初始化项目，由ModelHub XC社区提供模型
Model: anyreach-ai/semantic-turn-taking Source: Original Platform
This commit is contained in:
ModelHub XC
38
.gitattributes
vendored
Normal file
38
.gitattributes
vendored
Normal file
@@ -0,0 +1,38 @@
|
||||
*.7z filter=lfs diff=lfs merge=lfs -text
|
||||
*.arrow filter=lfs diff=lfs merge=lfs -text
|
||||
*.bin filter=lfs diff=lfs merge=lfs -text
|
||||
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
||||
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
||||
*.ftz filter=lfs diff=lfs merge=lfs -text
|
||||
*.gz filter=lfs diff=lfs merge=lfs -text
|
||||
*.h5 filter=lfs diff=lfs merge=lfs -text
|
||||
*.joblib filter=lfs diff=lfs merge=lfs -text
|
||||
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
||||
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
||||
*.model filter=lfs diff=lfs merge=lfs -text
|
||||
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
||||
*.npy filter=lfs diff=lfs merge=lfs -text
|
||||
*.npz filter=lfs diff=lfs merge=lfs -text
|
||||
*.onnx filter=lfs diff=lfs merge=lfs -text
|
||||
*.ot filter=lfs diff=lfs merge=lfs -text
|
||||
*.parquet filter=lfs diff=lfs merge=lfs -text
|
||||
*.pb filter=lfs diff=lfs merge=lfs -text
|
||||
*.pickle filter=lfs diff=lfs merge=lfs -text
|
||||
*.pkl filter=lfs diff=lfs merge=lfs -text
|
||||
*.pt filter=lfs diff=lfs merge=lfs -text
|
||||
*.pth filter=lfs diff=lfs merge=lfs -text
|
||||
*.rar filter=lfs diff=lfs merge=lfs -text
|
||||
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
||||
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
||||
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
||||
*.tar filter=lfs diff=lfs merge=lfs -text
|
||||
*.tflite filter=lfs diff=lfs merge=lfs -text
|
||||
*.tgz filter=lfs diff=lfs merge=lfs -text
|
||||
*.wasm filter=lfs diff=lfs merge=lfs -text
|
||||
*.xz filter=lfs diff=lfs merge=lfs -text
|
||||
*.zip filter=lfs diff=lfs merge=lfs -text
|
||||
*.zst filter=lfs diff=lfs merge=lfs -text
|
||||
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
||||
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
||||
onnx/model.onnx_data filter=lfs diff=lfs merge=lfs -text
|
||||
onnx/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
||||
186
README.md
Normal file
186
README.md
Normal file
@@ -0,0 +1,186 @@
|
||||
---
|
||||
license: apache-2.0
|
||||
language:
|
||||
- en
|
||||
library_name: transformers
|
||||
tags:
|
||||
- turn-taking
|
||||
- voice-ai
|
||||
- conversational-ai
|
||||
- dialogue
|
||||
- qwen2
|
||||
- onnx
|
||||
base_model: Qwen/Qwen2.5-0.5B-Instruct
|
||||
pipeline_tag: text-generation
|
||||
---
|
||||
|
||||
# Semantic Turn-Taking Model
|
||||
|
||||
A fine-tuned [Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct) model that predicts turn-taking actions in conversations. Given a conversation context, the model predicts what action a voice AI agent should take next.
|
||||
|
||||
Unlike acoustic-based approaches (VAD, silence detection), this model uses the **semantic content** of the conversation to make turn-taking decisions.
