Upload 9 files

config.json

@@ -0,0 +1,26 @@
+{
+  "architectures": [
+    "PathummaAudioModel"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_pathumma_audio.PathummaAudioConfig",
+    "AutoModel": "modeling_pathumma_audio.PathummaAudioModel"
+  },
+  "beats_path": "",
+  "init_from_scratch": true,
+  "llm_path": "Qwen/Qwen2-7B-Instruct",
+  "lora": true,
+  "lora_alpha": 32,
+  "lora_dropout": 0.1,
+  "lora_infer_mode": true,
+  "lora_rank": 8,
+  "model_type": "pathumma_audio",
+  "qformer_hidden_layers": 2,
+  "qformer_query_token": 1,
+  "second_per_window": 0.333333,
+  "second_stride": 0.333333,
+  "target_modules": null,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.44.2",
+  "whisper_path": "openai/whisper-large-v3"
+}

configuration_pathumma_audio.py

@@ -0,0 +1,55 @@
+import torch
+from typing import Dict, Tuple, List
+from transformers import PretrainedConfig
+
+class PathummaAudioConfig(PretrainedConfig):
+
+    model_type: str = "pathumma_audio"
+
+    def __init__(
+        self,
+        llm_path: str = "Qwen/Qwen2-7B-Instruct",
+        whisper_path: str = "openai/whisper-large-v3",
+        beats_path: str = "",
+        init_from_scratch: bool = True,
+        
+        lora: bool = True,
+        lora_infer_mode: bool = True,
+        lora_rank: int = 8,
+        lora_alpha: int = 32,
+        lora_dropout: float = 0.1,
+        target_modules: List[str] = ["q_proj", "v_proj"],
+        qformer_query_token: int = 1,
+        qformer_hidden_layers: int = 2,
+        second_per_window: float = 0.333333,
+        second_stride: float = 0.333333,
+        
+        torch_dtype: torch.dtype = torch.bfloat16,
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+        
+        self.architectures = kwargs.get("architectures", ["PathummaAudioModel"])
+        self.auto_map = kwargs.get("auto_map", {
+            "AutoConfig": "configuration_pathumma_audio.PathummaAudioConfig",
+            "AutoModel": "modeling_pathumma_audio.PathummaAudioModel"
+        })
+        
+        self.llm_path = llm_path
+        self.whisper_path = whisper_path
+        self.beats_path = beats_path
+        self.init_from_scratch = init_from_scratch
+
+        self.lora = lora
+        self.lora_infer_mode = lora_infer_mode
+        self.lora_rank = lora_rank
+        self.lora_alpha = lora_alpha
+        self.lora_dropout = lora_dropout
+        self.target_modules = target_modules
+
+        self.qformer_query_token = qformer_query_token
+        self.qformer_hidden_layers = qformer_hidden_layers
+        self.second_per_window = second_per_window
+        self.second_stride = second_stride        
+
+        self.torch_dtype = torch_dtype

generation_config.json

@@ -0,0 +1,4 @@
+{
+  "_from_model_config": true,
+  "transformers_version": "4.44.2"
+}

pytorch_model-00001-of-00004.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79692df1f14cd91e53518bf688fd8b4e54d72e4f983a96eea30463ad677d9982
+size 4880949349

pytorch_model-00002-of-00004.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b159089ab56607238ff6d8cfc93efc996ad6d4c338c4a4930bf072c1459a5f90
+size 4936404688

pytorch_model-00003-of-00004.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dc497d77f78e19581119f4c15cb88b3f093cc0d0aba378731d84e4a8828bcee
+size 4334153926

pytorch_model-00004-of-00004.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8170ad28032cb7cfee768f7d22f8a6de4ef323a6af296246478a90b7dd7da3f0
+size 2596849960