Sync from GitHub via hub-sync
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- README.md +1 -43
- chat_template.jinja +0 -117
- config.json +0 -44
- generation_config.json +0 -13
- merges.txt +0 -0
- model-00001-of-00040.safetensors +0 -3
- model-00002-of-00040.safetensors +0 -3
- model-00003-of-00040.safetensors +0 -3
- model-00004-of-00040.safetensors +0 -3
- model-00005-of-00040.safetensors +0 -3
- model-00006-of-00040.safetensors +0 -3
- model-00007-of-00040.safetensors +0 -3
- model-00008-of-00040.safetensors +0 -3
- model-00009-of-00040.safetensors +0 -3
- model-00010-of-00040.safetensors +0 -3
- model-00011-of-00040.safetensors +0 -3
- model-00012-of-00040.safetensors +0 -3
- model-00013-of-00040.safetensors +0 -3
- model-00014-of-00040.safetensors +0 -3
- model-00015-of-00040.safetensors +0 -3
- model-00016-of-00040.safetensors +0 -3
- model-00017-of-00040.safetensors +0 -3
- model-00018-of-00040.safetensors +0 -3
- model-00019-of-00040.safetensors +0 -3
- model-00020-of-00040.safetensors +0 -3
- model-00021-of-00040.safetensors +0 -3
- model-00022-of-00040.safetensors +0 -3
- model-00023-of-00040.safetensors +0 -3
- model-00024-of-00040.safetensors +0 -3
- model-00025-of-00040.safetensors +0 -3
- model-00026-of-00040.safetensors +0 -3
- model-00027-of-00040.safetensors +0 -3
- model-00028-of-00040.safetensors +0 -3
- model-00029-of-00040.safetensors +0 -3
- model-00030-of-00040.safetensors +0 -3
- model-00031-of-00040.safetensors +0 -3
- model-00032-of-00040.safetensors +0 -3
- model-00033-of-00040.safetensors +0 -3
- model-00034-of-00040.safetensors +0 -3
- model-00035-of-00040.safetensors +0 -3
- model-00036-of-00040.safetensors +0 -3
- model-00037-of-00040.safetensors +0 -3
- model-00038-of-00040.safetensors +0 -3
- model-00039-of-00040.safetensors +0 -3
- model-00040-of-00040.safetensors +0 -3
- model.safetensors.index.json +0 -0
- qwen3_coder_detector_sgl.py +0 -474
- qwen3coder_tool_parser_vllm.py +0 -690
- tokenizer.json +0 -0
- tokenizer_config.json +0 -239
README.md
CHANGED
|
@@ -1,43 +1 @@
|
|
| 1 |
-
---
|
| 2 |
-
library_name: transformers
|
| 3 |
-
license: apache-2.0
|
| 4 |
-
license_link: https://huggingface.co/Qwen/Qwen3-Coder-Next/blob/main/LICENSE
|
| 5 |
-
pipeline_tag: text-generation
|
| 6 |
-
---
|
| 7 |
-
|
| 8 |
-
# 🐈⬛ Caracat Code
|
| 9 |
-
|
| 10 |
-
**Caracat Code** is an AI coding model based on **Qwen3-Coder-Next**.
|
| 11 |
-
|
| 12 |
-
The goal of Caracat Code is to create a powerful and flexible coding model for software development, code generation, debugging, code analysis, and AI coding agents.
|
| 13 |
-
|
| 14 |
-
## 🚀 Features
|
| 15 |
-
|
| 16 |
-
- 💻 Code generation
|
| 17 |
-
- 🧠 Code understanding and analysis
|
| 18 |
-
- 🐛 Debugging and error fixing
|
| 19 |
-
- 🔧 Code refactoring and optimization
|
| 20 |
-
- 🤖 Support for coding-agent workflows
|
| 21 |
-
- 📚 Fine-tuning and further development
|
| 22 |
-
|
| 23 |
-
## 🧬 Base Model
|
| 24 |
-
|
| 25 |
-
Caracat Code is based on:
|
| 26 |
-
|
| 27 |
-
**Qwen3-Coder-Next**
|
| 28 |
-
|
| 29 |
-
The original model is provided by Qwen and is licensed under **Apache 2.0**.
|
| 30 |
-
|
| 31 |
-
## 🛠️ Development
|
| 32 |
-
|
| 33 |
-
Caracat Code is an ongoing project. The model may be further trained, fine-tuned, optimized, and modified over time.
|
| 34 |
-
|
| 35 |
-
## 📜 License
|
| 36 |
-
|
| 37 |
-
Caracat Code follows the licensing requirements of its base model and its own modifications.
|
| 38 |
-
|
| 39 |
-
For the original Qwen3-Coder-Next model, see the original model repository and its Apache 2.0 license.
|
| 40 |
-
|
| 41 |
-
---
|
| 42 |
-
|
| 43 |
-
🐈⬛ **Caracat Code — Coding, reimagined.**
|
|
|
|
| 1 |
+
# training-and-devoloping-caracat-code
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
chat_template.jinja
DELETED
|
@@ -1,117 +0,0 @@
|
|
| 1 |
-
{% macro render_extra_keys(json_dict, handled_keys) %}
|
| 2 |
-
{%- if json_dict is mapping %}
|
| 3 |
-
{%- for json_key in json_dict if json_key not in handled_keys %}
|
| 4 |
-
{%- if json_dict[json_key] is string %}
|
| 5 |
-
{{-'\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | string) ~ '</' ~ json_key ~ '>' }}
|
| 6 |
-
{%- else %}
|
| 7 |
-
{{- '\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | tojson | safe) ~ '</' ~ json_key ~ '>' }}
|
| 8 |
-
{%- endif %}
|
| 9 |
-
{%- endfor %}
|
| 10 |
-
{%- endif %}
|
| 11 |
-
{%- endmacro %}
|
| 12 |
-
|
| 13 |
-
{%- if messages[0]["role"] == "system" %}
|
| 14 |
-
{%- set system_message = messages[0]["content"] %}
|
| 15 |
-
{%- set loop_messages = messages[1:] %}
|
| 16 |
-
{%- else %}
|
| 17 |
-
{%- set loop_messages = messages %}
|
| 18 |
-
{%- endif %}
|
| 19 |
-
|
| 20 |
-
{%- if not tools is defined %}
|
| 21 |
-
{%- set tools = [] %}
|
| 22 |
-
{%- endif %}
|
| 23 |
-
|
| 24 |
-
{%- if system_message is defined %}
|
| 25 |
-
{{- "<|im_start|>system\n" + system_message }}
|
| 26 |
-
{%- else %}
|
| 27 |
-
{%- if tools is iterable and tools | length > 0 %}
|
| 28 |
-
{{- "<|im_start|>system\nYou are Qwen, a helpful AI assistant that can interact with a computer to solve tasks." }}
|
| 29 |
-
{%- endif %}
|
| 30 |
-
{%- endif %}
|
| 31 |
-
{%- if tools is iterable and tools | length > 0 %}
|
| 32 |
-
{{- "\n\n# Tools\n\nYou have access to the following functions:\n\n" }}
|
| 33 |
-
{{- "<tools>" }}
|
| 34 |
-
{%- for tool in tools %}
|
| 35 |
-
{%- if tool.function is defined %}
|
| 36 |
-
{%- set tool = tool.function %}
|
| 37 |
-
{%- endif %}
|
| 38 |
-
{{- "\n<function>\n<name>" ~ tool.name ~ "</name>" }}
|
| 39 |
-
{%- if tool.description is defined %}
|
| 40 |
-
{{- '\n<description>' ~ (tool.description | trim) ~ '</description>' }}
|
| 41 |
-
{%- endif %}
|
| 42 |
-
{{- '\n<parameters>' }}
|
| 43 |
-
{%- if tool.parameters is defined and tool.parameters is mapping and tool.parameters.properties is defined and tool.parameters.properties is mapping %}
|
| 44 |
-
{%- for param_name, param_fields in tool.parameters.properties|items %}
|
| 45 |
-
{{- '\n<parameter>' }}
|
| 46 |
-
{{- '\n<name>' ~ param_name ~ '</name>' }}
|
| 47 |
-
{%- if param_fields.type is defined %}
|
| 48 |
-
{{- '\n<type>' ~ (param_fields.type | string) ~ '</type>' }}
|
| 49 |
-
{%- endif %}
|
| 50 |
-
{%- if param_fields.description is defined %}
|
| 51 |
-
{{- '\n<description>' ~ (param_fields.description | trim) ~ '</description>' }}
|
| 52 |
-
{%- endif %}
|
| 53 |
-
{%- set handled_keys = ['name', 'type', 'description'] %}
|
| 54 |
-
{{- render_extra_keys(param_fields, handled_keys) }}
|
| 55 |
-
{{- '\n</parameter>' }}
|
| 56 |
-
{%- endfor %}
|
| 57 |
-
{%- endif %}
|
| 58 |
-
{%- set handled_keys = ['type', 'properties'] %}
|
| 59 |
-
{{- render_extra_keys(tool.parameters, handled_keys) }}
|
| 60 |
-
{{- '\n</parameters>' }}
|
| 61 |
-
{%- set handled_keys = ['type', 'name', 'description', 'parameters'] %}
|
| 62 |
-
{{- render_extra_keys(tool, handled_keys) }}
|
| 63 |
-
{{- '\n</function>' }}
|
| 64 |
-
{%- endfor %}
|
| 65 |
-
{{- "\n</tools>" }}
|
| 66 |
-
{{- '\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>' }}
|
| 67 |
-
{%- endif %}
|
| 68 |
-
{%- if system_message is defined %}
|
| 69 |
-
{{- '<|im_end|>\n' }}
|
| 70 |
-
{%- else %}
|
| 71 |
-
{%- if tools is iterable and tools | length > 0 %}
|
| 72 |
-
{{- '<|im_end|>\n' }}
|
| 73 |
-
{%- endif %}
|
| 74 |
-
{%- endif %}
|
| 75 |
-
{%- for message in loop_messages %}
|
| 76 |
-
{%- if message.role == "assistant" and message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}
|
| 77 |
-
{{- '<|im_start|>' + message.role }}
|
| 78 |
-
{%- if message.content is defined and message.content is string and message.content | trim | length > 0 %}
|
| 79 |
-
{{- '\n' + message.content | trim + '\n' }}
|
| 80 |
-
{%- endif %}
|
| 81 |
-
{%- for tool_call in message.tool_calls %}
|
| 82 |
-
{%- if tool_call.function is defined %}
|
| 83 |
-
{%- set tool_call = tool_call.function %}
|
| 84 |
-
{%- endif %}
|
| 85 |
-
{{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
|
| 86 |
-
{%- if tool_call.arguments is defined %}
|
| 87 |
-
{%- for args_name, args_value in tool_call.arguments|items %}
|
| 88 |
-
{{- '<parameter=' + args_name + '>\n' }}
|
| 89 |
-
{%- set args_value = args_value if args_value is string else args_value | tojson | safe %}
|
| 90 |
-
{{- args_value }}
|
| 91 |
-
{{- '\n</parameter>\n' }}
|
| 92 |
-
{%- endfor %}
|
| 93 |
-
{%- endif %}
|
| 94 |
-
{{- '</function>\n</tool_call>' }}
|
| 95 |
-
{%- endfor %}
|
| 96 |
-
{{- '<|im_end|>\n' }}
|
| 97 |
-
{%- elif message.role == "user" or message.role == "system" or message.role == "assistant" %}
|
| 98 |
-
{{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
|
| 99 |
-
{%- elif message.role == "tool" %}
|
| 100 |
-
{%- if loop.previtem and loop.previtem.role != "tool" %}
|
| 101 |
-
{{- '<|im_start|>user' }}
|
| 102 |
-
{%- endif %}
|
| 103 |
-
{{- '\n<tool_response>\n' }}
|
| 104 |
-
{{- message.content }}
|
| 105 |
-
{{- '\n</tool_response>' }}
|
| 106 |
-
{%- if not loop.last and loop.nextitem.role != "tool" %}
|
| 107 |
-
{{- '<|im_end|>\n' }}
|
| 108 |
-
{%- elif loop.last %}
|
| 109 |
-
{{- '<|im_end|>\n' }}
|
| 110 |
-
{%- endif %}
|
| 111 |
-
{%- else %}
|
| 112 |
-
{{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' }}
|
| 113 |
-
{%- endif %}
|
| 114 |
-
{%- endfor %}
|
| 115 |
-
{%- if add_generation_prompt %}
|
| 116 |
-
{{- '<|im_start|>assistant\n' }}
|
| 117 |
-
{%- endif %}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
config.json
DELETED
|
@@ -1,44 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"architectures": [
|
| 3 |
-
"Qwen3NextForCausalLM"
|
| 4 |
-
],
|
| 5 |
-
"attention_bias": false,
|
| 6 |
-
"attention_dropout": 0,
|
| 7 |
-
"bos_token_id": 151643,
|
| 8 |
-
"decoder_sparse_step": 1,
|
| 9 |
-
"eos_token_id": 151645,
|
| 10 |
-
"full_attention_interval": 4,
|
| 11 |
-
"head_dim": 256,
|
| 12 |
-
"hidden_act": "silu",
|
| 13 |
-
"hidden_size": 2048,
|
| 14 |
-
"initializer_range": 0.02,
|
| 15 |
-
"intermediate_size": 5120,
|
| 16 |
-
"linear_conv_kernel_dim": 4,
|
| 17 |
-
"linear_key_head_dim": 128,
|
| 18 |
-
"linear_num_key_heads": 16,
|
| 19 |
-
"linear_num_value_heads": 32,
|
| 20 |
-
"linear_value_head_dim": 128,
|
| 21 |
-
"max_position_embeddings": 262144,
|
| 22 |
-
"mlp_only_layers": [],
|
| 23 |
-
"model_type": "qwen3_next",
|
| 24 |
-
"moe_intermediate_size": 512,
|
| 25 |
-
"norm_topk_prob": true,
|
| 26 |
-
"num_attention_heads": 16,
|
| 27 |
-
"num_experts": 512,
|
| 28 |
-
"num_experts_per_tok": 10,
|
| 29 |
-
"num_hidden_layers": 48,
|
| 30 |
-
"num_key_value_heads": 2,
|
| 31 |
-
"output_router_logits": false,
|
| 32 |
-
"partial_rotary_factor": 0.25,
|
| 33 |
-
"rms_norm_eps": 1e-06,
|
| 34 |
-
"rope_scaling": null,
|
| 35 |
-
"rope_theta": 5000000,
|
| 36 |
-
"router_aux_loss_coef": 0.001,
|
| 37 |
-
"shared_expert_intermediate_size": 512,
|
