LLaVAOneVision1_5 Support

src/lmms_engine/models/__init__.py

@@ -4,6 +4,11 @@
 from .dream_dllm import DreamDLLMConfig, DreamDLLMForMaskedLM
 from .llada_dllm import LLaDADLLMConfig, LLaDADLLMForMaskedLM
 from .llava_onevision import apply_liger_kernel_to_llava_onevision
+from .llava_onevision1_5 import (
+    LLaVAOneVision1_5_ForConditionalGeneration,
+    Llavaonevision1_5Config,
+    apply_liger_kernel_to_llava_onevision1_5,
+)
 from .monkey_patch import MONKEY_PATCHER
 from .qwen2 import apply_liger_kernel_to_qwen2
 from .qwen2_5_omni import (
@@ -67,4 +72,7 @@
     "SiTModel",
     "SiTConfig",
     "SiT",
+    "Llavaonevision1_5Config",
+    "LLaVAOneVision1_5_ForConditionalGeneration",
+    "apply_liger_kernel_to_llava_onevision1_5",
 ]

src/lmms_engine/models/llava_onevision1_5/__init__.py

@@ -0,0 +1,18 @@
+from lmms_engine.mapping_func import register_model
+
+from .configuration_llavaonevision1_5 import Llavaonevision1_5Config
+from .modeling_llavaonevision1_5 import LLaVAOneVision1_5_ForConditionalGeneration
+from .monkey_patch import apply_liger_kernel_to_llava_onevision1_5
+
+register_model(
+    "llavaonevision1_5",
+    Llavaonevision1_5Config,
+    LLaVAOneVision1_5_ForConditionalGeneration,
+    model_general_type="image_text_to_text",
+)
+
+__all__ = [
+    "Llavaonevision1_5Config",
+    "LLaVAOneVision1_5_ForConditionalGeneration",
+    "apply_liger_kernel_to_llava_onevision1_5",
+]

