Text Classification
PEFT
Safetensors
Transformers
feedback-detection
user-satisfaction
lora
modernbert
mmbert
32k-context
Eval Results (legacy)
Instructions to use vllm-sr/mmbert32k-feedback-detector-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use vllm-sr/mmbert32k-feedback-detector-lora with PEFT:
from peft import PeftModel from transformers import AutoModelForSequenceClassification base_model = AutoModelForSequenceClassification.from_pretrained("vllm-sr/mmbert-32k-yarn") model = PeftModel.from_pretrained(base_model, "vllm-sr/mmbert32k-feedback-detector-lora") - Transformers
How to use vllm-sr/mmbert32k-feedback-detector-lora with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="vllm-sr/mmbert32k-feedback-detector-lora")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("vllm-sr/mmbert32k-feedback-detector-lora", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update Hugging Face org references: llm-semantic-router → vllm-sr
Browse files- README.md +12 -12
- adapter_config.json +1 -1
- checkpoint-4500/README.md +2 -2
- checkpoint-4500/adapter_config.json +1 -1
- checkpoint-6000/README.md +2 -2
- checkpoint-6000/adapter_config.json +1 -1
- training_config.json +1 -1
README.md
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
---
|
| 2 |
-
base_model:
|
| 3 |
library_name: peft
|
| 4 |
license: apache-2.0
|
| 5 |
language:
|
|
@@ -18,7 +18,7 @@ tags:
|
|
| 18 |
- mmbert
|
| 19 |
- 32k-context
|
| 20 |
datasets:
|
| 21 |
-
-
|
| 22 |
metrics:
|
| 23 |
- accuracy
|
| 24 |
- f1
|
|
@@ -31,7 +31,7 @@ model-index:
|
|
| 31 |
name: User Feedback Classification
|
| 32 |
dataset:
|
| 33 |
name: feedback-detector-dataset
|
| 34 |
-
type:
|
| 35 |
metrics:
|
| 36 |
- name: Accuracy
|
| 37 |
type: accuracy
|
|
@@ -43,7 +43,7 @@ model-index:
|
|
| 43 |
|
| 44 |
# mmBERT-32K Feedback Detector (LoRA)
|
| 45 |
|
| 46 |
-
A 4-class user feedback classifier fine-tuned from [mmbert-32k-yarn](https://huggingface.co/
|
| 47 |
|
| 48 |
## Model Description
|
| 49 |
|
|
@@ -85,13 +85,13 @@ from peft import PeftModel
|
|
| 85 |
|
| 86 |
# Load base model
|
| 87 |
base_model = AutoModelForSequenceClassification.from_pretrained(
|
| 88 |
-
"
|
| 89 |
num_labels=4
|
| 90 |
)
|
| 91 |
-
tokenizer = AutoTokenizer.from_pretrained("
|
| 92 |
|
| 93 |
# Load LoRA adapter
|
| 94 |
-
model = PeftModel.from_pretrained(base_model, "
|
| 95 |
model.eval()
|
| 96 |
|
| 97 |
# Inference
|
|
@@ -112,10 +112,10 @@ For easier deployment, use the merged version:
|
|
| 112 |
from transformers import AutoModelForSequenceClassification, AutoTokenizer
|
| 113 |
|
| 114 |
model = AutoModelForSequenceClassification.from_pretrained(
|
| 115 |
-
"
|
| 116 |
)
|
| 117 |
tokenizer = AutoTokenizer.from_pretrained(
|
| 118 |
-
"
|
| 119 |
)
|
| 120 |
```
|
| 121 |
|
|
@@ -125,7 +125,7 @@ tokenizer = AutoTokenizer.from_pretrained(
|
|
| 125 |
|
| 126 |
| Parameter | Value |
|
| 127 |
|-----------|-------|
|
| 128 |
-
| Base Model |
|
| 129 |
| LoRA Rank | 64 |
