Skip to content

CB Bug fix for Qwen3VL Dense and basic cleaning of example script and Model File #924

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
qcdipankar wants to merge 1 commit into from
Closed

CB Bug fix for Qwen3VL Dense and basic cleaning of example script and Model File #924

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 2 additions & 9 deletions QEfficient/transformers/models/qwen3_vl/modeling_qwen3_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -388,12 +388,6 @@ def forward(
key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)

# cos, sin = position_embeddings
# kv_seq_len = key_states.shape[-2]
# kv_seq_len = past_key_value.get_seq_length()
past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
# cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)

query_states, key_states = qeff_apply_rotary_pos_emb(
query_states,
key_states,
Expand All @@ -409,7 +403,6 @@ def forward(
"cos": cos_cached,
"batch_index": batch_index,
"position_ids": position_ids[0],
"past_seen_tokens": past_seen_tokens,
}
if comp_ctx_lengths is not None:
attention_mask = attention_mask[:, :, :, : comp_ctx_lengths.shape[-1]]
Expand Down Expand Up @@ -503,7 +496,7 @@ def forward(
residual = hidden_states
hidden_states = self.post_attention_layernorm(hidden_states)
hidden_states = self.mlp(hidden_states)
hidden_states = residual + hidden_states[0]
hidden_states = residual + hidden_states

outputs = (hidden_states,)

Expand Down Expand Up @@ -667,7 +660,7 @@ class QEffQwen3VLDecoderWrapper(nn.Module):
def __init__(self, model):
super().__init__()
self.model = model
self.language_model = self.model.model
self.language_model = self.model.model.language_model

def get_submodules_for_export(self) -> Type[nn.Module]:
"""
Expand Down
18 changes: 1 addition & 17 deletions examples/image_text_to_text/models/qwen3vl/qwen3_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,7 @@
aic_enable_depth_first=True,
skip_vision=True,
mos=1,
use_onnx_subfunctions=True,
use_onnx_subfunctions=False,
)

messages = [
Expand Down Expand Up @@ -84,8 +84,6 @@
num_devices=4,
height=354,
width=536,
# height=1024,
# width=1024,
mxfp6_matmul=True,
mxint8_kv_cache=True,
aic_enable_depth_first=True,
Expand All @@ -95,10 +93,6 @@

### IMAGE + TEXT ###
image_url = "https://picsum.photos/id/237/536/354"
# image_url = (
# "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/cat_style_layout.png"
# )

image = Image.open(requests.get(image_url, stream=True).raw)

messages_1 = [
Expand All @@ -111,16 +105,6 @@
},
]

# messages_2 = [
# {
# "role": "user",
# "content": [
# {"type": "image", "image": image},
# {"type": "text", "text": "Describe about the color of the dog."},
# ],
# },
# ]

messages = [messages_1] * batch_size

texts = [processor.apply_chat_template(msg, tokenize=False, add_generation_prompt=True) for msg in messages]
Expand Down
Loading