Update README.md

README.md

@@ -3,199 +3,231 @@ library_name: transformers
 tags: []
 ---
 
-# Model Card for Model ID
-
-<!-- Provide a quick summary of what the model is/does. -->
-
-
-
-## Model Details
-
-### Model Description
-
-<!-- Provide a longer summary of what this model is. -->
-
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
-
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-
-### Model Sources [optional]
-
-<!-- Provide the basic links for the model. -->
-
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-
-## Uses
-
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-
-### Direct Use
-
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-
-[More Information Needed]
-
-### Downstream Use [optional]
-
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-
-[More Information Needed]
-
-### Out-of-Scope Use
-
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-
-[More Information Needed]
-
-## Bias, Risks, and Limitations
-
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-
-[More Information Needed]
-
-### Recommendations
-
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-[More Information Needed]
-
-## Training Details
-
-### Training Data
-
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
-
-### Training Procedure 
-
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-
-#### Preprocessing [optional]
-
-[More Information Needed]
-
-
-#### Training Hyperparameters
-
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-
-#### Speeds, Sizes, Times [optional]
-
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-
-[More Information Needed]
-
-## Evaluation
-
-<!-- This section describes the evaluation protocols and provides the results. -->
-
-### Testing Data, Factors & Metrics
-
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
-
-#### Factors
-
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-
-[More Information Needed]
-
-#### Metrics
-
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-
-[More Information Needed]
-
-### Results
-
-[More Information Needed]
-
-#### Summary
-
-
-
-## Model Examination [optional]
-
-<!-- Relevant interpretability work for the model goes here -->
-
-[More Information Needed]
-
-## Environmental Impact
-
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-
-## Technical Specifications [optional]
-
-### Model Architecture and Objective
-
-[More Information Needed]
-
-### Compute Infrastructure
-
-[More Information Needed]
-
-#### Hardware
-
-[More Information Needed]
-
-#### Software
-
-[More Information Needed]
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-
-**BibTeX:**
-
-[More Information Needed]
-
-**APA:**
-
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-
-[More Information Needed]
-
-## More Information [optional]
-
-[More Information Needed]
-
-## Model Card Authors [optional]
-
-[More Information Needed]
-
-## Model Card Contact
-
-[More Information Needed]
-
-
+# Multi-images Multi-audio Multi-turn Malaysian TinyLlama
+
+WanDB https://wandb.ai/huseinzol05/vision-tinyllama?workspace=user-huseinzol05
+
+## how-to
+
+```python
+from modeling_combine import MM_LLMs, MM_LLMs_Config
+from transformers import AutoTokenizer, AutoProcessor
+from PIL import Image
+import requests
+
+def prepare_dataset(messages, images: List[str] = None, audio: List[str] = None, sr = 16000):
+    
+    if images is not None:
+        images = [Image.open(f).convert('RGB') for f in images]
+        image_output = image_processor(images=images, return_tensors='pt')['pixel_values']
+    else:
+        image_output = None
+        
+    if audio is not None:
+        audio = [librosa.load(f, sr=sr)[0] for f in audio]
+        audio_features = audio_processor(audio, sampling_rate=sr, return_tensors='pt',)['input_features']
+    else:
+        audio_features = None