|
||||
|
||||
## Action Classes
|
||||
|
||||
The model predicts one of 4 actions:
|
||||
|
||||
| Action | Token | Description |
|
||||
|--------|-------|-------------|
|
||||
| `start_speaking` | `<\|start_speaking\|>` | User finished their turn, agent should respond |
|
||||
| `continue_listening` | `<\|continue_listening\|>` | User is mid-utterance, keep listening |
|
||||
| `start_listening` | `<\|start_listening\|>` | User interrupted the agent, stop talking |
|
||||
| `continue_speaking` | `<\|continue_speaking\|>` | User gave a backchannel, agent keeps talking |
|
||||
|
||||
## Usage
|
||||
|
||||
### PyTorch
|
||||
|
||||
```python
|
||||
import torch
|
||||
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||||
|
||||
model_name = "anyreach-ai/semantic-turn-taking"
|
||||
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
||||
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16).cuda()
|
||||
model.eval()
|
||||
|
||||
# Format conversation as ChatML with <|predict|> trigger
|
||||
conversation = """<|im_start|>user
|
||||
I need help with my bill<|im_end|>
|
||||
<|im_start|>assistant
|
||||
Sure I can help with that what seems to be the issue<|im_end|>
|
||||
<|im_start|>user
|
||||
I was charged twice for the same order<|im_end|>
|
||||
<|predict|>"""
|
||||
|
||||
inputs = tokenizer(conversation, return_tensors="pt").to("cuda")
|
||||
|
||||
with torch.no_grad():
|
||||
logits = model(**inputs).logits[:, -1, :]
|
||||
|
||||
# Get action probabilities
|
||||
action_tokens = {
|
||||
"start_speaking": tokenizer.convert_tokens_to_ids("<|start_speaking|>"),
|
||||
"continue_listening": tokenizer.convert_tokens_to_ids("<|continue_listening|>"),
|
||||
"start_listening": tokenizer.convert_tokens_to_ids("<|start_listening|>"),
|
||||
"continue_speaking": tokenizer.convert_tokens_to_ids("<|continue_speaking|>"),
|
||||
}
|
||||
|
||||
action_logits = {name: logits[0, tid].item() for name, tid in action_tokens.items()}
|
||||
probs = torch.softmax(torch.tensor(list(action_logits.values())), dim=0)
|
||||
for (name, _), p in zip(action_logits.items(), probs):
|
||||
print(f" {name}: {p:.4f}")
|
||||
# → start_speaking: 0.95+ (user is done, agent should respond)
|
||||
```
|
||||
|
||||
### ONNX (CPU)
|
||||
|
||||
```python
|
||||
import numpy as np
|
||||
import onnxruntime as ort
|
||||
from transformers import AutoTokenizer
|
||||
|
||||
tokenizer = AutoTokenizer.from_pretrained("anyreach-ai/semantic-turn-taking")
|
||||
|
||||
sess_options = ort.SessionOptions()
|
||||
sess_options.intra_op_num_threads = 4
|
||||
session = ort.InferenceSession(
|
||||
"onnx/model_q8.onnx", # download from this repo
|
||||
providers=["CPUExecutionProvider"],
|
||||
sess_options=sess_options,
|
||||
)
|
||||
|
||||
# Tokenize
|
||||
conversation = "..." # ChatML format as above
|
||||
inputs = tokenizer(conversation, return_tensors="np")
|
||||
input_ids = inputs["input_ids"].astype("int64")
|
||||
seq_len = input_ids.shape[1]
|
||||
|
||||
# Build feed (empty KV cache for single forward pass)
|
||||
feed = {
|
||||
"input_ids": input_ids,
|
||||
"attention_mask": inputs["attention_mask"].astype("int64"),
|
||||
"position_ids": np.arange(seq_len, dtype="int64").reshape(1, -1),
|
||||
}
|
||||
for i in range(24):
|
||||
feed[f"past_key_values.{i}.key"] = np.zeros((1, 2, 0, 64), dtype="float32")
|
||||
feed[f"past_key_values.{i}.value"] = np.zeros((1, 2, 0, 64), dtype="float32")
|
||||
|
||||
# Run inference
|
||||
logits = session.run(None, feed)[0] # [1, seq_len, vocab_size]
|
||||
last_logits = logits[0, -1, :]
|
||||
|
||||
# Extract action probabilities
|
||||
ACTION_IDS = [151666, 151665, 151667, 151668] # SS, CL, SLi, CS
|
||||
action_logits = last_logits[ACTION_IDS]
|
||||
probs = np.exp(action_logits) / np.sum(np.exp(action_logits))
|
||||
```
|
||||
|
||||
## Benchmark Results
|
||||
|
||||
Evaluated on [anyreach-ai/semantic-turn-taking-benchmark](https://huggingface.co/datasets/anyreach-ai/semantic-turn-taking-benchmark).
|
||||
|
||||
### Binary (EOU vs Not-EOU)
|
||||
|
||||
Only `start_speaking` and `continue_listening` examples. Predictions mapped: SS/CS → EOU, CL/SLi → Not-EOU.
|
||||
|
||||
| Subset | N | Accuracy | F1 (macro) |
|
||||
|--------|--:|--:|--:|
|
||||
| TEN | 428 | 91.82% | 91.80% |
|
||||
| SwDA | 2,688 | 65.96% | 51.46% |
|
||||
| Synthetic | 36 | 86.11% | 85.57% |
|
||||
|
||||
### Multi-class
|
||||
|
||||
| Subset | N | Classes | Accuracy | F1 (macro) |
|
||||
|--------|--:|--------:|--:|--:|
|
||||
| TEN | 428 | 2 | 91.82% | 91.80% |
|
||||
| SwDA | 3,523 | 3 | 68.98% | 46.92% |
|
||||
| Synthetic | 60 | 4 | 76.67% | 72.07% |
|
||||
|
||||
## Latency
|
||||
|
||||
Measured on single examples, CPU (4 threads) and GPU (NVIDIA T4).
