huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-04-19 22:02:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix qwen vl (#3096)

Browse Source 
* Fixing qwen2.5 VL.

* Fixing the CI.
This commit is contained in:
Nicolas Patry 2025-03-11 11:00:41 +01:00 committed by GitHub
parent 094975c3a8
commit b447f7e821
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 2 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Cargo.lock generated
View File

@ -4754,7 +4754,7 @@ dependencies = [
"async-trait", "async-trait",
"bindgen 0.71.1", "bindgen 0.71.1",
"clap 4.5.30", "clap 4.5.30",
"hf-hub 0.3.2", "hf-hub 0.4.2",
"num_cpus", "num_cpus",
"pkg-config", "pkg-config",
"text-generation-router", "text-generation-router",

3
server/text_generation_server/models/custom_modeling/qwen2_5_vl.py
View File

@ -313,7 +313,6 @@ class Qwen2_5_VLVisionConfig(PretrainedConfig):
class Qwen2_5_VLConfig(PretrainedConfig): class Qwen2_5_VLConfig(PretrainedConfig):
def __init__( def __init__(
self, self,
vocab_size=152064, vocab_size=152064,
@ -702,7 +701,6 @@ class Qwen2_5VisionModel(nn.Module):
pixel_values: torch.Tensor, pixel_values: torch.Tensor,
grid_thw: Optional[torch.LongTensor] = None, grid_thw: Optional[torch.LongTensor] = None,
) -> torch.Tensor: ) -> torch.Tensor:
# reshape the input tensor for processing # reshape the input tensor for processing
shape = ( shape = (
-1, -1,
@ -967,6 +965,7 @@ class Qwen2_5VLForConditionalGeneration(nn.Module):
max_s=max_s, max_s=max_s,
true_max_s=max_s, true_max_s=max_s,
prefill_cache_indices=prefill_cache_indices, prefill_cache_indices=prefill_cache_indices,
adapter_data=adapter_data,
) )
if lm_head_indices is not None: if lm_head_indices is not None:
hidden_states = hidden_states[lm_head_indices] hidden_states = hidden_states[lm_head_indices]