Upload 3 files
Browse files- added_tokens.json +35 -0
- chat_template.jinja +122 -0
- config.json +552 -0
added_tokens.json
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"</think>": 151668,
|
| 3 |
+
"</tool_call>": 151658,
|
| 4 |
+
"</tool_response>": 151666,
|
| 5 |
+
"<think>": 151667,
|
| 6 |
+
"<tool_call>": 151657,
|
| 7 |
+
"<tool_response>": 151665,
|
| 8 |
+
"<tts_pad>": 151671,
|
| 9 |
+
"<tts_text_bos>": 151672,
|
| 10 |
+
"<tts_text_bos_single>": 151674,
|
| 11 |
+
"<tts_text_eod>": 151673,
|
| 12 |
+
"<|audio_end|>": 151670,
|
| 13 |
+
"<|audio_pad|>": 151675,
|
| 14 |
+
"<|audio_start|>": 151669,
|
| 15 |
+
"<|box_end|>": 151649,
|
| 16 |
+
"<|box_start|>": 151648,
|
| 17 |
+
"<|endoftext|>": 151643,
|
| 18 |
+
"<|file_sep|>": 151664,
|
| 19 |
+
"<|fim_middle|>": 151660,
|
| 20 |
+
"<|fim_pad|>": 151662,
|
| 21 |
+
"<|fim_prefix|>": 151659,
|
| 22 |
+
"<|fim_suffix|>": 151661,
|
| 23 |
+
"<|im_end|>": 151645,
|
| 24 |
+
"<|im_start|>": 151644,
|
| 25 |
+
"<|image_pad|>": 151655,
|
| 26 |
+
"<|object_ref_end|>": 151647,
|
| 27 |
+
"<|object_ref_start|>": 151646,
|
| 28 |
+
"<|quad_end|>": 151651,
|
| 29 |
+
"<|quad_start|>": 151650,
|
| 30 |
+
"<|repo_name|>": 151663,
|
| 31 |
+
"<|video_pad|>": 151656,
|
| 32 |
+
"<|vision_end|>": 151653,
|
| 33 |
+
"<|vision_pad|>": 151654,
|
| 34 |
+
"<|vision_start|>": 151652
|
| 35 |
+
}
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,122 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- if tools %}
|
| 2 |
+
{{- '<|im_start|>system\n' }}
|
| 3 |
+
{%- if messages[0].role == 'system' %}
|
| 4 |
+
{%- if messages[0].content is string %}
|
| 5 |
+
{{- messages[0].content }}
|
| 6 |
+
{%- else %}
|
| 7 |
+
{%- for content in messages[0].content %}
|
| 8 |
+
{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
|
| 9 |
+
{{- "<|vision_start|><|image_pad|><|vision_end|>" }}
|
| 10 |
+
{%- elif content.type == 'audio' or 'audio' in content or 'audio_url' in content %}
|
| 11 |
+
{{- "<|audio_start|><|audio_pad|><|audio_end|>" }}
|
| 12 |
+
{%- elif content.type == 'video' or 'video' in content %}
|
| 13 |
+
{{- "<|vision_start|><|video_pad|><|vision_end|>" }}
|
| 14 |
+
{%- elif content.type == 'text' %}
|
| 15 |
+
{{- content.text }}
|
| 16 |
+
{%- endif %}
|
| 17 |
+
{%- endfor %}
|
| 18 |
+
{%- endif %}
|
| 19 |
+
{%- endif %}
|
| 20 |
+
{{- '\n\n' }}
|
| 21 |
+
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 22 |
+
{%- for tool in tools %}
|
| 23 |
+
{{- "\n" }}
|
| 24 |
+
{{- tool | tojson }}
|
| 25 |
+
{%- endfor %}
|
| 26 |
+
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
| 27 |
+
{%- else %}
|
| 28 |
+
{%- if messages[0].role == 'system' %}
|
| 29 |
+
{%- if messages[0].content is string %}
|
| 30 |
+
{{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
|
| 31 |
+
{%- else %}
|
| 32 |
+
{%- for content in messages[0].content %}
|
| 33 |
+
{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
|
| 34 |
+
{{- '<|im_start|>system\n' +"<|vision_start|><|image_pad|><|vision_end|>"+ '<|im_end|>\n' }}
|
| 35 |
+
{%- elif content.type == 'audio' or 'audio' in content or 'audio_url' in content %}
|
| 36 |
+
{{- '<|im_start|>system\n' +"<|audio_start|><|audio_pad|><|audio_end|>"+ '<|im_end|>\n' }}
|
| 37 |
+
{%- elif content.type == 'video' or 'video' in content %}
|
| 38 |
+
