Implemented Qwen35ChatHandler for Qwen3.5#61
Conversation
JamePeng
force-pushed
the
main
branch
5 times, most recently
from
76d8272 to
68eacae
Compare
|
Detailed adaptation work can be done after Qwen3.5-9B-Instruct and Qwen3.5-35B-A3B-Instruct are released. |
|
I've tested pruned version of the Qwen 3.5 model. |
|
I'm updating and testing this with |
|
Closed the PR temporarily to update to the latest commit. |
|
I'm currently refactoring some logic locally, but the hybrid structure of qwen3-next and qwen3.5 is basically running. You can continue testing after I finish the initial implementation. |
|
I'm having some trouble testing this. I'm not sure if it's related to my OS, Python version, or this project. Last time I compiled it (about two days ago) it worked fine.
Does it crash because of this? |
|
No, it's because your compiled library is incompatible or the library file is missing. |
|
Okay, thanks! I'll try to fix it as soon as possible to continue testing. |
|
The initial implementation is now live. You can try out qwen3.5 and see how it performs. Looking forward to your feedback. |
|
I've tested the chat template with and without images, thinking mode, etc. I've only used The model seems to work fine without images, but hallucinates a lot when using images (can't describe them properly, etc.). Since the model is so big and I'm just asking for a simple description of the image, the quantization I'm using should not be a problem. Note Activating the thinking mode and looking at the reasoning content, the model does not see the image. I'll check the chat template to make sure it's not causing this problem. I'll continue to work on this. |
|
Looking at the console output, I find this: I have no idea if it's related to this bug, but is only printed when adding an image to the prompt. Comparing the Qwen35ChatHandler's template with the Qwen3VLChatHandler's template, the code to load an image is similar: Qwen3VLChatHandler: ...
{%- if 'image_url' in content -%}
{%- set image_count.value = image_count.value + 1 -%}
{%- if add_vision_id -%}
{{- 'Picture ' -}}
{{- image_count.value | string -}}
{{- ': ' -}}
{%- endif -%}
{{- '<|vision_start|>' -}}
{%- if content.image_url is string -%}
{{- content.image_url -}}
{%- else -%}
{{- content.image_url.url -}}
{%- endif -%}
{{- '<|vision_end|>' -}}
{%- endif -%}
...Qwen35ChatHandler: ...
{%- if 'image' in item or 'image_url' in item -%}
{%- if is_system_content -%}
{{- raise_exception('System message cannot contain images.') -}}
{%- endif -%}
{%- if do_vision_count -%}
{%- set image_count.value = image_count.value + 1 -%}
{%- endif -%}
{%- if add_vision_id -%}
{{- 'Picture ' ~ image_count.value ~ ': ' -}}
{%- endif -%}
{{- '<|vision_start|>' -}}
{%- if 'image' in item -%}
{%- if item.image is string -%}
{{- item.image -}}
{%- else -%}
{{- item.image.url -}}
{%- endif -%}
{%- elif 'image_url' in item -%}
{%- if item.image_url is string -%}
{{- item.image_url -}}
{%- else -%}
{{- item.image_url.url -}}
{%- endif -%}
{%- endif -%}
{{- '<|vision_end|>' -}}
{%- elif 'video' in item -%}
... |
# Clear state for multiple runs
llama.reset()
llama._ctx.memory_clear(True)
llama.n_tokens = 0These all need to be removed; the generation and eval processes should now manage them. |
You can check how the final format template it constructs is assembled, and whether there are any <media> elements mounted in it. |
I guess, even after you remove this part, an identical codes will be called by |
|
I've tested this again.
