From 87ab7fc01cdb257448bf3b36e39f2fe362f0fa2f Mon Sep 17 00:00:00 2001
From: BUAADreamer <1428195643@qq.com>
Date: Sun, 29 Sep 2024 22:00:01 +0800
Subject: [PATCH] fix constants

Former-commit-id: 485fc047169afd027ee65d05e3c5c08b371b6c4d
---
 src/llamafactory/extras/constants.py         |  4 ++--
 src/llamafactory/model/model_utils/misc.py   |  2 +-
 src/llamafactory/model/model_utils/visual.py | 16 ++++++++++------
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/src/llamafactory/extras/constants.py b/src/llamafactory/extras/constants.py
index 7dce012e..1be3940e 100644
--- a/src/llamafactory/extras/constants.py
+++ b/src/llamafactory/extras/constants.py
@@ -896,7 +896,7 @@ register_model_group(
         "LLaVA-NeXT-Video-7B-32k-Chat": {
             DownloadSource.DEFAULT: "llava-hf/LLaVA-NeXT-Video-7B-32K-hf",
         },
-        "LLaVA-NeXT-Video-7B-DPO": {
+        "LLaVA-NeXT-Video-7B-DPO-Chat": {
             DownloadSource.DEFAULT: "llava-hf/LLaVA-NeXT-Video-7B-DPO-hf",
         },
     },
@@ -910,7 +910,7 @@ register_model_group(
         "LLaVA-NeXT-Video-Yi-34B-Chat": {
             DownloadSource.DEFAULT: "llava-hf/LLaVA-NeXT-Video-34B-hf",
         },
-        "LLaVA-NeXT-Video-Yi-34B-DPO": {
+        "LLaVA-NeXT-Video-Yi-34B-DPO-Chat": {
             DownloadSource.DEFAULT: "llava-hf/LLaVA-NeXT-Video-34B-DPO-hf",
         },
     },
diff --git a/src/llamafactory/model/model_utils/misc.py b/src/llamafactory/model/model_utils/misc.py
index 12eafcac..4883fa23 100644
--- a/src/llamafactory/model/model_utils/misc.py
+++ b/src/llamafactory/model/model_utils/misc.py
@@ -34,7 +34,7 @@ def find_all_linear_modules(model: "PreTrainedModel", freeze_vision_tower: bool)
         forbidden_modules.add("output_layer")
     elif model_type == "internlm2":
         forbidden_modules.add("output")
-    elif model_type in ["llava", "paligemma", "video_llava"] or "llava_next" in model_type:
+    elif model_type in ["llava", "llava_next", "llava_next_video", "paligemma", "video_llava"]:
         forbidden_modules.add("multi_modal_projector")
     elif model_type == "qwen2_vl":
         forbidden_modules.add("merger")
diff --git a/src/llamafactory/model/model_utils/visual.py b/src/llamafactory/model/model_utils/visual.py
index 85f386de..c2aeb0dd 100644
--- a/src/llamafactory/model/model_utils/visual.py
+++ b/src/llamafactory/model/model_utils/visual.py
@@ -92,7 +92,7 @@ def autocast_projector_dtype(model: "PreTrainedModel", model_args: "ModelArgumen
 
     if getattr(model, "quantization_method", None):
         model_type = getattr(model.config, "model_type", None)
-        if model_type in ["llava", "paligemma", "video_llava"] or "llava_next" in model_type:
+        if model_type in ["llava", "llava_next", "llava_next_video", "paligemma", "video_llava"]:
             mm_projector: "torch.nn.Module" = getattr(model, "multi_modal_projector")
         elif model_type == "qwen2_vl":
             mm_projector: "torch.nn.Module" = getattr(getattr(model, "visual"), "merger")
@@ -108,9 +108,13 @@ def configure_visual_model(config: "PretrainedConfig") -> None:
     Patches VLMs before loading them.
     """
     model_type = getattr(config, "model_type", None)
-    if (
-        model_type in ["llava", "video_llava"] or "llava_next" in model_type
-    ):  # required for ds zero3 and valuehead models
+    if model_type in [
+        "llava",
+        "llava_next",
+        "llava_next_video",
+        "paligemma",
+        "video_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):
@@ -124,7 +128,7 @@ def get_forbidden_modules(config: "PretrainedConfig", finetuning_args: "Finetuni
     """
     model_type = getattr(config, "model_type", None)
     forbidden_modules = set()
-    if model_type in ["llava", "paligemma", "video_llava"] or "llava_next" in model_type:
+    if model_type in ["llava", "llava_next", "llava_next_video", "paligemma", "video_llava"]:
         if finetuning_args.freeze_vision_tower:
             forbidden_modules.add("vision_tower")
 
@@ -182,7 +186,7 @@ def patch_target_modules(
     """
     model_type = getattr(config, "model_type", None)
     if finetuning_args.freeze_vision_tower:
-        if model_type in ["llava", "paligemma", "video_llava"] or "llava_next" in model_type:
+        if model_type in ["llava", "llava_next", "llava_next_video", "paligemma", "video_llava"]:
             return "^(?!.*vision_tower).*(?:{}).*".format("|".join(target_modules))
         elif model_type == "qwen2_vl":
             return "^(?!.*visual).*(?:{}).*".format("|".join(target_modules))