| 38 |
-
"tie_word_embeddings": false,
|
| 39 |
-
"torch_dtype": "bfloat16",
|
| 40 |
-
"transformers_version": "4.57.0.dev0",
|
| 41 |
-
"use_cache": true,
|
| 42 |
-
"use_sliding_window": false,
|
| 43 |
-
"vocab_size": 151936
|
| 44 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
generation_config.json
DELETED
|
@@ -1,13 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"bos_token_id": 151643,
|
| 3 |
-
"do_sample": true,
|
| 4 |
-
"eos_token_id": [
|
| 5 |
-
151645,
|
| 6 |
-
151643
|
| 7 |
-
],
|
| 8 |
-
"pad_token_id": 151643,
|
| 9 |
-
"temperature": 1.0,
|
| 10 |
-
"top_k": 40,
|
| 11 |
-
"top_p": 0.95,
|
| 12 |
-
"transformers_version": "4.57.3"
|
| 13 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
merges.txt
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
model-00001-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8e9a517133bfbdc6806cf8b61793055a260efeb68e6e019fd90e4bbb1b665d0a
|
| 3 |
-
size 3999619288
|
|
|
|
|
|
|
|
|
|
|
|
model-00002-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0ab43285dec19a2eba2705538dfc914dd37b3c85477d3769054e44f300e659a3
|
| 3 |
-
size 3999841816
|
|
|
|
|
|
|
|
|
|
|
|
model-00003-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:68a5796b8bd75672e471daf4488bdb6c792ad895772efeacfdc41c09eeb3eb6b
|
| 3 |
-
size 3999515608
|
|
|
|
|
|
|
|
|
|
|
|
model-00004-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:93f07b2ae634bdaee995c35d16d1186e7481f202e05095d17b1fc162e62c45f6
|
| 3 |
-
size 3999842032
|
|
|
|
|
|
|
|
|
|
|
|
model-00005-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:9832c0217fc2ff64f1a051769aa51c004b39fcc12d785e336217087c968651f5
|
| 3 |
-
size 3999842240
|
|
|
|
|
|
|
|
|
|
|
|
model-00006-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:983dce04bba12c65e457feeeb37cfb13756690b63d75d93cea3b56c361cd6859
|
| 3 |
-
size 3999853248
|
|
|
|
|
|
|
|
|
|
|
|
model-00007-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:41f27ff915d0797f6016d32d30beb3ab092b53e56ad99461d130854b2db64bc1
|
| 3 |
-
size 3999841944
|
|
|
|
|
|
|
|
|
|
|
|
model-00008-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:7382d1f4621bcef01ebf2df6ce855692f48bc94184bbfe2afab30b8d976c87de
|
| 3 |
-
size 3999842032
|
|
|
|
|
|
|
|
|
|
|
|
model-00009-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8ba622e8d78aac210b56f250e3faa6c759ab858b3d683fc0520015a60d8674b2
|
| 3 |
-
size 3999843224
|
|
|
|
|
|
|
|
|
|
|
|
model-00010-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0b5ca0b7ddaa8d5f481753d8357db9ec5c66485d93732168a7e216e8bfda6f1f
|
| 3 |
-
size 3999517840
|
|
|
|
|
|
|
|
|
|
|
|
model-00011-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e85c1ce556170c785b9479b2e0c8302e2a5bc69ccf44bdc2d8b5477d15cfb356
|
| 3 |
-
size 4000181328
|
|
|
|
|
|
|
|
|
|
|
|
model-00012-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:91f8c5b1a133c191348ac8631474e3114b8f084a68dc4eb93664130efd758b79
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00013-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:a571b802c7946fc33c90f4c5728c53505b5b94cb002a4567541a9a1899c07495
|
| 3 |
-
size 3999517504
|
|
|
|
|
|
|
|
|
|
|
|
model-00014-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:32e918bdba2d2b9b7ff866e361f99626f7df92299df1e5e195f85f1ea22c8911
|
| 3 |
-
size 3999844016
|
|
|
|
|
|
|
|
|
|
|
|
model-00015-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:496865f049712fb4176154f9a2c7d48063fe855a17b2ad025466a3a316692fb4
|
| 3 |
-
size 4000181768
|
|
|
|
|
|
|
|
|
|
|
|
model-00016-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:9f1ff8098b81710184f274232b04ca4fbdd820afd28213bde60e4630ed57048f
|
| 3 |
-
size 3999517280
|
|
|
|
|
|
|
|
|
|
|
|
model-00017-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:04d03dd323e34ea5eb409a41685b515ef36d1c943c4546d31e9c73d96f8c0c7d
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00018-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:cc03f69e8d061957d9b5d08b7da0b40d79b27a34e8c4c3fb8d503b123b268df1
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00019-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:040ea97dd90ebb9a183fdbcc1bdfba1228f1fc56eeb35dbd09fe3dc807dbae41
|
| 3 |
-
size 3999844120
|
|
|
|
|
|
|
|
|
|
|
|
model-00020-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:036b86c2808d34d0d6972424937753ca434713defe433b04c1cf1ddde4c77685
|
| 3 |
-
size 3999855064
|
|
|
|
|
|
|
|
|
|
|
|
model-00021-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:f93091cfc1df5f0d290b43517382f06cc96709b71cc494431d36c217d4e05035
|
| 3 |
-
size 3999843824
|
|
|
|
|
|
|
|
|
|
|
|
model-00022-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:1173933660614dbcd56463c30573e4bc52f8b3e674217294149abd0368332366
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00023-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:d6c7d46d0b74aba6b7125c4797393900cf54c8545fd0da99485f24df7309e470
|
| 3 |
-
size 3999517496
|
|
|
|
|
|
|
|
|
|
|
|
model-00024-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e98c2798567def2ef801648b8e1ef5e0451a04dcb4586f19c7d6843c7e71160f
|
| 3 |
-
size 3999844296
|
|
|
|
|
|
|
|
|
|
|
|
model-00025-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:14b34f3e8899eb81a427320b7a17d94387f2e762af8dadcf667c28957c46b992
|
| 3 |
-
size 4000181328
|
|
|
|
|
|
|
|
|
|
|
|
model-00026-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8f0c2a1bae6835037643e18e3d964f3cd59679fe31ed2c57d3d3e0e2902ec3e0
|
| 3 |
-
size 3999517504
|
|
|
|
|
|
|
|
|
|
|
|
model-00027-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:2c2aefa4ae3e70a2275f151669076c857df1549c5d13458947a8f5459d5531f3
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00028-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8f3f40bafcd7bd2d902813ee3110746c8f46f2f646d11cf4c4611cf59d39001f
|
| 3 |
-
size 3999844016
|
|
|
|
|
|
|
|
|
|
|
|
model-00029-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:69f5a1c4f50c3d24fff41a7a0b4ef806c6ab1e2469f6523acbe0b6a7b6a840d2
|
| 3 |
-
size 3999855352
|
|
|
|
|
|
|
|
|
|
|
|
model-00030-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:518965232d096a23f01c9d955ad925828ed22e77e2a067013ba549820ad8231f
|
| 3 |
-
size 3999843704
|
|
|
|
|
|
|
|
|
|
|
|
model-00031-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:a24847e00ccfb3b510135d245083622f6f3337c016aa80c5bafceabf584e63c4
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00032-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:605e2609278cd172ab8e0fab95e89fed2e7884b2bad378a516d35275812c9031
|
| 3 |
-
size 3999843912
|
|
|
|
|
|
|
|
|
|
|
|
model-00033-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:3d2aa165f7b0d2a8acb66d869f823d7aa6344db75c6e1be833480acc2b252fdf
|
| 3 |
-
size 3999517720
|
|
|
|
|
|
|
|
|
|
|
|
model-00034-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:04bd78b03a131401169af0e67d96db5c6071631010ac19e9c0c9b7fb8c3f1871
|
| 3 |
-
size 4000181528
|
|
|
|
|
|
|
|
|
|
|
|
model-00035-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:eb6da0385f8ba56b98a6677bdc59f1c9cfff291024d61f26554b1881bba8edab
|
| 3 |
-
size 3999843824
|
|
|
|
|
|
|
|
|
|
|
|
model-00036-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:4925757ada6c754070fd6381c5e01225ca51c19213c83011ba6c39e541dda984
|
| 3 |
-
size 3999517504
|
|
|
|
|
|
|
|
|
|
|
|
model-00037-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:12528012534bba5c61b237f3441864f55707ec050e1372d8eaaf9fdce259cffd
|
| 3 |
-
size 3999843904
|
|
|
|
|
|
|
|
|
|
|
|
model-00038-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:58d4fd2e057fe619a690ee969bd79f85e45b0862df6a5a433abeff31abf03f7f
|
| 3 |
-
size 3999844296
|
|
|
|
|
|
|
|
|
|
|
|
model-00039-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:407d45b48d0682469ac555c84c6aa3ce6ee8c56759074a532670a4ab035fade3
|
| 3 |
-
size 3999854920
|
|
|
|
|
|
|
|
|
|
|
|
model-00040-of-00040.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:9606cbc3a087efaba8e17cb97b9d91cd8d25e6cba6958079d58c19871d272053
|
| 3 |
-
size 3365572528
|
|
|
|
|
|
|
|
|
|
|
|
model.safetensors.index.json
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
qwen3_coder_detector_sgl.py
DELETED
|
@@ -1,474 +0,0 @@
|
|
| 1 |
-
import ast
|
| 2 |
-
import json
|
| 3 |
-
import logging
|
| 4 |
-
import re
|
| 5 |
-
from typing import Any, List, Optional
|
| 6 |
-
|
| 7 |
-
from sglang.srt.entrypoints.openai.protocol import Tool
|
| 8 |
-
from sglang.srt.function_call.base_format_detector import BaseFormatDetector
|
| 9 |
-
from sglang.srt.function_call.core_types import (
|
| 10 |
-
StreamingParseResult,
|
| 11 |
-
ToolCallItem,
|
| 12 |
-
_GetInfoFunc,
|
| 13 |
-
)
|
| 14 |
-
|
| 15 |
-
logger = logging.getLogger(__name__)
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
class Qwen3CoderDetector(BaseFormatDetector):
|
| 19 |
-
def __init__(self):
|
| 20 |
-
super().__init__()
|
| 21 |
-
|
| 22 |
-
# Sentinel tokens
|
| 23 |
-
self.tool_call_start_token: str = "<tool_call>"
|
| 24 |
-
self.tool_call_end_token: str = "</tool_call>"
|
| 25 |
-
self.tool_call_prefix: str = "<function="
|
| 26 |
-
self.function_end_token: str = "</function>"
|
| 27 |
-
self.parameter_prefix: str = "<parameter="
|
| 28 |
-
self.parameter_end_token: str = "</parameter>"
|
| 29 |
-
|
| 30 |
-
# Regex for non-streaming fallback
|
| 31 |
-
self.tool_call_regex = re.compile(r"<tool_call>(.*?)</tool_call>", re.DOTALL)
|
| 32 |
-
self.tool_call_function_regex = re.compile(
|
| 33 |
-
r"<function=(.*?)</function>|<function=(.*)$", re.DOTALL
|
| 34 |
-
)
|
| 35 |
-
self.tool_call_parameter_regex = re.compile(
|
| 36 |
-
r"<parameter=(.*?)(?:</parameter>|(?=<parameter=)|(?=</function>)|$)",
|
| 37 |
-
re.DOTALL,
|
| 38 |
-
)
|
| 39 |
-
|
| 40 |
-
# Streaming State
|
| 41 |
-
# Base class already initializes _buffer, we just use it directly
|
| 42 |
-
# No need to check with hasattr - we control the lifecycle through inheritance
|
| 43 |
-
|
| 44 |
-
# Index pointing to the next character to be processed in buffer
|
| 45 |
-
self.parsed_pos: int = 0
|
| 46 |
-
# Parameter count inside the current tool being processed, used to determine whether to add comma
|
| 47 |
-
self.current_tool_param_count: int = 0
|
| 48 |
-
# Flag indicating whether current tool has already sent '{'
|
| 49 |
-
self.json_started: bool = False
|
| 50 |
-
|
| 51 |
-
# [FIX] New state flag: mark whether inside tool_call structure block
|
| 52 |
-
self.is_inside_tool_call: bool = False
|
| 53 |
-
|
| 54 |
-
# Initialize attributes that were missing in the original PR
|
| 55 |
-
self.current_func_name: Optional[str] = None
|
| 56 |
-
|
| 57 |
-
def has_tool_call(self, text: str) -> bool:
|
| 58 |
-
return self.tool_call_start_token in text
|
| 59 |
-
|
| 60 |
-
def _get_arguments_config(
|
| 61 |
-
self, func_name: str, tools: Optional[list[Tool]]
|
| 62 |
-
) -> dict:
|
| 63 |
-
"""Extract argument configuration for a function."""