Printing the chat template shows the I'll try to re-download the model and mmproj, maybe with another quantization. NOTE: According to the model's thoughts, sometimes it sees the image, but it's all black or just !!!!!!!!!! characters. Keep in mind that this is only according to the model's thoughts, it doesn't mean it can actually see anything. Also I'm using base64-encoded images, not URLs. |
|
This is normal. Currently, multimodal inputs are not entering the hybrid cache. I'll look into modifying the llava15chathandler into a temporary version this weekend. |
|
@alcoftTAO I've completed the local adaptation for qwen3.5 and tested your chat template; it seems to work fine. Please remove the changes from the qwen3-vl section, and then modify the chat_template in qwen3.5 as follows: class Qwen35ChatHandler(MTMDChatHandler):
CHAT_FORMAT = (
"{%- set image_count = namespace(value=0) -%}"
"{%- set video_count = namespace(value=0) -%}"
"{%- macro render_content(content, do_vision_count, is_system_content=false) -%}"
" {%- if content is string -%}"
" {{- content -}}"
" {%- elif content is iterable and content is not mapping -%}"
" {%- for item in content -%}"
" {%- if 'image_url' in item or item.type == 'image_url' -%}"
" {%- if is_system_content -%}"
" {{- raise_exception('System message cannot contain images.') -}}"
" {%- endif -%}"
" {%- if do_vision_count -%}"
" {%- set image_count.value = image_count.value + 1 -%}"
" {%- endif -%}"
" {%- if add_vision_id -%}"
" {{- 'Picture ' -}}"
" {{- image_count.value | string -}}"
" {{- ': ' -}}"
" {%- endif -%}"
" {{- '<|vision_start|>' -}}"
" {%- if item.image_url is string -%}"
" {{- item.image_url -}}"
" {%- else -%}"
" {{- item.image_url.url -}}"
" {%- endif -%}"
" {{- '<|vision_end|>' -}}"
" {%- elif 'video' in item -%}"
" {{- raise_exception('llama.cpp does not currently support video.') -}}" # Video not supported, raise exception
" {%- if is_system_content -%}"
" {{- raise_exception('System message cannot contain videos.') -}}"
" {%- endif -%}"
" {%- if do_vision_count -%}"
" {%- set video_count.value = video_count.value + 1 -%}"
" {%- endif -%}"
" {%- if add_vision_id -%}"
" {{- 'Video ' ~ video_count.value ~ ': ' -}}"
" {%- endif -%}"
" {{- '<|vision_start|>' -}}"
" {{- item.video -}}"
" {{- '<|vision_end|>' -}}"
" {%- elif 'text' in item -%}"
" {{- item.text -}}"
" {%- else -%}"
" {{- raise_exception('Unexpected item type in content.') -}}"
" {%- endif -%}"
" {%- endfor -%}"
" {%- elif content is none or content is undefined -%}"
" {{- '' -}}"
" {%- else -%}"
" {{- raise_exception('Unexpected content type.') -}}"
" {%- endif -%}"
"{%- endmacro -%}"
"{%- if not messages -%}"
" {{- raise_exception('No messages provided.') -}}"
"{%- endif -%}"
"{%- if tools and tools is iterable and tools is not mapping -%}"
" {{- '<|im_start|>system\n' -}}"
" {{- '# Tools\n\nYou have access to the following functions:\n\n<tools>' -}}"
" {%- for tool in tools -%}"
" {{- '\n' -}}"
" {{- tool | tojson -}}"
" {%- endfor -%}"
" {{- '\n</tools>' -}}"
" {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' -}}"
" {%- if messages[0].role == 'system' -%}"
" {%- set content = render_content(messages[0].content, false, true) | trim -%}"
" {%- if content -%}"
" {{- '\n\n' + content -}}"
" {%- endif -%}"
" {%- endif -%}"
" {{- '<|im_end|>\n' -}}"
"{%- elif messages[0].role == 'system' -%}"
" {%- set content = render_content(messages[0].content, false, true) -%}"
" {{- '<|im_start|>system\n' + content + '<|im_end|>\n' -}}"
"{%- endif -%}"
"{%- set ns = namespace(multi_step_tool=true, last_query_index=messages | length - 1) -%}"
"{%- for message in messages[::-1] -%}"
" {%- set index = messages | length - 1 - loop.index0 -%}"
" {%- if ns.multi_step_tool and message.role == 'user' -%}"
" {%- set content = render_content(message.content, false) | trim -%}"
" {%- if not (content.startswith('<tool_response>') and content.endswith('</tool_response>')) -%}"
" {%- set ns.multi_step_tool = false -%}"
" {%- set ns.last_query_index = index -%}"
" {%- endif -%}"
" {%- endif -%}"
"{%- endfor -%}"
"{%- if ns.multi_step_tool -%}"
" {{- raise_exception('No user query found in messages.') -}}"
"{%- endif -%}"
"{%- for message in messages -%}"
" {%- set content = render_content(message.content, true) | trim -%}"
" {%- if message.role == 'system' -%}"
" {%- if not loop.first -%}"
" {{- raise_exception('System message must be at the beginning.') -}}"
