Phase2 Training added.

app.py

@@ -1,55 +1,93 @@
+import gradio as gr
 import torch
+import wisperx
+from model import MainQLoraModel
+from configs import get_config_phase2
 from transformers import AutoTokenizer, AutoProcessor
-from model import CustomClipPhi2
-import gradio as gr
 
-clip_model_name = "openai/clip-vit-base-patch32"
-phi2_model_name  = "microsoft/phi-2"
+# get config
+config = get_config_phase2() 
+# tokenizer
+tokenizer  = AutoTokenizer.from_pretrained(config.get("phi2_model_name"), trust_remote_code=True)
+processor  = AutoProcessor.from_pretrained(config.get("clip_model_name"), trust_remote_code=True)
+llmModel = MainQLoraModel(tokenizer, config).to(config.get("device"))
+audio_model = wisperx.load_model('tiny', 'cpu', compute_type="float16")
 
-tokenizer  = AutoTokenizer.from_pretrained(phi2_model_name, trust_remote_code=True)
-processor  = AutoProcessor.from_pretrained(clip_model_name)
-tokenizer.pad_token = tokenizer.eos_token
 
-IMAGE_TOKEN_ID = 23903 # token for word Comments
-device = "cuda" if torch.cuda.is_available() else "cpu"
-# max_tokens = 30
-model = CustomClipPhi2(tokenizer, phi2_model_name, clip_model_name, clip_embed=768, phi_embed=2560)
+def generate_answers(img=None, aud = None, q = None, max_tokens = 30):
+          
+    batch_size = 1
+    start_iq = tokenizer.encode("<iQ>")
+    end_iq = tokenizer.encode("</iQ>")
+    start_iq_embeds = torch.tensor(start_iq).repeat(batch_size, 1)
+    end_iq_embeds = torch.tensor(end_iq).repeat(batch_size, 1)
+    start_iq_embeds = llmModel.phi2_model.model.model.embed_tokens(start_iq_embeds.to(config.get("device")))
+    end_iq_embeds = llmModel.phi2_model.model.model.embed_tokens(end_iq_embeds.to(config.get("device")))
+    
+    inputs_embeddings = []
+    inputs_embeddings.append(start_iq_embeds)
 
