README.md

OpenSeek-Small v1 Model Documentation

Overview

OpenSeek-Small-v1 is the initial production model of the OpenSeek project.

• Utilizes DeepSeek-V3-like MoE architecture.
• Comprises 1.4 billion total parameters, with 0.4 billion activated parameters.
• Trained on 720 billion tokens.
• Completely broken in stock form.

Key Fixes in this repository:

• Fixed Broken Position Embeddings
• Fixed Fundamental Incompatibilities Between Deepseekv3 Model and Qwen Tokenizer.

Usage Instructions

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Robertp423/OpenSeek-Fixed",trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("Robertp423/OpenSeek-Fixed",trust_remote_code=True)

inputs = tokenizer("The future of AI is", return_tensors="pt")
inputs.pop("token_type_ids", None)  # Critical fix
outputs = model.generate(**inputs, max_length=50)
print(tokenizer.decode(outputs[0]))