Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("KSP-NMAI/Boris-250M")
model = AutoModelForCausalLM.from_pretrained("KSP-NMAI/Boris-250M")
ids = tok("The ocean is", return_tensors="pt").input_ids
out = model.generate(ids, max_new_tokens=40, do_sample=True, top_p=0.95)
print(tok.decode(out[0], skip_special_tokens=True))
Details
Table | |
|---|
| Architecture | GPT-2 (pre-LN, learned positional embeddings, tied embeddings) |
| Layers / heads / d_model | 12 / 18 / 1152 |
| Context length | 1024 |
| Vocab | 50304 (GPT-NeoX-20B BPE, padded) |
| Tokenizer | EleutherAI/gpt-neox-20b |
| Precision | trained in bf16 autocast with fp32 master weights |
Base model training
Trained on 5.01B tokens for ~105:56:17 on one RTX 3060.
Table | |
|---|
| Final loss | 3.0693 |
| Final grad norm | 0.225 |
| Final learning rate | 6.00e-05 |

Limitations
A base model of this size will produce text that is frequently inaccurate,
inconsistent, or offensive. It has received no alignment or safety tuning and
should not be used for factual reference or deployed without supervision.
Copyright & License
Copyright 2026 Joseph Jones
This project and all associated files (the "Work") are licensed under the Apache
License, Version 2.0 (the "License"); you may not use this project except in
compliance with the License. You may obtain a copy of the License at:
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed
under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.