|
| 64 |
-
if tools is None:
|
| 65 |
-
return {}
|
| 66 |
-
for config in tools:
|
| 67 |
-
try:
|
| 68 |
-
config_type = config.type
|
| 69 |
-
config_function = config.function
|
| 70 |
-
config_function_name = config_function.name
|
| 71 |
-
except AttributeError:
|
| 72 |
-
continue
|
| 73 |
-
|
| 74 |
-
if config_type == "function" and config_function_name == func_name:
|
| 75 |
-
try:
|
| 76 |
-
params = config_function.parameters
|
| 77 |
-
except AttributeError:
|
| 78 |
-
return {}
|
| 79 |
-
|
| 80 |
-
if isinstance(params, dict) and "properties" in params:
|
| 81 |
-
return params["properties"]
|
| 82 |
-
elif isinstance(params, dict):
|
| 83 |
-
return params
|
| 84 |
-
else:
|
| 85 |
-
return {}
|
| 86 |
-
logger.warning(f"Tool '{func_name}' is not defined in the tools list.")
|
| 87 |
-
return {}
|
| 88 |
-
|
| 89 |
-
def _convert_param_value(
|
| 90 |
-
self, param_value: str, param_name: str, param_config: dict, func_name: str
|
| 91 |
-
) -> Any:
|
| 92 |
-
"""Convert parameter value based on its type in the schema."""
|
| 93 |
-
# Handle null value for any type
|
| 94 |
-
if param_value.lower() == "null":
|
| 95 |
-
return None
|
| 96 |
-
|
| 97 |
-
if param_name not in param_config:
|
| 98 |
-
if param_config != {}:
|
| 99 |
-
logger.warning(
|
| 100 |
-
f"Parsed parameter '{param_name}' is not defined in the tool "
|
| 101 |
-
f"parameters for tool '{func_name}', directly returning the string value."
|
| 102 |
-
)
|
| 103 |
-
return param_value
|
| 104 |
-
|
| 105 |
-
if (
|
| 106 |
-
isinstance(param_config[param_name], dict)
|
| 107 |
-
and "type" in param_config[param_name]
|
| 108 |
-
):
|
| 109 |
-
param_type = str(param_config[param_name]["type"]).strip().lower()
|
| 110 |
-
else:
|
| 111 |
-
param_type = "string"
|
| 112 |
-
if param_type in ["string", "str", "text", "varchar", "char", "enum"]:
|
| 113 |
-
return param_value
|
| 114 |
-
elif (
|
| 115 |
-
param_type.startswith("int")
|
| 116 |
-
or param_type.startswith("uint")
|
| 117 |
-
or param_type.startswith("long")
|
| 118 |
-
or param_type.startswith("short")
|
| 119 |
-
or param_type.startswith("unsigned")
|
| 120 |
-
):
|
| 121 |
-
try:
|
| 122 |
-
param_value = int(param_value)
|
| 123 |
-
except Exception:
|
| 124 |
-
logger.warning(
|
| 125 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' is not an integer in tool "
|
| 126 |
-
f"'{func_name}', degenerating to string."
|
| 127 |
-
)
|
| 128 |
-
return param_value
|
| 129 |
-
elif param_type.startswith("num") or param_type.startswith("float"):
|
| 130 |
-
try:
|
| 131 |
-
maybe_convert = (
|
| 132 |
-
False if "." in param_value or "e" in param_value.lower() else True
|
| 133 |
-
)
|
| 134 |
-
param_value: float = float(param_value)
|
| 135 |
-
if maybe_convert and param_value.is_integer():
|
| 136 |
-
param_value = int(param_value)
|
| 137 |
-
except Exception:
|
| 138 |
-
logger.warning(
|
| 139 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' is not a float in tool "
|
| 140 |
-
f"'{func_name}', degenerating to string."
|
| 141 |
-
)
|
| 142 |
-
return param_value
|
| 143 |
-
elif param_type in ["boolean", "bool", "binary"]:
|
| 144 |
-
param_value = param_value.lower()
|
| 145 |
-
if param_value not in ["true", "false"]:
|
| 146 |
-
logger.warning(
|
| 147 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' is not a boolean (`true` of `false`) in tool '{func_name}', degenerating to false."
|
| 148 |
-
)
|
| 149 |
-
return param_value == "true"
|
| 150 |
-
else:
|
| 151 |
-
if (
|
| 152 |
-
param_type in ["object", "array", "arr"]
|
| 153 |
-
or param_type.startswith("dict")
|
| 154 |
-
or param_type.startswith("list")
|
| 155 |
-
):
|
| 156 |
-
try:
|
| 157 |
-
param_value = json.loads(param_value)
|
| 158 |
-
return param_value
|
| 159 |
-
except Exception:
|
| 160 |
-
logger.warning(
|
| 161 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' cannot be parsed with json.loads in tool "
|
| 162 |
-
f"'{func_name}', will try other methods to parse it."
|
| 163 |
-
)
|
| 164 |
-
try:
|
| 165 |
-
param_value = ast.literal_eval(param_value) # safer
|
| 166 |
-
except Exception:
|
| 167 |
-
logger.warning(
|
| 168 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' cannot be converted via Python `ast.literal_eval()` in tool '{func_name}', degenerating to string."
|
| 169 |
-
)
|
| 170 |
-
return param_value
|
| 171 |
-
|
| 172 |
-
def detect_and_parse(self, text: str, tools: List[Tool]) -> StreamingParseResult:
|
| 173 |
-
"""One-shot parsing for non-streaming scenarios."""
|
| 174 |
-
if self.tool_call_start_token not in text:
|
| 175 |
-
return StreamingParseResult(normal_text=text)
|
| 176 |
-
|
| 177 |
-
calls = []
|
| 178 |
-
try:
|
| 179 |
-
# Simple cleanup of the text to find tool calls
|
| 180 |
-
# Note: This is a simplified regex approach consistent with vLLM
|
| 181 |
-
raw_tool_calls = self.tool_call_regex.findall(text)
|
| 182 |
-
if not raw_tool_calls:
|
| 183 |
-
# Fallback: maybe the whole text is inside the tag or tags are stripped
|
| 184 |
-
if self.tool_call_prefix in text:
|
| 185 |
-
raw_tool_calls = [text]
|
| 186 |
-
|
| 187 |
-
tool_idx = 0
|
| 188 |
-
for tool_content in raw_tool_calls:
|
| 189 |
-
# Find function calls
|
| 190 |
-
funcs = self.tool_call_function_regex.findall(tool_content)
|
| 191 |
-
for func_match in funcs:
|
| 192 |
-
func_body = func_match[0] or func_match[1]
|
| 193 |
-
if ">" not in func_body:
|
| 194 |
-
continue
|
| 195 |
-
|
| 196 |
-
name_end = func_body.index(">")
|
| 197 |
-
func_name = func_body[:name_end]
|
| 198 |
-
params_str = func_body[name_end + 1 :]
|
| 199 |
-
|
| 200 |
-
param_config = self._get_arguments_config(func_name, tools)
|
| 201 |
-
parsed_params = {}
|
| 202 |
-
|
| 203 |
-
for p_match in self.tool_call_parameter_regex.findall(params_str):
|
| 204 |
-
if ">" not in p_match:
|
| 205 |
-
continue
|
| 206 |
-
p_idx = p_match.index(">")
|
| 207 |
-
p_name = p_match[:p_idx]
|
| 208 |
-
p_val = p_match[p_idx + 1 :]
|
| 209 |
-
# Remove prefixing and trailing \n
|
| 210 |
-
if p_val.startswith("\n"):
|
| 211 |
-
p_val = p_val[1:]
|
| 212 |
-
if p_val.endswith("\n"):
|
| 213 |
-
p_val = p_val[:-1]
|
| 214 |
-
|
| 215 |
-
parsed_params[p_name] = self._convert_param_value(
|
| 216 |
-
p_val, p_name, param_config, func_name
|
| 217 |
-
)
|
| 218 |
-
|
| 219 |
-
calls.append(
|
| 220 |
-
ToolCallItem(
|
| 221 |
-
tool_index=tool_idx,
|
| 222 |
-
name=func_name,
|
| 223 |
-
parameters=json.dumps(parsed_params, ensure_ascii=False),
|
| 224 |
-
)
|
| 225 |
-
)
|
| 226 |
-
tool_idx += 1
|
| 227 |
-
|
| 228 |
-
# Determine normal text (text before the first tool call)
|
| 229 |
-
start_idx = text.find(self.tool_call_start_token)
|
| 230 |
-
if start_idx == -1:
|
| 231 |
-
start_idx = text.find(self.tool_call_prefix)
|
| 232 |
-
normal_text = text[:start_idx] if start_idx > 0 else ""
|
| 233 |
-
|
| 234 |
-
return StreamingParseResult(normal_text=normal_text, calls=calls)
|
| 235 |
-
|
| 236 |
-
except Exception as e:
|
| 237 |
-
logger.error(f"Error in detect_and_parse: {e}")
|
| 238 |
-
return StreamingParseResult(normal_text=text)
|
| 239 |
-
|
| 240 |
-
def parse_streaming_increment(
|
| 241 |
-
self, new_text: str, tools: List[Tool]
|
| 242 |
-
) -> StreamingParseResult:
|
| 243 |
-
"""
|
| 244 |
-
Robust cursor-based streaming parser.