+    
+    prompt = tokenizer.apply_chat_template(messages, tokenize = False)
+    outputs = tokenizer(
+                    prompt,
+                    return_tensors='pt',
+                    return_overflowing_tokens=False,
+                    return_length=False
+    )
+
+    outputs['images'] = image_output
+    outputs['audios'] = audio_features
+    
+    image_token = tokenizer.convert_tokens_to_ids('<image>')
+    audio_token = tokenizer.convert_tokens_to_ids('<audio>')
+    
+    if image_output is not None:
+        len_image = len(image_output)
+    else:
+        len_image = 0
+        
+    if audio_features is not None:
+        len_audio = len(audio_features)
+    else:
+        len_audio = 0
+        
+    outputs['image_index'] = torch.tensor([0] * len_image)
+    outputs['image_starts'] = torch.tensor([image_token] * (len_image + 1))
+    outputs['audio_index'] = torch.tensor([0] * len_audio)
+    outputs['audio_starts'] = torch.tensor([audio_token] * (len_audio + 1))
+        
+    where_is = torch.where((outputs['input_ids'] == image_token) | (outputs['input_ids'] == audio_token))
+    ls = []
+    for i in range(len(where_is[0])):
+        b, k = where_is[0][i], where_is[1][i]
+        l = int(outputs['input_ids'][b, k])
+        ls.append(l)
+
+    ls = torch.tensor(ls)
+    outputs['where_is_b'] = where_is[0]
+    outputs['where_is_k'] = where_is[1]
+    outputs['ls'] = ls
+        
+    return outputs
+
+model = MM_LLMs.from_pretrained(
+    'mesolitica/malaysian-tinyllama-1.1b-mmmmodal',
+    flash_attention = True,
+    dtype = torch.bfloat16,
+    torch_dtype = torch.bfloat16
+)
+_ = model.cuda()
+
+image_processor = AutoProcessor.from_pretrained('google/siglip-base-patch16-384')
+audio_processor = AutoProcessor.from_pretrained('mesolitica/malaysian-whisper-small')
+tokenizer = AutoTokenizer.from_pretrained('mesolitica/malaysian-tinyllama-1.1b-mmmmodal')
+
+with open('Persian-cat-breed.jpg', 'wb') as fopen:
+    fopen.write(requests.get('https://cdn.beautifulnara.net/wp-content/uploads/2017/12/10201620/Persian-cat-breed.jpg').content)
+
+with open('nasi-goreng-1-23.jpg', 'wb') as fopen:
+    fopen.write(requests.get('https://www.jocooks.com/wp-content/uploads/2023/09/nasi-goreng-1-23.jpg').content)
+
+with open('test.mp3', 'wb') as fopen:
+    fopen.write(requests.get('https://github.com/mesolitica/multimodal-LLM/raw/master/data/test.mp3').content)
+
+messages = [
+    {'role': 'user', 'content': '<image> </image> ini gambar apa'},
+]
+outputs = prepare_dataset(messages, images = ['Persian-cat-breed.jpg'])
+if outputs['images'] is not None:
+    outputs['images'] = outputs['images'].type(model.dtype)
+if outputs['audios'] is not None:
+    outputs['audios'] = outputs['audios'].type(model.dtype)
+for k in outputs.keys():
+    if outputs[k] is not None:
+        outputs[k] = outputs[k].cuda()
+
+with torch.no_grad():
+    model_inputs = model.prepare_inputs_for_generation(**outputs)
+r = model_inputs.pop('input_ids', None)
+
+generate_kwargs = dict(
+    model_inputs,
+    max_new_tokens=300,
+    top_p=0.95,
+    top_k=50,
+    temperature=0.1,
+    do_sample=True,
+    num_beams=1,
+)
+
+r = model.llm.generate(**generate_kwargs)
+print(tokenizer.decode(r[0]))
+```
+
+```
+<s>Imej itu menunjukkan seekor kucing putih yang comel duduk di atas sofa hitam.</s>
+```
+
+```python
+messages = [
+    {'role': 'user', 'content': '<image> </image> <image> </image> apa kaitan 2 gambar ni'},
+]
+outputs = prepare_dataset(messages, images = ['Persian-cat-breed.jpg', 'nasi-goreng-1-23.jpg'])
+if outputs['images'] is not None:
+    outputs['images'] = outputs['images'].type(model.dtype)
+if outputs['audios'] is not None:
+    outputs['audios'] = outputs['audios'].type(model.dtype)
+for k in outputs.keys():
+    if outputs[k] is not None:
