ahmedandaloes/CAMeLBERT-da-sentiment-CoreAI

CAMeLBERT-da-sentiment — Apple Core AI, Arabic dialect sentiment

A Core AI (.aimodel) build of CAMeL-Lab/bert-base-arabic-camelbert-da-sentiment for on-device dialectal Arabic sentiment on Apple Silicon.

Upstream ships pytorch_model.bin, flax_model.msgpack and tf_model.h5 — and no MLX, no ONNX, no Core AI. 81,297 downloads, and this is the Core AI build.

This is the dialect sibling of CAMeLBERT-mix-sentiment-CoreAI. Same three labels, trained on dialectal Arabic — and on the sentence below it disagrees with the MSA model, correctly.

Outputs

One graph, two outputs. Keeping both costs nothing at this size and means neither the trained classifier nor the reusable vectors are thrown away.

Output	Shape	What it is
logits	float32[1, 3]	the trained head, in id2label order: 0 = positive, 1 = negative, 2 = neutral
embeddings	float32[1, 768]	mean-pooled over the mask, L2-normalised — our addition, not an upstream objective

Read this before reading the outputs

Select outputs by shape, not by name. The runtime names tensors after the op that produced them. On this graph that is div_1 for the vectors and addmm_73 for the logits — the names in the wrapper's dict do not survive into the .aimodel. A one-line size == 3 / size == 768 pick is the robust way, and it is what the measurement script does.

Verified equivalence

Compared against the original PyTorch model on six fixed inputs at the exported shape:

Input	Predicted	logits max abs diff	embeddings max abs diff	cosine
الخدمة كانت زي الفل، أنصح بيهم جدا.	positive	7.15e-07	1.15e-07	1.00000000
تجربة زي الزفت، مش هكرر تاني.	negative	1.25e-06	2.31e-07	1.00000000
المنتج ووصل أمس.	neutral	4.77e-07	1.42e-07	1.00000000
الدعم رد عليا بسرعة وحلوا المشكلة.	positive	4.53e-06	3.65e-07	1.00000000
الطلب اتأخر أسبوعين وما حد رد عليا.	negative	3.46e-06	1.42e-07	1.00000000
كيف أستعيد كلمة المرور؟	neutral	2.32e-06	3.43e-07	1.00000000

Worst across all six: 4.53e-06, and 6/6 labels agree with PyTorch. That residual is relative ~1e-6 on logits of magnitude ~3 — float32 rounding accumulated through the classifier layer, not a difference in the model.

One dialect example, and what it does not show

الطلب اتأخر أسبوعين وما حد رد عليا. — an Egyptian-Arabic complaint — is classified negative here. The MSA model calls the same sentence neutral, which is the wrong answer for a complaint.

This is one sentence. It is the reason both bundles exist, but it is not a benchmark: no dialect sentiment corpus was run, and no accuracy figure for either model is claimed here.

Verified behaviour on six inputs, all matching PyTorch:

Input	Predicted
الخدمة كانت زي الفل، أنصح بيهم جدا.	positive
تجربة زي الزفت، مش هكرر تاني.	negative
المنتج ووصل أمس.	neutral
الدعم رد عليا بسرعة وحلوا المشكلة.	positive
الطلب اتأخر أسبوعين وما حد رد عليا.	negative (the MSA model says neutral)
كيف أستعيد كلمة المرور؟	neutral

Use

import asyncio, numpy as np
from coreai.runtime import AIModel, NDArray, SpecializationOptions
from transformers import AutoTokenizer

LABELS = ["positive", "negative", "neutral"]
SEQUENCE_LENGTH = 512          # fixed shape: pad to it

async def analyse(text: str):
    tok = AutoTokenizer.from_pretrained(".")
    model = await AIModel.load("camelbert-arabic-sentiment.aimodel", SpecializationOptions.default())
    fn = model.load_function("main")
    enc = tok([text], padding="max_length", truncation=True, max_length=SEQUENCE_LENGTH,
              return_tensors="np")
    out = await fn({"input_ids": NDArray(enc["input_ids"].astype("int32")),
                    "attention_mask": NDArray(enc["attention_mask"].astype("int32"))})
    flat = {k: np.asarray(v.numpy()).reshape(-1) for k, v in out.items()}
    logits = next(v for v in flat.values() if v.size == 3)
    vectors = next(v for v in flat.values() if v.size == 768)
    return LABELS[int(np.argmax(logits))], logits, vectors

print(asyncio.run(analyse("الخدمة كانت ممتازة وأنصح به بشدة."))[0])   # positive

What is in here

File
camelbert-arabic-sentiment.aimodel/	the graph
manifest.json	source, revision, label order, output shapes, precision
validation.json	PyTorch-to-Core AI equivalence for each case above (not the download count)
camelbert_sent_coreai.py	the exporter and measurer — --check, --export, --measure
vocab.txt, config.json, tokenizer files	so the tokenizer loads from this repository

Limits

• Fixed at 512 tokens; pad rather than passing a shorter sequence.
• Six inputs measured. Dialect variation, long documents and the 512-token boundary were not.
• No speed or memory benchmark. Equivalence only.
• The classification is 3-way and single-label; no confidence calibration was measured.

Licence and attribution

Apache-2.0, as declared by upstream. A derivative work of CAMeL-Lab/bert-base-arabic-camelbert-da-sentiment; credit for the model is theirs.