|
| 130 |
| LoRA Alpha | 128 |
|
| 131 |
| LoRA Dropout | 0.1 |
|
|
@@ -140,7 +140,7 @@ tokenizer = AutoTokenizer.from_pretrained(
|
|
| 140 |
|
| 141 |
### Training Data
|
| 142 |
|
| 143 |
-
Trained on [
|
| 144 |
- **Training samples**: 17,896 (balanced across 4 classes)
|
| 145 |
- **Validation samples**: 2,985
|
| 146 |
|
|
@@ -172,7 +172,7 @@ The model inherits multilingual capabilities from mmbert-32k-yarn (Glot500 token
|
|
| 172 |
author={LLM Semantic Router Team},
|
| 173 |
year={2026},
|
| 174 |
publisher={Hugging Face},
|
| 175 |
-
url={https://huggingface.co/
|
| 176 |
}
|
| 177 |
```
|
| 178 |
|
|
|
|
| 1 |
---
|
| 2 |
+
base_model: vllm-sr/mmbert-32k-yarn
|
| 3 |
library_name: peft
|
| 4 |
license: apache-2.0
|
| 5 |
language:
|
|
|
|
| 18 |
- mmbert
|
| 19 |
- 32k-context
|
| 20 |
datasets:
|
| 21 |
+
- vllm-sr/feedback-detector-dataset
|
| 22 |
metrics:
|
| 23 |
- accuracy
|
| 24 |
- f1
|
|
|
|
| 31 |
name: User Feedback Classification
|
| 32 |
dataset:
|
| 33 |
name: feedback-detector-dataset
|
| 34 |
+
type: vllm-sr/feedback-detector-dataset
|
| 35 |
metrics:
|
| 36 |
- name: Accuracy
|
| 37 |
type: accuracy
|
|
|
|
| 43 |
|
| 44 |
# mmBERT-32K Feedback Detector (LoRA)
|
| 45 |
|
| 46 |
+
A 4-class user feedback classifier fine-tuned from [mmbert-32k-yarn](https://huggingface.co/vllm-sr/mmbert-32k-yarn) using LoRA (Low-Rank Adaptation).
|
| 47 |
|
| 48 |
## Model Description
|
| 49 |
|
|
|
|
| 85 |
|
| 86 |
# Load base model
|
| 87 |
base_model = AutoModelForSequenceClassification.from_pretrained(
|
| 88 |
+
"vllm-sr/mmbert-32k-yarn",
|
| 89 |
num_labels=4
|
| 90 |
)
|
| 91 |
+
tokenizer = AutoTokenizer.from_pretrained("vllm-sr/mmbert-32k-yarn")
|
| 92 |
|
| 93 |
# Load LoRA adapter
|
| 94 |
+
model = PeftModel.from_pretrained(base_model, "vllm-sr/mmbert32k-feedback-detector-lora")
|
| 95 |
model.eval()
|
| 96 |
|
| 97 |
# Inference
|
|
|
|
| 112 |
from transformers import AutoModelForSequenceClassification, AutoTokenizer
|
| 113 |
|
| 114 |
model = AutoModelForSequenceClassification.from_pretrained(
|
| 115 |
+
"vllm-sr/mmbert32k-feedback-detector-merged"
|
| 116 |
)
|
| 117 |
tokenizer = AutoTokenizer.from_pretrained(
|
| 118 |
+
"vllm-sr/mmbert32k-feedback-detector-merged"
|
| 119 |
)
|
| 120 |
```
|
| 121 |
|
|
|
|
| 125 |
|
| 126 |
| Parameter | Value |
|
| 127 |
|-----------|-------|
|
| 128 |
+
| Base Model | vllm-sr/mmbert-32k-yarn |
|
| 129 |
| LoRA Rank | 64 |
|
| 130 |
| LoRA Alpha | 128 |
|
| 131 |
| LoRA Dropout | 0.1 |
|
|
|
|
| 140 |
|
| 141 |
### Training Data
|
| 142 |
|
| 143 |
+
Trained on [vllm-sr/feedback-detector-dataset](https://huggingface.co/datasets/vllm-sr/feedback-detector-dataset):
|
| 144 |
- **Training samples**: 17,896 (balanced across 4 classes)
|
| 145 |
- **Validation samples**: 2,985
|
| 146 |
|
|
|
|
| 172 |
author={LLM Semantic Router Team},
|
| 173 |
year={2026},
|
| 174 |
publisher={Hugging Face},
|
| 175 |
+
url={https://huggingface.co/vllm-sr/mmbert32k-feedback-detector-lora}