+        outputs[k] = outputs[k].cuda()
+
+with torch.no_grad():
+    model_inputs = model.prepare_inputs_for_generation(**outputs)
+r = model_inputs.pop('input_ids', None)
+
+generate_kwargs = dict(
+    model_inputs,
+    max_new_tokens=300,
+    top_p=0.95,
+    top_k=50,
+    temperature=0.1,
+    do_sample=True,
+    num_beams=1,
+)
+
+r = model.llm.generate(**generate_kwargs)
+print(tokenizer.decode(r[0]))
+```
+
+```
+<s>Tiada hubungan yang jelas antara gambar 1 (anak kucing putih duduk di atas sofa) dan gambar 2 (foto penutup mangkuk mi telur dengan nasi dan cili). Gambar pertama ialah imej haiwan, manakala gambar kedua ialah imej makanan. Mereka tergolong dalam kategori yang berbeza dan tidak mempunyai hubungan antara satu sama lain.</s>
+```
+
+```python
+messages = [
+    {'role': 'user', 'content': '<audio> </audio> apa isu audio ni'},
+]
+outputs = prepare_dataset(messages, audio = [audio])
+if outputs['images'] is not None:
+    outputs['images'] = outputs['images'].type(model.dtype)
+if outputs['audios'] is not None:
+    outputs['audios'] = outputs['audios'].type(model.dtype)
+for k in outputs.keys():
+    if outputs[k] is not None:
+        outputs[k] = outputs[k].cuda()
+
+with torch.no_grad():
+    model_inputs = model.prepare_inputs_for_generation(**outputs, inference = True)
+    
+r = model_inputs.pop('input_ids', None)
+generate_kwargs = dict(
+    model_inputs,
+    max_new_tokens=300,
+    top_p=0.95,
+    top_k=50,
+    temperature=0.9,
+    do_sample=True,
+    num_beams=1,
+)
+
+r = model.llm.generate(**generate_kwargs)
+print(tokenizer.decode(r[0]))
+```
+
+```
+<s>Isu audio ini berkisar tentang persepsi salah faham dan sikap bakhil berkenaan wang dalam konteks menggalakkan penggunaan e-dompet. Penceramah mencadangkan bahawa orang mungkin keberatan untuk menerima wang kerana tidak melihat manfaat atau nilai menggunakan e-dompet, dan kebimbangan tentang tidak dapat mengakses wang itu jika mereka memerlukannya segera. Penceramah juga menyebut isu ekonomi sistem dan kekurangan sistem yang berkesan di Malaysia. Secara keseluruhannya, isu ini menekankan keperluan untuk pemahaman dan kesedaran yang lebih baik tentang faedah menggunakan e-dompet, serta keperluan untuk pembaharuan sistemik untuk memastikan akses yang saksama kepada wang dan sumber lain.</s>
+```
+
+```python
+messages = [
+    {'role': 'user', 'content': '<image> </image> <audio> </audio> apa kaitan gambar dan audio ni'},
+]
+outputs = prepare_dataset(messages, images = [test_image], audio = [audio])
+if outputs['images'] is not None:
+    outputs['images'] = outputs['images'].type(model.dtype)
+if outputs['audios'] is not None:
+    outputs['audios'] = outputs['audios'].type(model.dtype)
+for k in outputs.keys():
+    if outputs[k] is not None:
+        outputs[k] = outputs[k].cuda()
+
+with torch.no_grad():
+    model_inputs = model.prepare_inputs_for_generation(**outputs, inference = True)
+    
+r = model_inputs.pop('input_ids', None)
+generate_kwargs = dict(
+    model_inputs,
+    max_new_tokens=300,
+    top_p=0.95,
+    top_k=50,
+    temperature=0.9,
+    do_sample=True,
+    num_beams=1,
+)
+
+r = model.llm.generate(**generate_kwargs)
+print(tokenizer.decode(r[0]))
+```
+
+```
+<s>Tidak jelas bagaimana gambar dan audio berkaitan antara satu sama lain. Gambar itu menunjukkan bas pelancongan dengan iklan yang menggalakkan orang ramai menggunakan e-dompet mereka, tetapi ia tidak menyatakan tujuan iklan itu. Audio itu membincangkan idea pembaziran dana sebanyak RM5 juta (kira-kira 1.2 juta USD) ke atas sesuatu projek, tetapi ia tidak menyebut secara langsung bas pelancongan atau e-dompet. Ada kemungkinan bahawa kedua-dua gambar dan audio sedang membincangkan topik yang sama, tetapi lebih banyak konteks diperlukan untuk membuat perkaitan yang pasti.</s>
+```