Upload 13 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,109 @@
+---
+language:
+- en
+tags:
+- llava
+- multimodal
+- qwen
+license: apache-2.0
+---
+# nanoLLaVA - Sub 1B Vision-Language Model
+
+<p align="center">
+  <img src="https://i.postimg.cc/d15k3YNG/nanollava.webp" alt="Logo" width="350">
+</p>
+
+## Description
+nanoLLaVA is a "small but mighty" 1B vision-language model designed to run efficiently on edge devices.
+- **Base LLM**: [Quyen-SE-v0.1](https://huggingface.co/vilm/Quyen-SE-v0.1) (Qwen1.5-0.5B)
+- **Vision Encoder**: [google/siglip-so400m-patch14-384](https://huggingface.co/google/siglip-so400m-patch14-384)
+
+| Model   | **VQA v2** | **TextVQA** | **ScienceQA** | **POPE** | **MMMU (Test)** | **MMMU (Eval)** | **GQA**  | **MM-VET** |
+|---------|--------|---------|-----------|------|-------------|-------------|------|--------|
+| Score   | 70.84  | 46.71   | 58.97     | 84.1 | 28.6        | 30.4        | 54.79| 23.9   |
+
+## Training Data
+Training Data will be released later as I am still writing a paper on this. Expect the final final to be much more powerful than the current one.
+
+## Finetuning Code
+Coming Soon!!!
+
+## Usage
+You can use with `transformers` with the following script:
+
+```bash
+pip install -U transformers accelerate flash_attn
+```
+
+```python
+import torch
+import transformers
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from PIL import Image
+import warnings
+
+# disable some warnings
+transformers.logging.set_verbosity_error()
+transformers.logging.disable_progress_bar()
+warnings.filterwarnings('ignore')
+
+# set device
+torch.set_default_device('cuda')  # or 'cpu'
+
+# create model
+model = AutoModelForCausalLM.from_pretrained(
+    'qnguyen3/nanoLLaVA',
+    torch_dtype=torch.float16,
+    device_map='auto',
+    trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained(
+    'qnguyen3/nanoLLaVA',
+    trust_remote_code=True)
+
+# text prompt
+prompt = 'Describe this image in detail'
+
+messages = [
+    {"role": "user", "content": f'<image>\n{prompt}'}
+]
+text = tokenizer.apply_chat_template(
+    messages,
+    tokenize=False,
+    add_generation_prompt=True
+)
+
+print(text)
+
+text_chunks = [tokenizer(chunk).input_ids for chunk in text.split('<image>')]
+input_ids = torch.tensor(text_chunks[0] + [-200] + text_chunks[1], dtype=torch.long).unsqueeze(0)
+
+# image, sample images can be found in images folder
+image = Image.open('/path/to/image.png')
+image_tensor = model.process_images([image], model.config).to(dtype=model.dtype)
+
+# generate
+output_ids = model.generate(
+    input_ids,
+    images=image_tensor,
+    max_new_tokens=2048,
+    use_cache=True)[0]
+
+print(tokenizer.decode(output_ids[input_ids.shape[1]:], skip_special_tokens=True).strip())
+```
+
+## Prompt Format
+The model follow the ChatML standard, however, without `\n` at the end of `<|im_end|>`:
+```
+<|im_start|>system
+Answer the question<|im_end|><|im_start|>user
+<image>
+What is the picture about?<|im_end|><|im_start|>assistant
+```
+
+---
+| Image                                | Example                                                                                     |
+|--------------------------------------|---------------------------------------------------------------------------------------------|
+| ![small](example_1.png)              | **What is the text saying?** <br> "Small but mighty". <br>**How does the text correlate to the context of the image?** <br> The text seems to be a playful or humorous representation of a small but mighty figure, possibly a mouse or a mouse toy, holding a weightlifting bar. |
+---
+
+Model is trained using a modified version from [Bunny](https://github.com/BAAI-DCAI/Bunny/tree/main/bunny)

added_tokens.json

@@ -0,0 +1,5 @@
+{
+  "<|endoftext|>": 151643,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644
+}

