rename package

Former-commit-id: 308edbc426

src/llamafactory/model/__init__.py

@@ -0,0 +1,12 @@
+from .loader import load_config, load_model, load_tokenizer
+from .utils.misc import find_all_linear_modules
+from .utils.valuehead import load_valuehead_params
+
+
+__all__ = [
+    "load_config",
+    "load_model",
+    "load_tokenizer",
+    "load_valuehead_params",
+    "find_all_linear_modules",
+]

src/llamafactory/model/adapter.py

@@ -0,0 +1,225 @@
+import re
+from typing import TYPE_CHECKING
+
+import torch
+from peft import LoraConfig, LoraModel, PeftModel, TaskType, get_peft_model
+from transformers.integrations import deepspeed_config, is_deepspeed_zero3_enabled
+from transformers.modeling_utils import is_fsdp_enabled
+
+from ..extras.logging import get_logger
+from .utils.misc import find_all_linear_modules, find_expanded_modules
+from .utils.quantization import QuantizationMethod
+from .utils.unsloth import get_unsloth_peft_model, load_unsloth_peft_model
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedModel
+
+    from ..hparams import FinetuningArguments, ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def init_adapter(
+    config: "PretrainedConfig",
+    model: "PreTrainedModel",
+    model_args: "ModelArguments",
+    finetuning_args: "FinetuningArguments",
+    is_trainable: bool,
+) -> "PreTrainedModel":
+    r"""
+    Initializes the adapters.
+
+    Support full-parameter, freeze and LoRA training.
+
+    Note that the trainable parameters must be cast to float32.
+    """
+
+    if (not is_trainable) and model_args.adapter_name_or_path is None:
+        logger.info("Adapter is not found at evaluation, load the base model.")
+        return model
+
+    if finetuning_args.finetuning_type != "lora" and getattr(model, "quantization_method", None):
+        raise ValueError("You can only use lora for quantized models.")
+
+    if deepspeed_config() is not None or is_fsdp_enabled() or finetuning_args.pure_bf16 or finetuning_args.use_badam:
+        logger.info("DeepSpeed/FSDP/PureBF16/BAdam detected, remaining trainable params in half precision.")
+        cast_trainable_params_to_fp32 = False
+    else:
+        logger.info("Upcasting trainable params to float32.")
+        cast_trainable_params_to_fp32 = True
+
+    if finetuning_args.finetuning_type == "full" and is_trainable:
+        logger.info("Fine-tuning method: Full")
+        if cast_trainable_params_to_fp32:
+            model = model.float()
+
+        if model_args.visual_inputs and hasattr(model, "vision_tower"):  # freeze vision model
+            model.vision_tower.requires_grad_(False)
+
+    if finetuning_args.finetuning_type == "freeze" and is_trainable:
+        logger.info("Fine-tuning method: Freeze")
+        num_layers = (
+            getattr(model.config, "num_hidden_layers", None)
+            or getattr(model.config, "num_layers", None)
+            or getattr(model.config, "n_layer", None)
+        )
+        if not num_layers:
+            raise ValueError("Current model does not support freeze tuning.")
+
+        if finetuning_args.use_llama_pro:
+            if num_layers % finetuning_args.freeze_trainable_layers != 0:
+                raise ValueError(
+                    "`num_layers` {} should be divisible by `num_layer_trainable` {}.".format(
+                        num_layers, finetuning_args.freeze_trainable_layers
+                    )
+                )
+
+            stride = num_layers // finetuning_args.freeze_trainable_layers
+            trainable_layer_ids = range(stride - 1, num_layers + stride - 1, stride)
+        elif finetuning_args.freeze_trainable_layers > 0:  # fine-tuning the last n layers if num_layer_trainable > 0
+            trainable_layer_ids = range(max(0, num_layers - finetuning_args.freeze_trainable_layers), num_layers)
+        else:  # fine-tuning the first n layers if num_layer_trainable < 0
+            trainable_layer_ids = range(min(-finetuning_args.freeze_trainable_layers, num_layers))
+
+        hidden_modules = set()
+        non_hidden_modules = set()
+        for name, _ in model.named_parameters():
+            if ".0." in name:
+                hidden_modules.add(name.split(".0.")[-1].split(".")[0])
+            elif ".1." in name:  # MoD starts from layer 1
+                hidden_modules.add(name.split(".1.")[-1].split(".")[0])
+
+            if re.search(r"\.\d+\.", name) is None:
+                non_hidden_modules.add(name.split(".")[-2])
+
+        trainable_layers = []
+        for module_name in finetuning_args.freeze_trainable_modules:
+            if module_name != "all" and module_name not in hidden_modules:
+                raise ValueError(
+                    "Module {} is not found, please choose from {}".format(module_name, ", ".join(hidden_modules))
+                )
+
+            for idx in trainable_layer_ids:
+                trainable_layers.append(".{:d}.{}".format(idx, module_name if module_name != "all" else ""))
+
+        if finetuning_args.freeze_extra_modules:
+            for module_name in finetuning_args.freeze_extra_modules:
+                if module_name not in non_hidden_modules:
+                    raise ValueError(
+                        "Module {} is not found, please choose from {}".format(
+                            module_name, ", ".join(non_hidden_modules)
+                        )
+                    )
+
+                trainable_layers.append(module_name)
+
+        for name, param in model.named_parameters():
+            if any(trainable_layer in name for trainable_layer in trainable_layers):
+                if cast_trainable_params_to_fp32:
+                    param.data = param.data.to(torch.float32)
+            else:
+                param.requires_grad_(False)
+
+        logger.info("Set trainable layers: {}".format(",".join(map(str, trainable_layer_ids))))
+
+    if finetuning_args.finetuning_type == "lora":
+        logger.info("Fine-tuning method: {}".format("DoRA" if finetuning_args.use_dora else "LoRA"))
+        adapter_to_resume = None
+
+        if model_args.adapter_name_or_path is not None:
+            is_mergeable = True
+            if getattr(model, "quantization_method", None):  # merge lora in quantized model is unstable
+                assert len(model_args.adapter_name_or_path) == 1, "Quantized model only accepts a single adapter."
+                is_mergeable = False
+
+            if is_deepspeed_zero3_enabled():
+                assert len(model_args.adapter_name_or_path) == 1, "Cannot use multiple adapters in DeepSpeed ZeRO-3."
+                is_mergeable = False
+
+            if model_args.use_unsloth:
+                assert len(model_args.adapter_name_or_path) == 1, "Unsloth model only accepts a single adapter."
+                is_mergeable = False
+
+            if (is_trainable and not finetuning_args.create_new_adapter) or (not is_mergeable):
+                adapter_to_merge = model_args.adapter_name_or_path[:-1]
+                adapter_to_resume = model_args.adapter_name_or_path[-1]
+            else:
+                adapter_to_merge = model_args.adapter_name_or_path
+
+            for adapter in adapter_to_merge:
+                model: "LoraModel" = PeftModel.from_pretrained(
+                    model, adapter, offload_folder=model_args.offload_folder
+                )
+                model = model.merge_and_unload()
+
+            if len(adapter_to_merge) > 0:
+                logger.info("Merged {} adapter(s).".format(len(adapter_to_merge)))
+
+            if adapter_to_resume is not None:  # resume lora training
+                if model_args.use_unsloth:
+                    model = load_unsloth_peft_model(config, model_args, is_trainable=is_trainable)
+                else:
+                    model = PeftModel.from_pretrained(
+                        model,
+                        adapter_to_resume,
+                        is_trainable=is_trainable,
+                        offload_folder=model_args.offload_folder,
+                    )
+
+        if is_trainable and adapter_to_resume is None:  # create new lora weights while training
+            if len(finetuning_args.lora_target) == 1 and finetuning_args.lora_target[0] == "all":
+                target_modules = find_all_linear_modules(model)
+            else:
+                target_modules = finetuning_args.lora_target
+
+            if finetuning_args.use_llama_pro:
+                target_modules = find_expanded_modules(model, target_modules, finetuning_args.num_layer_trainable)
+
+            if (
+                finetuning_args.use_dora
+                and getattr(model, "quantization_method", None) is not None
+                and getattr(model, "quantization_method", None) != QuantizationMethod.BITS_AND_BYTES
+            ):
+                raise ValueError("DoRA is not compatible with PTQ-quantized models.")
+
+            if model_args.resize_vocab and finetuning_args.additional_target is None:
+                input_embeddings = model.get_input_embeddings()
+                output_embeddings = model.get_output_embeddings()
+                module_names = set()
+                for name, module in model.named_modules():
+                    if module in [input_embeddings, output_embeddings]:
+                        module_names.add(name.split(".")[-1])
+
+                finetuning_args.additional_target = module_names
+                logger.warning("Vocab has been resized, add {} to trainable params.".format(",".join(module_names)))
+
+            peft_kwargs = {
+                "r": finetuning_args.lora_rank,
+                "target_modules": target_modules,
+                "lora_alpha": finetuning_args.lora_alpha,
+                "lora_dropout": finetuning_args.lora_dropout,
+                "use_rslora": finetuning_args.use_rslora,
+                "modules_to_save": finetuning_args.additional_target,
+            }
+
+            if model_args.use_unsloth:
+                model = get_unsloth_peft_model(model, model_args, peft_kwargs)
+            else:
+                lora_config = LoraConfig(
+                    task_type=TaskType.CAUSAL_LM,
+                    inference_mode=False,
+                    use_dora=finetuning_args.use_dora,
+                    **peft_kwargs,
+                )
+                model = get_peft_model(model, lora_config)
+
+        if cast_trainable_params_to_fp32:
+            for param in filter(lambda p: p.requires_grad, model.parameters()):
+                param.data = param.data.to(torch.float32)
+
+        if model_args.adapter_name_or_path is not None:
+            logger.info("Loaded adapter(s): {}".format(",".join(model_args.adapter_name_or_path)))
+
+    return model