|
| 245 |
-
"""
|
| 246 |
-
self._buffer += new_text
|
| 247 |
-
|
| 248 |
-
# Guard against empty buffer
|
| 249 |
-
if not self._buffer:
|
| 250 |
-
return StreamingParseResult()
|
| 251 |
-
|
| 252 |
-
calls = []
|
| 253 |
-
normal_text_chunks = []
|
| 254 |
-
|
| 255 |
-
while True:
|
| 256 |
-
# Working text slice
|
| 257 |
-
current_slice = self._buffer[self.parsed_pos :]
|
| 258 |
-
|
| 259 |
-
# Optimization: If almost empty, wait for more
|
| 260 |
-
if not current_slice:
|
| 261 |
-
break
|
| 262 |
-
|
| 263 |
-
# -------------------------------------------------------
|
| 264 |
-
# 1. Priority detection: check if it's the start of Tool Call
|
| 265 |
-
# -------------------------------------------------------
|
| 266 |
-
if current_slice.startswith(self.tool_call_start_token):
|
| 267 |
-
self.parsed_pos += len(self.tool_call_start_token)
|
| 268 |
-
self.is_inside_tool_call = True
|
| 269 |
-
continue
|
| 270 |
-
|
| 271 |
-
# -------------------------------------------------------
|
| 272 |
-
# 2. Function Name: <function=name>
|
| 273 |
-
# -------------------------------------------------------
|
| 274 |
-
if current_slice.startswith(self.tool_call_prefix):
|
| 275 |
-
end_angle = current_slice.find(">")
|
| 276 |
-
if end_angle != -1:
|
| 277 |
-
func_name = current_slice[len(self.tool_call_prefix) : end_angle]
|
| 278 |
-
|
| 279 |
-
self.current_tool_id += 1
|
| 280 |
-
self.current_tool_name_sent = True
|
| 281 |
-
self.current_tool_param_count = 0
|
| 282 |
-
self.json_started = False
|
| 283 |
-
self.current_func_name = func_name
|
| 284 |
-
|
| 285 |
-
calls.append(
|
| 286 |
-
ToolCallItem(
|
| 287 |
-
tool_index=self.current_tool_id,
|
| 288 |
-
name=func_name,
|
| 289 |
-
parameters="",
|
| 290 |
-
)
|
| 291 |
-
)
|
| 292 |
-
|
| 293 |
-
self.parsed_pos += end_angle + 1
|
| 294 |
-
continue
|
| 295 |
-
else:
|
| 296 |
-
# Incomplete tag
|
| 297 |
-
break
|
| 298 |
-
|
| 299 |
-
# -------------------------------------------------------
|
| 300 |
-
# 3. Parameter: <parameter=name>value...
|
| 301 |
-
# -------------------------------------------------------
|
| 302 |
-
if current_slice.startswith(self.parameter_prefix):
|
| 303 |
-
name_end = current_slice.find(">")
|
| 304 |
-
if name_end != -1:
|
| 305 |
-
value_start_idx = name_end + 1
|
| 306 |
-
rest_of_slice = current_slice[value_start_idx:]
|
| 307 |
-
|
| 308 |
-
# A parameter can end in multiple ways:
|
| 309 |
-
# 1. [Normal] Encounter </parameter>
|
| 310 |
-
# 2. [Abnormal] Encounter next <parameter=
|
| 311 |
-
# 3. [Abnormal] Encounter </function>
|
| 312 |
-
# So we need to find the smallest one as the parameter end position.
|
| 313 |
-
cand_end_param = rest_of_slice.find(self.parameter_end_token)
|
| 314 |
-
cand_next_param = rest_of_slice.find(self.parameter_prefix)
|
| 315 |
-
cand_end_func = rest_of_slice.find(self.function_end_token)
|
| 316 |
-
|
| 317 |
-
candidates = []
|
| 318 |
-
if cand_end_param != -1:
|
| 319 |
-
candidates.append(
|
| 320 |
-
(cand_end_param, len(self.parameter_end_token))
|
| 321 |
-
)
|
| 322 |
-
if cand_next_param != -1:
|
| 323 |
-
candidates.append((cand_next_param, 0))
|
| 324 |
-
if cand_end_func != -1:
|
| 325 |
-
candidates.append((cand_end_func, 0))
|
| 326 |
-
|
| 327 |
-
if candidates:
|
| 328 |
-
best_cand = min(candidates, key=lambda x: x[0])
|
| 329 |
-
end_pos = best_cand[0]
|
| 330 |
-
end_token_len = best_cand[1]
|
| 331 |
-
|
| 332 |
-
param_name = current_slice[
|
| 333 |
-
len(self.parameter_prefix) : name_end
|
| 334 |
-
]
|
| 335 |
-
raw_value = rest_of_slice[:end_pos]
|
| 336 |
-
|
| 337 |
-
# Cleanup value
|
| 338 |
-
if raw_value.startswith("\n"):
|
| 339 |
-
raw_value = raw_value[1:]
|
| 340 |
-
if raw_value.endswith("\n"):
|
| 341 |
-
raw_value = raw_value[:-1]
|
| 342 |
-
|
| 343 |
-
# JSON Construction
|
| 344 |
-
if not self.json_started:
|
| 345 |
-
calls.append(
|
| 346 |
-
ToolCallItem(
|
| 347 |
-
tool_index=self.current_tool_id, parameters="{"
|
| 348 |
-
)
|
| 349 |
-
)
|
| 350 |
-
self.json_started = True
|
| 351 |
-
|
| 352 |
-
param_config = self._get_arguments_config(
|
| 353 |
-
self.current_func_name, tools
|
| 354 |
-
)
|
| 355 |
-
converted_val = self._convert_param_value(
|
| 356 |
-
raw_value, param_name, param_config, self.current_func_name
|
| 357 |
-
)
|
| 358 |
-
|
| 359 |
-
# Construct JSON fragment: "key": value
|
| 360 |
-
# Note: We must be careful with json.dumps to ensure valid JSON streaming
|
| 361 |
-
json_key_val = f"{json.dumps(param_name)}: {json.dumps(converted_val, ensure_ascii=False)}"
|
| 362 |
-
|
| 363 |
-
if self.current_tool_param_count > 0:
|
| 364 |
-
fragment = f", {json_key_val}"
|
| 365 |
-
else:
|
| 366 |
-
fragment = json_key_val
|
| 367 |
-
|
| 368 |
-
calls.append(
|
| 369 |
-
ToolCallItem(
|
| 370 |
-
tool_index=self.current_tool_id, parameters=fragment
|
| 371 |
-
)
|
| 372 |
-
)
|
| 373 |
-
self.current_tool_param_count += 1
|
| 374 |
-
|
| 375 |
-
# Advance cursor
|
| 376 |
-
total_len = (name_end + 1) + end_pos + end_token_len
|
| 377 |
-
self.parsed_pos += total_len
|
| 378 |
-
continue
|
| 379 |
-
|
| 380 |
-
# Incomplete parameter tag or value
|
| 381 |
-
break
|
| 382 |
-
|
| 383 |
-
# -------------------------------------------------------
|
| 384 |
-
# 4. Function End: </function>
|
| 385 |
-
# -------------------------------------------------------
|
| 386 |
-
if current_slice.startswith(self.function_end_token):
|
| 387 |
-
if not self.json_started:
|
| 388 |
-
calls.append(
|
| 389 |
-
ToolCallItem(tool_index=self.current_tool_id, parameters="{")
|
| 390 |
-
)
|
| 391 |
-
self.json_started = True
|
| 392 |
-
|
| 393 |
-
calls.append(
|
| 394 |
-
ToolCallItem(tool_index=self.current_tool_id, parameters="}")
|
| 395 |
-
)
|
| 396 |
-
self.parsed_pos += len(self.function_end_token)
|
| 397 |
-
self.current_func_name = None
|
| 398 |
-
continue
|
| 399 |
-
|
| 400 |
-
# -------------------------------------------------------
|
| 401 |
-
# 5. Tool Call End: </tool_call>
|
| 402 |
-
# -------------------------------------------------------
|
| 403 |
-
if current_slice.startswith(self.tool_call_end_token):
|
| 404 |
-
self.parsed_pos += len(self.tool_call_end_token)
|
| 405 |
-
self.is_inside_tool_call = False # [FIX] Exit tool call region
|
| 406 |
-
continue
|
| 407 |
-
|
| 408 |
-
# -------------------------------------------------------
|
| 409 |
-
# 6. Handling content / whitespace / normal text
|
| 410 |
-
# -------------------------------------------------------
|
| 411 |
-
# If current position is not the start of a tag (i.e., doesn't start with <), it might be plain text,
|
| 412 |
-
# or a newline between two tags.
|
| 413 |
-
# But we need to be careful not to output truncated tags like "<fun" as text.
|
| 414 |
-
|
| 415 |
-
next_open_angle = current_slice.find("<")
|
| 416 |
-
|
| 417 |
-
if next_open_angle == -1:
|
| 418 |
-
# This entire segment is plain text
|
| 419 |
-
if not self.is_inside_tool_call:
|
| 420 |
-
normal_text_chunks.append(current_slice)
|
| 421 |
-
# [FIX] If inside tool call, discard this text (usually \n), don't append
|
| 422 |
-
self.parsed_pos += len(current_slice)
|
| 423 |
-
continue
|
| 424 |
-
|
| 425 |
-
elif next_open_angle == 0:
|
| 426 |
-
# Looks like a Tag, but doesn't match any known Tag above
|
| 427 |
-
|
| 428 |
-
possible_tags = [
|
| 429 |
-
self.tool_call_start_token,
|
| 430 |
-
self.tool_call_end_token,
|
| 431 |
-
self.tool_call_prefix,
|
| 432 |
-
self.function_end_token,
|
| 433 |
-
self.parameter_prefix,
|
| 434 |
-
self.parameter_end_token,
|
| 435 |
-
]
|
| 436 |
-
|
| 437 |
-
is_potential_tag = False
|
| 438 |
-
for tag in possible_tags:
|
| 439 |
-
if tag.startswith(current_slice):
|
| 440 |
-
is_potential_tag = True
|
| 441 |
-
break
|
| 442 |
-
|
| 443 |
-
if is_potential_tag:
|
| 444 |
-
break # Wait for more
|
| 445 |
-
else:
|
| 446 |
-
# Just a plain '<' symbol
|
| 447 |
-
if not self.is_inside_tool_call:
|
| 448 |
-
normal_text_chunks.append("<")
|
| 449 |
-
self.parsed_pos += 1
|
| 450 |
-
continue
|
| 451 |
-
|
| 452 |
-
else:
|
| 453 |
-
# '<' is in the middle
|
| 454 |
-
text_segment = current_slice[:next_open_angle]
|
| 455 |
-
if not self.is_inside_tool_call:
|
| 456 |
-
normal_text_chunks.append(text_segment)
|
| 457 |
-
# [FIX] If inside tool call, discard whitespace/text before Tag
|
| 458 |
-
self.parsed_pos += next_open_angle
|
| 459 |
-
continue
|
| 460 |
-
|
| 461 |
-
# Memory Cleanup: Slice the buffer
|
| 462 |
-
# Keep unparsed part, discard parsed part
|
| 463 |
-
if self.parsed_pos > 0:
|
| 464 |
-
self._buffer = self._buffer[self.parsed_pos :]
|
| 465 |
-
self.parsed_pos = 0
|
| 466 |
-
|
| 467 |
-
normal_text = "".join(normal_text_chunks) if normal_text_chunks else ""
|
| 468 |
-
return StreamingParseResult(calls=calls, normal_text=normal_text)
|
| 469 |
-
|
| 470 |
-
def supports_structural_tag(self) -> bool:
|
| 471 |
-
return False
|
| 472 |
-
|
| 473 |
-
def structure_info(self) -> _GetInfoFunc:
|
| 474 |
-
raise NotImplementedError
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
qwen3coder_tool_parser_vllm.py
DELETED
|
@@ -1,690 +0,0 @@
|
|
| 1 |
-
# SPDX-License-Identifier: Apache-2.0
|
| 2 |
-
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
| 3 |
-
import ast
|
| 4 |
-
import json
|
| 5 |
-
import uuid
|
| 6 |
-
from collections.abc import Sequence
|
| 7 |
-
from typing import Any, List, Optional, Union
|
| 8 |
-
|
| 9 |
-
import regex as re
|
| 10 |
-
|
| 11 |
-
from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
|
| 12 |
-
ChatCompletionToolsParam,
|
| 13 |
-
DeltaFunctionCall, DeltaMessage,
|
| 14 |
-
DeltaToolCall,
|
| 15 |
-
ExtractedToolCallInformation,
|
| 16 |
-
FunctionCall, ToolCall)
|
| 17 |
-
from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
|
| 18 |
-
ToolParser, ToolParserManager)
|
| 19 |
-
from vllm.logger import init_logger
|
| 20 |
-
from vllm.transformers_utils.tokenizer import AnyTokenizer
|
| 21 |
-
|
| 22 |
-
logger = init_logger(__name__)
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
@ToolParserManager.register_module("qwen3_coder")
|
| 26 |
-
class Qwen3CoderToolParser(ToolParser):
|
| 27 |
-
|
| 28 |
-
def __init__(self, tokenizer: AnyTokenizer):
|
| 29 |
-
super().__init__(tokenizer)
|
| 30 |
-
|
| 31 |
-
self.current_tool_name_sent: bool = False
|
| 32 |
-
self.prev_tool_call_arr: list[dict] = []
|
| 33 |
-
self.current_tool_id: int = -1
|
| 34 |
-
self.streamed_args_for_tool: list[str] = []
|
| 35 |
-
|
| 36 |
-
# Sentinel tokens for streaming mode
|
| 37 |
-
self.tool_call_start_token: str = "<tool_call>"
|
| 38 |
-
self.tool_call_end_token: str = "</tool_call>"
|
| 39 |
-
self.tool_call_prefix: str = "<function="
|
| 40 |
-
self.function_end_token: str = "</function>"
|
| 41 |
-
self.parameter_prefix: str = "<parameter="
|
| 42 |
-
self.parameter_end_token: str = "</parameter>"
|
| 43 |
-
self.is_tool_call_started: bool = False
|
| 44 |
-
self.failed_count: int = 0
|
| 45 |
-
|
| 46 |
-
# Enhanced streaming state - reset for each new message
|
| 47 |
-
self._reset_streaming_state()
|
| 48 |
-
|
| 49 |
-
# Regex patterns
|
| 50 |
-
self.tool_call_complete_regex = re.compile(
|
| 51 |
-
r"<tool_call>(.*?)</tool_call>", re.DOTALL)
|
| 52 |
-
self.tool_call_regex = re.compile(
|
| 53 |
-
r"<tool_call>(.*?)</tool_call>|<tool_call>(.*?)$", re.DOTALL)
|
| 54 |
-
self.tool_call_function_regex = re.compile(
|
| 55 |
-
r"<function=(.*?)</function>|<function=(.*)$", re.DOTALL)
|
| 56 |
-
self.tool_call_parameter_regex = re.compile(
|
| 57 |
-
r"<parameter=(.*?)(?:</parameter>|(?=<parameter=)|(?=</function>)|$)",
|
| 58 |
-
re.DOTALL)
|
| 59 |
-
|
| 60 |
-
if not self.model_tokenizer:
|
| 61 |
-
raise ValueError(
|
| 62 |
-
"The model tokenizer must be passed to the ToolParser "
|
| 63 |
-
"constructor during construction.")