{{- '<|im_start|>system\n' +"<|vision_start|><|video_pad|><|vision_end|>"+ '<|im_end|>\n' }}
|
| 39 |
+
{%- elif content.type == 'text' %}
|
| 40 |
+
{{- '<|im_start|>system\n' +content.text+ '<|im_end|>\n' }}
|
| 41 |
+
{%- endif %}
|
| 42 |
+
{%- endfor %}
|
| 43 |
+
{%- endif %}
|
| 44 |
+
{%- endif %}
|
| 45 |
+
{%- endif %}
|
| 46 |
+
{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
|
| 47 |
+
{%- for message in messages[::-1] %}
|
| 48 |
+
{%- set index = (messages|length - 1) - loop.index0 %}
|
| 49 |
+
{%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
|
| 50 |
+
{%- set ns.multi_step_tool = false %}
|
| 51 |
+
{%- set ns.last_query_index = index %}
|
| 52 |
+
{%- endif %}
|
| 53 |
+
{%- endfor %}
|
| 54 |
+
{%- for message in messages %}
|
| 55 |
+
{%- if message.content is string %}
|
| 56 |
+
{%- set content = message.content %}
|
| 57 |
+
{%- else %}
|
| 58 |
+
{%- set content = namespace(text="") %}
|
| 59 |
+
{%- for mcontent in message.content %}
|
| 60 |
+
{%- if mcontent.type == 'image' or 'image' in mcontent or 'image_url' in mcontent %}
|
| 61 |
+
{%- set content.text = content.text~"<|vision_start|><|image_pad|><|vision_end|>" %}
|
| 62 |
+
{%- elif mcontent.type == 'audio' or 'audio' in mcontent or 'audio_url' in mcontent %}
|
| 63 |
+
{%- set content.text = content.text~"<|audio_start|><|audio_pad|><|audio_end|>" %}
|
| 64 |
+
{%- elif mcontent.type == 'video' or 'video' in mcontent %}
|
| 65 |
+
{%- set content.text = content.text~"<|vision_start|><|video_pad|><|vision_end|>" %}
|
| 66 |
+
{%- elif mcontent.type == 'text' %}
|
| 67 |
+
{%- set content.text = content.text~mcontent.text %}
|
| 68 |
+
{%- endif %}
|
| 69 |
+
{%- endfor %}
|
| 70 |
+
{%- set content = content.text %}
|
| 71 |
+
{%- endif %}
|
| 72 |
+
{%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
|
| 73 |
+
{{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
|
| 74 |
+
{%- elif message.role == "assistant" %}
|
| 75 |
+
{%- set reasoning_content = "" %}
|
| 76 |
+
{%- if message.reasoning_content is string %}
|
| 77 |
+
{%- set reasoning_content = message.reasoning_content %}
|
| 78 |
+
{%- else %}
|
| 79 |
+
{%- if '</think>' in content %}
|
| 80 |
+
{%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
|
| 81 |
+
{%- set content = content.split('</think>')[-1].lstrip('\n') %}
|
| 82 |
+
{%- endif %}
|
| 83 |
+
{%- endif %}
|
| 84 |
+
{%- if loop.index0 > ns.last_query_index %}
|
| 85 |
+
{%- if loop.last or (not loop.last and reasoning_content) %}
|
| 86 |
+
{{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip("\n") + '\n</think>\n\n' + content.lstrip('\n') }}
|
| 87 |
+
{%- else %}
|
| 88 |
+
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 89 |
+
{%- endif %}
|
| 90 |
+
{%- else %}
|
| 91 |
+
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 92 |
+
{%- endif %}
|
| 93 |
+
{%- if message.tool_calls %}
|
| 94 |
+
{%- for tool_call in message.tool_calls %}
|
| 95 |
+
{%- if (loop.first and content) or (not loop.first) %}{{- '\n' }}{%- endif %}
|
| 96 |
+
{%- if tool_call.function %}
|
| 97 |
+
{%- set tool_call = tool_call.function %}
|
| 98 |
+
{%- endif %}
|
| 99 |
+
{{- '<tool_call>\n{"name": "' }}
|
| 100 |
+
{{- tool_call.name }}
|
| 101 |
+
{{- '", "arguments": ' }}
|
| 102 |
+
{%- if tool_call.arguments is string %}
|
| 103 |
+
{{- tool_call.arguments }}