src/llamafactory/model/loader.py

@@ -0,0 +1,183 @@
+from typing import TYPE_CHECKING, Any, Dict, Optional, TypedDict
+
+from transformers import AutoConfig, AutoModelForCausalLM, AutoModelForVision2Seq, AutoProcessor, AutoTokenizer
+from trl import AutoModelForCausalLMWithValueHead
+
+from ..extras.logging import get_logger
+from ..extras.misc import count_parameters, try_download_model_from_ms
+from .adapter import init_adapter
+from .patcher import patch_config, patch_model, patch_tokenizer, patch_valuehead_model
+from .utils.misc import register_autoclass
+from .utils.mod import convert_pretrained_model_to_mod, load_mod_pretrained_model
+from .utils.unsloth import load_unsloth_pretrained_model
+from .utils.valuehead import load_valuehead_params
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedModel, PreTrainedTokenizer, ProcessorMixin
+
+    from ..hparams import FinetuningArguments, ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+class TokenizerModule(TypedDict):
+    tokenizer: "PreTrainedTokenizer"
+    processor: Optional["ProcessorMixin"]
+
+
+def _get_init_kwargs(model_args: "ModelArguments") -> Dict[str, Any]:
+    r"""
+    Gets arguments to load config/tokenizer/model.
+
+    Note: including inplace operation of model_args.
+    """
+    model_args.model_name_or_path = try_download_model_from_ms(model_args)
+    return {
+        "trust_remote_code": True,
+        "cache_dir": model_args.cache_dir,
+        "revision": model_args.model_revision,
+        "token": model_args.hf_hub_token,
+    }
+
+
+def load_tokenizer(model_args: "ModelArguments") -> "TokenizerModule":
+    r"""
+    Loads pretrained tokenizer.
+
+    Note: including inplace operation of model_args.
+    """
+    init_kwargs = _get_init_kwargs(model_args)
+    try:
+        tokenizer = AutoTokenizer.from_pretrained(
+            model_args.model_name_or_path,
+            use_fast=model_args.use_fast_tokenizer,
+            split_special_tokens=model_args.split_special_tokens,
+            padding_side="right",
+            **init_kwargs,
+        )
+    except ValueError:  # try the fast one
+        tokenizer = AutoTokenizer.from_pretrained(
+            model_args.model_name_or_path,
+            use_fast=True,
+            padding_side="right",
+            **init_kwargs,
+        )
+
+    if model_args.new_special_tokens is not None:
+        num_added_tokens = tokenizer.add_special_tokens(
+            dict(additional_special_tokens=model_args.new_special_tokens),
+            replace_additional_special_tokens=False,
+        )
+        logger.info("Add {} to special tokens.".format(",".join(model_args.new_special_tokens)))
+        if num_added_tokens > 0 and not model_args.resize_vocab:
+            model_args.resize_vocab = True
+            logger.warning("New tokens have been added, changed `resize_vocab` to True.")
+
+    patch_tokenizer(tokenizer)
+
+    if model_args.visual_inputs:
+        try:
+            processor = AutoProcessor.from_pretrained(model_args.model_name_or_path, **init_kwargs)
+            setattr(processor, "tokenizer", tokenizer)
+        except Exception:
+            raise ValueError(
+                "This multimodal LLM is not supported.\n"
+                "Download LLaVA-1.5 models from: https://huggingface.co/llava-hf\n"
+                "Download Yi-VL models from: https://huggingface.co/BUAADreamer"
+            )
+    else:
+        processor = None
+
+    return {"tokenizer": tokenizer, "processor": processor}
+
+
+def load_config(model_args: "ModelArguments") -> "PretrainedConfig":
+    r"""
+    Loads model config.
+    """
+    init_kwargs = _get_init_kwargs(model_args)
+    return AutoConfig.from_pretrained(model_args.model_name_or_path, **init_kwargs)
+
+
+def load_model(
+    tokenizer: "PreTrainedTokenizer",
+    model_args: "ModelArguments",
+    finetuning_args: "FinetuningArguments",
+    is_trainable: bool = False,
+    add_valuehead: bool = False,
+) -> "PreTrainedModel":
+    r"""
+    Loads pretrained model.
+    """
+    init_kwargs = _get_init_kwargs(model_args)
+    config = load_config(model_args)
+    patch_config(config, tokenizer, model_args, init_kwargs, is_trainable)
+
+    model = None
+    lazy_load = False
+    if model_args.use_unsloth:
+        if model_args.adapter_name_or_path is not None:
+            lazy_load = True
+        elif is_trainable:
+            model = load_unsloth_pretrained_model(config, model_args)
+
+    if model is None and not lazy_load:
+        init_kwargs["config"] = config
+        init_kwargs["pretrained_model_name_or_path"] = model_args.model_name_or_path
+
+        if model_args.mixture_of_depths == "load":
+            model = load_mod_pretrained_model(**init_kwargs)
+        elif model_args.visual_inputs:
+            model = AutoModelForVision2Seq.from_pretrained(**init_kwargs)
+        else:
+            model = AutoModelForCausalLM.from_pretrained(**init_kwargs)
+
+        if model_args.mixture_of_depths == "convert":
+            model = convert_pretrained_model_to_mod(model, config, model_args)
+
+    if not lazy_load:
+        patch_model(model, tokenizer, model_args, is_trainable, add_valuehead)
+        register_autoclass(config, model, tokenizer)
+
+    model = init_adapter(config, model, model_args, finetuning_args, is_trainable)
+
+    if add_valuehead:
+        model = AutoModelForCausalLMWithValueHead.from_pretrained(model)
+        patch_valuehead_model(model)
+
+        if model_args.adapter_name_or_path is not None:
+            vhead_path = model_args.adapter_name_or_path[-1]
+        else:
+            vhead_path = model_args.model_name_or_path
+
+        vhead_params = load_valuehead_params(vhead_path, model_args)
+        if vhead_params is not None:
+            model.load_state_dict(vhead_params, strict=False)
+            logger.info("Loaded valuehead from checkpoint: {}".format(vhead_path))
+
+    if not is_trainable:
+        model.requires_grad_(False)
+        model.eval()
+    else:
+        model.train()
+
+    trainable_params, all_param = count_parameters(model)
+    if is_trainable:
+        param_stats = "trainable params: {:d} || all params: {:d} || trainable%: {:.4f}".format(
+            trainable_params, all_param, 100 * trainable_params / all_param
+        )
+    else:
+        param_stats = "all params: {:d}".format(all_param)
+    logger.info(param_stats)
+
+    if model_args.print_param_status:
+        for name, param in model.named_parameters():
+            print(
+                "name: {}, dtype: {}, device: {}, trainable: {}".format(
+                    name, param.dtype, param.device, param.requires_grad
+                )
+            )
+
+    return model