config.json

@@ -0,0 +1,43 @@
+{
+  "_name_or_path": "/home/ea/work/my_optimum_intel/optimum-intel/tiny-random-nanollava",
+  "architectures": [
+    "LlavaQwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_llava_qwen2.LlavaQwen2Config",
+    "AutoModelForCausalLM": "modeling_llava_qwen2.LlavaQwen2ForCausalLM"
+  },
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "freeze_mm_mlp_adapter": false,
+  "hidden_act": "silu",
+  "hidden_size": 8,
+  "image_aspect_ratio": "pad",
+  "initializer_range": 0.02,
+  "intermediate_size": 32,
+  "language_model": "fxmarty/tiny-dummy-qwen2",
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "mm_hidden_size": 144,
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_vision_tower": "katuni4ka/tiny-random-siglip",
+  "model_type": "llava-qwen2",
+  "num_attention_heads": 4,
+  "num_hidden_layers": 2,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": 32768,
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 4096,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "float32",
+  "transformers_version": "4.45.2",
+  "tune_mm_mlp_adapter": false,
+  "use_cache": false,
+  "use_mm_proj": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

configuration_llava_qwen2.py

@@ -0,0 +1,205 @@
+# coding=utf-8
+# Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Qwen2 model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+QWEN2_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "Qwen/Qwen2-7B-beta": "https://huggingface.co/Qwen/Qwen2-7B-beta/resolve/main/config.json",
+}
+
+
+class Qwen2Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Qwen2Model`]. It is used to instantiate a
+    Qwen2 model according to the specified arguments, defining the model architecture. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of
+    Qwen2-7B-beta [Qwen/Qwen2-7B-beta](https://huggingface.co/Qwen/Qwen2-7B-beta).
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 151936):
+            Vocabulary size of the Qwen2 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Qwen2Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 22016):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 32):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `32`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 32768):
+            The maximum sequence length that this model might ever be used with.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        use_sliding_window (`bool`, *optional*, defaults to `False`):
+            Whether to use sliding window attention.
+        sliding_window (`int`, *optional*, defaults to 4096):
+            Sliding window attention (SWA) window size. If not specified, will default to `4096`.
+        max_window_layers (`int`, *optional*, defaults to 28):
+            The number of layers that use SWA (Sliding Window Attention). The bottom layers use SWA while the top use full attention.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+
+    ```python
+    >>> from transformers import Qwen2Model, Qwen2Config
+
+    >>> # Initializing a Qwen2 style configuration
+    >>> configuration = Qwen2Config()
+
+    >>> # Initializing a model from the Qwen2-7B style configuration
+    >>> model = Qwen2Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "qwen2"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=4096,
+        intermediate_size=22016,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=32,
+        hidden_act="silu",
+        max_position_embeddings=32768,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        use_sliding_window=False,
+        sliding_window=4096,
+        max_window_layers=28,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window
+        self.max_window_layers = max_window_layers
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+
+import os
+from typing import Union
+
+from transformers import PretrainedConfig
+
+
+class SigLipVisionConfig(PretrainedConfig):
+    model_type = "siglip_vision_model"
+
+    def __init__(
+        self,
+        hidden_size=1152,
+        image_mean=(0.5, 0.5, 0.5),
+        intermediate_size=4304,
+        num_hidden_layers=27,
+        num_attention_heads=16,
+        num_channels=3,
+        image_size=384,
+        patch_size=14,
+        hidden_act="gelu_pytorch_tanh",
+        layer_norm_eps=1e-6,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.attention_dropout = attention_dropout
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+        self.image_mean = image_mean
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> "PretrainedConfig":
+        cls._set_token_in_kwargs(kwargs)
+
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        # get the vision config dict if we are loading from SigLipConfig
+        if config_dict.get("model_type") == "siglip":
+            config_dict = config_dict["vision_config"]
+
+        if "model_type" in config_dict and hasattr(cls, "model_type") and config_dict["model_type"] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                f"{cls.model_type}. This is not supported for all configurations of models and can yield errors."
+            )
+
+        return cls.from_dict(config_dict, **kwargs)
+
+
+class LlavaQwen2Config(Qwen2Config):
+    model_type = "llava-qwen2"

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "transformers_version": "4.45.2",
+  "use_cache": false
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6fc99e04347fab21e43cd0c47844360434ed8c3ee13429893162cf840273e26
+size 9739928

special_tokens_map.json

@@ -0,0 +1,20 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcfe42da0a4497e8b2b172c1f9f4ec423a46dc12907f4349c55025f670422ba9
+size 11418266

tokenizer_config.json

@@ -0,0 +1,43 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}