src/lmms_engine/models/llava_onevision1_5/configuration_llavaonevision1_5.py

@@ -0,0 +1,276 @@
+"""LLaVALLaVAOneVision1_5 model configuration"""
+from transformers.configuration_utils import PretrainedConfig, layer_type_validation
+from transformers.modeling_rope_utils import rope_config_validation
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+
+class RiceConfig(PretrainedConfig):
+    model_type = "rice_vit"
+    base_config_key = "vision_config"
+
+    def __init__(
+        self,
+        depth=24,
+        embed_dim=1024,
+        hidden_size=1024,
+        hidden_act="gelu",
+        intermediate_size=4096,
+        num_heads=16,
+        in_channels=3,
+        patch_size=14,
+        spatial_merge_size=2,
+        temporal_patch_size=1,
+        initializer_range=0.02,
+        layer_norm_eps=1e-05,
+        text_hidden_size=2560,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.depth = depth
+        self.embed_dim = embed_dim
+        self.hidden_size = hidden_size
+        self.hidden_act = hidden_act
+        self.intermediate_size = intermediate_size
+        self.num_heads = num_heads
+        self.in_channels = in_channels
+        self.patch_size = patch_size
+        self.spatial_merge_size = spatial_merge_size
+        self.temporal_patch_size = temporal_patch_size
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
+        self.text_hidden_size = text_hidden_size
+
+
+class LLaVAOneVision1_5_TextConfig(PretrainedConfig):
+    r"""
+    Args:
+        vocab_size (`int`, *optional*, defaults to 152064):
+            Vocabulary size of the Qwen2VL model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Qwen2VLModel`]
+        hidden_size (`int`, *optional*, defaults to 8192):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 29568):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 80):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 64):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 8):
+            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-05):
+            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 1000000.0):
+            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 80):
+            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.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type
+            and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value
+            accordingly.
+            Expected contents:
+                `rope_type` (`str`):
+                    The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope',
+                    'llama3'], with 'default' being the original RoPE implementation.
+                `factor` (`float`, *optional*):
+                    Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In
+                    most scaling types, a `factor` of x will enable the model to handle sequences of length x *
+                    original maximum pre-trained length.
+                `original_max_position_embeddings` (`int`, *optional*):
+                    Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during
+                    pretraining.
+                `attention_factor` (`float`, *optional*):
+                    Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention
+                    computation. If unspecified, it defaults to value recommended by the implementation, using the
+                    `factor` field to infer the suggested value.
+                `beta_fast` (`float`, *optional*):
+                    Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear
+                    ramp function. If unspecified, it defaults to 32.
+                `beta_slow` (`float`, *optional*):
+                    Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
+                    ramp function. If unspecified, it defaults to 1.
+                `short_factor` (`List[float]`, *optional*):
+                    Only used with 'longrope'. The scaling factor to be applied to short contexts (<
+                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
+                    size divided by the number of attention heads divided by 2
+                `long_factor` (`List[float]`, *optional*):
+                    Only used with 'longrope'. The scaling factor to be applied to long contexts (<
+                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
+                    size divided by the number of attention heads divided by 2
+                `low_freq_factor` (`float`, *optional*):
+                    Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE
+                `high_freq_factor` (`float`, *optional*):
+                    Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE
+        image_token_id (`int`, *optional*):
+            Token index used as placeholder for image embeddings.
+        video_token_id (`int`, *optional*):
+            Token index used as placeholder for video embeddings.
+    """
+
+    model_type = "LLaVAOneVision1_5_text"
+    base_config_key = "text_config"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    # Default tensor parallel plan for base model `Qwen2VL`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=4096,
+        intermediate_size=12288,
+        num_hidden_layers=36,
+        num_attention_heads=32,
+        num_key_value_heads=8,
+        head_dim=128,
+        hidden_act="silu",
+        max_position_embeddings=32768,
+        initializer_range=0.02,
+        rms_norm_eps=1e-06,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=1000000.0,
+        attention_bias=False,
+        use_sliding_window=False,
+        sliding_window=None,
+        max_window_layers=36,
+        attention_dropout=0.0,
+        rope_scaling=None,
+        layer_types=None,
+        image_token_id=None,
+        video_token_id=None,
+        **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.head_dim = head_dim
+        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
+        self.rope_scaling = rope_scaling
+        self.attention_bias = attention_bias
+        self.tie_word_embeddings = tie_word_embeddings
+
+        # Validate the correctness of rotary position embeddings parameters
+        # BC: if there is a 'type' field, move it to 'rope_type'.
+        # and change type from 'mrope' to 'default' because `mrope` does default RoPE calculations
+        # one can set it to "linear"/"dynamic" etc. to have scaled RoPE
+        # TODO: @raushan update config in the hub
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            if self.rope_scaling["type"] == "mrope":
+                self.rope_scaling["type"] = "default"
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self, ignore_keys={"mrope_section"})
+        self.image_token_id = image_token_id
+        self.video_token_id = video_token_id
+
+        self.layer_types = layer_types
+        if self.layer_types is None:
+            self.layer_types = [
+                "sliding_attention"
+                if self.sliding_window is not None and i >= self.max_window_layers
+                else "full_attention"
+                for i in range(self.num_hidden_layers)
+            ]
+        layer_type_validation(self.layer_types)
+
+        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
+
+
+class Llavaonevision1_5Config(PretrainedConfig):
+    r"""
+    Args:
+        text_config (`Union[PreTrainedConfig, dict]`, *optional*, defaults to `LLaVAOneVision1_5_TextConfig`):
+            The config object or dictionary of the text backbone.
+        vision_config (`Union[PreTrainedConfig, dict]`,  *optional*, defaults to `LLaVAOneVision1_5_VisionConfig`):
+            The config object or dictionary of the vision backbone.
+        image_token_id (`int`, *optional*, defaults to 151655):
+            The image token index to encode the image prompt.
+        video_token_id (`int`, *optional*, defaults to 151656):
+            The video token index to encode the image prompt.
+    """
+
+    model_type = "llavaonevision1_5"
+    sub_configs = {"vision_config": RiceConfig, "text_config": LLaVAOneVision1_5_TextConfig}
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        text_config=None,
+        vision_config=None,
+        image_token_id=151655,
+        video_token_id=151656,
+        vocab_size=152064,
+        **kwargs,
+    ):
+        if isinstance(vision_config, dict):
+            self.vision_config = self.sub_configs["vision_config"](**vision_config)
+        elif vision_config is None:
+            self.vision_config = self.sub_configs["vision_config"]()
+
+        if isinstance(text_config, dict):
+            self.text_config = self.sub_configs["text_config"](**text_config)
+        elif text_config is None:
+            # For BC use all kwargs to init `TextConfig`
+            self.text_config = self.sub_configs["text_config"](**kwargs)
+
+        self.image_token_id = image_token_id
+        self.video_token_id = video_token_id
+        self.vocab_size = vocab_size
+
+        super().__init__(**kwargs)
+
+
+__all__ = ["Llavaonevision1_5Config", "LLaVAOneVision1_5_TextConfig"]