src/llamafactory/model/patcher.py

@@ -0,0 +1,140 @@
+import os
+from types import MethodType
+from typing import TYPE_CHECKING, Any, Dict
+
+import torch
+from peft import PeftModel
+from transformers import PreTrainedModel, PreTrainedTokenizerBase, is_torch_npu_available
+from transformers.integrations import is_deepspeed_zero3_enabled
+from transformers.modeling_utils import is_fsdp_enabled
+
+from ..extras.logging import get_logger
+from ..extras.misc import infer_optim_dtype
+from .utils.attention import configure_attn_implementation, print_attn_implementation
+from .utils.checkpointing import prepare_model_for_training
+from .utils.embedding import resize_embedding_layer
+from .utils.longlora import configure_longlora
+from .utils.moe import add_z3_leaf_module, configure_moe
+from .utils.quantization import configure_quantization
+from .utils.rope import configure_rope
+from .utils.valuehead import prepare_valuehead_model
+from .utils.visual import autocast_projector_dtype, configure_visual_model
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedTokenizer
+    from trl import AutoModelForCausalLMWithValueHead
+
+    from ..hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def patch_tokenizer(tokenizer: "PreTrainedTokenizer") -> None:
+    if "PreTrainedTokenizerBase" not in str(tokenizer._pad.__func__):
+        tokenizer._pad = MethodType(PreTrainedTokenizerBase._pad, tokenizer)
+
+
+def patch_config(
+    config: "PretrainedConfig",
+    tokenizer: "PreTrainedTokenizer",
+    model_args: "ModelArguments",
+    init_kwargs: Dict[str, Any],
+    is_trainable: bool,
+) -> None:
+    if model_args.compute_dtype is None:  # priority: bf16 > fp16 > fp32
+        model_args.compute_dtype = infer_optim_dtype(model_dtype=getattr(config, "torch_dtype", None))
+
+    if is_torch_npu_available():
+        use_jit_compile = os.environ.get("JIT_COMPILE", "0").lower() in ["true", "1"]
+        torch.npu.set_compile_mode(jit_compile=use_jit_compile)
+
+    configure_attn_implementation(config, model_args)
+    configure_rope(config, model_args, is_trainable)
+    configure_longlora(config, model_args, is_trainable)
+    configure_quantization(config, tokenizer, model_args, init_kwargs)
+    configure_moe(config, model_args, is_trainable)
+    configure_visual_model(config)
+
+    if model_args.use_cache and not is_trainable:
+        setattr(config, "use_cache", True)
+        logger.info("Using KV cache for faster generation.")
+
+    if getattr(config, "model_type", None) == "qwen":
+        setattr(config, "use_flash_attn", model_args.flash_attn == "fa2")
+        for dtype_name, dtype in [("fp16", torch.float16), ("bf16", torch.bfloat16), ("fp32", torch.float32)]:
+            setattr(config, dtype_name, model_args.compute_dtype == dtype)
+
+    if getattr(config, "model_type", None) == "qwen2" and is_trainable and model_args.flash_attn:
+        setattr(config, "use_cache", False)  # qwen2 does not support use_cache when using flashattn
+
+    init_kwargs["torch_dtype"] = model_args.compute_dtype
+    if not is_deepspeed_zero3_enabled() and not is_fsdp_enabled():
+        init_kwargs["low_cpu_mem_usage"] = model_args.low_cpu_mem_usage
+        if init_kwargs["low_cpu_mem_usage"]:
+            if "device_map" not in init_kwargs and model_args.device_map:
+                init_kwargs["device_map"] = model_args.device_map
+
+            if init_kwargs["device_map"] == "auto":
+                init_kwargs["offload_folder"] = model_args.offload_folder
+
+
+def patch_model(
+    model: "PreTrainedModel",
+    tokenizer: "PreTrainedTokenizer",
+    model_args: "ModelArguments",
+    is_trainable: bool,
+    add_valuehead: bool,
+) -> None:
+    gen_config = model.generation_config  # check and fix generation config
+    if not gen_config.do_sample and (
+        (gen_config.temperature is not None and gen_config.temperature != 1.0)
+        or (gen_config.top_p is not None and gen_config.top_p != 1.0)
+        or (gen_config.typical_p is not None and gen_config.typical_p != 1.0)
+    ):
+        gen_config.do_sample = True
+
+    if "GenerationMixin" not in str(model.generate.__func__):
+        model.generate = MethodType(PreTrainedModel.generate, model)
+
+    if add_valuehead:
+        prepare_valuehead_model(model)
+
+    if model_args.resize_vocab:
+        resize_embedding_layer(model, tokenizer)
+
+    if model_args.visual_inputs:
+        autocast_projector_dtype(model, model_args)
+
+    if is_trainable:
+        prepare_model_for_training(model, model_args)
+        add_z3_leaf_module(model)
+
+    if not model_args.use_unsloth:
+        print_attn_implementation(model.config)
+
+    try:
+        model.add_model_tags(["llama-factory"])
+    except Exception:
+        logger.warning("Cannot properly tag the model.")
+
+
+def patch_valuehead_model(model: "AutoModelForCausalLMWithValueHead") -> None:
+    def tie_weights(self: "AutoModelForCausalLMWithValueHead") -> None:
+        if isinstance(self.pretrained_model, PreTrainedModel):
+            self.pretrained_model.tie_weights()
+
+    def get_input_embeddings(self: "AutoModelForCausalLMWithValueHead") -> torch.nn.Module:
+        if isinstance(self.pretrained_model, PreTrainedModel):
+            return self.pretrained_model.get_input_embeddings()
+
+    def create_or_update_model_card(self: "AutoModelForCausalLMWithValueHead", output_dir: str) -> None:
+        if isinstance(self.pretrained_model, PeftModel):
+            self.pretrained_model.create_or_update_model_card(output_dir)
+
+    ignore_modules = [name for name, _ in model.named_parameters() if "pretrained_model" in name]
+    setattr(model, "_keys_to_ignore_on_save", ignore_modules)
+    setattr(model, "tie_weights", MethodType(tie_weights, model))
+    setattr(model, "get_input_embeddings", MethodType(get_input_embeddings, model))
+    setattr(model, "create_or_update_model_card", MethodType(create_or_update_model_card, model))

src/llamafactory/model/utils/attention.py

@@ -0,0 +1,55 @@
+from typing import TYPE_CHECKING
+
+from ...extras.logging import get_logger
+from ...extras.packages import is_flash_attn2_available, is_sdpa_available
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def configure_attn_implementation(config: "PretrainedConfig", model_args: "ModelArguments") -> None:
+    if model_args.flash_attn == "auto":
+        return
+
+    elif model_args.flash_attn == "off":
+        requested_attn_implementation = "eager"
+
+    elif model_args.flash_attn == "sdpa":
+        if not is_sdpa_available():
+            logger.warning("torch>=2.1.1 is required for SDPA attention.")
+            return
+
+        requested_attn_implementation = "sdpa"
+    elif model_args.flash_attn == "fa2":
+        if not is_flash_attn2_available():
+            logger.warning("FlashAttention-2 is not installed.")
+            return
+
+        requested_attn_implementation = "flash_attention_2"
+    else:
+        raise NotImplementedError("Unknown attention type: {}".format(model_args.flash_attn))
+
+    if getattr(config, "model_type", None) == "internlm2":  # special case for custom models
+        setattr(config, "attn_implementation", requested_attn_implementation)
+    else:
+        setattr(config, "_attn_implementation", requested_attn_implementation)
+
+
+def print_attn_implementation(config: "PretrainedConfig") -> None:
+    if getattr(config, "model_type", None) == "internlm2":  # special case for custom models
+        attn_implementation = getattr(config, "attn_implementation", None)
+    else:
+        attn_implementation = getattr(config, "_attn_implementation", None)
+
+    if attn_implementation == "flash_attention_2":
+        logger.info("Using FlashAttention-2 for faster training and inference.")
+    elif attn_implementation == "sdpa":
+        logger.info("Using torch SDPA for faster training and inference.")
+    else:
+        logger.info("Using vanilla attention implementation.")