|
| 104 |
+
{%- else %}
|
| 105 |
+
{{- tool_call.arguments | tojson }}
|
| 106 |
+
{%- endif %}
|
| 107 |
+
{{- '}\n</tool_call>' }}
|
| 108 |
+
{%- endfor %}
|
| 109 |
+
{%- endif %}
|
| 110 |
+
{{- '<|im_end|>\n' }}
|
| 111 |
+
{%- elif message.role == "tool" %}
|
| 112 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}{{- '<|im_start|>user' }}{%- endif %}
|
| 113 |
+
{{- '\n<tool_response>\n' }}
|
| 114 |
+
{{- content }}
|
| 115 |
+
{{- '\n</tool_response>' }}
|
| 116 |
+
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}{{- '<|im_end|>\n' }}{%- endif %}
|
| 117 |
+
{%- endif %}
|
| 118 |
+
{%- endfor %}
|
| 119 |
+
{%- if add_generation_prompt %}
|
| 120 |
+
{{- '<|im_start|>assistant\n' }}
|
| 121 |
+
{%- if enable_thinking is defined and enable_thinking is false %}{{- '<think>\n\n</think>\n\n' }}{%- endif %}
|
| 122 |
+
{%- endif %}
|
config.json
ADDED
|
@@ -0,0 +1,552 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3OmniMoeForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"assistant_token_id": 77091,
|
| 6 |
+
"code2wav_config": {
|
| 7 |
+
"attention_bias": false,
|
| 8 |
+
"attention_dropout": 0.0,
|
| 9 |
+
"codebook_dim": 512,
|
| 10 |
+
"codebook_size": 2048,
|
| 11 |
+
"decoder_dim": 1536,
|
| 12 |
+
"hidden_act": "silu",
|
| 13 |
+
"hidden_size": 1024,
|
| 14 |
+
"intermediate_size": 3072,
|
| 15 |
+
"layer_scale_initial_scale": 0.01,
|
| 16 |
+
"max_position_embeddings": 8000,
|
| 17 |
+
"model_type": "",
|
| 18 |
+
"num_attention_heads": 16,
|
| 19 |
+
"num_hidden_layers": 8,
|
| 20 |
+
"num_key_value_heads": 16,
|
| 21 |
+
"num_quantizers": 16,
|
| 22 |
+
"num_semantic_quantizers": 1,
|
| 23 |
+
"rms_norm_eps": 1e-05,
|
| 24 |
+
"rope_theta": 10000,
|
| 25 |
+
"semantic_codebook_size": 4096,
|
| 26 |
+
"sliding_window": 72,
|
| 27 |
+
"upsample_rates": [
|
| 28 |
+
8,
|
| 29 |
+
5,
|
| 30 |
+
4,
|
| 31 |
+
3
|
| 32 |
+
],
|
| 33 |
+
"upsampling_ratios": [
|
| 34 |
+
2,
|
| 35 |
+
2
|
| 36 |
+
],
|
| 37 |
+
"vector_quantization_hidden_dimension": 512
|
| 38 |
+
},
|
| 39 |
+
"dtype": "bfloat16",
|
| 40 |
+
"enable_audio_output": true,
|
| 41 |
+
"hidden_size": 1024,
|
| 42 |
+
"im_end_token_id": 151645,
|
| 43 |
+
"im_start_token_id": 151644,
|
| 44 |
+
"keys_to_ignore_at_inference": [
|
| 45 |
+
"past_key_values"
|
| 46 |
+
],
|
| 47 |
+
"model_type": "qwen3_omni_moe",
|
| 48 |
+
"system_token_id": 8948,
|
| 49 |
+
"talker_config": {
|
| 50 |
+
"accept_hidden_layer": 24,
|
| 51 |
+
"audio_end_token_id": 151670,
|
| 52 |
+
"audio_start_token_id": 151669,
|
| 53 |
+
"audio_token_id": 151675,
|
| 54 |
+
"code_predictor_config": {
|
| 55 |
+
"_name_or_path": "",
|
| 56 |
+
"add_cross_attention": false,
|
| 57 |
+
"architectures": null,
|
| 58 |
+
"attention_bias": false,
|
| 59 |
+
"attention_dropout": 0,
|
| 60 |
+
"bad_words_ids": null,
|
| 61 |
+
"begin_suppress_tokens": null,
|
| 62 |
+
"bos_token_id": null,
|
| 63 |
+
"chunk_size_feed_forward": 0,
|
| 64 |
+
"cross_attention_hidden_size": null,
|
| 65 |
+
"decoder_start_token_id": null,
|
| 66 |
+
"diversity_penalty": 0.0,
|
| 67 |
+
"do_sample": false,
|
| 68 |
+
"dtype": null,
|
| 69 |
+
"early_stopping": false,
|
| 70 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 71 |
+
"eos_token_id": null,
|
| 72 |
+
"exponential_decay_length_penalty": null,
|
| 73 |
+