|
||||
|
||||
| Format | Size | Short (8 tok) | Medium (28 tok) | Long (54 tok) |
|
||||
|--------|-----:|--:|--:|--:|
|
||||
| PyTorch GPU (fp16) | 942 MB | 26 ms | 30 ms | 34 ms |
|
||||
| PyTorch CPU (fp32) | 942 MB | 165 ms | 247 ms | 289 ms |
|
||||
| ONNX CPU (q8) | 473 MB | 128 ms | 151 ms | 191 ms |
|
||||
|
||||
## Model Details
|
||||
|
||||
- **Base model**: [Qwen/Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct) (494M parameters)
|
||||
- **Training**: Full fine-tuning on ~154K synthetic conversation examples
|
||||
- **Input format**: Qwen ChatML with `<|predict|>` trigger token
|
||||
- **Max sequence length**: 1024 tokens (left truncation)
|
||||
- **Special tokens**: 5 added (`<|predict|>`, 4 action tokens)
|
||||
|
||||
## Files
|
||||
|
||||
| File | Description |
|
||||
|------|-------------|
|
||||
| `model.safetensors` | PyTorch model weights (fp32) |
|
||||
| `onnx/model_q8.onnx` | ONNX INT8 quantized (dynamic quantization) |
|
||||
| `config.json` | Model configuration |
|
||||
| `tokenizer.json` | Tokenizer |
|
||||
|
||||
## Citation
|
||||
|
||||
```bibtex
|
||||
@misc{semantic-turn-taking-2026,
|
||||
title={Semantic Turn-Taking Model},
|
||||
author={Shangeth Rajaa},
|
||||
year={2026},
|
||||
publisher={Hugging Face},
|
||||
url={https://huggingface.co/anyreach-ai/semantic-turn-taking}
|
||||
}
|
||||
```
|
||||
|
||||
## Authors
|
||||
|
||||
- [**Shangeth Rajaa**](https://github.com/shangeth)
|
||||
|
||||
## License
|
||||
|
||||
Apache 2.0
|
||||
29
added_tokens.json
Normal file
29
added_tokens.json
Normal file
@@ -0,0 +1,29 @@
|
||||
{
|
||||
"</tool_call>": 151658,
|
||||
"<tool_call>": 151657,
|
||||
"<|box_end|>": 151649,
|
||||
"<|box_start|>": 151648,
|
||||
"<|continue_listening|>": 151665,
|
||||
"<|continue_speaking|>": 151668,
|
||||
"<|endoftext|>": 151643,
|
||||
"<|file_sep|>": 151664,
|
||||
"<|fim_middle|>": 151660,
|
||||
"<|fim_pad|>": 151662,
|
||||
"<|fim_prefix|>": 151659,
|
||||
"<|fim_suffix|>": 151661,
|
||||
"<|im_end|>": 151645,
|
||||
"<|im_start|>": 151644,
|
||||
"<|image_pad|>": 151655,
|
||||
"<|object_ref_end|>": 151647,
|
||||
"<|object_ref_start|>": 151646,
|
||||
"<|predict|>": 151669,
|
||||
"<|quad_end|>": 151651,
|
||||
"<|quad_start|>": 151650,
|
||||
"<|repo_name|>": 151663,
|
||||
"<|start_listening|>": 151667,
|
||||
"<|start_speaking|>": 151666,
|
||||
"<|video_pad|>": 151656,
|
||||
"<|vision_end|>": 151653,
|
||||
"<|vision_pad|>": 151654,
|
||||
"<|vision_start|>": 151652
|
||||
}
|
||||
54
chat_template.jinja
Normal file
54
chat_template.jinja
Normal file
@@ -0,0 +1,54 @@
|
||||
{%- if tools %}
|
||||
{{- '<|im_start|>system\n' }}
|
||||
{%- if messages[0]['role'] == 'system' %}
|
||||
{{- messages[0]['content'] }}
|
||||
{%- else %}
|
||||
{{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
|
||||
{%- endif %}
|
||||
{{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
||||
{%- for tool in tools %}
|
||||
{{- "\n" }}
|
||||
{{- tool | tojson }}
|
||||
{%- endfor %}
|
||||
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
||||
{%- else %}
|
||||
{%- if messages[0]['role'] == 'system' %}
|
||||
{{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
|
||||
{%- else %}
|
||||
{{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
|
||||
{%- endif %}
|
||||
{%- endif %}
|
||||
{%- for message in messages %}
|
||||
{%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
|
||||
{{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
|
||||
{%- elif message.role == "assistant" %}
|
||||
{{- '<|im_start|>' + message.role }}
|
||||
{%- if message.content %}
|
||||
{{- '\n' + message.content }}
|
||||
{%- endif %}
|
||||
{%- for tool_call in message.tool_calls %}
|
||||
{%- if tool_call.function is defined %}
|
||||
{%- set tool_call = tool_call.function %}
|
||||
{%- endif %}
|
||||
{{- '\n<tool_call>\n{"name": "' }}
|
||||
{{- tool_call.name }}
|
||||
{{- '", "arguments": ' }}
|
||||
{{- tool_call.arguments | tojson }}
|
||||
{{- '}\n</tool_call>' }}
|
||||
{%- endfor %}
|
||||
{{- '<|im_end|>\n' }}
|
||||
{%- elif message.role == "tool" %}
|
||||
{%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
|
||||
{{- '<|im_start|>user' }}
|
||||
