Octen-Embedding-8B W4A16

Overview

octen-embedding-8b-w4a16 is a weight-quantized checkpoint intended for efficient GPU inference, published by groxaxo. It is intended for open-source evaluation, reproducible experimentation, and compatible local or hosted inference workflows. The wording below is deliberately limited to what can be verified from this repository's metadata and artifacts.

At a glance

Field	Details
Format	AWQ / AutoRound
Source / base	`Octen/Octen-Embedding-8B`
Intended task	`feature-extraction`
License	`apache-2.0`

What is included

*.safetensors (2 files)
config.json
generation_config.json
tokenizer.json
tokenizer_config.json
chat_template.jinja
quantization_config.json
Additional configuration, tokenizer, processor, or shard files (10 visible artifacts total)

Quick start

vLLM (AWQ-compatible runtimes)

vllm serve groxaxo/octen-embedding-8b-w4a16 \
  --quantization awq_marlin \
  --dtype float16 \
  --trust-remote-code

The exact kernel and flags depend on the quantizer and architecture. Check the files and source model card before selecting a production serving configuration.

Compatibility and responsible use

Use a runtime that explicitly supports this format, architecture, and modality.
Keep configuration, tokenizer, processor, projection, and weight files from the same revision together.
Review the source model card and license before redistribution or deployment.
Hardware needs depend on parameter count, context length, cache precision, quantization, and concurrency.
Report reproducible issues with the runtime version, hardware, launch command, and a minimal example.

Quantization or conversion changes numerical behavior, memory use, and throughput relative to the source checkpoint; validate quality on your own workload.

Generated outputs may be inaccurate or unsuitable for a given use case. Users are responsible for testing behavior, applying appropriate safeguards, and complying with applicable licenses and laws.

This repo contains a W4A16 quantized version of Octen/Octen-Embedding-8B in the validated auto-round-auto-gptq format.

Quantization

Item	Value
Base model	`Octen/Octen-Embedding-8B`
Quantization	W4A16, 4-bit weights / 16-bit activations
Tooling	AutoRound 0.12.2, transformers 5.6.2, torch 2.6.0+cu124
Calibration	8 samples, seqlen 512, 200 iterations, float32 tuning
Quantized size	8.1 GB, 2 shards
Base size	15.0 GB
Compression	~1.9x
Embedding dim	4096
Layers quantized	252/253; `lm_head` skipped

Validation vs base model

Evaluation used a small retrieval set of 5 query-document pairs, last-token pooling, L2 normalization, and cosine similarity.

Metric	Base	W4A16	Delta
Recall@1	0.8	1.0	+0.2
Recall@5	1.0	1.0	0.0
Mean query cosine, base vs quant	—	0.9840	—
Mean doc cosine, base vs quant	—	0.9820	—

Assessment: this model passed all validation gates cleanly, with >0.98 mean cosine to the base model and no retrieval degradation on the validation set.

See validation-8b-auto-round-auto-gptq.json for the raw metrics.

RTX 3060 smoke test

This quantized model was loaded and run on an RTX 3060 12GB GPU.

Result	Value
VRAM after load	4.53 GB
Single short-query forward pass	0.9s smoke test; later benchmark ~612ms
Output shape	`[1, 4, 4096]`
Embeddings	Valid normalized vectors; no NaNs observed

Recommended usage

import torch
from transformers import AutoModel, AutoTokenizer

model_id = "groxaxo/octen-embedding-8b-w4a16"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModel.from_pretrained(
    model_id,
    trust_remote_code=True,
    torch_dtype=torch.float16,
).cuda().eval()

texts = ["how to implement binary search"]
tokens = tokenizer(texts, padding=True, truncation=True, max_length=512, return_tensors="pt")
tokens = {k: v.cuda() for k, v in tokens.items()}

with torch.no_grad():
    out = model(**tokens)

emb = torch.nn.functional.normalize(out.last_hidden_state[:, -1, :], p=2, dim=-1)

Note: the model card records local validation and smoke-test results. For production use, evaluate on your own retrieval distribution.