|
| 64 |
-
|
| 65 |
-
self.tool_call_start_token_id = self.vocab.get(
|
| 66 |
-
self.tool_call_start_token)
|
| 67 |
-
self.tool_call_end_token_id = self.vocab.get(self.tool_call_end_token)
|
| 68 |
-
|
| 69 |
-
if self.tool_call_start_token_id is None or self.tool_call_end_token_id is None:
|
| 70 |
-
raise RuntimeError(
|
| 71 |
-
"Qwen3 XML Tool parser could not locate tool call start/end "
|
| 72 |
-
"tokens in the tokenizer!")
|
| 73 |
-
|
| 74 |
-
logger.info(
|
| 75 |
-
f"vLLM Successfully import tool parser {self.__class__.__name__} !"
|
| 76 |
-
)
|
| 77 |
-
|
| 78 |
-
def _generate_tool_call_id(self) -> str:
|
| 79 |
-
"""Generate a unique tool call ID."""
|
| 80 |
-
return f"call_{uuid.uuid4().hex[:24]}"
|
| 81 |
-
|
| 82 |
-
def _reset_streaming_state(self):
|
| 83 |
-
"""Reset all streaming state."""
|
| 84 |
-
self.current_tool_index = 0
|
| 85 |
-
self.is_tool_call_started = False
|
| 86 |
-
self.header_sent = False
|
| 87 |
-
self.current_tool_id = None
|
| 88 |
-
self.current_function_name = None
|
| 89 |
-
self.current_param_name = None
|
| 90 |
-
self.current_param_value = ""
|
| 91 |
-
self.param_count = 0
|
| 92 |
-
self.in_param = False
|
| 93 |
-
self.in_function = False
|
| 94 |
-
self.accumulated_text = ""
|
| 95 |
-
self.json_started = False
|
| 96 |
-
self.json_closed = False
|
| 97 |
-
# Store accumulated parameters for type conversion
|
| 98 |
-
self.accumulated_params = {}
|
| 99 |
-
self.streaming_request = None
|
| 100 |
-
|
| 101 |
-
def _get_arguments_config(
|
| 102 |
-
self, func_name: str,
|
| 103 |
-
tools: Optional[list[ChatCompletionToolsParam]]) -> dict:
|
| 104 |
-
"""Extract argument configuration for a function."""
|
| 105 |
-
if tools is None:
|
| 106 |
-
return {}
|
| 107 |
-
for config in tools:
|
| 108 |
-
if not hasattr(config, "type") or not (hasattr(
|
| 109 |
-
config, "function") and hasattr(config.function, "name")):
|
| 110 |
-
continue
|
| 111 |
-
if config.type == "function" and config.function.name == func_name:
|
| 112 |
-
if not hasattr(config.function, "parameters"):
|
| 113 |
-
return {}
|
| 114 |
-
params = config.function.parameters
|
| 115 |
-
if isinstance(params, dict) and "properties" in params:
|
| 116 |
-
return params["properties"]
|
| 117 |
-
elif isinstance(params, dict):
|
| 118 |
-
return params
|
| 119 |
-
else:
|
| 120 |
-
return {}
|
| 121 |
-
logger.warning(f"Tool '{func_name}' is not defined in the tools list.")
|
| 122 |
-
return {}
|
| 123 |
-
|
| 124 |
-
def _convert_param_value(self, param_value: str, param_name: str,
|
| 125 |
-
param_config: dict, func_name: str) -> Any:
|
| 126 |
-
"""Convert parameter value based on its type in the schema."""
|
| 127 |
-
# Handle null value for any type
|
| 128 |
-
if param_value.lower() == "null":
|
| 129 |
-
return None
|
| 130 |
-
|
| 131 |
-
if param_name not in param_config:
|
| 132 |
-
if param_config != {}:
|
| 133 |
-
logger.warning(
|
| 134 |
-
f"Parsed parameter '{param_name}' is not defined in the tool "
|
| 135 |
-
f"parameters for tool '{func_name}', directly returning the string value."
|
| 136 |
-
)
|
| 137 |
-
return param_value
|
| 138 |
-
|
| 139 |
-
if isinstance(param_config[param_name],
|
| 140 |
-
dict) and "type" in param_config[param_name]:
|
| 141 |
-
param_type = str(param_config[param_name]["type"]).strip().lower()
|
| 142 |
-
else:
|
| 143 |
-
param_type = "string"
|
| 144 |
-
if param_type in ["string", "str", "text", "varchar", "char", "enum"]:
|
| 145 |
-
return param_value
|
| 146 |
-
elif param_type.startswith("int") or param_type.startswith(
|
| 147 |
-
"uint") or param_type.startswith(
|
| 148 |
-
"long") or param_type.startswith(
|
| 149 |
-
"short") or param_type.startswith("unsigned"):
|
| 150 |
-
try:
|
| 151 |
-
param_value = int(param_value)
|
| 152 |
-
except:
|
| 153 |
-
logger.warning(
|
| 154 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' is not an integer in tool "
|
| 155 |
-
f"'{func_name}', degenerating to string.")
|
| 156 |
-
return param_value
|
| 157 |
-
elif param_type.startswith("num") or param_type.startswith("float"):
|
| 158 |
-
try:
|
| 159 |
-
maybe_convert = False if "." in param_value or "e" in param_value.lower() else True
|
| 160 |
-
param_value: float = float(param_value)
|
| 161 |
-
if maybe_convert and param_value.is_integer():
|
| 162 |
-
param_value = int(param_value)
|
| 163 |
-
except:
|
| 164 |
-
logger.warning(
|
| 165 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' is not a float in tool "
|
| 166 |
-
f"'{func_name}', degenerating to string.")
|
| 167 |
-
return param_value
|
| 168 |
-
elif param_type in ["boolean", "bool", "binary"]:
|
| 169 |
-
param_value = param_value.lower()
|
| 170 |
-
if param_value not in ["true", "false"]:
|
| 171 |
-
logger.warning(
|
| 172 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' is not a boolean (`true` of `false`) in tool '{func_name}', degenerating to false."
|
| 173 |
-
)
|
| 174 |
-
return param_value == "true"
|
| 175 |
-
else:
|
| 176 |
-
if param_type in ["object", "array", "arr"
|
| 177 |
-
] or param_type.startswith(
|
| 178 |
-
"dict") or param_type.startswith("list"):
|
| 179 |
-
try:
|
| 180 |
-
param_value = json.loads(param_value)
|
| 181 |
-
return param_value
|
| 182 |
-
except:
|
| 183 |
-
logger.warning(
|
| 184 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' cannot be parsed with json.loads in tool "
|
| 185 |
-
f"'{func_name}', will try other methods to parse it.")
|
| 186 |
-
try:
|
| 187 |
-
param_value = ast.literal_eval(param_value) # safer
|
| 188 |
-
except:
|
| 189 |
-
logger.warning(
|
| 190 |
-
f"Parsed value '{param_value}' of parameter '{param_name}' cannot be converted via Python `ast.literal_eval()` in tool '{func_name}', degenerating to string."
|
| 191 |
-
)
|
| 192 |
-
return param_value
|
| 193 |
-
|
| 194 |
-
def _parse_xml_function_call(
|
| 195 |
-
self, function_call_str: str,
|
| 196 |
-
tools: Optional[list[ChatCompletionToolsParam]]
|
| 197 |
-
) -> Optional[ToolCall]:
|
| 198 |
-
|
| 199 |
-
# Extract function name
|
| 200 |
-
end_index = function_call_str.index(">")
|
| 201 |
-
function_name = function_call_str[:end_index]
|
| 202 |
-
param_config = self._get_arguments_config(function_name, tools)
|
| 203 |
-
parameters = function_call_str[end_index + 1:]
|
| 204 |
-
param_dict = {}
|
| 205 |
-
for match_text in self.tool_call_parameter_regex.findall(parameters):
|
| 206 |
-
idx = match_text.index(">")
|
| 207 |
-
param_name = match_text[:idx]
|
| 208 |
-
param_value = str(match_text[idx + 1:])
|
| 209 |
-
# Remove prefix and trailing \n
|
| 210 |
-
if param_value.startswith("\n"):
|
| 211 |
-
param_value = param_value[1:]
|
| 212 |
-
if param_value.endswith("\n"):
|
| 213 |
-
param_value = param_value[:-1]
|
| 214 |
-
|
| 215 |
-
param_dict[param_name] = self._convert_param_value(
|
| 216 |
-
param_value, param_name, param_config, function_name)
|
| 217 |
-
return ToolCall(
|
| 218 |
-
type="function",
|
| 219 |
-
function=FunctionCall(name=function_name,
|
| 220 |
-
arguments=json.dumps(param_dict,
|
| 221 |
-
ensure_ascii=False)),
|
| 222 |
-
)
|
| 223 |
-
|
| 224 |
-
def _get_function_calls(self, model_output: str) -> List[str]:
|
| 225 |
-
# Find all tool calls
|
| 226 |
-
matched_ranges = self.tool_call_regex.findall(model_output)
|
| 227 |
-
raw_tool_calls = [
|
| 228 |
-
match[0] if match[0] else match[1] for match in matched_ranges
|
| 229 |
-
]
|
| 230 |
-
|
| 231 |
-
# Back-off strategy if no tool_call tags found
|
| 232 |
-
if len(raw_tool_calls) == 0:
|
| 233 |
-
raw_tool_calls = [model_output]
|
| 234 |
-
|
| 235 |
-
raw_function_calls = []
|
| 236 |
-
for tool_call in raw_tool_calls:
|
| 237 |
-
raw_function_calls.extend(
|
| 238 |
-
self.tool_call_function_regex.findall(tool_call))
|
| 239 |
-
|
| 240 |
-
function_calls = [
|
| 241 |
-
match[0] if match[0] else match[1] for match in raw_function_calls
|
| 242 |
-
]
|
| 243 |
-
return function_calls
|
| 244 |
-
|
| 245 |
-
def extract_tool_calls(
|
| 246 |
-
self,
|
| 247 |
-
model_output: str,
|
| 248 |
-
request: ChatCompletionRequest,
|
| 249 |
-
) -> ExtractedToolCallInformation:
|
| 250 |
-
# Quick check to avoid unnecessary processing
|
| 251 |
-
if self.tool_call_prefix not in model_output:
|
| 252 |
-
return ExtractedToolCallInformation(tools_called=False,
|
| 253 |
-
tool_calls=[],
|
| 254 |
-
content=model_output)
|
| 255 |
-
|
| 256 |
-
try:
|
| 257 |
-
function_calls = self._get_function_calls(model_output)
|
| 258 |
-
if len(function_calls) == 0:
|
| 259 |
-
return ExtractedToolCallInformation(tools_called=False,
|
| 260 |
-
tool_calls=[],
|
| 261 |
-
content=model_output)
|
| 262 |
-
|
| 263 |
-
tool_calls = [
|
| 264 |
-
self._parse_xml_function_call(function_call_str, request.tools)
|
| 265 |
-
for function_call_str in function_calls
|
| 266 |
-
]
|
| 267 |
-
|
| 268 |
-
# Populate prev_tool_call_arr for serving layer to set finish_reason
|
| 269 |
-
self.prev_tool_call_arr.clear() # Clear previous calls
|
| 270 |
-
for tool_call in tool_calls:
|
| 271 |
-
if tool_call:
|
| 272 |
-
self.prev_tool_call_arr.append({
|
| 273 |
-
"name":
|
| 274 |
-
tool_call.function.name,
|
| 275 |
-
"arguments":
|
| 276 |
-
tool_call.function.arguments,
|
| 277 |
-
})
|
| 278 |
-
|
| 279 |
-
# Extract content before tool calls
|
| 280 |
-
content_index = model_output.find(self.tool_call_start_token)
|
| 281 |
-
content_index = content_index if content_index >= 0 else model_output.find(
|
| 282 |
-
self.tool_call_prefix)
|
| 283 |
-
content = model_output[:content_index] # .rstrip()
|
| 284 |
-
|
| 285 |
-
return ExtractedToolCallInformation(
|
| 286 |
-
tools_called=(len(tool_calls) > 0),
|
| 287 |
-
tool_calls=tool_calls,
|
| 288 |
-
content=content if content else None,
|
| 289 |
-
)
|
| 290 |
-
|
| 291 |
-
except Exception:
|
| 292 |
-
logger.exception("Error in extracting tool call from response.")