"finetuning_task": null,
|
| 74 |
+
"forced_bos_token_id": null,
|
| 75 |
+
"forced_eos_token_id": null,
|
| 76 |
+
"head_dim": 128,
|
| 77 |
+
"hidden_act": "silu",
|
| 78 |
+
"hidden_size": 1024,
|
| 79 |
+
"id2label": {
|
| 80 |
+
"0": "LABEL_0",
|
| 81 |
+
"1": "LABEL_1"
|
| 82 |
+
},
|
| 83 |
+
"initializer_range": 0.02,
|
| 84 |
+
"intermediate_size": 3072,
|
| 85 |
+
"is_decoder": false,
|
| 86 |
+
"is_encoder_decoder": false,
|
| 87 |
+
"label2id": {
|
| 88 |
+
"LABEL_0": 0,
|
| 89 |
+
"LABEL_1": 1
|
| 90 |
+
},
|
| 91 |
+
"layer_types": [
|
| 92 |
+
"full_attention",
|
| 93 |
+
"full_attention",
|
| 94 |
+
"full_attention",
|
| 95 |
+
"full_attention",
|
| 96 |
+
"full_attention"
|
| 97 |
+
],
|
| 98 |
+
"length_penalty": 1.0,
|
| 99 |
+
"max_length": 20,
|
| 100 |
+
"max_position_embeddings": 32768,
|
| 101 |
+
"max_window_layers": 28,
|
| 102 |
+
"min_length": 0,
|
| 103 |
+
"model_type": "qwen3_omni_moe_talker_code_predictor",
|
| 104 |
+
"no_repeat_ngram_size": 0,
|
| 105 |
+
"num_attention_heads": 16,
|
| 106 |
+
"num_beam_groups": 1,
|
| 107 |
+
"num_beams": 1,
|
| 108 |
+
"num_code_groups": 16,
|
| 109 |
+
"num_hidden_layers": 5,
|
| 110 |
+
"num_key_value_heads": 8,
|
| 111 |
+
"num_return_sequences": 1,
|
| 112 |
+
"output_attentions": false,
|
| 113 |
+
"output_hidden_states": false,
|
| 114 |
+
"output_scores": false,
|
| 115 |
+
"pad_token_id": null,
|
| 116 |
+
"prefix": null,
|
| 117 |
+
"problem_type": null,
|
| 118 |
+
"pruned_heads": {},
|
| 119 |
+
"remove_invalid_values": false,
|
| 120 |
+
"repetition_penalty": 1.0,
|
| 121 |
+
"return_dict": true,
|
| 122 |
+
"return_dict_in_generate": false,
|
| 123 |
+
"rms_norm_eps": 1e-06,
|
| 124 |
+
"rope_scaling": null,
|
| 125 |
+
"rope_theta": 1000000,
|
| 126 |
+
"sep_token_id": null,
|
| 127 |
+
"sliding_window": null,
|
| 128 |
+
"suppress_tokens": null,
|
| 129 |
+
"task_specific_params": null,
|
| 130 |
+
"temperature": 1.0,
|
| 131 |
+
"tf_legacy_loss": false,
|
| 132 |
+
"tie_encoder_decoder": false,
|
| 133 |
+
"tie_word_embeddings": false,
|
| 134 |
+
"tokenizer_class": null,
|
| 135 |
+
"top_k": 50,
|
| 136 |
+
"top_p": 1.0,
|
| 137 |
+
"torchscript": false,
|
| 138 |
+
"typical_p": 1.0,
|
| 139 |
+
"use_bfloat16": false,
|
| 140 |
+
"use_cache": true,
|
| 141 |
+
"use_sliding_window": false,
|
| 142 |
+
"vocab_size": 2048
|
| 143 |
+
},
|
| 144 |
+
"codec_bos_id": 2149,
|
| 145 |
+
"codec_eos_token_id": 2150,
|
| 146 |
+
"codec_nothink_id": 2155,
|
| 147 |
+
"codec_pad_id": 2148,
|
| 148 |
+
"codec_think_bos_id": 2156,
|
| 149 |
+
"codec_think_eos_id": 2157,
|
| 150 |
+
"image_token_id": 151655,
|
| 151 |
+
"model_type": "",
|
| 152 |
+
"num_code_groups": 16,
|
| 153 |
+
"output_router_logits": false,
|
| 154 |
+
"position_id_per_seconds": 13,
|
| 155 |
+
"seconds_per_chunk": 2,
|
| 156 |
+
"spatial_merge_size": 2,
|
| 157 |
+
"speaker_id": {
|
| 158 |
+
"aiden": 2303,
|
| 159 |
+
"chelsie": 2301,
|
| 160 |
+
"ethan": 2302
|
| 161 |
+
},
|
| 162 |
+
"text_config": {
|
| 163 |
+
"_name_or_path": "",
|
| 164 |
+
"add_cross_attention": false,
|
| 165 |
+
"architectures": null,
|
| 166 |
+
"attention_bias": false,
|
| 167 |
+
"attention_dropout": 0,
|
| 168 |
+
"bad_words_ids": null,
|
| 169 |
+
"begin_suppress_tokens": null,
|
| 170 |
+
"bos_token_id": null,
|
| 171 |
+
"chunk_size_feed_forward": 0,