{%- endif %}
|
||||
{{- '\n<tool_response>\n' }}
|
||||
{{- message.content }}
|
||||
{{- '\n</tool_response>' }}
|
||||
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
||||
{{- '<|im_end|>\n' }}
|
||||
{%- endif %}
|
||||
{%- endif %}
|
||||
{%- endfor %}
|
||||
{%- if add_generation_prompt %}
|
||||
{{- '<|im_start|>assistant\n' }}
|
||||
{%- endif %}
|
||||
55
config.json
Normal file
55
config.json
Normal file
@@ -0,0 +1,55 @@
|
||||
{
|
||||
"architectures": [
|
||||
"Qwen2ForCausalLM"
|
||||
],
|
||||
"attention_dropout": 0.0,
|
||||
"bos_token_id": 151643,
|
||||
"dtype": "bfloat16",
|
||||
"eos_token_id": 151645,
|
||||
"hidden_act": "silu",
|
||||
"hidden_size": 896,
|
||||
"initializer_range": 0.02,
|
||||
"intermediate_size": 4864,
|
||||
"layer_types": [
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention"
|
||||
],
|
||||
"max_position_embeddings": 32768,
|
||||
"max_window_layers": 21,
|
||||
"model_type": "qwen2",
|
||||
"num_attention_heads": 14,
|
||||
"num_hidden_layers": 24,
|
||||
"num_key_value_heads": 2,
|
||||
"pad_token_id": 151643,
|
||||
"rms_norm_eps": 1e-06,
|
||||
"rope_scaling": null,
|
||||
"rope_theta": 1000000.0,
|
||||
"sliding_window": null,
|
||||
"tie_word_embeddings": true,
|
||||
"transformers_version": "4.57.6",
|
||||
"use_cache": true,
|
||||
"use_sliding_window": false,
|
||||
"vocab_size": 151670
|
||||
}
|
||||
14
generation_config.json
Normal file
14
generation_config.json
Normal file
@@ -0,0 +1,14 @@
|
||||
{
|
||||
"bos_token_id": 151643,
|
||||
"do_sample": true,
|
||||
"eos_token_id": [
|
||||
151645,
|
||||
151643
|
||||
],
|
||||
"pad_token_id": 151643,
|
||||
"repetition_penalty": 1.1,
|
||||
"temperature": 0.7,
|
||||
"top_k": 20,
|
||||
"top_p": 0.8,
|
||||
"transformers_version": "4.57.6"
|
||||
}
|
||||
151388
merges.txt
Normal file
151388
merges.txt
Normal file
File diff suppressed because it is too large
Load Diff
3
model.safetensors
Normal file
3
model.safetensors
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:7bb1d46d3bfb9b60f62ce3e3ae0f444df4cfa12ed9e709ba713d87719126c21e
|
||||
size 987621152
|
||||
29
onnx/added_tokens.json
Normal file
29
onnx/added_tokens.json
Normal file
@@ -0,0 +1,29 @@
|
||||
{
|
||||
"</tool_call>": 151658,
|
||||
"<tool_call>": 151657,
|
||||
"<|box_end|>": 151649,
|
||||
"<|box_start|>": 151648,
|
||||
"<|continue_listening|>": 151665,
|
||||
"<|continue_speaking|>": 151668,
|
||||
"<|endoftext|>": 151643,
|
||||
"<|file_sep|>": 151664,
|
||||
"<|fim_middle|>": 151660,
|
||||
"<|fim_pad|>": 151662,
|
||||
"<|fim_prefix|>": 151659,
|
||||
"<|fim_suffix|>": 151661,
|
||||
"<|im_end|>": 151645,
|
||||
"<|im_start|>": 151644,
|
||||
"<|image_pad|>": 151655,
|
||||
"<|object_ref_end|>": 151647,
|
||||
"<|object_ref_start|>": 151646,
|
||||
"<|predict|>": 151669,
|
||||
"<|quad_end|>": 151651,
|
||||
"<|quad_start|>": 151650,
|
||||
"<|repo_name|>": 151663,
|
||||
"<|start_listening|>": 151667,
|
||||
"<|start_speaking|>": 151666,
|
||||
"<|video_pad|>": 151656,
|
||||
"<|vision_end|>": 151653,
|
||||
"<|vision_pad|>": 151654,
|
||||
"<|vision_start|>": 151652
|
||||
}
|
||||
54
onnx/chat_template.jinja
Normal file
54
onnx/chat_template.jinja
Normal file
@@ -0,0 +1,54 @@
|
||||
{%- if tools %}
|
||||
{{- '<|im_start|>system\n' }}
|
||||
{%- if messages[0]['role'] == 'system' %}
|
||||
{{- messages[0]['content'] }}
|
||||
{%- else %}
|
||||
{{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
|
||||
{%- endif %}
|
||||
{{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
||||
{%- for tool in tools %}
|
||||
{{- "\n" }}
|
||||
{{- tool | tojson }}
|
||||
{%- endfor %}
|
||||
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
||||
{%- else %}
|
||||
{%- if messages[0]['role'] == 'system' %}
|
||||
{{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
|
||||
{%- else %}
|
||||
{{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
|
||||
{%- endif %}
|
||||
{%- endif %}
|
||||
{%- for message in messages %}
|
||||
{%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
|
||||
{{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
|
||||
{%- elif message.role == "assistant" %}
|
||||
{{- '<|im_start|>' + message.role }}
|
||||
