Upload folder using huggingface_hub

inference.py

@@ -0,0 +1,100 @@
+import numpy as np
+import torch
+import librosa
+from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
+import soundfile as sf
+import os
+
+
+class InferenceRecipe:
+    def __init__(self, cache_dir='./models', device='cuda'):
+        self.device = device
+        self.asr_pipeline = None
+        self.chat_tokenizer = None
+        self.chat_model = None
+        self.tts_model = None
+        self.tts_sample_rate = 22050  # TTS output sample rate
+        self.cache_dir = cache_dir
+        self.initialize_models()
+
+    def initialize_models(self):
+        """Initialize models from local cache"""
+        # ASR: OpenAI Whisper
+        self.asr_pipeline = pipeline(
+            "automatic-speech-recognition",
+            model=os.path.join(self.cache_dir, "asr"),
+            device=self.device,
+            torch_dtype=torch.float32
+        )
+
+        # Chat: DialoGPT
+        dialogpt_path = os.path.join(self.cache_dir, "llm")
+        self.chat_tokenizer = AutoTokenizer.from_pretrained(dialogpt_path)
+        self.chat_model = AutoModelForCausalLM.from_pretrained(dialogpt_path)
+        self.chat_model = self.chat_model.to(self.device)
+
+
+        # TTS: Facebook MMS
+        self.tts_model = pipeline(
+            "text-to-speech",
+            model=os.path.join(self.cache_dir, "tts"),
+            device=self.device,
+            torch_dtype=torch.float32
+        )
+
+    def inference(self, audio_array, sample_rate):
+        """Updated inference pipeline"""
+        # Speech-to-Text using Whisper
+        text = self.asr_pipeline({"raw": audio_array, "sampling_rate": sample_rate})["text"]
+
+        # Generate response with proper attention mask
+        input_ids = self.chat_tokenizer.encode(text + self.chat_tokenizer.eos_token, return_tensors="pt")
+        attention_mask = torch.ones_like(input_ids)
+        chat_output = self.chat_model.generate(
+            input_ids.to(self.device),
+            attention_mask=attention_mask.to(self.device),
+            max_length=1000,
+            pad_token_id=self.chat_tokenizer.eos_token_id
+        )
+        reply = self.chat_tokenizer.decode(chat_output[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
+
+        # Text-to-Speech using HF Pipeline
+        tts_output = self.tts_model(reply)
+        audio_array = tts_output['audio']
+        
+        # Ensure audio is in correct format
+        audio_array = audio_array.astype(np.float32)
+        audio_array = np.clip(audio_array, -1.0, 1.0)
+        
+        # Resample to match input rate
+        if sample_rate != self.tts_sample_rate:
+            audio_array = librosa.resample(
+                audio_array,
+                orig_sr=self.tts_sample_rate,
+                target_sr=sample_rate
+            )
+
+        # Ensure the audio is 1D
+        if len(audio_array.shape) > 1:
+            audio_array = audio_array.squeeze()
+
+        return audio_array
+
+if __name__ == "__main__":
+    recipe = InferenceRecipe(cache_dir="./models")  # Specify your cache directory here
+    # Test with realistic input (silent audio)
+    sr = 16000
+    duration = 3
+    audio = np.zeros(int(sr * duration))  # Silent input
+    response_audio = recipe.inference(audio, sr)
+    
+    print(f"Audio shape: {response_audio.shape}, Range: [{response_audio.min()}, {response_audio.max()}]")
+    
+    # Save with explicit format
+    sf.write(
+        "response.wav", 
+        response_audio, 
+        sr, 
+        format='WAV',
+        subtype='FLOAT'
+    )

models/asr/config.json

@@ -0,0 +1,138 @@
+{
+  "_name_or_path": "openai/whisper-small.en",
+  "activation_dropout": 0.0,
+  "activation_function": "gelu",
+  "architectures": [
+    "WhisperForConditionalGeneration"
+  ],
+  "attention_dropout": 0.0,
+  "begin_suppress_tokens": [
+    220,
+    50256
+  ],
+  "bos_token_id": 50257,
+  "d_model": 768,
+  "decoder_attention_heads": 12,
+  "decoder_ffn_dim": 3072,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 12,
+  "decoder_start_token_id": 50257,
+  "dropout": 0.0,
+  "encoder_attention_heads": 12,
+  "encoder_ffn_dim": 3072,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 12,
+  "eos_token_id": 50256,
+  "forced_decoder_ids": [
+    [
+      1,
+      50362
+    ]
+  ],
+  "init_std": 0.02,
+  "is_encoder_decoder": true,
+  "max_length": 448,
+  "max_source_positions": 1500,
+  "max_target_positions": 448,
+  "model_type": "whisper",
+  "num_hidden_layers": 12,
+  "num_mel_bins": 80,
+  "pad_token_id": 50256,
+  "scale_embedding": false,
+  "suppress_tokens": [
+    1,
+    2,
+    7,
+    8,
+    9,
+    10,
+    14,
+    25,
+    26,
+    27,
+    28,
+    29,
+    31,
+    58,
+    59,
+    60,
+    61,
+    62,
+    63,
+    90,
+    91,
+    92,
+    93,
+    357,
+    366,
+    438,
+    532,
+    685,
+    705,
+    796,
+    930,
+    1058,
+    1220,
+    1267,
+    1279,
+    1303,
+    1343,
+    1377,
+    1391,
+    1635,
+    1782,
+    1875,
+    2162,
+    2361,
+    2488,
+    3467,
+    4008,
+    4211,
+    4600,
+    4808,
+    5299,
+    5855,
+    6329,
+    7203,
+    9609,
+    9959,
+    10563,
+    10786,
+    11420,
+    11709,
+    11907,
+    13163,
+    13697,
+    13700,
+    14808,
+    15306,
+    16410,
+    16791,
+    17992,
+    19203,
+    19510,
+    20724,
+    22305,
+    22935,
+    27007,
+    30109,
+    30420,
+    33409,
+    34949,
+    40283,
+    40493,
+    40549,
+    47282,
+    49146,
+    50257,
+    50357,
+    50358,
+    50359,
+    50360,
+    50361
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.27.0.dev0",
+  "use_cache": true,
+  "vocab_size": 51864
+}