|
| 172 |
+
"cross_attention_hidden_size": null,
|
| 173 |
+
"decoder_sparse_step": 1,
|
| 174 |
+
"decoder_start_token_id": null,
|
| 175 |
+
"diversity_penalty": 0.0,
|
| 176 |
+
"do_sample": false,
|
| 177 |
+
"dtype": "bfloat16",
|
| 178 |
+
"early_stopping": false,
|
| 179 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 180 |
+
"eos_token_id": null,
|
| 181 |
+
"exponential_decay_length_penalty": null,
|
| 182 |
+
"finetuning_task": null,
|
| 183 |
+
"forced_bos_token_id": null,
|
| 184 |
+
"forced_eos_token_id": null,
|
| 185 |
+
"head_dim": 128,
|
| 186 |
+
"hidden_act": "silu",
|
| 187 |
+
"hidden_size": 1024,
|
| 188 |
+
"id2label": {
|
| 189 |
+
"0": "LABEL_0",
|
| 190 |
+
"1": "LABEL_1"
|
| 191 |
+
},
|
| 192 |
+
"initializer_range": 0.02,
|
| 193 |
+
"intermediate_size": 2048,
|
| 194 |
+
"is_decoder": false,
|
| 195 |
+
"is_encoder_decoder": false,
|
| 196 |
+
"label2id": {
|
| 197 |
+
"LABEL_0": 0,
|
| 198 |
+
"LABEL_1": 1
|
| 199 |
+
},
|
| 200 |
+
"length_penalty": 1.0,
|
| 201 |
+
"max_length": 20,
|
| 202 |
+
"max_position_embeddings": 65536,
|
| 203 |
+
"min_length": 0,
|
| 204 |
+
"mlp_only_layers": [],
|
| 205 |
+
"model_type": "qwen3_omni_moe_talker_text",
|
| 206 |
+
"moe_intermediate_size": 384,
|
| 207 |
+
"no_repeat_ngram_size": 0,
|
| 208 |
+
"norm_topk_prob": true,
|
| 209 |
+
"num_attention_heads": 16,
|
| 210 |
+
"num_beam_groups": 1,
|
| 211 |
+
"num_beams": 1,
|
| 212 |
+
"num_experts": 128,
|
| 213 |
+
"num_experts_per_tok": 6,
|
| 214 |
+
"num_hidden_layers": 20,
|
| 215 |
+
"num_key_value_heads": 2,
|
| 216 |
+
"num_return_sequences": 1,
|
| 217 |
+
"output_attentions": false,
|
| 218 |
+
"output_hidden_states": false,
|
| 219 |
+
"output_router_logits": false,
|
| 220 |
+
"output_scores": false,
|
| 221 |
+
"pad_token_id": null,
|
| 222 |
+
"prefix": null,
|
| 223 |
+
"problem_type": null,
|
| 224 |
+
"pruned_heads": {},
|
| 225 |
+
"remove_invalid_values": false,
|
| 226 |
+
"repetition_penalty": 1.0,
|
| 227 |
+
"return_dict": true,
|
| 228 |
+
"return_dict_in_generate": false,
|
| 229 |
+
"rms_norm_eps": 1e-06,
|
| 230 |
+
"rope_scaling": {
|
| 231 |
+
"interleaved": true,
|
| 232 |
+
"mrope_section": [
|
| 233 |
+
24,
|
| 234 |
+
20,
|
| 235 |
+
20
|
| 236 |
+
],
|
| 237 |
+
"rope_type": "default",
|
| 238 |
+
"type": "default"
|
| 239 |
+
},
|
| 240 |
+
"rope_theta": 1000000,
|
| 241 |
+
"router_aux_loss_coef": 0.001,
|
| 242 |
+
"sep_token_id": null,
|
| 243 |
+
"shared_expert_intermediate_size": 768,
|
| 244 |
+
"sliding_window": null,
|
| 245 |
+
"suppress_tokens": null,
|
| 246 |
+
"task_specific_params": null,
|
| 247 |
+
"temperature": 1.0,
|
| 248 |
+
"tf_legacy_loss": false,
|
| 249 |
+
"tie_encoder_decoder": false,
|
| 250 |
+
"tie_word_embeddings": false,
|
| 251 |
+
"tokenizer_class": null,
|
| 252 |
+
"top_k": 50,
|
| 253 |
+
"top_p": 1.0,
|
| 254 |
+
"torchscript": false,
|
| 255 |
+
"typical_p": 1.0,
|
| 256 |
+
"use_bfloat16": false,
|
| 257 |
+
"use_cache": true,
|
| 258 |
+
"use_sliding_window": false,
|
| 259 |
+
"vocab_size": 3072
|
| 260 |
+
},
|
| 261 |
+
"thinker_hidden_size": 2048,
|
| 262 |
+
"video_token_id": 151656,
|
| 263 |
+
"vision_start_token_id": 151652
|
| 264 |
+
},
|
| 265 |
+
"thinker_config": {
|
| 266 |
+
"audio_config": {
|
| 267 |
+
"_name_or_path": "",
|
| 268 |
+