{%- if message.content %}
|
||||
{{- '\n' + message.content }}
|
||||
{%- endif %}
|
||||
{%- for tool_call in message.tool_calls %}
|
||||
{%- if tool_call.function is defined %}
|
||||
{%- set tool_call = tool_call.function %}
|
||||
{%- endif %}
|
||||
{{- '\n<tool_call>\n{"name": "' }}
|
||||
{{- tool_call.name }}
|
||||
{{- '", "arguments": ' }}
|
||||
{{- tool_call.arguments | tojson }}
|
||||
{{- '}\n</tool_call>' }}
|
||||
{%- endfor %}
|
||||
{{- '<|im_end|>\n' }}
|
||||
{%- elif message.role == "tool" %}
|
||||
{%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
|
||||
{{- '<|im_start|>user' }}
|
||||
{%- endif %}
|
||||
{{- '\n<tool_response>\n' }}
|
||||
{{- message.content }}
|
||||
{{- '\n</tool_response>' }}
|
||||
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
||||
{{- '<|im_end|>\n' }}
|
||||
{%- endif %}
|
||||
{%- endif %}
|
||||
{%- endfor %}
|
||||
{%- if add_generation_prompt %}
|
||||
{{- '<|im_start|>assistant\n' }}
|
||||
{%- endif %}
|
||||
55
onnx/config.json
Normal file
55
onnx/config.json
Normal file
@@ -0,0 +1,55 @@
|
||||
{
|
||||
"architectures": [
|
||||
"Qwen2ForCausalLM"
|
||||
],
|
||||
"attention_dropout": 0.0,
|
||||
"bos_token_id": 151643,
|
||||
"dtype": "bfloat16",
|
||||
"eos_token_id": 151645,
|
||||
"hidden_act": "silu",
|
||||
"hidden_size": 896,
|
||||
"initializer_range": 0.02,
|
||||
"intermediate_size": 4864,
|
||||
"layer_types": [
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention",
|
||||
"full_attention"
|
||||
],
|
||||
"max_position_embeddings": 32768,
|
||||
"max_window_layers": 21,
|
||||
"model_type": "qwen2",
|
||||
"num_attention_heads": 14,
|
||||
"num_hidden_layers": 24,
|
||||
"num_key_value_heads": 2,
|
||||
"pad_token_id": 151643,
|
||||
"rms_norm_eps": 1e-06,
|
||||
"rope_scaling": null,
|
||||
"rope_theta": 1000000.0,
|
||||
"sliding_window": null,
|
||||
"tie_word_embeddings": true,
|
||||
"transformers_version": "4.57.6",
|
||||
"use_cache": true,
|
||||
"use_sliding_window": false,
|
||||
"vocab_size": 151670
|
||||
}
|
||||
14
onnx/generation_config.json
Normal file
14
onnx/generation_config.json
Normal file
@@ -0,0 +1,14 @@
|
||||
{
|
||||
"bos_token_id": 151643,
|
||||
"do_sample": true,
|
||||
"eos_token_id": [
|
||||
151645,
|
||||
151643
|
||||
],
|
||||
"pad_token_id": 151643,
|
||||
"repetition_penalty": 1.1,
|
||||
"temperature": 0.7,
|
||||
"top_k": 20,
|
||||
"top_p": 0.8,
|
||||
"transformers_version": "4.57.6"
|
||||
}
|
||||
151388
onnx/merges.txt
Normal file
151388
onnx/merges.txt
Normal file
File diff suppressed because it is too large
Load Diff
3
onnx/model.onnx
Normal file
3
onnx/model.onnx
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:46a669b8d19c01c7ce6147394c421f2fa2904d0ab948b7275ed10b19148170c2
|
||||
size 1125454
|
||||
3
onnx/model.onnx_data
Normal file
3
onnx/model.onnx_data
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:25418b69c1666c540a7117db39689bb7949aaebb1f323e39b5fbddb4f4d0e6fb
|
||||
size 1975174272
|
||||
3
onnx/model_q8.onnx
Normal file
3
onnx/model_q8.onnx
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:1974c790d2aa557642b0906eafbb593af8ca48fc045e5b9d187d7fa95faf88fe
|
||||
size 495715962
|
||||
53
onnx/special_tokens_map.json
Normal file
53
onnx/special_tokens_map.json
Normal file
@@ -0,0 +1,53 @@
|
||||
{
|
||||
"additional_special_tokens": [
|
||||
{
|
||||
"content": "<|continue_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|start_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|start_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|continue_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|predict|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
}
|
||||
],
|
||||
"eos_token": {
|
||||
"content": "<|im_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
"pad_token": {
|
||||
"content": "<|endoftext|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
}
|
||||
}
|
||||
3
onnx/tokenizer.json
Normal file
3
onnx/tokenizer.json
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:cb8544e622fff1dd549fddc8273bd5fd327f9642604868434f594f19ac2a49cc
|
||||
size 11422970
|
||||
239
onnx/tokenizer_config.json
Normal file
239
onnx/tokenizer_config.json
Normal file
