_tokenized_ids
+        ], 1).to(combined_embeds.device).to(torch.int64)
+        return combined_embeds, atts, label_ids
+
+    def forward(self, wav_tensor, pre_tokenized_ids, post_tokenized_ids, output_tokenized_ids, attention_mask=None):
+        combined_embeds, atts, label_ids = self.encode(wav_tensor, pre_tokenized_ids, post_tokenized_ids, output_tokenized_ids)
+        outputs = self.llm_model(inputs_embeds=combined_embeds, attention_mask=attention_mask)
+        return outputs
+    
+    def generate_meta(self, audio_path, instruction="Give me the following information about the audio [Transcript]", max_new_tokens=2000):
+        device = self.audio_encoder.return_device()
+        pre_speech_prompt = f'''Instruction:
+{instruction}
+
+Input: 
+<speech>'''
+        post_speech_prompt = f'''</speech>
+
+Output:'''
+        output_prompt = '\n<s>'
+
+        with torch.no_grad():
+            wav_tensor, sr = torchaudio.load(audio_path)
+            wav_tensor = self.audio_processor(wav_tensor.squeeze(), return_tensors="pt", sampling_rate=16000).input_values
+
+            pre_tokenized_ids = self.llm_tokenizer(pre_speech_prompt, padding="do_not_pad", return_tensors='pt', truncation=False, add_special_tokens=False)["input_ids"]
+            post_tokenized_ids = self.llm_tokenizer(post_speech_prompt, padding="do_not_pad", return_tensors='pt', truncation=False, add_special_tokens=False)["input_ids"]
+            output_tokenized_ids = self.llm_tokenizer(output_prompt, padding="do_not_pad", return_tensors='pt', truncation=False, add_special_tokens=False)["input_ids"]
+
+            combined_embeds, atts, label_ids = self.encode(wav_tensor.to(device), pre_tokenized_ids.to(device), post_tokenized_ids.to(device), output_tokenized_ids.to(device))
+
+            out = self.llm_model.generate(
+                inputs_embeds=combined_embeds,
+                max_new_tokens=max_new_tokens,
+            ).cpu().tolist()[0]
+
+            output_text = self.llm_tokenizer.decode(out, skip_special_tokens=True)
+            return output_text
+

model.safetensors.index.json

@@ -0,0 +1,702 @@
+{
+  "metadata": {
+    "total_size": 5808827136
+  },
+  "weight_map": {
+    "audio_encoder.encoder.encoder.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.rel_attn_embed.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.10.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.11.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.12.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.13.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.14.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.15.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.16.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.17.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.18.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.19.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.2.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.20.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.21.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.22.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.23.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.3.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.4.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.5.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.6.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.7.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.8.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.gru_rel_pos_const": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.gru_rel_pos_linear.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.gru_rel_pos_linear.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.k_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.k_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.out_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.out_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.q_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.q_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.v_proj.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.attention.v_proj.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.feed_forward.intermediate_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.feed_forward.intermediate_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.feed_forward.output_dense.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.feed_forward.output_dense.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.final_layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.final_layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.layers.9.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.pos_conv_embed.conv.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.pos_conv_embed.conv.weight_g": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.encoder.pos_conv_embed.conv.weight_v": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.0.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.0.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.1.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.1.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.2.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.2.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.2.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.3.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.3.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.3.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.4.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.4.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.4.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.5.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.5.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.5.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.6.conv.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.6.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_extractor.conv_layers.6.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_projection.layer_norm.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_projection.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_projection.projection.bias": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.feature_projection.projection.weight": "model-00001-of-00002.safetensors",
+    "audio_encoder.encoder.masked_spec_embed": "model-00001-of-00002.safetensors",
+    "connector.layer.1.bias": "model-00001-of-00002.safetensors",
+    "connector.layer.1.weight": "model-00001-of-00002.safetensors",
+    "connector.layer.3.bias": "model-00001-of-00002.safetensors",
+    "connector.layer.3.weight": "model-00001-of-00002.safetensors",
+    "connector.layer.5.bias": "model-00001-of-00002.safetensors",
+    "connector.layer.5.weight": "model-00001-of-00002.safetensors",
+    "llm_model.lm_head.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.18.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.18.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.18.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.18.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.18.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.18.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.19.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.19.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.20.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "llm_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "llm_model.model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}