"activation_dropout": 0,
|
| 269 |
+
"activation_function": "gelu",
|
| 270 |
+
"add_cross_attention": false,
|
| 271 |
+
"architectures": null,
|
| 272 |
+
"attention_dropout": 0,
|
| 273 |
+
"bad_words_ids": null,
|
| 274 |
+
"begin_suppress_tokens": null,
|
| 275 |
+
"bos_token_id": null,
|
| 276 |
+
"chunk_size_feed_forward": 0,
|
| 277 |
+
"conv_chunksize": 500,
|
| 278 |
+
"cross_attention_hidden_size": null,
|
| 279 |
+
"d_model": 1280,
|
| 280 |
+
"decoder_start_token_id": null,
|
| 281 |
+
"diversity_penalty": 0.0,
|
| 282 |
+
"do_sample": false,
|
| 283 |
+
"downsample_hidden_size": 480,
|
| 284 |
+
"dropout": 0,
|
| 285 |
+
"dtype": "bfloat16",
|
| 286 |
+
"early_stopping": false,
|
| 287 |
+
"encoder_attention_heads": 20,
|
| 288 |
+
"encoder_ffn_dim": 5120,
|
| 289 |
+
"encoder_layers": 32,
|
| 290 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 291 |
+
"eos_token_id": null,
|
| 292 |
+
"exponential_decay_length_penalty": null,
|
| 293 |
+
"finetuning_task": null,
|
| 294 |
+
"forced_bos_token_id": null,
|
| 295 |
+
"forced_eos_token_id": null,
|
| 296 |
+
"id2label": {
|
| 297 |
+
"0": "LABEL_0",
|
| 298 |
+
"1": "LABEL_1"
|
| 299 |
+
},
|
| 300 |
+
"initializer_range": 0.02,
|
| 301 |
+
"is_decoder": false,
|
| 302 |
+
"is_encoder_decoder": false,
|
| 303 |
+
"label2id": {
|
| 304 |
+
"LABEL_0": 0,
|
| 305 |
+
"LABEL_1": 1
|
| 306 |
+
},
|
| 307 |
+
"length_penalty": 1.0,
|
| 308 |
+
"max_length": 20,
|
| 309 |
+
"max_source_positions": 1500,
|
| 310 |
+
"min_length": 0,
|
| 311 |
+
"model_type": "qwen3_omni_moe_audio_encoder",
|
| 312 |
+
"n_window": 50,
|
| 313 |
+
"n_window_infer": 800,
|
| 314 |
+
"no_repeat_ngram_size": 0,
|
| 315 |
+
"num_beam_groups": 1,
|
| 316 |
+
"num_beams": 1,
|
| 317 |
+
"num_hidden_layers": 32,
|
| 318 |
+
"num_mel_bins": 128,
|
| 319 |
+
"num_return_sequences": 1,
|
| 320 |
+
"output_attentions": false,
|
| 321 |
+
"output_dim": 2048,
|
| 322 |
+
"output_hidden_states": false,
|
| 323 |
+
"output_scores": false,
|
| 324 |
+
"pad_token_id": null,
|
| 325 |
+
"prefix": null,
|
| 326 |
+
"problem_type": null,
|
| 327 |
+
"pruned_heads": {},
|
| 328 |
+
"remove_invalid_values": false,
|
| 329 |
+
"repetition_penalty": 1.0,
|
| 330 |
+
"return_dict": true,
|
| 331 |
+
"return_dict_in_generate": false,
|
| 332 |
+
"scale_embedding": false,
|
| 333 |
+
"sep_token_id": null,
|
| 334 |
+
"suppress_tokens": null,
|
| 335 |
+
"task_specific_params": null,
|
| 336 |
+
"temperature": 1.0,
|
| 337 |
+
"tf_legacy_loss": false,
|
| 338 |
+
"tie_encoder_decoder": false,
|
| 339 |
+
"tie_word_embeddings": true,
|
| 340 |
+
"tokenizer_class": null,
|
| 341 |
+
"top_k": 50,
|
| 342 |
+
"top_p": 1.0,
|
| 343 |
+
"torchscript": false,
|
| 344 |
+
"typical_p": 1.0,
|
| 345 |
+
"use_bfloat16": false
|
| 346 |
+
},
|
| 347 |
+
"audio_end_token_id": 151670,
|
| 348 |
+
"audio_start_token_id": 151669,
|
| 349 |
+
"audio_token_id": 151675,
|
| 350 |
+
"dtype": "bfloat16",
|
| 351 |
+
"image_token_id": 151655,
|
| 352 |
+
"initializer_range": 0.02,
|
| 353 |
+
"model_type": "qwen3_omni_moe_thinker",
|
| 354 |
+
"position_id_per_seconds": 13,
|
| 355 |
+
"seconds_per_chunk": 2,
|
| 356 |
+
"text_config": {
|
| 357 |
+
"_name_or_path": "",
|
| 358 |
+
"add_cross_attention": false,
|
| 359 |
+
"architectures": null,
|
| 360 |
+
"attention_bias": false,