@@ -0,0 +1,239 @@
|
||||
{
|
||||
"add_bos_token": false,
|
||||
"add_prefix_space": false,
|
||||
"added_tokens_decoder": {
|
||||
"151643": {
|
||||
"content": "<|endoftext|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151644": {
|
||||
"content": "<|im_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151645": {
|
||||
"content": "<|im_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151646": {
|
||||
"content": "<|object_ref_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151647": {
|
||||
"content": "<|object_ref_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151648": {
|
||||
"content": "<|box_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151649": {
|
||||
"content": "<|box_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151650": {
|
||||
"content": "<|quad_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151651": {
|
||||
"content": "<|quad_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151652": {
|
||||
"content": "<|vision_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151653": {
|
||||
"content": "<|vision_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151654": {
|
||||
"content": "<|vision_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151655": {
|
||||
"content": "<|image_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151656": {
|
||||
"content": "<|video_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151657": {
|
||||
"content": "<tool_call>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151658": {
|
||||
"content": "</tool_call>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151659": {
|
||||
"content": "<|fim_prefix|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151660": {
|
||||
"content": "<|fim_middle|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151661": {
|
||||
"content": "<|fim_suffix|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151662": {
|
||||
"content": "<|fim_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151663": {
|
||||
"content": "<|repo_name|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151664": {
|
||||
"content": "<|file_sep|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151665": {
|
||||
"content": "<|continue_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151666": {
|
||||
"content": "<|start_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151667": {
|
||||
"content": "<|start_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151668": {
|
||||
"content": "<|continue_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151669": {
|
||||
"content": "<|predict|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
}
|
||||
},
|
||||
"additional_special_tokens": [
|
||||
"<|continue_listening|>",
|
||||
"<|start_speaking|>",
|
||||
"<|start_listening|>",
|
||||
"<|continue_speaking|>",
|
||||
"<|predict|>"
|
||||
],
|
||||
"bos_token": null,
|
||||
"clean_up_tokenization_spaces": false,
|
||||
"eos_token": "<|im_end|>",
|
||||
"errors": "replace",
|
||||
"extra_special_tokens": {},
|
||||
"model_max_length": 131072,
|
||||
"pad_token": "<|endoftext|>",
|
||||
"split_special_tokens": false,
|
||||
"tokenizer_class": "Qwen2Tokenizer",
|
||||
"unk_token": null
|
||||
}
|
||||
1
onnx/vocab.json
Normal file
1
onnx/vocab.json
Normal file
File diff suppressed because one or more lines are too long
53
special_tokens_map.json
Normal file
53
special_tokens_map.json
Normal file
@@ -0,0 +1,53 @@
|
||||
{
|
||||
"additional_special_tokens": [
|
||||
{
|
||||
"content": "<|continue_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|start_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|start_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|continue_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
{
|
||||
"content": "<|predict|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
}
|
||||
],
|
||||
"eos_token": {
|
||||
"content": "<|im_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
},
|
||||
"pad_token": {
|
||||
"content": "<|endoftext|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false
|
||||
}
|
||||
}
|
||||
76
test_metrics.json
Normal file
76
test_metrics.json
Normal file
@@ -0,0 +1,76 @@
|
||||
{
|
||||
"model": "semantic-turn-taking/production-v1/ckpt-7000",