src/llamafactory/model/utils/checkpointing.py

@@ -0,0 +1,94 @@
+import inspect
+from functools import partial
+from types import MethodType
+from typing import TYPE_CHECKING, Any, Dict, Optional, Tuple
+
+import torch
+
+from ...extras.constants import LAYERNORM_NAMES
+from ...extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedModel
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def _gradient_checkpointing_enable(
+    self: "PreTrainedModel", gradient_checkpointing_kwargs: Optional[Dict[str, Any]] = None
+) -> None:
+    r"""
+    Activates gradient checkpointing for the current model.
+
+    Modification of the original method to enable gradient checkpointing for block-wise optimizer.
+    """
+    from torch.utils.checkpoint import checkpoint
+
+    if not self.supports_gradient_checkpointing:
+        raise ValueError("{} does not support gradient checkpointing.".format(self.__class__.__name__))
+
+    if gradient_checkpointing_kwargs is None:
+        gradient_checkpointing_kwargs = {"use_reentrant": True}
+
+    gradient_checkpointing_func = partial(checkpoint, **gradient_checkpointing_kwargs)
+
+    def custom_gradient_checkpointing_func(func, *args, **kwargs):
+        module: "torch.nn.Module" = func.__self__
+
+        if any(param.requires_grad for param in module.parameters()):
+            for arg in args:
+                if torch.is_tensor(arg) and torch.is_floating_point(arg):
+                    arg.requires_grad_(True)
+
+        return gradient_checkpointing_func(func, *args, **kwargs)
+
+    if "value" in inspect.signature(self._set_gradient_checkpointing).parameters:  # old GC format
+        self.apply(partial(self._set_gradient_checkpointing, value=True))
+        self.enable_input_require_grads()
+        logger.warning("You are using the old GC format, some features (e.g. BAdam) will be invalid.")
+    else:  # have already enabled input require gradients
+        self._set_gradient_checkpointing(enable=True, gradient_checkpointing_func=custom_gradient_checkpointing_func)
+
+
+def _fp32_forward_post_hook(
+    module: "torch.nn.Module", args: Tuple["torch.Tensor"], output: "torch.Tensor"
+) -> "torch.Tensor":
+    return output.to(torch.float32)
+
+
+def prepare_model_for_training(
+    model: "PreTrainedModel", model_args: "ModelArguments", output_layer_name: str = "lm_head"
+) -> None:
+    r"""
+    Includes:
+        (1) cast the layernorm in fp32
+        (2) make output embedding layer require grads
+        (3) add the upcasting of the lm_head in fp32
+    Inspired by: https://github.com/huggingface/peft/blob/v0.7.1/src/peft/utils/other.py#L72
+    """
+    if model_args.upcast_layernorm:
+        logger.info("Upcasting layernorm weights in float32.")
+        for name, param in model.named_parameters():
+            if param.ndim == 1 and any(ln_name in name for ln_name in LAYERNORM_NAMES):
+                param.data = param.data.to(torch.float32)
+
+    if not model_args.disable_gradient_checkpointing:
+        if not getattr(model, "supports_gradient_checkpointing", False):
+            logger.warning("Current model does not support gradient checkpointing.")
+        else:
+            # use_reentrant=False might increase VRAM usage (have not been empirically verified yet)
+            # According to: https://github.com/huggingface/transformers/issues/28339
+            model.gradient_checkpointing_enable = MethodType(_gradient_checkpointing_enable, model)
+            model.gradient_checkpointing_enable(gradient_checkpointing_kwargs={"use_reentrant": True})
+            setattr(model.config, "use_cache", False)  # turn off when gradient checkpointing is enabled
+            logger.info("Gradient checkpointing enabled.")
+
+    if hasattr(model, output_layer_name) and model_args.upcast_lmhead_output:
+        logger.info("Upcasting lm_head outputs in float32.")
+        output_layer = getattr(model, output_layer_name)
+        if isinstance(output_layer, torch.nn.Linear) and output_layer.weight.dtype != torch.float32:
+            output_layer.register_forward_hook(_fp32_forward_post_hook)

src/llamafactory/model/utils/embedding.py

@@ -0,0 +1,58 @@
+import math
+from contextlib import nullcontext
+from typing import TYPE_CHECKING
+
+import torch
+from transformers.integrations import is_deepspeed_zero3_enabled
+
+from ...extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedModel, PreTrainedTokenizer
+
+
+logger = get_logger(__name__)
+
+
+def _noisy_mean_initialization(embed_weight: torch.Tensor, num_new_tokens: int) -> None:
+    embedding_dim = embed_weight.size(1)
+    avg_weight = embed_weight[:-num_new_tokens].mean(dim=0, keepdim=True)
+    noise_weight = torch.empty_like(embed_weight[-num_new_tokens:])
+    noise_weight.normal_(mean=0, std=(1.0 / math.sqrt(embedding_dim)))
+    embed_weight[-num_new_tokens:] = avg_weight + noise_weight
+
+
+def resize_embedding_layer(model: "PreTrainedModel", tokenizer: "PreTrainedTokenizer") -> None:
+    r"""
+    Resize token embeddings.
+    """
+    if is_deepspeed_zero3_enabled():
+        import deepspeed  # type: ignore
+
+        params = [model.get_input_embeddings().weight]
+        if model.get_output_embeddings() is not None and not model.config.tie_word_embeddings:
+            params.append(model.get_output_embeddings().weight)
+
+        context_maybe_zero3 = deepspeed.zero.GatheredParameters(params, modifier_rank=0)
+    else:
+        context_maybe_zero3 = nullcontext()
+
+    with context_maybe_zero3:
+        current_embedding_size = model.get_input_embeddings().weight.size(0)
+
+    if len(tokenizer) > current_embedding_size:
+        if getattr(model, "quantization_method", None):
+            raise ValueError("Cannot resize embedding layers of a quantized model.")
+
+        if not isinstance(model.get_output_embeddings(), torch.nn.Linear):
+            raise ValueError("Current model does not support resizing embedding layers.")
+
+        model.resize_token_embeddings(len(tokenizer), pad_to_multiple_of=64)
+        with context_maybe_zero3:
+            new_embedding_size = model.get_input_embeddings().weight.size(0)
+            num_new_tokens = new_embedding_size - current_embedding_size
+            _noisy_mean_initialization(model.get_input_embeddings().weight.data, num_new_tokens)
+            _noisy_mean_initialization(model.get_output_embeddings().weight.data, num_new_tokens)
+
+        logger.info("Resized token embeddings from {} to {}.".format(current_embedding_size, new_embedding_size))

