From 3f246d21a40ef14c15a31ea44111d9e07595d1d3 Mon Sep 17 00:00:00 2001 From: MukundaKatta Date: Wed, 15 Apr 2026 01:33:25 -0700 Subject: [PATCH] Fix 'seperate' typo in qwen3/glm video-model docstrings --- src/transformers/models/glm46v/modeling_glm46v.py | 4 ++-- src/transformers/models/glm4v/modeling_glm4v.py | 4 ++-- src/transformers/models/glm4v/modular_glm4v.py | 4 ++-- src/transformers/models/glm4v_moe/modeling_glm4v_moe.py | 4 ++-- src/transformers/models/glm_ocr/modeling_glm_ocr.py | 4 ++-- src/transformers/models/qwen3_5/modeling_qwen3_5.py | 4 ++-- src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py | 4 ++-- src/transformers/models/qwen3_vl/modeling_qwen3_vl.py | 4 ++-- src/transformers/models/qwen3_vl/modular_qwen3_vl.py | 4 ++-- src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py | 4 ++-- 10 files changed, 20 insertions(+), 20 deletions(-) diff --git a/src/transformers/models/glm46v/modeling_glm46v.py b/src/transformers/models/glm46v/modeling_glm46v.py index 81207e4c8608..1a4da925e3bc 100644 --- a/src/transformers/models/glm46v/modeling_glm46v.py +++ b/src/transformers/models/glm46v/modeling_glm46v.py @@ -172,7 +172,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - GLM46V uses timestamps to seperate each video frame, so the video_grid_thw should also be split too. + - GLM46V uses timestamps to separate each video frame, so the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -195,7 +195,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/glm4v/modeling_glm4v.py b/src/transformers/models/glm4v/modeling_glm4v.py index 6121dc8d3fe8..9c20581ef245 100644 --- a/src/transformers/models/glm4v/modeling_glm4v.py +++ b/src/transformers/models/glm4v/modeling_glm4v.py @@ -1015,7 +1015,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - GLM4V uses timestamps to seperate each video frame, so the video_grid_thw should also be split too. + - GLM4V uses timestamps to separate each video frame, so the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -1038,7 +1038,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/glm4v/modular_glm4v.py b/src/transformers/models/glm4v/modular_glm4v.py index d4a34a1952ad..0b06bded536e 100644 --- a/src/transformers/models/glm4v/modular_glm4v.py +++ b/src/transformers/models/glm4v/modular_glm4v.py @@ -884,7 +884,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - GLM4V uses timestamps to seperate each video frame, so the video_grid_thw should also be split too. + - GLM4V uses timestamps to separate each video frame, so the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -907,7 +907,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/glm4v_moe/modeling_glm4v_moe.py b/src/transformers/models/glm4v_moe/modeling_glm4v_moe.py index b3f5118a3d67..45c50380998c 100644 --- a/src/transformers/models/glm4v_moe/modeling_glm4v_moe.py +++ b/src/transformers/models/glm4v_moe/modeling_glm4v_moe.py @@ -1184,7 +1184,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - GLM4V_MOE uses timestamps to seperate each video frame, so the video_grid_thw should also be split too. + - GLM4V_MOE uses timestamps to separate each video frame, so the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -1207,7 +1207,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/glm_ocr/modeling_glm_ocr.py b/src/transformers/models/glm_ocr/modeling_glm_ocr.py index 828a99a705b5..78ffaf0727fd 100644 --- a/src/transformers/models/glm_ocr/modeling_glm_ocr.py +++ b/src/transformers/models/glm_ocr/modeling_glm_ocr.py @@ -931,7 +931,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - GLM_OCR uses timestamps to seperate each video frame, so the video_grid_thw should also be split too. + - GLM_OCR uses timestamps to separate each video frame, so the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -954,7 +954,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/qwen3_5/modeling_qwen3_5.py b/src/transformers/models/qwen3_5/modeling_qwen3_5.py index 2c4eba9597dc..23c505db3ad0 100644 --- a/src/transformers/models/qwen3_5/modeling_qwen3_5.py +++ b/src/transformers/models/qwen3_5/modeling_qwen3_5.py @@ -1396,7 +1396,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - Since Qwen3.5 use timestamps to seperate videos, like , the video_grid_thw should also be split too. + - Since Qwen3.5 use timestamps to separate videos, like , the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -1419,7 +1419,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py b/src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py index 0b2a6a06aa85..9d79279d6257 100644 --- a/src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py +++ b/src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py @@ -1521,7 +1521,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - Since Qwen3.5 use timestamps to seperate videos, like , the video_grid_thw should also be split too. + - Since Qwen3.5 use timestamps to separate videos, like , the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -1544,7 +1544,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/qwen3_vl/modeling_qwen3_vl.py b/src/transformers/models/qwen3_vl/modeling_qwen3_vl.py index 9522cb354789..dfc02a500d79 100644 --- a/src/transformers/models/qwen3_vl/modeling_qwen3_vl.py +++ b/src/transformers/models/qwen3_vl/modeling_qwen3_vl.py @@ -1041,7 +1041,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - Since Qwen3.5 use timestamps to seperate videos, like , the video_grid_thw should also be split too. + - Since Qwen3.5 use timestamps to separate videos, like , the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -1064,7 +1064,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/qwen3_vl/modular_qwen3_vl.py b/src/transformers/models/qwen3_vl/modular_qwen3_vl.py index 74d887726119..3f61715a0737 100644 --- a/src/transformers/models/qwen3_vl/modular_qwen3_vl.py +++ b/src/transformers/models/qwen3_vl/modular_qwen3_vl.py @@ -741,7 +741,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - Since Qwen3.5 use timestamps to seperate videos, like , the video_grid_thw should also be split too. + - Since Qwen3.5 use timestamps to separate videos, like , the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -764,7 +764,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1 diff --git a/src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py b/src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py index be248a160e7d..d92784da03a8 100644 --- a/src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py +++ b/src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py @@ -1170,7 +1170,7 @@ def get_rope_index( ) -> tuple[torch.Tensor, torch.Tensor]: """ Difference from Qwen2VL/Qwen2.5VL's get_rope_index: - - Since Qwen3.5 use timestamps to seperate videos, like , the video_grid_thw should also be split too. + - Since Qwen3.5 use timestamps to separate videos, like , the video_grid_thw should also be split too. Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): @@ -1193,7 +1193,7 @@ def get_rope_index( mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`) """ - # Separate video grid thw into multiple grids because timestamps are used to seperate videos. + # Separate video grid thw into multiple grids because timestamps are used to separate videos. if video_grid_thw is not None: video_grid_thw = torch.repeat_interleave(video_grid_thw, video_grid_thw[:, 0], dim=0) video_grid_thw[:, 0] = 1