|
||||
"base_model": "Qwen/Qwen2.5-0.5B-Instruct",
|
||||
"checkpoint_step": 7000,
|
||||
"benchmarks": {
|
||||
"ten": {
|
||||
"dataset": "TEN Turn Detection",
|
||||
"examples": 528,
|
||||
"binary_accuracy": 0.869,
|
||||
"binary_f1_macro": 0.868,
|
||||
"finished_recall": 0.88,
|
||||
"unfinished_recall": 0.96,
|
||||
"wait_recall": 0.66
|
||||
},
|
||||
"swda_v2": {
|
||||
"dataset": "SwDA (improved heuristic v2)",
|
||||
"examples": 4087,
|
||||
"four_class_accuracy": 0.6337,
|
||||
"four_class_f1_macro": 0.4364,
|
||||
"binary_accuracy": 0.7149,
|
||||
"binary_f1_macro": 0.7122,
|
||||
"per_class_accuracy": {
|
||||
"start_speaking": 0.618,
|
||||
"continue_listening": 0.479,
|
||||
"start_listening": 0.0,
|
||||
"continue_speaking": 0.847
|
||||
},
|
||||
"note": "v2 fixes false start_listening from agent backchannels"
|
||||
},
|
||||
"internal_test_en": {
|
||||
"dataset": "Hand-crafted Internal Test (English)",
|
||||
"examples": 48,
|
||||
"four_class_accuracy": 0.7708,
|
||||
"four_class_f1_macro": 0.7421,
|
||||
"binary_accuracy": 0.9583,
|
||||
"binary_f1_macro": 0.9583,
|
||||
"per_class_accuracy": {
|
||||
"start_speaking": 0.833,
|
||||
"continue_listening": 1.0,
|
||||
"start_listening": 0.25,
|
||||
"continue_speaking": 1.0
|
||||
}
|
||||
},
|
||||
"internal_test_es": {
|
||||
"dataset": "Hand-crafted Internal Test (Spanish)",
|
||||
"examples": 48,
|
||||
"four_class_accuracy": 0.6667,
|
||||
"four_class_f1_macro": 0.631,
|
||||
"binary_accuracy": 0.8542,
|
||||
"binary_f1_macro": 0.8536,
|
||||
"per_class_accuracy": {
|
||||
"start_speaking": 0.667,
|
||||
"continue_listening": 1.0,
|
||||
"start_listening": 0.167,
|
||||
"continue_speaking": 0.833
|
||||
},
|
||||
"note": "Model was never trained on Spanish — cross-lingual transfer from Qwen2.5"
|
||||
},
|
||||
"synthetic_eval": {
|
||||
"dataset": "Synthetic validation set",
|
||||
"eval_accuracy": 0.900,
|
||||
"eval_f1_macro": 0.767
|
||||
}
|
||||
},
|
||||
"livekit_comparison": {
|
||||
"model": "livekit/turn-detector@v0.4.1-intl",
|
||||
"ten_binary_accuracy": 0.667,
|
||||
"ten_binary_f1_macro": 0.592,
|
||||
"swda_v2_binary_accuracy": 0.367,
|
||||
"swda_v2_binary_f1_macro": 0.289,
|
||||
"internal_en_binary_accuracy": 0.708,
|
||||
"internal_en_binary_f1_macro": 0.681,
|
||||
"internal_es_binary_accuracy": 0.625,
|
||||
"internal_es_binary_f1_macro": 0.578
|
||||
}
|
||||
}
|
||||
3
tokenizer.json
Normal file
3
tokenizer.json
Normal file
@@ -0,0 +1,3 @@
|
||||
version https://git-lfs.github.com/spec/v1
|
||||
oid sha256:cb8544e622fff1dd549fddc8273bd5fd327f9642604868434f594f19ac2a49cc
|
||||
size 11422970
|
||||
239
tokenizer_config.json
Normal file
239
tokenizer_config.json
Normal file
@@ -0,0 +1,239 @@
|
||||
{
|
||||
"add_bos_token": false,
|
||||
"add_prefix_space": false,
|
||||
"added_tokens_decoder": {
|
||||
"151643": {
|
||||
"content": "<|endoftext|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151644": {
|
||||
"content": "<|im_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151645": {
|
||||
"content": "<|im_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151646": {
|
||||
"content": "<|object_ref_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151647": {
|
||||
"content": "<|object_ref_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151648": {
|
||||
"content": "<|box_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151649": {
|
||||
"content": "<|box_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151650": {
|
||||
"content": "<|quad_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151651": {
|
||||
"content": "<|quad_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151652": {
|
||||
"content": "<|vision_start|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151653": {
|
||||
"content": "<|vision_end|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151654": {
|
||||
"content": "<|vision_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151655": {
|
||||
"content": "<|image_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151656": {
|
||||
"content": "<|video_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151657": {
|
||||