src/llamafactory/model/utils/longlora.py

@@ -0,0 +1,323 @@
+import math
+from typing import TYPE_CHECKING, Optional, Tuple
+
+import torch
+import torch.nn as nn
+from transformers.models.llama.modeling_llama import (
+    Cache,
+    LlamaAttention,
+    LlamaFlashAttention2,
+    LlamaSdpaAttention,
+    apply_rotary_pos_emb,
+    repeat_kv,
+)
+from transformers.utils import logging
+from transformers.utils.versions import require_version
+
+from ...extras.constants import SUPPORTED_CLASS_FOR_S2ATTN
+from ...extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig
+
+    from ...hparams import ModelArguments
+
+
+logger = logging.get_logger(__name__)
+
+
+# Modified from:
+# https://github.com/huggingface/transformers/blob/v4.40.0/src/transformers/models/llama/modeling_llama.py
+def llama_attention_forward(
+    self: "LlamaAttention",
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.LongTensor] = None,
+    past_key_value: Optional["Cache"] = None,
+    output_attentions: bool = False,
+    cache_position: Optional[torch.LongTensor] = None,
+    **kwargs,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    bsz, q_len, _ = hidden_states.size()
+
+    query_states: "torch.Tensor" = self.q_proj(hidden_states)
+    key_states: "torch.Tensor" = self.k_proj(hidden_states)
+    value_states: "torch.Tensor" = self.v_proj(hidden_states)
+
+    query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+    key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+    value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+    cos, sin = self.rotary_emb(value_states, position_ids)
+    query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+    past_key_value = getattr(self, "past_key_value", past_key_value)
+
+    if past_key_value is not None:
+        cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+        key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+    if getattr(self.config, "group_size_ratio", None) and self.training:  # shift
+        groupsz = int(q_len * getattr(self.config, "group_size_ratio"))
+        assert q_len % groupsz == 0, "q_len {} should be divisible by group size {}.".format(q_len, groupsz)
+        num_groups = q_len // groupsz
+
+        def shift(state: torch.Tensor) -> torch.Tensor:
+            state = state.transpose(1, 2)  # output: (bsz, seq_len, n_heads, head_dim)
+            state = torch.cat(
+                (state[:, :, : self.num_heads // 2], state[:, :, self.num_heads // 2 :].roll(-groupsz // 2, dims=1)),
+                dim=2,
+            )
+            return state.reshape(bsz * num_groups, groupsz, self.num_heads, self.head_dim).transpose(1, 2)
+
+        query_states, key_states, value_states = shift(query_states), shift(key_states), shift(value_states)
+        if attention_mask is not None:
+            attention_mask = attention_mask[:, :, :groupsz, :groupsz].repeat(num_groups, 1, 1, 1)
+
+    attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+
+    if attention_mask is not None:  # no matter the length, we just slice it
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    # upcast attention to fp32
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+    attn_output = torch.matmul(attn_weights, value_states)  # (bsz, :, seq_len, :) or (bsz * n_group, :, groupsz, :)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    if getattr(self.config, "group_size_ratio", None) and self.training:  # shift back
+        attn_output.reshape(bsz, q_len, self.num_heads, self.head_dim)
+        attn_output = torch.cat(
+            (
+                attn_output[:, :, : self.num_heads // 2],
+                attn_output[:, :, self.num_heads // 2 :].roll(groupsz // 2, dims=1),
+            )
+        )
+
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+    attn_output = self.o_proj(attn_output)
+
+    if not output_attentions:
+        attn_weights = None
+
+    return attn_output, attn_weights, past_key_value
+
+
+# Modified from:
+# https://github.com/huggingface/transformers/blob/v4.40.0/src/transformers/models/llama/modeling_llama.py
+def llama_flash_attention_2_forward(
+    self: "LlamaFlashAttention2",
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.LongTensor] = None,
+    past_key_value: Optional["Cache"] = None,
+    output_attentions: bool = False,
+    cache_position: Optional[torch.LongTensor] = None,
+    **kwargs,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    # LlamaFlashAttention2 attention does not support output_attentions
+    output_attentions = False
+
+    bsz, q_len, _ = hidden_states.size()
+
+    query_states: "torch.Tensor" = self.q_proj(hidden_states)
+    key_states: "torch.Tensor" = self.k_proj(hidden_states)
+    value_states: "torch.Tensor" = self.v_proj(hidden_states)
+
+    query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+    key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+    value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+    cos, sin = self.rotary_emb(value_states, position_ids)
+    query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+    past_key_value = getattr(self, "past_key_value", past_key_value)
+
+    if past_key_value is not None:
+        cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+        key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+    # FlashAttention requires the input to have the shape (bsz, seq_len, n_heads, head_dim)
+    query_states = query_states.transpose(1, 2)
+    key_states = key_states.transpose(1, 2)
+    value_states = value_states.transpose(1, 2)
+
+    dropout_rate = self.attention_dropout if self.training else 0.0
+
+    input_dtype = query_states.dtype
+    if input_dtype == torch.float32:
+        if torch.is_autocast_enabled():
+            target_dtype = torch.get_autocast_gpu_dtype()
+        elif hasattr(self.config, "_pre_quantization_dtype"):
+            target_dtype = self.config._pre_quantization_dtype
+        else:
+            target_dtype = self.q_proj.weight.dtype
+
+        logger.warning_once("The input hidden states seems to be silently casted in float32.")
+        query_states = query_states.to(target_dtype)
+        key_states = key_states.to(target_dtype)
+        value_states = value_states.to(target_dtype)
+
+    if getattr(self.config, "group_size_ratio", None) and self.training:  # shift
+        groupsz = int(q_len * getattr(self.config, "group_size_ratio"))
+        assert q_len % groupsz == 0, "q_len {} should be divisible by group size {}.".format(q_len, groupsz)
+        num_groups = q_len // groupsz
+
+        def shift(state: torch.Tensor) -> torch.Tensor:
+            state = torch.cat(
+                (state[:, :, : self.num_heads // 2], state[:, :, self.num_heads // 2 :].roll(-groupsz // 2, dims=1)),
+                dim=2,
+            )
+            return state.reshape(bsz * num_groups, groupsz, self.num_heads, self.head_dim)
+
+        query_states, key_states, value_states = shift(query_states), shift(key_states), shift(value_states)
+        if attention_mask is not None:
+            attention_mask = attention_mask[:, :groupsz].repeat(num_groups, 1)
+    else:
+        groupsz = q_len
+
+    attn_output: torch.Tensor = self._flash_attention_forward(
+        query_states, key_states, value_states, attention_mask, groupsz, dropout=dropout_rate
+    )
+
+    if getattr(self.config, "group_size_ratio", None) and self.training:  # shift back
+        attn_output.reshape(bsz, q_len, self.num_heads, self.head_dim)
+        attn_output = torch.cat(
+            (
+                attn_output[:, :, : self.num_heads // 2],
+                attn_output[:, :, self.num_heads // 2 :].roll(groupsz // 2, dims=1),
+            )
+        )
+
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+    attn_output = self.o_proj(attn_output)
+
+    if not output_attentions:
+        attn_weights = None
+
+    return attn_output, attn_weights, past_key_value
+
+
+# Modified from:
+# https://github.com/huggingface/transformers/blob/v4.40.0/src/transformers/models/llama/modeling_llama.py
+def llama_sdpa_attention_forward(
+    self: "LlamaSdpaAttention",
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.LongTensor] = None,
+    past_key_value: Optional["Cache"] = None,
+    output_attentions: bool = False,
+    cache_position: Optional[torch.LongTensor] = None,
+    **kwargs,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    if output_attentions:
+        logger.warning_once("SDPA does not support `output_attentions=True`. Falling back to the vanilla attention")
+        return llama_attention_forward(
+            self,
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+    bsz, q_len, _ = hidden_states.size()
+
+    query_states: "torch.Tensor" = self.q_proj(hidden_states)
+    key_states: "torch.Tensor" = self.k_proj(hidden_states)
+    value_states: "torch.Tensor" = self.v_proj(hidden_states)
+
+    query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+    key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+    value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+    cos, sin = self.rotary_emb(value_states, position_ids)
+    query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+    if past_key_value is not None:
+        cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+        key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+    if getattr(self.config, "group_size_ratio", None) and self.training:  # shift
+        groupsz = int(q_len * getattr(self.config, "group_size_ratio"))
+        assert q_len % groupsz == 0, "q_len {} should be divisible by group size {}.".format(q_len, groupsz)
+        num_groups = q_len // groupsz
+
+        def shift(state: torch.Tensor) -> torch.Tensor:
+            state = state.transpose(1, 2)  # output: (bsz, seq_len, n_heads, head_dim)
+            state = torch.cat(
+                (state[:, :, : self.num_heads // 2], state[:, :, self.num_heads // 2 :].roll(-groupsz // 2, dims=1)),
+                dim=2,
+            )
+            return state.reshape(bsz * num_groups, groupsz, self.num_heads, self.head_dim).transpose(1, 2)
+
+        query_states, key_states, value_states = shift(query_states), shift(key_states), shift(value_states)
+        if attention_mask is not None:
+            attention_mask = attention_mask[:, :, :groupsz, :groupsz].repeat(num_groups, 1, 1, 1)
+
+    causal_mask = attention_mask
+    if attention_mask is not None:
+        causal_mask = causal_mask[:, :, :, : key_states.shape[-2]]
+
+    if query_states.device.type == "cuda" and causal_mask is not None:
+        query_states = query_states.contiguous()
+        key_states = key_states.contiguous()
+        value_states = value_states.contiguous()
+
+    attn_output = torch.nn.functional.scaled_dot_product_attention(
+        query_states,
+        key_states,
+        value_states,
+        attn_mask=causal_mask,
+        dropout_p=self.attention_dropout if self.training else 0.0,
+        is_causal=causal_mask is None and q_len > 1,
+    )
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    if getattr(self.config, "group_size_ratio", None) and self.training:  # shift back
+        attn_output.reshape(bsz, q_len, self.num_heads, self.head_dim)
+        attn_output = torch.cat(
+            (
+                attn_output[:, :, : self.num_heads // 2],
+                attn_output[:, :, self.num_heads // 2 :].roll(groupsz // 2, dims=1),
+            )
+        )
+
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+    attn_output = self.o_proj(attn_output)
+
+    return attn_output, None, past_key_value
+
+
+def _apply_llama_patch() -> None:
+    require_version("transformers==4.40.2", "To fix: pip install transformers==4.40.2")
+    LlamaAttention.forward = llama_attention_forward
+    LlamaFlashAttention2.forward = llama_flash_attention_2_forward
+    LlamaSdpaAttention.forward = llama_sdpa_attention_forward
+
+
+def configure_longlora(config: "PretrainedConfig", model_args: "ModelArguments", is_trainable: bool) -> None:
+    if not is_trainable or not model_args.shift_attn:
+        return
+
+    logger = get_logger(__name__)
+
+    if getattr(config, "model_type", None) in SUPPORTED_CLASS_FOR_S2ATTN:
+        setattr(config, "group_size_ratio", 0.25)
+        _apply_llama_patch()
+        logger.info("Using shift short attention with group_size_ratio=1/4.")
+    else:
+        logger.warning("Current model does not support shift short attention.")