models/asr/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6014ac49b506df900f66f4aca6b0801eed7245594ace97bcaf73e0ae5b863066
+size 966992008

models/llm/config.json

@@ -0,0 +1,30 @@
+{
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 768,
+  "n_head": 12,
+  "n_layer": 12,
+  "n_positions": 1024,
+  "resid_pdrop": 0.1,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "conversational": {
+      "max_length": 1000
+    }
+  },
+  "vocab_size": 50257
+}

models/llm/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea50e39b0e9368b9110cc5ab48012ddbc7bd90f2b17410aed643a7404acac567
+size 351256598

models/llm/tokenizer_config.json

@@ -0,0 +1,23 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "50256": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "chat_template": "{% for message in messages %}{{ message.content }}{{ eos_token }}{% endfor %}",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 1024,
+  "pad_token": null,
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

models/tts/config.json

@@ -0,0 +1,82 @@
+{
+  "activation_dropout": 0.1,
+  "architectures": [
+    "VitsModel"
+  ],
+  "attention_dropout": 0.1,
+  "depth_separable_channels": 2,
+  "depth_separable_num_layers": 3,
+  "duration_predictor_dropout": 0.5,
+  "duration_predictor_filter_channels": 256,
+  "duration_predictor_flow_bins": 10,
+  "duration_predictor_kernel_size": 3,
+  "duration_predictor_num_flows": 4,
+  "duration_predictor_tail_bound": 5.0,
+  "ffn_dim": 768,
+  "ffn_kernel_size": 3,
+  "flow_size": 192,
+  "hidden_act": "relu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 192,
+  "initializer_range": 0.02,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "leaky_relu_slope": 0.1,
+  "model_type": "vits",
+  "noise_scale": 0.667,
+  "noise_scale_duration": 0.8,
+  "num_attention_heads": 2,
+  "num_hidden_layers": 6,
+  "num_speakers": 1,
+  "posterior_encoder_num_wavenet_layers": 16,
+  "prior_encoder_num_flows": 4,
+  "prior_encoder_num_wavenet_layers": 4,
+  "resblock_dilation_sizes": [
+    [
+      1,
+      3,
+      5
+    ],
+    [
+      1,
+      3,
+      5
+    ],
+    [
+      1,
+      3,
+      5
+    ]
+  ],
+  "resblock_kernel_sizes": [
+    3,
+    7,
+    11
+  ],
+  "sampling_rate": 16000,
+  "speaker_embedding_size": 0,
+  "speaking_rate": 1.0,
+  "spectrogram_bins": 513,
+  "torch_dtype": "float32",
+  "transformers_version": "4.33.0.dev0",
+  "upsample_initial_channel": 512,
+  "upsample_kernel_sizes": [
+    16,
+    16,
+    4,
+    4
+  ],
+  "upsample_rates": [
+    8,
+    8,
+    2,
+    2
+  ],
+  "use_bias": true,
+  "use_stochastic_duration_prediction": true,
+  "vocab_size": 38,
+  "wavenet_dilation_rate": 1,
+  "wavenet_dropout": 0.0,
+  "wavenet_kernel_size": 5,
+  "window_size": 4
+}

models/tts/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69cf8b651c1493f1801dfd2311c298d694a38357bc9a1e41f410491ea6f0e1be
+size 145227512

models/tts/special_tokens_map.json

@@ -0,0 +1,4 @@
+{
+  "pad_token": "k",
+  "unk_token": "<unk>"
+}

models/tts/tokenizer_config.json

@@ -0,0 +1,12 @@
+{
+  "add_blank": true,
+  "clean_up_tokenization_spaces": true,
+  "is_uroman": false,
+  "language": "eng",
+  "model_max_length": 1000000000000000019884624838656,
+  "normalize": true,
+  "pad_token": "k",
+  "phonemize": false,
+  "tokenizer_class": "VitsTokenizer",
+  "unk_token": "<unk>"
+}

models/tts/vocab.json

@@ -0,0 +1,40 @@
+{
+  " ": 19,
+  "'": 1,
+  "-": 14,
+  "0": 23,
+  "1": 15,
+  "2": 28,
+  "3": 11,
+  "4": 27,
+  "5": 35,
+  "6": 36,
+  "_": 30,
+  "a": 26,
+  "b": 24,
+  "c": 12,
+  "d": 5,
+  "e": 7,
+  "f": 20,
+  "g": 37,
+  "h": 6,
+  "i": 18,
+  "j": 16,
+  "k": 0,
+  "l": 21,
+  "m": 17,
+  "n": 29,
+  "o": 22,
+  "p": 13,
+  "q": 34,
+  "r": 25,
+  "s": 8,
+  "t": 33,
+  "u": 4,
+  "v": 32,
+  "w": 9,
+  "x": 31,
+  "y": 3,
+  "z": 2,
+  "–": 10
+}

requirements.txt

@@ -0,0 +1,3 @@
+transformers==4.48.1
+torch==2.5.1
+librosa==0.10.2.post1