|
| 361 |
+
"attention_dropout": 0.0,
|
| 362 |
+
"bad_words_ids": null,
|
| 363 |
+
"begin_suppress_tokens": null,
|
| 364 |
+
"bos_token_id": null,
|
| 365 |
+
"chunk_size_feed_forward": 0,
|
| 366 |
+
"cross_attention_hidden_size": null,
|
| 367 |
+
"decoder_sparse_step": 1,
|
| 368 |
+
"decoder_start_token_id": null,
|
| 369 |
+
"diversity_penalty": 0.0,
|
| 370 |
+
"do_sample": false,
|
| 371 |
+
"dtype": "bfloat16",
|
| 372 |
+
"early_stopping": false,
|
| 373 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 374 |
+
"eos_token_id": null,
|
| 375 |
+
"exponential_decay_length_penalty": null,
|
| 376 |
+
"finetuning_task": null,
|
| 377 |
+
"forced_bos_token_id": null,
|
| 378 |
+
"forced_eos_token_id": null,
|
| 379 |
+
"head_dim": 128,
|
| 380 |
+
"hidden_act": "silu",
|
| 381 |
+
"hidden_size": 2048,
|
| 382 |
+
"id2label": {
|
| 383 |
+
"0": "LABEL_0",
|
| 384 |
+
"1": "LABEL_1"
|
| 385 |
+
},
|
| 386 |
+
"initializer_range": 0.02,
|
| 387 |
+
"intermediate_size": 768,
|
| 388 |
+
"is_decoder": false,
|
| 389 |
+
"is_encoder_decoder": false,
|
| 390 |
+
"label2id": {
|
| 391 |
+
"LABEL_0": 0,
|
| 392 |
+
"LABEL_1": 1
|
| 393 |
+
},
|
| 394 |
+
"length_penalty": 1.0,
|
| 395 |
+
"max_length": 20,
|
| 396 |
+
"max_position_embeddings": 65536,
|
| 397 |
+
"min_length": 0,
|
| 398 |
+
"mlp_only_layers": [],
|
| 399 |
+
"model_type": "qwen3_omni_moe_text",
|
| 400 |
+
"moe_intermediate_size": 768,
|
| 401 |
+
"no_repeat_ngram_size": 0,
|
| 402 |
+
"norm_topk_prob": true,
|
| 403 |
+
"num_attention_heads": 32,
|
| 404 |
+
"num_beam_groups": 1,
|
| 405 |
+
"num_beams": 1,
|
| 406 |
+
"num_experts": 128,
|
| 407 |
+
"num_experts_per_tok": 8,
|
| 408 |
+
"num_hidden_layers": 48,
|
| 409 |
+
"num_key_value_heads": 4,
|
| 410 |
+
"num_return_sequences": 1,
|
| 411 |
+
"output_attentions": false,
|
| 412 |
+
"output_hidden_states": false,
|
| 413 |
+
"output_router_logits": false,
|
| 414 |
+
"output_scores": false,
|
| 415 |
+
"pad_token_id": null,
|
| 416 |
+
"prefix": null,
|
| 417 |
+
"problem_type": null,
|
| 418 |
+
"pruned_heads": {},
|
| 419 |
+
"remove_invalid_values": false,
|
| 420 |
+
"repetition_penalty": 1.0,
|
| 421 |
+
"return_dict": true,
|
| 422 |
+
"return_dict_in_generate": false,
|
| 423 |
+
"rms_norm_eps": 1e-06,
|
| 424 |
+
"rope_scaling": {
|
| 425 |
+
"interleaved": true,
|
| 426 |
+
"mrope_interleaved": true,
|
| 427 |
+
"mrope_section": [
|
| 428 |
+
24,
|
| 429 |
+
20,
|
| 430 |
+
20
|
| 431 |
+
],
|
| 432 |
+
"rope_type": "default",
|
| 433 |
+
"type": "default"
|
| 434 |
+
},
|
| 435 |
+
"rope_theta": 1000000,
|
| 436 |
+
"router_aux_loss_coef": 0.001,
|
| 437 |
+
"sep_token_id": null,
|
| 438 |
+
"shared_expert_intermediate_size": 0,
|
| 439 |
+
"sliding_window": null,
|
| 440 |
+
"suppress_tokens": null,
|
| 441 |
+
"task_specific_params": null,
|
| 442 |
+
"temperature": 1.0,
|
| 443 |
+
"tf_legacy_loss": false,
|
| 444 |
+
"tie_encoder_decoder": false,
|
| 445 |
+
"tie_word_embeddings": false,
|
| 446 |
+
"tokenizer_class": null,
|
| 447 |
+
"top_k": 50,
|
| 448 |
+
"top_p": 1.0,
|
| 449 |
+
"torchscript": false,
|
| 450 |
+
"typical_p": 1.0,
|
| 451 |
+
"use_bfloat16": false,
|
| 452 |
+
"use_cache": true,
|
| 453 |
+
"use_qk_norm": true,
|
| 454 |
+
"use_sliding_window": false,
|
| 455 |
+
"vocab_size": 152064
|
| 456 |
+
},
|
