Hugging Face's logo

Collab-uniba
/
fprime-binary-setfit

Text Classification
setfit
Safetensors
sentence-transformers
mpnet
generated_from_setfit_trainer
Model card Files Community
fprime-binary-setfit / README.md
PeppoCola's picture
PeppoCola
Upload folder using huggingface_hub
c8c7e15 verified
preview code
|
raw
history blame contribute delete
52.6 kB
metadata
library_name: setfit
metrics:
  - accuracy
pipeline_tag: text-classification
tags:
  - setfit
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
widget:
  - text: "Throttle send frame EVR\n| | |\r\n|:---|:---|\r\n|**_F´ Version_**|commit d3fa31c |\r\n|**_Affected Component_**| ?  |\r\n---\r\n## Problem Description\r\n\r\nA description of the problem with sufficient detail to understand the issue.\r\n\r\nIf there is no ground system, the interface continuously sends this pair of EVRs:\r\n\r\n```\r\n0x201db690 (TV_TLM): [ERROR] Failed to send framed data: 0\r\n0x201db690 (TV_TLM): [ERROR] Failed to send framed data: 0\r\n0x201db690 (TV_TLM): [ERROR] Failed to send framed data: 0\r\n0x201db690 (TV_TLM): [ERROR] Failed to send framed data: 0\r\n0x201db690 (TV_TLM): [ERROR] Failed to send framed data: 0\r\n0x201db690 (TV_TLM): [ERROR] Failed to send framed data: 0\r\n0x202236f0 (TV_ReceiveTask): [WARNING] Failed to open port with status 61 and errno 0\r\n```\r\n\r\n## How to Reproduce\r\n\r\n1. Run Ref without the ground system\r\n2.\r\n3.\r\n\r\n## Expected Behavior\r\n\r\nIMHO the EVR should throttle, and perhaps the throttle is reset when the connection is make.\r\n"
  - text: "Color-coding interlaced Events in the API Log\n| | |\r\n|:---|:---|\r\n|**_F´ Version_**| |\r\n|**_Affected Component_**|  |\r\n---\r\nOne feature that wasn't completed this summer was to color-code interlaced event logs based on severity. Presently, interlacing events are implemented by making the API a consumer of the event decoder in the GDS and then filtering events. Modifying the color of these log messages can be done [here](https://github.com/nasa/fprime/blob/717bc6fab85c53680108fc961cad6338e779816f/Gds/src/fprime_gds/common/testing_fw/api.py#L1258).\r\n"
  - text: "Switch Framer and Deframer to use Mallocator Pattern\n| | |\r\n|:---|:---|\r\n|**_F´ Version_**| |\r\n|**_Affected Component_**|   |\r\n---\r\n## Problem Description\r\n\r\nMallocator pattern is preferred over member-allocated buffers."
  - text: "Ninja support for fprime-tools\n| | |\r\n|:---|:---|\r\n|**_F´ Version_**| |\r\n|**_Affected Component_**|   |\r\n---\r\n## Problem Description\r\n\r\nThere are a couple places in fprime-tools where things would break if Ninja was used instead of Make. We need to fix that, as Ninja is usually much faster.\r\ne.g. [this](https://github.com/fprime-community/fprime-tools/blob/0a9fdf58ce4b428d407ab264f7266041808237c8/src/fprime/fbuild/cmake.py#L133) is Make-specific output, Ninja formats it differently\r\n\r\n## Expected Behavior\r\n\r\nSupport Ninja with fprime-tools. Add a convenient option to chose which one to use.\r\n"
  - text: "Build A Frame Reassembler\n| | |\r\n|:---|:---|\r\n|**_F´ Version_**| |\r\n|**_Affected Component_**|  |\r\n---\r\n## Feature Description\r\n\r\nBuild a component that can be used to reassemble communication frames given protocol information. This will break-off this functionality from the Deframer.\r\n\r\nBasic requirements:\r\n1. Accept incoming Fw::Buffers of any size\r\n2. Accumulate buffers in-order\r\n3. Call frame detector helper class\r\n4. On \"NO_FRAME\" discard first byte and try again\r\n5. On \"NEED DATA\" continue to accumulate data\r\n6. On \"FRAME\" allocate buffer, copy-out frame\r\n\r\nHelper class requirements:\r\n1. Must implement `Enum detect_frame(const CircularBuffer& buffer, FwSizeType& size_output)` method\r\n2. Cannot alter circular buffer (uses peeks)\r\n3. Must set `size_output` when data is needed and when frame detected\r\n "
inference: true