|
| 293 |
-
return ExtractedToolCallInformation(tools_called=False,
|
| 294 |
-
tool_calls=[],
|
| 295 |
-
content=model_output)
|
| 296 |
-
|
| 297 |
-
def extract_tool_calls_streaming(
|
| 298 |
-
self,
|
| 299 |
-
previous_text: str,
|
| 300 |
-
current_text: str,
|
| 301 |
-
delta_text: str,
|
| 302 |
-
previous_token_ids: Sequence[int],
|
| 303 |
-
current_token_ids: Sequence[int],
|
| 304 |
-
delta_token_ids: Sequence[int],
|
| 305 |
-
request: ChatCompletionRequest,
|
| 306 |
-
) -> Union[DeltaMessage, None]:
|
| 307 |
-
# Store request for type conversion
|
| 308 |
-
if not previous_text:
|
| 309 |
-
self._reset_streaming_state()
|
| 310 |
-
self.streaming_request = request
|
| 311 |
-
|
| 312 |
-
# If no delta text, return None unless it's an EOS token after tool calls
|
| 313 |
-
if not delta_text:
|
| 314 |
-
# Check if this is an EOS token after all tool calls are complete
|
| 315 |
-
# We check for tool calls in the text even if is_tool_call_started is False
|
| 316 |
-
# because it might have been reset after processing all tools
|
| 317 |
-
if delta_token_ids and self.tool_call_end_token_id not in delta_token_ids:
|
| 318 |
-
# Count complete tool calls
|
| 319 |
-
complete_calls = len(
|
| 320 |
-
self.tool_call_complete_regex.findall(current_text))
|
| 321 |
-
|
| 322 |
-
# If we have completed tool calls and populated prev_tool_call_arr
|
| 323 |
-
if complete_calls > 0 and len(self.prev_tool_call_arr) > 0:
|
| 324 |
-
# Check if all tool calls are closed
|
| 325 |
-
open_calls = current_text.count(
|
| 326 |
-
self.tool_call_start_token) - current_text.count(
|
| 327 |
-
self.tool_call_end_token)
|
| 328 |
-
if open_calls == 0:
|
| 329 |
-
# Return empty delta message to allow finish_reason processing
|
| 330 |
-
return DeltaMessage(content="")
|
| 331 |
-
elif not self.is_tool_call_started and current_text:
|
| 332 |
-
# This is a regular content response that's now complete
|
| 333 |
-
return DeltaMessage(content="")
|
| 334 |
-
return None
|
| 335 |
-
|
| 336 |
-
# Update accumulated text
|
| 337 |
-
self.accumulated_text = current_text
|
| 338 |
-
|
| 339 |
-
# Check if we need to advance to next tool
|
| 340 |
-
if self.json_closed and not self.in_function:
|
| 341 |
-
# Check if this tool call has ended
|
| 342 |
-
tool_ends = current_text.count(self.tool_call_end_token)
|
| 343 |
-
if tool_ends > self.current_tool_index:
|
| 344 |
-
# This tool has ended, advance to next
|
| 345 |
-
self.current_tool_index += 1
|
| 346 |
-
self.header_sent = False
|
| 347 |
-
self.param_count = 0
|
| 348 |
-
self.json_started = False
|
| 349 |
-
self.json_closed = False
|
| 350 |
-
self.accumulated_params = {}
|
| 351 |
-
|
| 352 |
-
# Check if there are more tool calls
|
| 353 |
-
tool_starts = current_text.count(self.tool_call_start_token)
|
| 354 |
-
if self.current_tool_index >= tool_starts:
|
| 355 |
-
# No more tool calls
|
| 356 |
-
self.is_tool_call_started = False
|
| 357 |
-
# Continue processing next tool
|
| 358 |
-
return None
|
| 359 |
-
|
| 360 |
-
# Handle normal content before tool calls
|
| 361 |
-
if not self.is_tool_call_started:
|
| 362 |
-
# Check if tool call is starting
|
| 363 |
-
if self.tool_call_start_token_id in delta_token_ids or self.tool_call_start_token in delta_text:
|
| 364 |
-
self.is_tool_call_started = True
|
| 365 |
-
# Return any content before the tool call
|
| 366 |
-
if self.tool_call_start_token in delta_text:
|
| 367 |
-
content_before = delta_text[:delta_text.index(
|
| 368 |
-
self.tool_call_start_token)]
|
| 369 |
-
if content_before:
|
| 370 |
-
return DeltaMessage(content=content_before)
|
| 371 |
-
return None
|
| 372 |
-
else:
|
| 373 |
-
# Check if we're between tool calls - skip whitespace
|
| 374 |
-
if current_text.rstrip().endswith(self.tool_call_end_token):
|
| 375 |
-
# We just ended a tool call, skip whitespace
|
| 376 |
-
if delta_text.strip() == "":
|
| 377 |
-
return None
|
| 378 |
-
# Normal content, no tool call
|
| 379 |
-
return DeltaMessage(content=delta_text)
|
| 380 |
-
|
| 381 |
-
# Check if we're between tool calls (waiting for next one)
|
| 382 |
-
# Count tool calls we've seen vs processed
|
| 383 |
-
tool_starts_count = current_text.count(self.tool_call_start_token)
|
| 384 |
-
if self.current_tool_index >= tool_starts_count:
|
| 385 |
-
# We're past all tool calls, shouldn't be here
|
| 386 |
-
return None
|
| 387 |
-
|
| 388 |
-
# We're in a tool call, find the current tool call portion
|
| 389 |
-
# Need to find the correct tool call based on current_tool_index
|
| 390 |
-
tool_starts = []
|
| 391 |
-
idx = 0
|
| 392 |
-
while True:
|
| 393 |
-
idx = current_text.find(self.tool_call_start_token, idx)
|
| 394 |
-
if idx == -1:
|
| 395 |
-
break
|
| 396 |
-
tool_starts.append(idx)
|
| 397 |
-
idx += len(self.tool_call_start_token)
|
| 398 |
-
|
| 399 |
-
if self.current_tool_index >= len(tool_starts):
|
| 400 |
-
# No more tool calls to process yet
|
| 401 |
-
return None
|
| 402 |
-
|
| 403 |
-
tool_start_idx = tool_starts[self.current_tool_index]
|
| 404 |
-
# Find where this tool call ends (or current position if not ended yet)
|
| 405 |
-
tool_end_idx = current_text.find(self.tool_call_end_token,
|
| 406 |
-
tool_start_idx)
|
| 407 |
-
if tool_end_idx == -1:
|
| 408 |
-
tool_text = current_text[tool_start_idx:]
|
| 409 |
-
else:
|
| 410 |
-
tool_text = current_text[tool_start_idx:tool_end_idx +
|
| 411 |
-
len(self.tool_call_end_token)]
|
| 412 |
-
|
| 413 |
-
# Looking for function header
|
| 414 |
-
if not self.header_sent:
|
| 415 |
-
if self.tool_call_prefix in tool_text:
|
| 416 |
-
func_start = tool_text.find(self.tool_call_prefix) + len(
|
| 417 |
-
self.tool_call_prefix)
|
| 418 |
-
func_end = tool_text.find(">", func_start)
|
| 419 |
-
|
| 420 |
-
if func_end != -1:
|
| 421 |
-
# Found complete function name
|
| 422 |
-
self.current_function_name = tool_text[func_start:func_end]
|
| 423 |
-
self.current_tool_id = self._generate_tool_call_id()
|
| 424 |
-
self.header_sent = True
|
| 425 |
-
self.in_function = True
|
| 426 |
-
|
| 427 |
-
# IMPORTANT: Add to prev_tool_call_arr immediately when we detect a tool call
|
| 428 |
-
# This ensures finish_reason="tool_calls" even if parsing isn't complete
|
| 429 |
-
already_added = any(
|
| 430 |
-
tool.get("name") == self.current_function_name
|
| 431 |
-
for tool in self.prev_tool_call_arr)
|
| 432 |
-
if not already_added:
|
| 433 |
-
self.prev_tool_call_arr.append({
|
| 434 |
-
"name": self.current_function_name,
|
| 435 |
-
"arguments":
|
| 436 |
-
"{}", # Placeholder, will be updated later
|
| 437 |
-
})
|
| 438 |
-
|
| 439 |
-
# Send header with function info
|
| 440 |
-
return DeltaMessage(tool_calls=[
|
| 441 |
-
DeltaToolCall(
|
| 442 |
-
index=self.current_tool_index,
|
| 443 |
-
id=self.current_tool_id,
|
| 444 |
-
function=DeltaFunctionCall(
|
| 445 |
-
name=self.current_function_name, arguments=""),
|
| 446 |
-
type="function",
|
| 447 |
-
)
|
| 448 |
-
])
|
| 449 |
-
return None
|
| 450 |
-
|
| 451 |
-
# We've sent header, now handle function body
|
| 452 |
-
if self.in_function:
|
| 453 |
-
# Send opening brace if not sent yet
|
| 454 |
-
if not self.json_started and self.parameter_prefix not in delta_text:
|
| 455 |
-
self.json_started = True
|
| 456 |
-
return DeltaMessage(tool_calls=[
|
| 457 |
-
DeltaToolCall(
|
| 458 |
-
index=self.current_tool_index,
|
| 459 |
-
function=DeltaFunctionCall(arguments="{"),
|
| 460 |
-
)
|
| 461 |
-
])
|
| 462 |
-
|
| 463 |
-
# Make sure json_started is set if we're processing parameters
|
| 464 |
-
if not self.json_started:
|
| 465 |
-
self.json_started = True
|
| 466 |
-
|
| 467 |
-
# Check for function end in accumulated text
|
| 468 |
-
if not self.json_closed and self.function_end_token in tool_text:
|
| 469 |
-
# Close JSON
|
| 470 |
-
self.json_closed = True
|
| 471 |
-
|
| 472 |
-
# Extract the complete tool call to update prev_tool_call_arr with final arguments
|
| 473 |
-
# Find the function content
|
| 474 |
-
func_start = tool_text.find(self.tool_call_prefix) + len(
|
| 475 |
-
self.tool_call_prefix)
|
| 476 |
-
func_content_end = tool_text.find(self.function_end_token,
|
| 477 |
-
func_start)
|
| 478 |
-
if func_content_end != -1:
|
| 479 |
-
func_content = tool_text[func_start:func_content_end]
|
| 480 |
-
# Parse to get the complete arguments
|
| 481 |
-
try:
|
| 482 |
-
parsed_tool = self._parse_xml_function_call(
|
| 483 |
-
func_content, self.streaming_request.tools
|
| 484 |
-
if self.streaming_request else None)
|
| 485 |
-
if parsed_tool:
|
| 486 |
-
# Update existing entry in prev_tool_call_arr with complete arguments
|
| 487 |
-
for i, tool in enumerate(self.prev_tool_call_arr):
|
| 488 |
-
if tool.get(
|
| 489 |
-
"name") == parsed_tool.function.name:
|
| 490 |
-
self.prev_tool_call_arr[i][
|
| 491 |
-
"arguments"] = parsed_tool.function.arguments
|
| 492 |
-
break
|
| 493 |
-
except Exception:
|
| 494 |
-
pass # Ignore parsing errors during streaming
|
| 495 |
-
|
| 496 |
-
result = DeltaMessage(tool_calls=[
|
| 497 |
-
DeltaToolCall(
|
| 498 |
-
index=self.current_tool_index,
|
| 499 |
-
function=DeltaFunctionCall(arguments="}"),
|
| 500 |
-
)
|
| 501 |
-
])
|
| 502 |
-
|
| 503 |
-
# Reset state for next tool
|
| 504 |
-
self.in_function = False
|
| 505 |
-
self.json_closed = True
|
| 506 |
-
self.accumulated_params = {}
|
| 507 |
-
|
| 508 |
-