|
| 176 |
}
|
| 177 |
```
|
| 178 |
|
adapter_config.json
CHANGED
|
@@ -3,7 +3,7 @@
|
|
| 3 |
"alpha_pattern": {},
|
| 4 |
"arrow_config": null,
|
| 5 |
"auto_mapping": null,
|
| 6 |
-
"base_model_name_or_path": "
|
| 7 |
"bias": "none",
|
| 8 |
"corda_config": null,
|
| 9 |
"ensure_weight_tying": false,
|
|
|
|
| 3 |
"alpha_pattern": {},
|
| 4 |
"arrow_config": null,
|
| 5 |
"auto_mapping": null,
|
| 6 |
+
"base_model_name_or_path": "vllm-sr/mmbert-32k-yarn",
|
| 7 |
"bias": "none",
|
| 8 |
"corda_config": null,
|
| 9 |
"ensure_weight_tying": false,
|
checkpoint-4500/README.md
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
---
|
| 2 |
-
base_model:
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
-
- base_model:adapter:
|
| 6 |
- lora
|
| 7 |
- transformers
|
| 8 |
---
|
|
|
|
| 1 |
---
|
| 2 |
+
base_model: vllm-sr/mmbert-32k-yarn
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
+
- base_model:adapter:vllm-sr/mmbert-32k-yarn
|
| 6 |
- lora
|
| 7 |
- transformers
|
| 8 |
---
|
checkpoint-4500/adapter_config.json
CHANGED
|
@@ -3,7 +3,7 @@
|
|
| 3 |
"alpha_pattern": {},
|
| 4 |
"arrow_config": null,
|
| 5 |
"auto_mapping": null,
|
| 6 |
-
"base_model_name_or_path": "
|
| 7 |
"bias": "none",
|
| 8 |
"corda_config": null,
|
| 9 |
"ensure_weight_tying": false,
|
|
|
|
| 3 |
"alpha_pattern": {},
|
| 4 |
"arrow_config": null,
|
| 5 |
"auto_mapping": null,
|
| 6 |
+
"base_model_name_or_path": "vllm-sr/mmbert-32k-yarn",
|
| 7 |
"bias": "none",
|
| 8 |
"corda_config": null,
|
| 9 |
"ensure_weight_tying": false,
|
checkpoint-6000/README.md
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
---
|
| 2 |
-
base_model:
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
-
- base_model:adapter:
|
| 6 |
- lora
|
| 7 |
- transformers
|
| 8 |
---
|
|
|
|
| 1 |
---
|
| 2 |
+
base_model: vllm-sr/mmbert-32k-yarn
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
+
- base_model:adapter:vllm-sr/mmbert-32k-yarn
|
| 6 |
- lora
|
| 7 |
- transformers
|
| 8 |
---
|
checkpoint-6000/adapter_config.json
CHANGED
|
@@ -3,7 +3,7 @@
|
|
| 3 |
"alpha_pattern": {},
|
| 4 |
"arrow_config": null,
|
| 5 |
"auto_mapping": null,
|
| 6 |
-
"base_model_name_or_path": "
|
| 7 |
"bias": "none",
|
| 8 |
"corda_config": null,
|
| 9 |
"ensure_weight_tying": false,
|
|
|
|
| 3 |
"alpha_pattern": {},
|
| 4 |
"arrow_config": null,
|
| 5 |
"auto_mapping": null,
|
| 6 |
+
"base_model_name_or_path": "vllm-sr/mmbert-32k-yarn",
|
| 7 |
"bias": "none",
|
| 8 |
"corda_config": null,
|
| 9 |
"ensure_weight_tying": false,
|
training_config.json
CHANGED
|
@@ -18,7 +18,7 @@
|
|
| 18 |
"2": "WRONG_ANSWER",
|
| 19 |
"3": "WANT_DIFFERENT"
|
| 20 |
},
|
| 21 |
-
"base_model": "
|
| 22 |
"max_length": 512,
|
| 23 |
"use_lora": true,
|
| 24 |
"lora_rank": 64,
|
|
|
|
| 18 |
"2": "WRONG_ANSWER",
|
| 19 |
"3": "WANT_DIFFERENT"
|
| 20 |
},
|
| 21 |
+
"base_model": "vllm-sr/mmbert-32k-yarn",
|
| 22 |
"max_length": 512,
|
| 23 |
"use_lora": true,
|
| 24 |
"lora_rank": 64,
|