| 457 |
+
"user_token_id": 872,
|
| 458 |
+
"video_token_id": 151656,
|
| 459 |
+
"vision_config": {
|
| 460 |
+
"_name_or_path": "",
|
| 461 |
+
"add_cross_attention": false,
|
| 462 |
+
"apply_vit_abs_pos_embed": true,
|
| 463 |
+
"architectures": null,
|
| 464 |
+
"bad_words_ids": null,
|
| 465 |
+
"begin_suppress_tokens": null,
|
| 466 |
+
"bos_token_id": null,
|
| 467 |
+
"chunk_size_feed_forward": 0,
|
| 468 |
+
"cross_attention_hidden_size": null,
|
| 469 |
+
"decoder_start_token_id": null,
|
| 470 |
+
"deepstack_visual_indexes": [
|
| 471 |
+
8,
|
| 472 |
+
16,
|
| 473 |
+
24
|
| 474 |
+
],
|
| 475 |
+
"depth": 27,
|
| 476 |
+
"diversity_penalty": 0.0,
|
| 477 |
+
"do_sample": false,
|
| 478 |
+
"dtype": "bfloat16",
|
| 479 |
+
"early_stopping": false,
|
| 480 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 481 |
+
"eos_token_id": null,
|
| 482 |
+
"exponential_decay_length_penalty": null,
|
| 483 |
+
"finetuning_task": null,
|
| 484 |
+
"forced_bos_token_id": null,
|
| 485 |
+
"forced_eos_token_id": null,
|
| 486 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 487 |
+
"hidden_size": 1152,
|
| 488 |
+
"id2label": {
|
| 489 |
+
"0": "LABEL_0",
|
| 490 |
+
"1": "LABEL_1"
|
| 491 |
+
},
|
| 492 |
+
"image_size": 768,
|
| 493 |
+
"in_channels": 3,
|
| 494 |
+
"in_chans": 3,
|
| 495 |
+
"initializer_range": 0.02,
|
| 496 |
+
"intermediate_size": 4304,
|
| 497 |
+
"is_decoder": false,
|
| 498 |
+
"is_encoder_decoder": false,
|
| 499 |
+
"label2id": {
|
| 500 |
+
"LABEL_0": 0,
|
| 501 |
+
"LABEL_1": 1
|
| 502 |
+
},
|
| 503 |
+
"length_penalty": 1.0,
|
| 504 |
+
"max_length": 20,
|
| 505 |
+
"min_length": 0,
|
| 506 |
+
"model_type": "qwen3_omni_moe_vision_encoder",
|
| 507 |
+
"no_repeat_ngram_size": 0,
|
| 508 |
+
"num_beam_groups": 1,
|
| 509 |
+
"num_beams": 1,
|
| 510 |
+
"num_heads": 16,
|
| 511 |
+
"num_position_embeddings": 2304,
|
| 512 |
+
"num_return_sequences": 1,
|
| 513 |
+
"out_hidden_size": 2048,
|
| 514 |
+
"output_attentions": false,
|
| 515 |
+
"output_hidden_states": false,
|
| 516 |
+
"output_scores": false,
|
| 517 |
+
"pad_token_id": null,
|
| 518 |
+
"patch_size": 16,
|
| 519 |
+
"prefix": null,
|
| 520 |
+
"problem_type": null,
|
| 521 |
+
"pruned_heads": {},
|
| 522 |
+
"remove_invalid_values": false,
|
| 523 |
+
"repetition_penalty": 1.0,
|
| 524 |
+
"return_dict": true,
|
| 525 |
+
"return_dict_in_generate": false,
|
| 526 |
+
"sep_token_id": null,
|
| 527 |
+
"spatial_merge_size": 2,
|
| 528 |
+
"spatial_patch_size": 16,
|
| 529 |
+
"suppress_tokens": null,
|
| 530 |
+
"task_specific_params": null,
|
| 531 |
+
"temperature": 1.0,
|
| 532 |
+
"temporal_patch_size": 2,
|
| 533 |
+
"tf_legacy_loss": false,
|
| 534 |
+
"tie_encoder_decoder": false,
|
| 535 |
+
"tie_word_embeddings": true,
|
| 536 |
+
"tokenizer_class": null,
|
| 537 |
+
"tokens_per_second": 2,
|
| 538 |
+
"top_k": 50,
|
| 539 |
+
"top_p": 1.0,
|
| 540 |
+
"torchscript": false,
|
| 541 |
+
"typical_p": 1.0,
|
| 542 |
+
"use_bfloat16": false
|
| 543 |
+
},
|
| 544 |
+
"vision_end_token_id": 151653,
|
| 545 |
+
"vision_start_token_id": 151652
|
| 546 |
+
},
|
| 547 |
+
"transformers_version": "4.57.6",
|
| 548 |
+
"tts_bos_token_id": 151672,
|
| 549 |
+
"tts_eos_token_id": 151673,
|
| 550 |
+
"tts_pad_token_id": 151671,
|
| 551 |
+
"user_token_id": 872
|
| 552 |
+
}
|