" {%- endif -%}"
" {%- elif message.role == 'user' -%}"
" {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>\n' -}}"
" {%- elif message.role == 'assistant' -%}"
" {%- set reasoning_content = '' -%}"
" {%- if message.reasoning_content is string -%}"
" {%- set reasoning_content = message.reasoning_content -%}"
" {%- elif '</think>' in content -%}"
" {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') -%}"
" {%- set content = content.split('</think>')[-1].lstrip('\n') -%}"
" {%- endif -%}"
" {%- set reasoning_content = reasoning_content | trim -%}"
" {%- if loop.index0 > ns.last_query_index -%}"
" {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content -}}"
" {%- else -%}"
" {{- '<|im_start|>' + message.role + '\n' + content -}}"
" {%- endif -%}"
" {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping -%}"
" {%- for tool_call in message.tool_call -%}"
" {%- if tool_call.function is defined -%}"
" {%- set tool_call = tool_call.function -%}"
" {%- endif -%}"
" {%- if loop.first -%}"
" {%- if content | trim -%}"
" {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' -}}"
" {%- else -%}"
" {{- '<tool_call>\n<function=' + tool_call.name + '>\n' -}}"
" {%- endif -%}"
" {%- else -%}"
" {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' -}}"
" {%- endif -%}"
" {%- if tool_call.arguments is defined -%}"
" {%- for (args_name, args_value) in tool_calls.arguments | items -%}"
" {{- '<parameter=' + args.name + '>\n' -}}"
" {%- set args_value = args_value | tojson | safe if args_value is mapping or args_value is sequence and args_value is not string else args_value | string -%}"
" {{- args_value -}}"
" {{- '\n</parameter>' -}}"
" {%- endfor -%}"
" {%- endif -%}"
" {{- '</function>\n</tool_call>' -}}"
" {%- endfor -%}"
" {%- endif -%}"
" {{- '<|im_end|>\n' -}}"
" {%- elif message.role == 'tool' -%}"
" {%- if loop.previtem and loop.previtem.role != 'tool' -%}"
" {{- '<|im_start|>user' -}}"
" {%- endif -%}"
" {{- '\n<tool_response>\n' -}}"
" {{- content -}}"
" {{- '\n</tool_response>' -}}"
" {%- if not loop.last and loop.nextitem.role != 'tool' -%}"
" {{- '<|im_end|>\n' -}}"
" {%- elif loop.last -%}"
" {{- '<|im_end|>\n' -}}"
" {%- endif -%}"
" {%- else -%}"
" {{- raise_exception('Unexpected message role.') -}}"
" {%- endif -%}"
"{%- endfor -%}"
"{%- if add_generation_prompt -%}"
" {{- '<|im_start|>assistant\n' -}}"
" {%- if enable_thinking is false -%}"
" {{- '<think>\n\n</think>\n\n' -}}"
" {%- else -%}"
" {{- '<think>\n' -}}"
" {%- endif -%}"
"{%- endif -%}"
)
def __init__(
self,
enable_thinking: bool = True,
add_vision_id: bool = True,
**kwargs,
):
"""
Parameters:
- enable_thinking (bool):
- True (default): Enables reasoning for better results.
- False: Disables reasoning for faster results.
- add_vision_id (bool):
- True (default): Count all the images. Recommended for multi-image.
- False: Doesn't count the images. Can save tokens with single-image.
"""
super().__init__(**kwargs)
self.enable_thinking = enable_thinking
self.extra_template_arguments["enable_thinking"] = enable_thinking
self.extra_template_arguments["add_vision_id"] = add_vision_id
def __call__(self, **kwargs):
llama = kwargs['llama']
if hasattr(llama, 'input_ids'):
llama.input_ids.fill(0)
if self.verbose:
print(f"{self.log_prefix}(enable_thinking={self.enable_thinking}) - Start processing")
# Use parent implementation
return super().__call__(**kwargs) |
|
Test result: |
|
The new refactoring code has been committed; you can try it again after syncing. After this PR is merged, it will be almost time to update the new version code :) |
|
I've tested it and works fine! |
JamePeng
changed the title
|
@alcoftTAO If interested, you can also adapt multimodal models of LFM2.5-VL or other Hybrid/Recurrent/SWA structures. Let's see how well other models perform when adapted and run. I've also added a multi-threaded image-to-bitmap conversion function this time. You can also try passing in multi-image bitmap video flags to see how efficient it is. |
|
Thanks! I'll look forward to implement LFM2.5-VL soon! |
3 participants
This PR implements Qwen3.5 models. Not tested yet due to lack of compute power on my end.
This PR is going to be a draft for now until I can test it with smaller models and also check and fix the chat template and parameters of the
Qwen35ChatHandlerclass.I still need to decide which parameters are useful.