add model

README.md

@@ -0,0 +1,134 @@
+---
+language:
+- en
+thumbnail: null
+tags:
+- response-generation
+- llama2
+- pytorch
+- speechbrain
+license: apache-2.0
+datasets:
+- multiwoz
+metrics:
+- name: Test PPL
+  type: ppl
+  value: ' 2.90'
+---
+
+<iframe src="https://ghbtns.com/github-btn.html?user=speechbrain&repo=speechbrain&type=star&count=true&size=large&v=2" frameborder="0" scrolling="0" width="170" height="30" title="GitHub"></iframe>
+<br/><br/>
+
+# Llama2 trained on MultiWOZ.2.1
+### Notice: “Llama 2 is licensed under the LLAMA 2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.” ###
+
+This repository provides all the necessary tools to perform response generation from an end-to-end system within
+SpeechBrain. For a better experience, we encourage you to learn more about
+[SpeechBrain](https://speechbrain.github.io). 
+The performance of the model is the following:
+
+| Release | Test PPL | Test BLEU 4 | GPUs |
+|:-------------:|:--------------:|:--------------:| :--------:|
+| 2023-10-15 |  2.90 | 7.45e-04 | 1xV100 32GB |
+
+## Credits
+The model is provided by [vitas.ai](https://www.vitas.ai/).
+
+## Pipeline description
+This dialogue system is composed of 2 different but linked blocks:
+
+- Pretrained Llama2 Tokenizer that transforms words into subwords.
+- Llama2 to generate the next sentence given the history of the dialogue.
+
+The system is trained with dialogue from the MultiWOZ corpus.
+
+
+## Install SpeechBrain
+First of all, please install SpeechBrain with the following command:
+
+```
+git clone https://github.com/speechbrain/speechbrain
+cd speechbrain
+pip install -r requirements.txt
+pip install -e .
+pip install -r recipes/MultiWOZ/response_generation/llama2/extra_requirements.txt
+```
+
+Please notice that we encourage you to read our tutorials and learn more about
+[SpeechBrain](https://speechbrain.github.io).
+
+### Generating your Own Dialogue
+
+```python
+from speechbrain.inference.text import Llama2ResponseGenerator
+res_gen_model = Llama2ResponseGenerator.from_hparams(source="speechbrain/MultiWOZ-Llama2-Response_Generation", savedir="pretrained_models/MultiWOZ-Llama2-Response_Generation", pymodule_file="custom.py")
+print("Hi,How could I help you today?", end="\n")
+while True:
+  turn = input()
+  response = res_gen_model.generate_response(turn)
+  print(response, end="\n")
+```
+
+### Inference on GPU
+
+To perform inference on the GPU, add  `run_opts={"device":"cuda"}`  when calling the `from_hparams` method.
+
+## Parallel Inference on a Batch
+
+Please, [see this Colab notebook](https://colab.research.google.com/drive/1hX5ZI9S4jHIjahFCZnhwwQmFoGAi3tmu?usp=sharing) to figure out how to transcribe in parallel a batch of input sentences using a pre-trained model.
+
+### Training
+
+The model was trained with SpeechBrain (986a2175).
+To train it from scratch follow these steps:
+
+1. Clone SpeechBrain:
+
+```bash
+git clone https://github.com/speechbrain/speechbrain/
+```
+
+2. Install it:
+
+```
+cd speechbrain
+pip install -r requirements.txt
+pip install -e .
+
+```
+
+3. Run Training:
+
+```
+cd recipes/MultiWOZ/response_generation/llama2
+pip install -r extra_requirements.txt
+python train_with_llama2.py hparams/train_llama2.yaml --data_folder=your_data_folder
+```
+
+You can find our training results (models, logs, etc) [here](ttps://www.dropbox.com/sh/d093vsje1d7ijj9/AAA-nHEd_MwNEFJfBGLmXxJra?dl=0)
+
+### Limitations
+
+The SpeechBrain team does not provide any warranty on the performance achieved by this model when used on other datasets.
+
+# **About SpeechBrain**
+
+- Website: https://speechbrain.github.io/
+- Code: https://github.com/speechbrain/speechbrain/
+- HuggingFace: https://huggingface.co/speechbrain/
+
+# **Citing SpeechBrain**
+
+Please, cite SpeechBrain if you use it for your research or business.
+
+```bibtex
+@misc{speechbrain,
+  title={{SpeechBrain}: A General-Purpose Speech Toolkit},
+  author={Mirco Ravanelli and Titouan Parcollet and Peter Plantinga and Aku Rouhe and Samuele Cornell and Loren Lugosch and Cem Subakan and Nauman Dawalatabad and Abdelwahab Heba and Jianyuan Zhong and Ju-Chieh Chou and Sung-Lin Yeh and Szu-Wei Fu and Chien-Feng Liao and Elena Rastorgueva and François Grondin and William Aris and Hwidong Na and Yan Gao and Renato De Mori and Yoshua Bengio},
+  year={2021},
+  eprint={2106.04624},
+  archivePrefix={arXiv},
+  primaryClass={eess.AS},
+  note={arXiv:2106.04624}
+}
+```

