Update app.py
Browse files
app.py
CHANGED
@@ -7,10 +7,10 @@
|
|
7 |
| - Import Python | | - Define interface | | - Transcribe audio | | - XTTS model generates |
|
8 |
| libraries | | components | | to text using | | spoken response from |
|
9 |
| - Initialize models: |--------> - Configure audio and |------->| Faster Whisper ASR |------->| LLM's text response |
|
10 |
-
| Whisper,
|
11 |
| XTTS | | - Launch interface | | is added to | | |
|
12 |
| | | | | chatbot's history | | |
|
13 |
-
| | | | | -
|
14 |
| | | | | processes chatbot | | |
|
15 |
| | | | | history to generate | | |
|
16 |
| | | | | response | | |
|
@@ -48,7 +48,7 @@ whisper_model = WhisperModel("large-v3", device="cuda", compute_type="float16")
|
|
48 |
print("Loading Saul-Instruct-v1-GGUF.Q4_K_M")
|
49 |
hf_hub_download(repo_id="MaziyarPanahi/Saul-Instruct-v1-GGUF", local_dir=".", filename="Saul-Instruct-v1.Q4_K_M.gguf")
|
50 |
saul_model_path="./Saul-Instruct-v1.Q4_K_M.gguf"
|
51 |
-
saul_instruct_llm = Llama(model_path=saul_model_path,n_gpu_layers=35,max_new_tokens=256, context_window=
|
52 |
|
53 |
# Load XTTS Model
|
54 |
print("Loading XTTS model")
|
@@ -170,7 +170,6 @@ with gr.Blocks(title="Voice chat with Saul-Instruct-v1-GGUF") as demo:
|
|
170 |
It relies on the following models :
|
171 |
- Speech to Text Model: [Faster-Whisper-large-v3](https://huggingface.co/Systran/faster-whisper-large-v3) an ASR model, to transcribe recorded audio to text.
|
172 |
- Legal Large Language Model: [MaziyarPanahi/Saul-Instruct-v1-GGUF](https://huggingface.co/MaziyarPanahi/Saul-Instruct-v1-GGUF/blob/main/Saul-Instruct-v1.Q4_K_M.gguf) a LLM to generate legal chatbot responses.
|
173 |
-
- Large Language Model: [Mistral-7b-instruct-v0.1-quantized](https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF) a LLM to generate the chatbot responses.
|
174 |
- Text to Speech Model: [XTTS-v2](https://huggingface.co/spaces/coqui/xtts) a TTS model, to generate the voice of the chatbot.
|
175 |
|
176 |
Note:
|
|
|
7 |
| - Import Python | | - Define interface | | - Transcribe audio | | - XTTS model generates |
|
8 |
| libraries | | components | | to text using | | spoken response from |
|
9 |
| - Initialize models: |--------> - Configure audio and |------->| Faster Whisper ASR |------->| LLM's text response |
|
10 |
+
| Whisper, SaulLM, | | text interaction | | - Transcribed text | | |
|
11 |
| XTTS | | - Launch interface | | is added to | | |
|
12 |
| | | | | chatbot's history | | |
|
13 |
+
| | | | | - SaulLM LLM | | |
|
14 |
| | | | | processes chatbot | | |
|
15 |
| | | | | history to generate | | |
|
16 |
| | | | | response | | |
|
|
|
48 |
print("Loading Saul-Instruct-v1-GGUF.Q4_K_M")
|
49 |
hf_hub_download(repo_id="MaziyarPanahi/Saul-Instruct-v1-GGUF", local_dir=".", filename="Saul-Instruct-v1.Q4_K_M.gguf")
|
50 |
saul_model_path="./Saul-Instruct-v1.Q4_K_M.gguf"
|
51 |
+
saul_instruct_llm = Llama(model_path=saul_model_path,n_gpu_layers=35,max_new_tokens=256, context_window=16384, n_ctx=16384, n_batch=128,verbose=False)
|
52 |
|
53 |
# Load XTTS Model
|
54 |
print("Loading XTTS model")
|
|
|
170 |
It relies on the following models :
|
171 |
- Speech to Text Model: [Faster-Whisper-large-v3](https://huggingface.co/Systran/faster-whisper-large-v3) an ASR model, to transcribe recorded audio to text.
|
172 |
- Legal Large Language Model: [MaziyarPanahi/Saul-Instruct-v1-GGUF](https://huggingface.co/MaziyarPanahi/Saul-Instruct-v1-GGUF/blob/main/Saul-Instruct-v1.Q4_K_M.gguf) a LLM to generate legal chatbot responses.
|
|
|
173 |
- Text to Speech Model: [XTTS-v2](https://huggingface.co/spaces/coqui/xtts) a TTS model, to generate the voice of the chatbot.
|
174 |
|
175 |
Note:
|