src/llamafactory/model/utils/misc.py

@@ -0,0 +1,78 @@
+from typing import TYPE_CHECKING, List
+
+import torch
+
+from ...extras.logging import get_logger
+from .quantization import QuantizationMethod
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedModel, PreTrainedTokenizer
+
+
+logger = get_logger(__name__)
+
+
+def find_all_linear_modules(model: "PreTrainedModel") -> List[str]:
+    r"""
+    Finds all available modules to apply lora or galore.
+    """
+    quantization_method = getattr(model, "quantization_method", None)
+    if quantization_method is None:
+        linear_cls = torch.nn.Linear
+    elif quantization_method == QuantizationMethod.BITS_AND_BYTES:
+        import bitsandbytes as bnb
+
+        linear_cls = bnb.nn.Linear4bit if getattr(model, "is_loaded_in_4bit", False) else bnb.nn.Linear8bitLt
+    else:
+        raise ValueError("Finding linear modules for {} models is not supported.".format(quantization_method))
+
+    output_layer_names = ["lm_head"]
+    if model.config.model_type == "chatglm":
+        output_layer_names.append("output_layer")
+    elif model.config.model_type == "internlm2":
+        output_layer_names.append("output")
+
+    module_names = set()
+    for name, module in model.named_modules():
+        if isinstance(module, linear_cls) and not any(output_layer in name for output_layer in output_layer_names):
+            module_names.add(name.split(".")[-1])
+
+    logger.info("Found linear modules: {}".format(",".join(module_names)))
+    return list(module_names)
+
+
+def find_expanded_modules(model: "PreTrainedModel", target_modules: List[str], num_layer_trainable: int) -> List[str]:
+    r"""
+    Finds the modules in the expanded blocks to apply lora.
+    """
+    num_layers = getattr(model.config, "num_hidden_layers", None)
+    if not num_layers:
+        raise ValueError("Model was not supported.")
+
+    if num_layers % num_layer_trainable != 0:
+        raise ValueError(
+            "`num_layers` {} should be divisible by `num_layer_trainable` {}.".format(num_layers, num_layer_trainable)
+        )
+
+    stride = num_layers // num_layer_trainable
+    trainable_layer_ids = range(stride - 1, num_layers + stride - 1, stride)
+    trainable_layers = [".{:d}.".format(idx) for idx in trainable_layer_ids]
+    module_names = []
+    for name, _ in model.named_modules():
+        if any(target_module in name for target_module in target_modules) and any(
+            trainable_layer in name for trainable_layer in trainable_layers
+        ):
+            module_names.append(name)
+
+    logger.info("Apply lora to layers: {}".format(",".join(map(str, trainable_layer_ids))))
+    return module_names
+
+
+def register_autoclass(config: "PretrainedConfig", model: "PreTrainedModel", tokenizer: "PreTrainedTokenizer"):
+    if "AutoConfig" in getattr(config, "auto_map", {}):
+        config.__class__.register_for_auto_class()
+    if "AutoModelForCausalLM" in getattr(config, "auto_map", {}):
+        model.__class__.register_for_auto_class()
+    if "AutoTokenizer" in tokenizer.init_kwargs.get("auto_map", {}):
+        tokenizer.__class__.register_for_auto_class()

src/llamafactory/model/utils/mod.py

@@ -0,0 +1,28 @@
+from typing import TYPE_CHECKING
+
+from ...extras.constants import MOD_SUPPORTED_MODELS
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedModel
+
+    from ...hparams import ModelArguments
+
+
+def load_mod_pretrained_model(**init_kwargs) -> "PreTrainedModel":
+    from MoD import AutoMoDModelForCausalLM
+
+    return AutoMoDModelForCausalLM.from_pretrained(**init_kwargs)
+
+
+def convert_pretrained_model_to_mod(
+    model: "PreTrainedModel", config: "PretrainedConfig", model_args: "ModelArguments"
+) -> "PreTrainedModel":
+    from MoD import apply_mod_to_hf
+
+    if getattr(config, "model_type", None) not in MOD_SUPPORTED_MODELS:
+        raise ValueError("Current model is not supported by mixture-of-depth.")
+
+    model = apply_mod_to_hf(model)
+    model = model.to(model_args.compute_dtype)
+    return model

src/llamafactory/model/utils/moe.py

@@ -0,0 +1,53 @@
+from typing import TYPE_CHECKING
+
+from transformers.integrations import is_deepspeed_zero3_enabled
+from transformers.utils.versions import require_version
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedModel
+
+    from ...hparams import ModelArguments
+
+
+def add_z3_leaf_module(model: "PreTrainedModel") -> None:
+    r"""
+    Sets module as a leaf module to skip partitioning in deepspeed zero3.
+    """
+    if not is_deepspeed_zero3_enabled():
+        return
+
+    require_version("deepspeed>=0.13.0", "To fix: pip install deepspeed>=0.13.0")
+    from deepspeed.utils import set_z3_leaf_modules  # type: ignore
+
+    if getattr(model.config, "model_type", None) == "mixtral":
+        from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
+
+        set_z3_leaf_modules(model, [MixtralSparseMoeBlock])
+
+    if getattr(model.config, "model_type", None) == "qwen2moe":
+        from transformers.models.qwen2_moe.modeling_qwen2_moe import Qwen2MoeSparseMoeBlock
+
+        set_z3_leaf_modules(model, [Qwen2MoeSparseMoeBlock])
+
+    if getattr(model.config, "model_type", None) == "jamba":
+        from transformers.models.jamba.modeling_jamba import JambaSparseMoeBlock
+
+        set_z3_leaf_modules(model, [JambaSparseMoeBlock])
+
+    if getattr(model.config, "model_type", None) == "dbrx":
+        from transformers.models.dbrx.modeling_dbrx import DbrxFFN
+
+        set_z3_leaf_modules(model, [DbrxFFN])
+
+
+def configure_moe(config: "PretrainedConfig", model_args: "ModelArguments", is_trainable: bool) -> None:
+    if model_args.moe_aux_loss_coef is not None:
+        if getattr(config, "model_type", None) in ["jamba", "mixtral", "qwen2_moe"]:
+            setattr(config, "router_aux_loss_coef", model_args.moe_aux_loss_coef)
+
+        elif getattr(config, "model_type", None) == "deepseek":
+            setattr(config, "aux_loss_alpha", model_args.moe_aux_loss_coef)
+
+    if getattr(config, "model_type", None) in ["dbrx", "jamba", "mixtral", "qwen2_moe"]:
+        setattr(config, "output_router_logits", is_trainable)