"content": "<tool_call>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151658": {
|
||||
"content": "</tool_call>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151659": {
|
||||
"content": "<|fim_prefix|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151660": {
|
||||
"content": "<|fim_middle|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151661": {
|
||||
"content": "<|fim_suffix|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151662": {
|
||||
"content": "<|fim_pad|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151663": {
|
||||
"content": "<|repo_name|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151664": {
|
||||
"content": "<|file_sep|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": false
|
||||
},
|
||||
"151665": {
|
||||
"content": "<|continue_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151666": {
|
||||
"content": "<|start_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151667": {
|
||||
"content": "<|start_listening|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151668": {
|
||||
"content": "<|continue_speaking|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
},
|
||||
"151669": {
|
||||
"content": "<|predict|>",
|
||||
"lstrip": false,
|
||||
"normalized": false,
|
||||
"rstrip": false,
|
||||
"single_word": false,
|
||||
"special": true
|
||||
}
|
||||
},
|
||||
"additional_special_tokens": [
|
||||
"<|continue_listening|>",
|
||||
"<|start_speaking|>",
|
||||
"<|start_listening|>",
|
||||
"<|continue_speaking|>",
|
||||
"<|predict|>"
|
||||
],
|
||||
"bos_token": null,
|
||||
"clean_up_tokenization_spaces": false,
|
||||
"eos_token": "<|im_end|>",
|
||||
"errors": "replace",
|
||||
"extra_special_tokens": {},
|
||||
"model_max_length": 131072,
|
||||
"pad_token": "<|endoftext|>",
|
||||
"split_special_tokens": false,
|
||||
"tokenizer_class": "Qwen2Tokenizer",
|
||||
"unk_token": null
|
||||
}
|
||||
78
training_config.yaml
Normal file
78
training_config.yaml
Normal file
@@ -0,0 +1,78 @@
|
||||
# Training configuration for checkpoint-7000 (synth-only-v2-fixed-run2-removedlabelaug)
|
||||
# This is the config used to train the published model.
|
||||
|
||||
model:
|
||||
name: "Qwen/Qwen2.5-0.5B-Instruct"
|
||||
max_length: 1024
|
||||
predict_token: "<|predict|>"
|
||||
|
||||
action_tokens:
|
||||
- "<|continue_listening|>"
|
||||
- "<|start_speaking|>"
|
||||
- "<|start_listening|>"
|
||||
- "<|continue_speaking|>"
|
||||
|
||||
prediction_tokens:
|
||||
- "<|continue_listening|>"
|
||||
- "<|start_speaking|>"
|
||||
- "<|start_listening|>"
|
||||
- "<|continue_speaking|>"
|
||||
|
||||
data:
|
||||
train:
|
||||
- "data/context_action/synthetic/train.csv" # ~140K
|
||||
- "data/context_action/synthetic/supplementary_start_listening.csv" # ~14K
|
||||
val: "data/context_action/synthetic/val.csv"
|
||||
test: "data/context_action/synthetic/test.csv"
|
||||
|
||||
training:
|
||||
batch_size: 8
|
||||
gradient_accumulation_steps: 4 # effective batch size = 32
|
||||
learning_rate: 5.0e-5
|
||||
num_epochs: 50
|
||||
warmup_steps: 100
|
||||
weight_decay: 0.01
|
||||
max_grad_norm: 1.0
|
||||
bf16: true
|
||||
lr_scheduler: "cosine"
|
||||
|
||||
loss:
|
||||
ntp_weight: 0.1 # 90% action CE + 10% NTP auxiliary
|
||||
|
||||
early_stopping:
|
||||
enabled: true
|
||||
patience: 10
|
||||
metric: "eval_f1_macro"
|
||||
|
||||
augmentation:
|
||||
enabled: true
|
||||
context_truncation:
|
||||
enabled: true
|
||||
probability: 0.2
|
||||
min_turns: 1
|
||||
asr_styles:
|
||||
- "pure"
|
||||
- "punctuated"
|
||||
- "mixed"
|
||||
filler_injection:
|
||||
enabled: true
|
||||
probability: 0.2
|
||||
max_fillers: 3
|
||||
disfluency:
|
||||
enabled: true
|
||||
probability: 0.2
|
||||
# Label-changing augmentations DISABLED (broke balanced sampling)
|
||||
streaming_crop:
|
||||
enabled: false
|
||||
backchannel_inject:
|
||||
enabled: false
|
||||
|
||||
output:
|
||||
dir: "models_longer_training"
|
||||
save_steps: 1000
|
||||
eval_steps: 1000
|
||||
logging_steps: 100
|
||||
save_total_limit: 5
|
||||
|
||||
evaluation:
|
||||
batch_size: 8
|
||||
1
vocab.json
Normal file
1
vocab.json
Normal file
File diff suppressed because one or more lines are too long
Reference in New Issue
Block a user