config.json

@@ -0,0 +1,23 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 2048,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "pad_token_id": 0,
+  "rms_norm_eps": 1e-06,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.29.0.dev0",
+  "use_cache": true,
+  "vocab_size": 32000
+}
+

custom.py

@@ -0,0 +1,70 @@
+"""This lobe enables the integration of huggingface pretrained Llama2 Model model plus the expanding embedding layer for additional PAD tokens .
+
+Transformer from HuggingFace needs to be installed:
+https://huggingface.co/transformers/installation.html
+
+Authors
+ * Pooneh Mousavi 2023
+"""
+
+import logging
+from torch import Tensor
+import torch
+import torch.nn as nn
+from speechbrain.lobes.models.huggingface_transformers.llama2 import LLAMA2
+
+
+logger = logging.getLogger(__name__)
+
+
+class LLAMA2_expanded(LLAMA2):
+    """This lobe enables the integration of HuggingFace pretrained LLAMA2 model.
+     Source paper LLAMA2:
+       https://arxiv.org/abs/2307.09288
+    Transformer from HuggingFace needs to be installed:
+        https://huggingface.co/transformers/installation.html
+
+    The model can be finetuned. It will download automatically the model from
+    HuggingFace or use a local path.
+
+    Arguments
+    ---------
+    source : str
+        HuggingFace hub name: e.g "meta-llama/Llama-2-7b-chat-hf"
+    save_path : str
+        Path (dir) of the downloaded model.
+    freeze : bool (default: False)
+        If True, the model is frozen. If False, the model will be trained
+        alongside with the rest of the pipeline.
+    Example
+    -------
+    >>> model_hub = "meta-llama/Llama-2-7b-chat-hf"
+    >>> save_path = "savedir"
+    >>> model = LLAMA2(model_hub, save_path)
+    >>> tokens = torch.tensor([[1, 1]])
+    >>> attention_mask = torch.tensor([[1, 1]])
+    >>> outputs = model(tokens, attention_mask)
+    """
+    def __init__(
+        self, *args, **kwrds
+    ) -> None:
+        super().__init__( *args, **kwrds)
+    # Load tokenizer and add special tokens
+        #  # Add special tokens to the tokenizer and resize model embedding
+        # Special tokens
+     
+        self.add_special_tokens_(
+                   {"pad_token": "<pad>"}
+                )
+
+    def add_special_tokens_(self, attr_to_special_token,) -> None:
+        orig_num_tokens = len(self.tokenizer)
+        num_added_tokens = self.tokenizer.add_special_tokens(
+            attr_to_special_token  # type: ignore
+        )  # doesn't add if they are already there
+        if num_added_tokens > 0:
+            self.model.resize_token_embeddings(
+                new_num_tokens=orig_num_tokens + num_added_tokens
+            )
+
+   

hyperparams.yaml

@@ -0,0 +1,44 @@
+# ################################
+# Model: Llama2 dModel +  NLL
+# Authors:
+    # Pooneh Mousavi 2023
+# ################################
+
+
+# URL for the gpt2 model
+model_hub: "/local_disk/clytie/hnguyen/speechbrain_recipes/HF_PR/MultiWOZ-Llama2-Response_Generation" #meta-llama/Llama-2-7b-chat-hf
+llama2_folder: recipes/MultiWOZ/response_generation/llama2/results/train_with_llama2/1995/save/llama2_checkpoint/
+
+
+# history_window, i.e. how many user-system exchanges consider as context.
+max_history: 2
+
+# decoder setting
+freeze_model: True
+num_beams: 8
+max_new_tokens: 50
+top_k: 45
+top_p: 0.9
+
+#LLAMA2 model
+model: !new:custom.LLAMA2_expanded
+    source: !ref <model_hub>
+    freeze: !ref <freeze_model>
+    save_path: !ref <llama2_folder>
+    max_new_tokens: !ref <max_new_tokens>
+    num_beams: !ref <num_beams>
+    top_k: !ref  <top_k>
+    top_p: !ref <top_p>
+    with_peft: True
+
+
+# Masks
+padding_mask: !name:speechbrain.lobes.models.transformer.Transformer.get_key_padding_mask
+
+pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
+    loadables:
+        model: !ref <model>
+
+modules:
+    model: !ref <model>
+

llama2_model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78575f6cd2df7db958e2bd696fcd535322664e222a499f740560de8fdf6411a9
+size 4827154534

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

tokenizer_config.json

@@ -0,0 +1,33 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "bos_token": {
+    "__type": "AddedToken",
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "clean_up_tokenization_spaces": false,
+  "eos_token": {
+    "__type": "AddedToken",
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": null,
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": {
+    "__type": "AddedToken",
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}