src/llamafactory/model/utils/quantization.py

@@ -0,0 +1,147 @@
+import os
+import random
+from enum import Enum, unique
+from typing import TYPE_CHECKING, Any, Dict, List
+
+import torch
+from datasets import load_dataset
+from transformers import BitsAndBytesConfig, GPTQConfig
+from transformers.integrations import is_deepspeed_zero3_enabled
+from transformers.modeling_utils import is_fsdp_enabled
+from transformers.utils.versions import require_version
+
+from ...extras.constants import FILEEXT2TYPE
+from ...extras.logging import get_logger
+from ...extras.misc import get_current_device
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedTokenizer
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+@unique
+class QuantizationMethod(str, Enum):
+    r"""
+    Borrowed from `transformers.utils.quantization_config.QuantizationMethod`.
+    """
+
+    BITS_AND_BYTES = "bitsandbytes"
+    GPTQ = "gptq"
+    AWQ = "awq"
+    AQLM = "aqlm"
+    QUANTO = "quanto"
+
+
+def _get_quantization_dataset(tokenizer: "PreTrainedTokenizer", model_args: "ModelArguments") -> List[str]:
+    r"""
+    Inspired by: https://github.com/huggingface/optimum/blob/v1.16.0/optimum/gptq/data.py#L133
+    TODO: remove tokenizer.decode() https://github.com/huggingface/optimum/pull/1600
+    """
+    if os.path.isfile(model_args.export_quantization_dataset):
+        data_path = FILEEXT2TYPE.get(model_args.export_quantization_dataset.split(".")[-1], None)
+        data_files = model_args.export_quantization_dataset
+    else:
+        data_path = model_args.export_quantization_dataset
+        data_files = None
+
+    dataset = load_dataset(path=data_path, data_files=data_files, split="train", cache_dir=model_args.cache_dir)
+    maxlen = model_args.export_quantization_maxlen
+
+    samples = []
+    for _ in range(model_args.export_quantization_nsamples):
+        while True:
+            sample_idx = random.randint(0, len(dataset) - 1)
+            sample: Dict[str, torch.Tensor] = tokenizer(dataset[sample_idx]["text"], return_tensors="pt")
+            if sample["input_ids"].size(1) >= maxlen:
+                break  # TODO: fix large maxlen
+
+        word_idx = random.randint(0, sample["input_ids"].size(1) - maxlen - 1)
+        input_ids = sample["input_ids"][:, word_idx : word_idx + maxlen]
+        samples.append(tokenizer.decode(input_ids[0].tolist(), skip_special_tokens=True))
+
+    return samples
+
+
+def configure_quantization(
+    config: "PretrainedConfig",
+    tokenizer: "PreTrainedTokenizer",
+    model_args: "ModelArguments",
+    init_kwargs: Dict[str, Any],
+) -> None:
+    r"""
+    Priority: PTQ-quantized (training) > AutoGPTQ (export) > Bitsandbytes (training)
+    """
+    if getattr(config, "quantization_config", None):  # ptq
+        if is_deepspeed_zero3_enabled():
+            raise ValueError("DeepSpeed ZeRO-3 is incompatible with quantized models.")
+
+        if model_args.quantization_device_map != "auto":
+            init_kwargs["device_map"] = {"": get_current_device()}
+
+        quantization_config: Dict[str, Any] = getattr(config, "quantization_config", None)
+        quant_method = quantization_config.get("quant_method", "")
+
+        if quant_method == QuantizationMethod.GPTQ:
+            require_version("auto_gptq>=0.5.0", "To fix: pip install auto_gptq>=0.5.0")
+            quantization_config.pop("disable_exllama", None)  # remove deprecated args
+            quantization_config["use_exllama"] = False  # disable exllama
+
+        if quant_method == QuantizationMethod.AWQ:
+            require_version("autoawq", "To fix: pip install autoawq")
+
+        if quant_method == QuantizationMethod.AQLM:
+            require_version("transformers>=4.39.0", "To fix: pip install transformers>=4.39.0")
+            require_version("aqlm>=1.1.0", "To fix: pip install aqlm[gpu]>=1.1.0")
+            quantization_config["bits"] = 2
+
+        quant_bits = quantization_config.get("bits", "?")
+        logger.info("Loading {}-bit {}-quantized model.".format(quant_bits, quant_method.upper()))
+
+    elif model_args.export_quantization_bit is not None:  # auto-gptq
+        require_version("optimum>=1.16.0", "To fix: pip install optimum>=1.16.0")
+        require_version("auto_gptq>=0.5.0", "To fix: pip install auto_gptq>=0.5.0")
+        from accelerate.utils import get_max_memory
+
+        if getattr(config, "model_type", None) == "chatglm":
+            raise ValueError("ChatGLM model is not supported.")
+
+        init_kwargs["quantization_config"] = GPTQConfig(
+            bits=model_args.export_quantization_bit,
+            tokenizer=tokenizer,
+            dataset=_get_quantization_dataset(tokenizer, model_args),
+        )
+        init_kwargs["device_map"] = "auto"
+        init_kwargs["max_memory"] = get_max_memory()
+        logger.info("Quantizing model to {} bit.".format(model_args.export_quantization_bit))
+
+    elif model_args.quantization_bit is not None:  # bnb
+        if model_args.quantization_bit == 8:
+            require_version("bitsandbytes>=0.37.0", "To fix: pip install bitsandbytes>=0.37.0")
+            init_kwargs["quantization_config"] = BitsAndBytesConfig(load_in_8bit=True)
+
+        elif model_args.quantization_bit == 4:
+            require_version("bitsandbytes>=0.39.0", "To fix: pip install bitsandbytes>=0.39.0")
+            init_kwargs["quantization_config"] = BitsAndBytesConfig(
+                load_in_4bit=True,
+                bnb_4bit_compute_dtype=model_args.compute_dtype,
+                bnb_4bit_use_double_quant=model_args.double_quantization,
+                bnb_4bit_quant_type=model_args.quantization_type,
+                bnb_4bit_quant_storage=model_args.compute_dtype,  # crucial for fsdp qlora
+            )
+
+        if is_deepspeed_zero3_enabled() or is_fsdp_enabled() or model_args.quantization_device_map == "auto":
+            if model_args.quantization_bit != 4:
+                raise ValueError("Only 4-bit quantized model can use auto device map.")
+
+            require_version("transformers>=4.39.0", "To fix: pip install transformers>=4.39.0")
+            require_version("accelerate>=0.28.0", "To fix: pip install accelerate>=0.28.0")
+            require_version("bitsandbytes>=0.43.0", "To fix: pip install bitsandbytes>=0.43.0")
+        else:
+            init_kwargs["device_map"] = {"": get_current_device()}
+
+        logger.info("Quantizing model to {} bit.".format(model_args.quantization_bit))

src/llamafactory/model/utils/rope.py

@@ -0,0 +1,47 @@
+import math
+from typing import TYPE_CHECKING
+
+from ...extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def configure_rope(config: "PretrainedConfig", model_args: "ModelArguments", is_trainable: bool) -> None:
+    if model_args.rope_scaling is None:
+        return
+
+    if not hasattr(config, "rope_scaling"):
+        logger.warning("Current model does not support RoPE scaling.")
+        return
+
+    if is_trainable:
+        if model_args.rope_scaling == "dynamic":
+            logger.warning(
+                "Dynamic NTK scaling may not work well with fine-tuning. "
+                "See: https://github.com/huggingface/transformers/pull/24653"
+            )
+
+        current_max_length = getattr(config, "max_position_embeddings", None)
+        if current_max_length and model_args.model_max_length > current_max_length:
+            logger.info(
+                "Enlarge max model length from {} to {}.".format(current_max_length, model_args.model_max_length)
+            )
+            setattr(config, "max_position_embeddings", model_args.model_max_length)
+            scaling_factor = float(math.ceil(model_args.model_max_length / current_max_length))
+        else:
+            logger.warning("Input length is smaller than max length. Consider increase input length.")
+            scaling_factor = 1.0
+    else:
+        scaling_factor = 2.0
+
+    setattr(config, "rope_scaling", {"type": model_args.rope_scaling, "factor": scaling_factor})
+    logger.info(
+        "Using {} scaling strategy and setting scaling factor to {}".format(model_args.rope_scaling, scaling_factor)
+    )