SetFit

This is a SetFit model that can be used for Text Classification. A LogisticRegression instance is used for classification.

The model has been trained using an efficient few-shot learning technique that involves:

  1. Fine-tuning a Sentence Transformer with contrastive learning.
  2. Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

  • Model Type: SetFit
  • Classification head: a LogisticRegression instance
  • Maximum Sequence Length: 384 tokens
  • Number of Classes: 2 classes

Model Sources

Model Labels

Label Examples
bug
  • "A global-buffer-overflow error in BufferQueueTest.cpp line 126\n
non-bug
  • "Install Upgraded PIP in new Project\n

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("setfit_model_id")
# Run inference
preds = model("Switch Framer and Deframer to use Mallocator Pattern
| | |
|:---|:---|
|**_F´ Version_**| |
|**_Affected Component_**|   |
---
## Problem Description

Mallocator pattern is preferred over member-allocated buffers.")

Training Details

Training Set Metrics

Training set Min Median Max
Word count 4 124.1383 2486
Label Training Sample Count
bug 296
non-bug 304

Training Hyperparameters

  • batch_size: (16, 2)
  • num_epochs: (1, 1)
  • max_steps: -1
  • sampling_strategy: oversampling
  • num_iterations: 20
  • body_learning_rate: (2e-05, 1e-05)
  • head_learning_rate: 0.01
  • loss: CosineSimilarityLoss
  • distance_metric: cosine_distance
  • margin: 0.25
  • end_to_end: False
  • use_amp: False
  • warmup_proportion: 0.1
  • l2_weight: 0.01
  • seed: 42
  • eval_max_steps: -1
  • load_best_model_at_end: False

Training Results

Epoch Step Training Loss Validation Loss
0.0007 1 0.447 -
0.0333 50 0.2333 -
0.0667 100 0.083 -
0.1 150 0.039 -
0.1333 200 0.0354 -
0.1667 250 0.0177 -
0.2 300 0.0053 -
0.2333 350 0.0004 -
0.2667 400 0.0027 -
0.3 450 0.0015 -
0.3333 500 0.002 -
0.3667 550 0.0003 -
0.4 600 0.0001 -
0.4333 650 0.0001 -
0.4667 700 0.0001 -
0.5 750 0.0001 -
0.5333 800 0.0001 -
0.5667 850 0.0001 -
0.6 900 0.0001 -
0.6333 950 0.0001 -
0.6667 1000 0.0001 -
0.7 1050 0.0 -
0.7333 1100 0.0 -
0.7667 1150 0.0001 -
0.8 1200 0.0 -
0.8333 1250 0.0001 -
0.8667 1300 0.0 -
0.9 1350 0.0 -
0.9333 1400 0.0001 -
0.9667 1450 0.0 -
1.0 1500 0.0 -

Framework Versions

  • Python: 3.11.6
  • SetFit: 1.1.0
  • Sentence Transformers: 3.0.1
  • Transformers: 4.44.2
  • PyTorch: 2.4.1+cu121
  • Datasets: 2.21.0
  • Tokenizers: 0.19.1

Citation

BibTeX 

@article{https://doi.org/10.48550/arxiv.2209.11055,
    doi = {10.48550/ARXIV.2209.11055},
    url = {https://arxiv.org/abs/2209.11055},
    author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
    keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Efficient Few-Shot Learning Without Prompts},
    publisher = {arXiv},
    year = {2022},
    copyright = {Creative Commons Attribution 4.0 International}
}