return result
|
| 509 |
-
|
| 510 |
-
# Look for parameters
|
| 511 |
-
# Find all parameter starts
|
| 512 |
-
param_starts = []
|
| 513 |
-
idx = 0
|
| 514 |
-
while True:
|
| 515 |
-
idx = tool_text.find(self.parameter_prefix, idx)
|
| 516 |
-
if idx == -1:
|
| 517 |
-
break
|
| 518 |
-
param_starts.append(idx)
|
| 519 |
-
idx += len(self.parameter_prefix)
|
| 520 |
-
|
| 521 |
-
# Check if we should start a new parameter
|
| 522 |
-
if not self.in_param and self.param_count < len(param_starts):
|
| 523 |
-
|
| 524 |
-
if len(param_starts) > self.param_count:
|
| 525 |
-
# Process the next parameter
|
| 526 |
-
param_idx = param_starts[self.param_count]
|
| 527 |
-
param_start = param_idx + len(self.parameter_prefix)
|
| 528 |
-
remaining = tool_text[param_start:]
|
| 529 |
-
|
| 530 |
-
if ">" in remaining:
|
| 531 |
-
# We have the complete parameter name
|
| 532 |
-
name_end = remaining.find(">")
|
| 533 |
-
self.current_param_name = remaining[:name_end]
|
| 534 |
-
|
| 535 |
-
# Find the parameter value
|
| 536 |
-
value_start = param_start + name_end + 1
|
| 537 |
-
value_text = tool_text[value_start:]
|
| 538 |
-
if value_text.startswith("\n"):
|
| 539 |
-
value_text = value_text[1:]
|
| 540 |
-
|
| 541 |
-
# Find where this parameter ends
|
| 542 |
-
param_end_idx = value_text.find(
|
| 543 |
-
self.parameter_end_token)
|
| 544 |
-
if param_end_idx == -1:
|
| 545 |
-
# No closing tag, look for next parameter or function end
|
| 546 |
-
next_param_idx = value_text.find(
|
| 547 |
-
self.parameter_prefix)
|
| 548 |
-
func_end_idx = value_text.find(
|
| 549 |
-
self.function_end_token)
|
| 550 |
-
|
| 551 |
-
if next_param_idx != -1 and (func_end_idx == -1
|
| 552 |
-
or next_param_idx
|
| 553 |
-
< func_end_idx):
|
| 554 |
-
param_end_idx = next_param_idx
|
| 555 |
-
elif func_end_idx != -1:
|
| 556 |
-
param_end_idx = func_end_idx
|
| 557 |
-
else:
|
| 558 |
-
# Neither found, check if tool call is complete
|
| 559 |
-
if self.tool_call_end_token in tool_text:
|
| 560 |
-
# Tool call is complete, so parameter must be complete too
|
| 561 |
-
# Use all remaining text before function end as value
|
| 562 |
-
param_end_idx = len(value_text)
|
| 563 |
-
else:
|
| 564 |
-
# Still streaming, wait for more content
|
| 565 |
-
return None
|
| 566 |
-
|
| 567 |
-
if param_end_idx != -1:
|
| 568 |
-
# Complete parameter found
|
| 569 |
-
param_value = value_text[:param_end_idx]
|
| 570 |
-
if param_value.endswith("\n"):
|
| 571 |
-
param_value = param_value[:-1]
|
| 572 |
-
|
| 573 |
-
# Store raw value for later processing
|
| 574 |
-
self.accumulated_params[
|
| 575 |
-
self.current_param_name] = param_value
|
| 576 |
-
|
| 577 |
-
# Get parameter configuration for type conversion
|
| 578 |
-
param_config = self._get_arguments_config(
|
| 579 |
-
self.current_function_name,
|
| 580 |
-
self.streaming_request.tools
|
| 581 |
-
if self.streaming_request else None)
|
| 582 |
-
|
| 583 |
-
# Convert the parameter value to the appropriate type
|
| 584 |
-
converted_value = self._convert_param_value(
|
| 585 |
-
param_value, self.current_param_name,
|
| 586 |
-
param_config, self.current_function_name)
|
| 587 |
-
|
| 588 |
-
# Build JSON fragment based on the converted type
|
| 589 |
-
# Use json.dumps to properly serialize the value
|
| 590 |
-
serialized_value = json.dumps(converted_value,
|
| 591 |
-
ensure_ascii=False)
|
| 592 |
-
|
| 593 |
-
if self.param_count == 0:
|
| 594 |
-
json_fragment = f'"{self.current_param_name}": {serialized_value}'
|
| 595 |
-
else:
|
| 596 |
-
json_fragment = f', "{self.current_param_name}": {serialized_value}'
|
| 597 |
-
|
| 598 |
-
self.param_count += 1
|
| 599 |
-
|
| 600 |
-
return DeltaMessage(tool_calls=[
|
| 601 |
-
DeltaToolCall(
|
| 602 |
-
index=self.current_tool_index,
|
| 603 |
-
function=DeltaFunctionCall(
|
| 604 |
-
arguments=json_fragment),
|
| 605 |
-
)
|
| 606 |
-
])
|
| 607 |
-
|
| 608 |
-
# Continue parameter value - Not used in the current implementation
|
| 609 |
-
# since we process complete parameters above
|
| 610 |
-
if self.in_param:
|
| 611 |
-
if self.parameter_end_token in delta_text:
|
| 612 |
-
# End of parameter
|
| 613 |
-
end_idx = delta_text.find(self.parameter_end_token)
|
| 614 |
-
value_chunk = delta_text[:end_idx]
|
| 615 |
-
|
| 616 |
-
# Skip past > if at start
|
| 617 |
-
if not self.current_param_value and ">" in value_chunk:
|
| 618 |
-
gt_idx = value_chunk.find(">")
|
| 619 |
-
value_chunk = value_chunk[gt_idx + 1:]
|
| 620 |
-
|
| 621 |
-
if not self.current_param_value and value_chunk.startswith(
|
| 622 |
-
"\n"):
|
| 623 |
-
value_chunk = value_chunk[1:]
|
| 624 |
-
|
| 625 |
-
# Store complete value
|
| 626 |
-
full_value = self.current_param_value + value_chunk
|
| 627 |
-
self.accumulated_params[
|
| 628 |
-
self.current_param_name] = full_value
|
| 629 |
-
|
| 630 |
-
# Get parameter configuration for type conversion
|
| 631 |
-
param_config = self._get_arguments_config(
|
| 632 |
-
self.current_function_name,
|
| 633 |
-
self.streaming_request.tools
|
| 634 |
-
if self.streaming_request else None)
|
| 635 |
-
|
| 636 |
-
# Convert the parameter value to the appropriate type
|
| 637 |
-
converted_value = self._convert_param_value(
|
| 638 |
-
full_value, self.current_param_name, param_config,
|
| 639 |
-
self.current_function_name)
|
| 640 |
-
|
| 641 |
-
# Serialize the converted value
|
| 642 |
-
serialized_value = json.dumps(converted_value,
|
| 643 |
-
ensure_ascii=False)
|
| 644 |
-
|
| 645 |
-
# Since we've been streaming the quoted version, we need to close it properly
|
| 646 |
-
# This is complex - for now just complete the value
|
| 647 |
-
self.in_param = False
|
| 648 |
-
self.current_param_value = ""
|
| 649 |
-
|
| 650 |
-
# Just close the current parameter string
|
| 651 |
-
return DeltaMessage(tool_calls=[
|
| 652 |
-
DeltaToolCall(
|
| 653 |
-
index=self.current_tool_index,
|
| 654 |
-
function=DeltaFunctionCall(
|
| 655 |
-
arguments='"'), # Close the string quote
|
| 656 |
-
)
|
| 657 |
-
])
|
| 658 |
-
else:
|
| 659 |
-
# Continue accumulating value
|
| 660 |
-
value_chunk = delta_text
|
| 661 |
-
|
| 662 |
-
# Handle first chunk after param name
|
| 663 |
-
if not self.current_param_value and ">" in value_chunk:
|
| 664 |
-
gt_idx = value_chunk.find(">")
|
| 665 |
-
value_chunk = value_chunk[gt_idx + 1:]
|
| 666 |
-
|
| 667 |
-
if not self.current_param_value and value_chunk.startswith(
|
| 668 |
-
"\n"):
|
| 669 |
-
value_chunk = value_chunk[1:]
|
| 670 |
-
|
| 671 |
-
if value_chunk:
|
| 672 |
-
# Stream the escaped delta
|
| 673 |
-
prev_escaped = json.dumps(
|
| 674 |
-
self.current_param_value, ensure_ascii=False
|
| 675 |
-
)[1:-1] if self.current_param_value else ""
|
| 676 |
-
self.current_param_value += value_chunk
|
| 677 |
-
full_escaped = json.dumps(self.current_param_value,
|
| 678 |
-
ensure_ascii=False)[1:-1]
|
| 679 |
-
delta_escaped = full_escaped[len(prev_escaped):]
|
| 680 |
-
|
| 681 |
-
if delta_escaped:
|
| 682 |
-
return DeltaMessage(tool_calls=[
|
| 683 |
-
DeltaToolCall(
|
| 684 |
-
index=self.current_tool_index,
|
| 685 |
-
function=DeltaFunctionCall(
|
| 686 |
-
arguments=delta_escaped),
|
| 687 |
-
)
|
| 688 |
-
])
|
| 689 |
-
|
| 690 |
-
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tokenizer.json
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
DELETED
|
@@ -1,239 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"add_prefix_space": false,
|
| 3 |
-
"added_tokens_decoder": {
|
| 4 |
-
"151643": {
|
| 5 |
-
"content": "<|endoftext|>",
|
| 6 |
-
"lstrip": false,
|
| 7 |
-
"normalized": false,
|
| 8 |
-
"rstrip": false,
|
| 9 |
-
"single_word": false,
|
| 10 |
-
"special": true
|
| 11 |
-
},
|
| 12 |
-
"151644": {
|
| 13 |
-
"content": "<|im_start|>",
|
| 14 |
-
"lstrip": false,
|
| 15 |
-
"normalized": false,
|
| 16 |
-
"rstrip": false,
|
| 17 |
-
"single_word": false,
|
| 18 |
-
"special": true
|
| 19 |
-
},
|
| 20 |
-
"151645": {
|
| 21 |
-
"content": "<|im_end|>",
|
| 22 |
-
"lstrip": false,
|
| 23 |
-
"normalized": false,
|
| 24 |
-
"rstrip": false,
|
| 25 |
-
"single_word": false,
|
| 26 |
-
"special": true
|
| 27 |
-
},
|
| 28 |
-
"151646": {
|
| 29 |
-
"content": "<|object_ref_start|>",
|
| 30 |
-
"lstrip": false,
|
| 31 |
-
"normalized": false,
|
| 32 |
-
"rstrip": false,
|
| 33 |
-
"single_word": false,
|
| 34 |
-
"special": true
|
| 35 |
-
},
|
| 36 |
-
"151647": {
|
| 37 |
-
"content": "<|object_ref_end|>",
|
| 38 |
-
"lstrip": false,
|
| 39 |
-
"normalized": false,
|
| 40 |
-
"rstrip": false,
|
| 41 |
-
"single_word": false,
|
| 42 |
-
"special": true
|
| 43 |
-
},
|
| 44 |
-
"151648": {
|
| 45 |
-
"content": "<|box_start|>",
|
| 46 |
-
"lstrip": false,
|
| 47 |
-
"normalized": false,
|
| 48 |
-
"rstrip": false,
|
| 49 |
-
"single_word": false,
|
| 50 |
-
"special": true
|
| 51 |
-
},
|
| 52 |
-
"151649": {
|
| 53 |
-
"content": "<|box_end|>",
|
| 54 |
-
"lstrip": false,
|
| 55 |
-
"normalized": false,
|
| 56 |
-
"rstrip": false,
|
| 57 |
-
"single_word": false,
|
| 58 |
-
"special": true
|
| 59 |
-
},
|
| 60 |
-
"151650": {
|
| 61 |
-
"content": "<|quad_start|>",
|
| 62 |
-
"lstrip": false,
|
| 63 |
-
"normalized": false,
|
| 64 |
-
"rstrip": false,
|
| 65 |
-
"single_word": false,
|
| 66 |
-
"special": true
|
| 67 |
-
},
|
| 68 |
-
"151651": {
|
| 69 |
-
"content": "<|quad_end|>",
|
| 70 |
-
"lstrip": false,
