[model] Add Ministral3 (#9582)

Co-authored-by: kingsley <kingsleydodonow@gmail.com>

src/llamafactory/model/model_utils/quantization.py

@@ -22,7 +22,7 @@ from typing import TYPE_CHECKING, Any
 
 import torch
 from datasets import load_dataset
-from transformers import BitsAndBytesConfig, EetqConfig, GPTQConfig, HqqConfig
+from transformers import BitsAndBytesConfig, EetqConfig, FineGrainedFP8Config, GPTQConfig, HqqConfig
 from transformers.integrations import is_deepspeed_zero3_enabled
 from transformers.modeling_utils import is_fsdp_enabled
 
@@ -83,6 +83,7 @@ def configure_quantization(
     config: "PretrainedConfig",
     tokenizer: "PreTrainedTokenizer",
     model_args: "ModelArguments",
+    is_trainable: bool,
     init_kwargs: dict[str, Any],
 ) -> None:
     r"""Priority: PTQ-quantized (train/infer) > AutoGPTQ (export) > On-the-fly quantization (train/infer)."""
@@ -109,6 +110,10 @@ def configure_quantization(
             check_version("aqlm>=1.1.0", mandatory=True)
             quantization_config["bits"] = 2
 
+        if quant_method == QuantizationMethod.FP8 and is_trainable:
+            quant_config = FineGrainedFP8Config(dequantize=True)
+            init_kwargs["quantization_config"] = quant_config
+
         quant_bits = quantization_config.get("bits", "?")
         logger.info_rank0(f"Loading {quant_bits}-bit {quant_method.upper()}-quantized model.")