huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-09 11:24:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix the image_token_id issue

Browse Source 
Signed-off-by: yuanwu <yuan.wu@intel.com>
This commit is contained in:
yuanwu 2025-05-11 22:11:42 +00:00
parent 50ecfc625a
commit f5aaa18d8e
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
backends/gaudi/server/text_generation_server/models/custom_modeling/flash_llama4_modeling.py
View File

@ -1397,7 +1397,9 @@ class Llama4ForConditionalGeneration(nn.Module):
vision_flat = image_features.view(-1, image_features.size(-1)) vision_flat = image_features.view(-1, image_features.size(-1))
projected_vision_flat = self.multi_modal_projector(vision_flat) projected_vision_flat = self.multi_modal_projector(vision_flat)
special_image_mask = (input_ids == self.config.image_token_id).unsqueeze(-1) special_image_mask = (input_ids == self.config.image_token_index).unsqueeze(
-1
)
final_mask = special_image_mask.to(inputs_embeds.device) final_mask = special_image_mask.to(inputs_embeds.device)
inputs_embeds = inputs_embeds.view(-1, inputs_embeds.size(-1)) inputs_embeds = inputs_embeds.view(-1, inputs_embeds.size(-1))