-def generate(images, max_tokens):
-        image  = processor(images=images, return_tensors="pt").to(device)
-        clip_outputs = model.clip_model(**image)
+    predicted_caption = torch.full((batch_size, max_tokens), llmModel.EOS_TOKEN_ID, dtype=torch.long, device=config.get('device'))
+    
+    if images is not None:
+        images = processor(images=img, return_tensors="pt").to(config.get("device"))
+        images = {'pixel_values': images.to(config.get("device"))}
+        clip_outputs = llmModel.clip_model(**images)
         # remove cls token
         images = clip_outputs.last_hidden_state[:, 1:, :]
-        image_embeddings = model.projection_layer(images).to(torch.float16)
-
-        batch_size = images.size()[0]
-        predicted_caption = torch.full((batch_size, max_tokens), model.EOS_TOKEN_ID, dtype=torch.long, device=device)
-        img_token_tensor = torch.tensor(IMAGE_TOKEN_ID).repeat(batch_size, 1)
-        img_token_embeds = model.phi2_model.model.embed_tokens(img_token_tensor.to(image_embeddings.device))
-        combined_embeds  = torch.cat([image_embeddings, img_token_embeds], dim=1)
-
-        for pos in range(max_tokens - 1):
-            model_output_logits = model.phi2_model.forward(inputs_embeds = combined_embeds)['logits']
-            predicted_word_token_logits = model_output_logits[:, -1, :].unsqueeze(1)
-            predicted_word_token = torch.argmax(predicted_word_token_logits, dim = -1)
-            predicted_caption[:, pos] = predicted_word_token.view(1,-1).to('cpu')
-            next_token_embeds = model.phi2_model.model.embed_tokens(predicted_word_token)
-            combined_embeds   = torch.cat([combined_embeds, next_token_embeds], dim=1)
-        predicted_captions_decoded = tokenizer.batch_decode(predicted_caption,ignore_index = tokenizer.eos_token_id)
-        return predicted_captions_decoded
-
-# Create a Gradio interface
-iface = gr.Interface(fn=generate,  # Function to be called on user input
-    inputs=[gr.Image(
-            width=416, height=416, 
-            type="pil", image_mode='RGB', label="Upload Image"
-        ), 
-    gr.Slider(1, 50, value = 10, step=1, label="Max Length")],
-    outputs=gr.Textbox(
-            label="Response from AI Model: ", 
-        )
-    # examples = ['car.jpg']
-)
-
-# Launch the Gradio app
-iface.launch()
+        image_embeddings = llmModel.projection_layer(images).to(torch.float16)
+        inputs_embeddings.append(image_embeddings)
+    
+    if aud is not None:
+        trans = audio_model.transcribe(aud)
+        audio_res = ""
+        for seg in trans['segments']:
+            audio_res += seg['text']
+        audio_res = audio_res.strip()
+        audio_tokens = tokenizer(q,return_tensors="pt", return_attention_mask=False)['input_ids']
+        audio_embeds = llmModel.phi2_model.model.model.embed_tokens(audio_tokens.to(config.get("device")))
+        inputs_embeddings.append(audio_embeds)
+        
+    if q is not None:
+        ques = tokenizer(q, return_tensors="pt", return_attention_mask=False)['input_ids']
+        q_embeds = llmModel.phi2_model.model.model.embed_tokens(ques.to(config.get("device")))
+        inputs_embeddings.append(q_embeds)
+        
+    inputs_embeddings.append(end_iq_embeds)
+    # Combine embeddings
+    combined_embeds  = torch.cat(inputs_embeddings, dim=1)
+
+    for pos in range(max_tokens - 1):
+        model_output_logits = llmModel.phi2_model.forward(inputs_embeds = combined_embeds)['logits']
+        predicted_word_token_logits = model_output_logits[:, -1, :].unsqueeze(1)
+        predicted_word_token = torch.argmax(predicted_word_token_logits, dim = -1)
+        predicted_caption[:, pos] = predicted_word_token.view(1,-1).to('cpu')
+        next_token_embeds = llmModel.phi2_model.model.model.embed_tokens(predicted_word_token)
+        combined_embeds   = torch.cat([combined_embeds, next_token_embeds], dim=1)
+    predicted_captions_decoded = tokenizer.batch_decode(predicted_caption,ignore_index = 50256)[0]
+    return predicted_captions_decoded
+
+
+with gr.Blocks() as demo:
+
+    gr.Markdown(
+    """
+    # TAI2T Model(Text, Audio, Image to Text Model)
+    Multimodel GPT with inputs as Image, Audio, Text with output as Text.
+    """
+    )
+
+    with gr.Row():
+            image = gr.Image(label="Image", type="pil")
+            audio_q = gr.Audio(label="Audio Question", sources=['microphone', 'upload'], type='filepath')
+    with gr.Row():
+        question = gr.Text(label ='Question?')
+    with gr.Row():
+        max_tokens = gr.Slider(1, 50, value = 10, step=1, label="Maximum length of tokens in asnwer.")
+    submit = gr.Button("Submit")
+    with gr.Row():
+        answer   = gr.Text(label ='Answer')
+    submit.click(generate_answers, inputs=[image,audio_q,question, max_tokens], outputs=[answer])
+    
+if __name__ == "__main__":
+    demo.launch(share=True)

ckpts/Qlora_adaptor/README.md

@@ -0,0 +1,204 @@
+---
+library_name: peft
+base_model: microsoft/phi-2
+---
+
+# 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. -->
+
+
+
+- **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]
+
+
+### Framework versions
+
+- PEFT 0.7.2.dev0

ckpts/Qlora_adaptor/adapter_config.json

@@ -0,0 +1,31 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "microsoft/phi-2",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.1,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "fc1",
+    "fc2",
+    "q_proj",
+    "k_proj",
+    "dense",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_rslora": false
+}

ckpts/Qlora_adaptor/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:667821e30438c00dd61ef1b2d55f0cb07411c608e5c375ba6723d90aa2695f7d
+size 377538512

configs.py

@@ -1,6 +1,4 @@
 import torch
-import multiprocessing
-
 def get_config_phase1():
     return {
         "data_dir": "./data",
@@ -13,24 +11,23 @@ def get_config_phase1():
         "max_tokens": 20,
         "clip_embed": 768,
         "phi_embed": 2560,
-        "num_workers": 32, 
+        "num_workers": 4, 
         "ckpts": "./ckpts"
     }
 
 def get_config_phase2():
     return {
-        "i150k_json": "./data/llava_instruct_150k.json",
-        "QA_datasetName": "OpenAssistant/oasst1",
+        "data_dir": "./data",
         "clip_model_name": "openai/clip-vit-base-patch16",
         "phi2_model_name": "microsoft/phi-2",
         "train_batch_size": 1,
         "val_batch_size": 1,
         "device": torch.device("cuda" if torch.cuda.is_available() else "cpu"),
-        "epochs": 2,
-        "max_tokens": 20,
+        "epochs": 10,
+        "max_tokens": 100,
         "clip_embed": 768,
         "phi_embed": 2560,
-        "num_workers": 1, 
+        "num_workers": 0, 
         "ckpts": "./ckpts",
         "vocab_size": 51200
     }

requirements.txt

@@ -1,6 +1,6 @@
 torch
-torchvision
 git+https://github.com/huggingface/peft.git
 accelerate
 transformers
-einops
+einops
+git+https://github.com/m-bain/whisperx.git