|
| 71 |
-
"normalized": false,
|
| 72 |
-
"rstrip": false,
|
| 73 |
-
"single_word": false,
|
| 74 |
-
"special": true
|
| 75 |
-
},
|
| 76 |
-
"151652": {
|
| 77 |
-
"content": "<|vision_start|>",
|
| 78 |
-
"lstrip": false,
|
| 79 |
-
"normalized": false,
|
| 80 |
-
"rstrip": false,
|
| 81 |
-
"single_word": false,
|
| 82 |
-
"special": true
|
| 83 |
-
},
|
| 84 |
-
"151653": {
|
| 85 |
-
"content": "<|vision_end|>",
|
| 86 |
-
"lstrip": false,
|
| 87 |
-
"normalized": false,
|
| 88 |
-
"rstrip": false,
|
| 89 |
-
"single_word": false,
|
| 90 |
-
"special": true
|
| 91 |
-
},
|
| 92 |
-
"151654": {
|
| 93 |
-
"content": "<|vision_pad|>",
|
| 94 |
-
"lstrip": false,
|
| 95 |
-
"normalized": false,
|
| 96 |
-
"rstrip": false,
|
| 97 |
-
"single_word": false,
|
| 98 |
-
"special": true
|
| 99 |
-
},
|
| 100 |
-
"151655": {
|
| 101 |
-
"content": "<|image_pad|>",
|
| 102 |
-
"lstrip": false,
|
| 103 |
-
"normalized": false,
|
| 104 |
-
"rstrip": false,
|
| 105 |
-
"single_word": false,
|
| 106 |
-
"special": true
|
| 107 |
-
},
|
| 108 |
-
"151656": {
|
| 109 |
-
"content": "<|video_pad|>",
|
| 110 |
-
"lstrip": false,
|
| 111 |
-
"normalized": false,
|
| 112 |
-
"rstrip": false,
|
| 113 |
-
"single_word": false,
|
| 114 |
-
"special": true
|
| 115 |
-
},
|
| 116 |
-
"151657": {
|
| 117 |
-
"content": "<tool_call>",
|
| 118 |
-
"lstrip": false,
|
| 119 |
-
"normalized": false,
|
| 120 |
-
"rstrip": false,
|
| 121 |
-
"single_word": false,
|
| 122 |
-
"special": false
|
| 123 |
-
},
|
| 124 |
-
"151658": {
|
| 125 |
-
"content": "</tool_call>",
|
| 126 |
-
"lstrip": false,
|
| 127 |
-
"normalized": false,
|
| 128 |
-
"rstrip": false,
|
| 129 |
-
"single_word": false,
|
| 130 |
-
"special": false
|
| 131 |
-
},
|
| 132 |
-
"151659": {
|
| 133 |
-
"content": "<|fim_prefix|>",
|
| 134 |
-
"lstrip": false,
|
| 135 |
-
"normalized": false,
|
| 136 |
-
"rstrip": false,
|
| 137 |
-
"single_word": false,
|
| 138 |
-
"special": false
|
| 139 |
-
},
|
| 140 |
-
"151660": {
|
| 141 |
-
"content": "<|fim_middle|>",
|
| 142 |
-
"lstrip": false,
|
| 143 |
-
"normalized": false,
|
| 144 |
-
"rstrip": false,
|
| 145 |
-
"single_word": false,
|
| 146 |
-
"special": false
|
| 147 |
-
},
|
| 148 |
-
"151661": {
|
| 149 |
-
"content": "<|fim_suffix|>",
|
| 150 |
-
"lstrip": false,
|
| 151 |
-
"normalized": false,
|
| 152 |
-
"rstrip": false,
|
| 153 |
-
"single_word": false,
|
| 154 |
-
"special": false
|
| 155 |
-
},
|
| 156 |
-
"151662": {
|
| 157 |
-
"content": "<|fim_pad|>",
|
| 158 |
-
"lstrip": false,
|
| 159 |
-
"normalized": false,
|
| 160 |
-
"rstrip": false,
|
| 161 |
-
"single_word": false,
|
| 162 |
-
"special": false
|
| 163 |
-
},
|
| 164 |
-
"151663": {
|
| 165 |
-
"content": "<|repo_name|>",
|
| 166 |
-
"lstrip": false,
|
| 167 |
-
"normalized": false,
|
| 168 |
-
"rstrip": false,
|
| 169 |
-
"single_word": false,
|
| 170 |
-
"special": false
|
| 171 |
-
},
|
| 172 |
-
"151664": {
|
| 173 |
-
"content": "<|file_sep|>",
|
| 174 |
-
"lstrip": false,
|
| 175 |
-
"normalized": false,
|
| 176 |
-
"rstrip": false,
|
| 177 |
-
"single_word": false,
|
| 178 |
-
"special": false
|
| 179 |
-
},
|
| 180 |
-
"151665": {
|
| 181 |
-
"content": "<tool_response>",
|
| 182 |
-
"lstrip": false,
|
| 183 |
-
"normalized": false,
|
| 184 |
-
"rstrip": false,
|
| 185 |
-
"single_word": false,
|
| 186 |
-
"special": false
|
| 187 |
-
},
|
| 188 |
-
"151666": {
|
| 189 |
-
"content": "</tool_response>",
|
| 190 |
-
"lstrip": false,
|
| 191 |
-
"normalized": false,
|
| 192 |
-
"rstrip": false,
|
| 193 |
-
"single_word": false,
|
| 194 |
-
"special": false
|
| 195 |
-
},
|
| 196 |
-
"151667": {
|
| 197 |
-
"content": "<think>",
|
| 198 |
-
"lstrip": false,
|
| 199 |
-
"normalized": false,
|
| 200 |
-
"rstrip": false,
|
| 201 |
-
"single_word": false,
|
| 202 |
-
"special": false
|
| 203 |
-
},
|
| 204 |
-
"151668": {
|
| 205 |
-
"content": "</think>",
|
| 206 |
-
"lstrip": false,
|
| 207 |
-
"normalized": false,
|
| 208 |
-
"rstrip": false,
|
| 209 |
-
"single_word": false,
|
| 210 |
-
"special": false
|
| 211 |
-
}
|
| 212 |
-
},
|
| 213 |
-
"additional_special_tokens": [
|
| 214 |
-
"<|im_start|>",
|
| 215 |
-
"<|im_end|>",
|
| 216 |
-
"<|object_ref_start|>",
|
| 217 |
-
"<|object_ref_end|>",
|
| 218 |
-
"<|box_start|>",
|
| 219 |
-
"<|box_end|>",
|
| 220 |
-
"<|quad_start|>",
|
| 221 |
-
"<|quad_end|>",
|
| 222 |
-
"<|vision_start|>",
|
| 223 |
-
"<|vision_end|>",
|
| 224 |
-
"<|vision_pad|>",
|
| 225 |
-
"<|image_pad|>",
|
| 226 |
-
"<|video_pad|>"
|
| 227 |
-
],
|
| 228 |
-
"bos_token": null,
|
| 229 |
-
"chat_template": "{% macro render_extra_keys(json_dict, handled_keys) %}\n {%- if json_dict is mapping %}\n {%- for json_key in json_dict if json_key not in handled_keys %}\n {%- if json_dict[json_key] is string %}\n {{-'\\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | string) ~ '</' ~ json_key ~ '>' }}\n {%- else %}\n {{- '\\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | tojson | safe) ~ '</' ~ json_key ~ '>' }}\n {%- endif %}\n {%- endfor %}\n {%- endif %}\n{%- endmacro %}\n\n{%- if messages[0][\"role\"] == \"system\" %}\n {%- set system_message = messages[0][\"content\"] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n\n{%- if not tools is defined %}\n {%- set tools = [] %}\n{%- endif %}\n\n{%- if system_message is defined %}\n {{- \"<|im_start|>system\\n\" + system_message }}\n{%- else %}\n {%- if tools is iterable and tools | length > 0 %}\n {{- \"<|im_start|>system\\nYou are Qwen, a helpful AI assistant that can interact with a computer to solve tasks.\" }}\n {%- endif %}\n{%- endif %}\n{%- if tools is iterable and tools | length > 0 %}\n {{- \"\\n\\n# Tools\\n\\nYou have access to the following functions:\\n\\n\" }}\n {{- \"<tools>\" }}\n {%- for tool in tools %}\n {%- if tool.function is defined %}\n {%- set tool = tool.function %}\n {%- endif %}\n {{- \"\\n<function>\\n<name>\" ~ tool.name ~ \"</name>\" }}\n {%- if tool.description is defined %}\n {{- '\\n<description>' ~ (tool.description | trim) ~ '</description>' }}\n {%- endif %}\n {{- '\\n<parameters>' }}\n {%- if tool.parameters is defined and tool.parameters is mapping and tool.parameters.properties is defined and tool.parameters.properties is mapping %}\n {%- for param_name, param_fields in tool.parameters.properties|items %}\n {{- '\\n<parameter>' }}\n {{- '\\n<name>' ~ param_name ~ '</name>' }}\n {%- if param_fields.type is defined %}\n {{- '\\n<type>' ~ (param_fields.type | string) ~ '</type>' }}\n {%- endif %}\n {%- if param_fields.description is defined %}\n {{- '\\n<description>' ~ (param_fields.description | trim) ~ '</description>' }}\n {%- endif %}\n {%- set handled_keys = ['name', 'type', 'description'] %}\n {{- render_extra_keys(param_fields, handled_keys) }}\n {{- '\\n</parameter>' }}\n {%- endfor %}\n {%- endif %}\n {%- set handled_keys = ['type', 'properties'] %}\n {{- render_extra_keys(tool.parameters, handled_keys) }}\n {{- '\\n</parameters>' }}\n {%- set handled_keys = ['type', 'name', 'description', 'parameters'] %}\n {{- render_extra_keys(tool, handled_keys) }}\n {{- '\\n</function>' }}\n {%- endfor %}\n {{- \"\\n</tools>\" }}\n {{- '\\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>' }}\n{%- endif %}\n{%- if system_message is defined %}\n {{- '<|im_end|>\\n' }}\n{%- else %}\n {%- if tools is iterable and tools | length > 0 %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- for message in loop_messages %}\n {%- if message.role == \"assistant\" and message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}\n {{- '<|im_start|>' + message.role }}\n {%- if message.content is defined and message.content is string and message.content | trim | length > 0 %}\n {{- '\\n' + message.content | trim + '\\n' }}\n {%- endif %}\n {%- for tool_call in message.tool_calls %}\n {%- if tool_call.function is defined %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {{- '\\n<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n {%- if tool_call.arguments is defined %}\n {%- for args_name, args_value in tool_call.arguments|items %}\n {{- '<parameter=' + args_name + '>\\n' }}\n {%- set args_value = args_value if args_value is string else args_value | tojson | safe %}\n {{- args_value }}\n {{- '\\n</parameter>\\n' }}\n {%- endfor %}\n {%- endif %}\n {{- '</function>\\n</tool_call>' }}\n {%- endfor %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"user\" or message.role == \"system\" or message.role == \"assistant\" %}\n {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if loop.previtem and loop.previtem.role != \"tool\" %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- message.content }}\n {{- '\\n</tool_response>' }}\n {%- if not loop.last and loop.nextitem.role != \"tool\" %}\n {{- '<|im_end|>\\n' }}\n {%- elif loop.last %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>\\n' }}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n{%- endif %}",
|
| 230 |
-
"clean_up_tokenization_spaces": false,
|
| 231 |
-
"eos_token": "<|im_end|>",
|
| 232 |
-
"errors": "replace",
|
| 233 |
-
"model_max_length": 1048576,
|
| 234 |
-
"pad_token": "<|endoftext|>",
|
| 235 |
-
"split_special_tokens": false,
|
| 236 |
-
"tokenizer_class": "Qwen2Tokenizer",
|
| 237 |
-
"unk_token": null,
|
| 238 |
-
"add_bos_token": false
|
| 239 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|