src/llamafactory/model/utils/unsloth.py

@@ -0,0 +1,88 @@
+from typing import TYPE_CHECKING, Any, Dict, Optional
+
+from ...extras.logging import get_logger
+from ...extras.misc import get_current_device
+
+
+if TYPE_CHECKING:
+    from transformers import PretrainedConfig, PreTrainedModel
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def _get_unsloth_kwargs(
+    config: "PretrainedConfig", model_name_or_path: str, model_args: "ModelArguments"
+) -> Dict[str, Any]:
+    return {
+        "model_name": model_name_or_path,
+        "max_seq_length": model_args.model_max_length or 4096,
+        "dtype": model_args.compute_dtype,
+        "load_in_4bit": model_args.quantization_bit == 4,
+        "token": model_args.hf_hub_token,
+        "device_map": {"": get_current_device()},
+        "rope_scaling": getattr(config, "rope_scaling", None),
+        "fix_tokenizer": False,
+        "trust_remote_code": True,
+        "use_gradient_checkpointing": "unsloth",
+    }
+
+
+def load_unsloth_pretrained_model(
+    config: "PretrainedConfig", model_args: "ModelArguments"
+) -> Optional["PreTrainedModel"]:
+    r"""
+    Optionally loads pretrained model with unsloth. Used in training.
+    """
+    from unsloth import FastLanguageModel
+
+    unsloth_kwargs = _get_unsloth_kwargs(config, model_args.model_name_or_path, model_args)
+    try:
+        model, _ = FastLanguageModel.from_pretrained(**unsloth_kwargs)
+    except NotImplementedError:
+        logger.warning("Unsloth does not support model type {}.".format(getattr(config, "model_type", None)))
+        model = None
+        model_args.use_unsloth = False
+
+    return model
+
+
+def get_unsloth_peft_model(
+    model: "PreTrainedModel", model_args: "ModelArguments", peft_kwargs: Dict[str, Any]
+) -> "PreTrainedModel":
+    r"""
+    Gets the peft model for the pretrained model with unsloth. Used in training.
+    """
+    from unsloth import FastLanguageModel
+
+    unsloth_peft_kwargs = {
+        "model": model,
+        "max_seq_length": model_args.model_max_length,
+        "use_gradient_checkpointing": "unsloth",
+    }
+    return FastLanguageModel.get_peft_model(**peft_kwargs, **unsloth_peft_kwargs)
+
+
+def load_unsloth_peft_model(
+    config: "PretrainedConfig", model_args: "ModelArguments", is_trainable: bool
+) -> "PreTrainedModel":
+    r"""
+    Loads peft model with unsloth. Used in both training and inference.
+    """
+    from unsloth import FastLanguageModel
+
+    unsloth_kwargs = _get_unsloth_kwargs(config, model_args.adapter_name_or_path[0], model_args)
+    try:
+        if not is_trainable:
+            unsloth_kwargs["use_gradient_checkpointing"] = False
+
+        model, _ = FastLanguageModel.from_pretrained(**unsloth_kwargs)
+    except NotImplementedError:
+        raise ValueError("Unsloth does not support model type {}.".format(getattr(config, "model_type", None)))
+
+    if not is_trainable:
+        FastLanguageModel.for_inference(model)
+
+    return model

src/llamafactory/model/utils/valuehead.py

@@ -0,0 +1,58 @@
+from typing import TYPE_CHECKING, Dict
+
+import torch
+from transformers.utils import cached_file
+
+from ...extras.constants import V_HEAD_SAFE_WEIGHTS_NAME, V_HEAD_WEIGHTS_NAME
+from ...extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedModel
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+def load_valuehead_params(path_or_repo_id: str, model_args: "ModelArguments") -> Dict[str, torch.Tensor]:
+    r"""
+    Loads value head parameters from Hugging Face Hub or local disk.
+
+    Returns: dict with keys `v_head.summary.weight` and `v_head.summary.bias`.
+    """
+    kwargs = {"path_or_repo_id": path_or_repo_id, "cache_dir": model_args.cache_dir, "token": model_args.hf_hub_token}
+
+    try:
+        from safetensors import safe_open
+
+        vhead_file = cached_file(filename=V_HEAD_SAFE_WEIGHTS_NAME, **kwargs)
+        with safe_open(vhead_file, framework="pt", device="cpu") as f:
+            return {key: f.get_tensor(key) for key in f.keys()}
+    except Exception as err:
+        logger.info("Failed to load {}: {}".format(V_HEAD_SAFE_WEIGHTS_NAME, str(err)))
+
+    try:
+        vhead_file = cached_file(filename=V_HEAD_WEIGHTS_NAME, **kwargs)
+        return torch.load(vhead_file, map_location="cpu")
+    except Exception as err:
+        logger.info("Failed to load {}: {}".format(V_HEAD_WEIGHTS_NAME, str(err)))
+
+    logger.info("Provided path ({}) does not contain value head weights.".format(path_or_repo_id))
+    logger.info("Ignore these messages if you are not resuming the training of a value head model.")
+    return None
+
+
+def prepare_valuehead_model(model: "PreTrainedModel") -> None:
+    if getattr(model.config, "model_type", None) == "llava":
+        setattr(model, "lm_head", model.language_model.get_output_embeddings())
+        setattr(model, "_keys_to_ignore_on_save", ["lm_head.weight"])
+
+    if getattr(model.config, "model_type", None) == "chatglm":
+        setattr(model, "lm_head", model.transformer.output_layer)
+        setattr(model, "_keys_to_ignore_on_save", ["lm_head.weight"])
+
+    if getattr(model.config, "model_type", None) == "internlm2":
+        setattr(model, "lm_head", model.output)
+        setattr(model, "_keys_to_ignore_on_save", ["lm_head.weight"])

src/llamafactory/model/utils/visual.py

@@ -0,0 +1,84 @@
+from typing import TYPE_CHECKING, Tuple
+
+import torch
+import transformers.models
+from transformers.activations import ACT2FN
+
+from ...extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from transformers import LlavaConfig, PretrainedConfig, PreTrainedModel
+
+    from ...hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+class LlavaMultiModalProjectorForYiVL(torch.nn.Module):
+    def __init__(self, config: "LlavaConfig") -> None:
+        super().__init__()
+
+        self.config = config
+        if config is None:
+            return
+
+        self.linear_1 = torch.nn.Linear(config.vision_config.hidden_size, config.text_config.hidden_size, bias=True)
+        self.linear_2 = torch.nn.LayerNorm(config.text_config.hidden_size, bias=True)
+        self.linear_3 = torch.nn.Linear(config.text_config.hidden_size, config.text_config.hidden_size, bias=True)
+        self.linear_4 = torch.nn.LayerNorm(config.text_config.hidden_size, bias=True)
+        self.act = ACT2FN[config.projector_hidden_act]
+
+    def forward(self, image_features: "torch.Tensor") -> "torch.Tensor":
+        hidden_states = self.linear_1(image_features)
+        hidden_states = self.linear_2(hidden_states)
+        hidden_states = self.act(hidden_states)
+        hidden_states = self.linear_3(hidden_states)
+        hidden_states = self.linear_4(hidden_states)
+        if hidden_states.dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            elif hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.linear_1.weight.dtype
+
+            logger.warning_once("The hidden states seems to be silently casted in float32.")
+            hidden_states = hidden_states.to(target_dtype)
+
+        return hidden_states
+
+
+class LlavaMultiModalProjectorForYiVLForVLLM(LlavaMultiModalProjectorForYiVL):
+    def __init__(self, vision_hidden_size: int, text_hidden_size: int, projector_hidden_act: str) -> None:
+        super().__init__(config=None)
+
+        self.linear_1 = torch.nn.Linear(vision_hidden_size, text_hidden_size, bias=True)
+        self.linear_2 = torch.nn.LayerNorm(text_hidden_size, bias=True)
+        self.linear_3 = torch.nn.Linear(text_hidden_size, text_hidden_size, bias=True)
+        self.linear_4 = torch.nn.LayerNorm(text_hidden_size, bias=True)
+        self.act = ACT2FN[projector_hidden_act]
+
+
+def autocast_projector_dtype(
+    model: "PreTrainedModel", model_args: "ModelArguments", mm_projector_name: str = "multi_modal_projector"
+) -> None:
+    def _mm_projector_forward_post_hook(
+        module: "torch.nn.Module", args: Tuple["torch.Tensor"], output: "torch.Tensor"
+    ) -> "torch.Tensor":
+        return output.to(model_args.compute_dtype)
+
+    if hasattr(model, mm_projector_name) and getattr(model.config, "quantization_method", None):
+        logger.info("Casting multimodal projector outputs in {}.".format(model_args.compute_dtype))
+        mm_projector: "torch.nn.Module" = getattr(model, mm_projector_name)
+        mm_projector.register_forward_hook(_mm_projector_forward_post_hook)
+
+
+def configure_visual_model(config: "PretrainedConfig") -> None:
+    if getattr(config, "model_type", None) == "llava":  # required for ds zero3 and valuehead models
+        setattr(config, "hidden_size", getattr(config.text_config, "hidden_size", None))
+
+    if getattr(config, "is_yi_vl_derived_model", None):
+        logger.info("Detected Yi-VL model, applying projector patch.")
+        transformers.models.llava.modeling_llava.LlavaMultiModalProjector = LlavaMultiModalProjectorForYiVL