Downloads · 30 days
88
14% of all-time downloads
opencerebral/Boris-75M-Instruct
Boris-75M-Instruct is a text generation model from opencerebral. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
Downloads · 30 days
88
14% of all-time downloads
All-time downloads
634
Public
Parameters
77.4M
311 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors310 MB · 99%
From the Hugging Face model README

Boris-75M-Instruct is the instruction-tuned variant of KSP-NMAI/Boris-75M, a 75 million-parameter language model created by New Millennium Artificial Intelligence (NMAI). It was fine-tuned on tatsu-lab/alpaca.
This model uses the Alpaca format. A chat template is included in
tokenizer_config.json, so apply_chat_template produces the correct prompt
automatically:
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("KSP-NMAI/Boris-75M-Instruct")
model = AutoModelForCausalLM.from_pretrained("KSP-NMAI/Boris-75M-Instruct")
messages = [{"role": "user", "content": "What is the capital of France?"}]
prompt = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
ids = tok(prompt, return_tensors="pt").input_ids
out = model.generate(ids, max_new_tokens=64, do_sample=True, top_p=0.95, temperature=0.7)
print(tok.decode(out[0][ids.shape[1]:], skip_special_tokens=True))
If you are building the prompt by hand, the layout is:
### Instruction:
{your instruction}
### Response:
Generation should stop at ### Instruction: (or end-of-text, token id 0).
| Architecture | GPT-2 (pre-LN, learned positional embeddings, tied embeddings) |
| Layers / heads / d_model | 12 / 9 / 576 |
| Context length | 1024 |
| Vocab | 50304 (GPT-NeoX-20B BPE, padded) |
| Tokenizer | EleutherAI/gpt-neox-20b |
| Precision | trained in bf16 autocast with fp32 master weights |
Trained on 1.55B tokens for 14:49:08 on one RTX 3060.
| Final loss | 3.6356 |
| Final grad norm | 0.328 |
| Final learning rate | 6.00e-05 |

The table above describes the base model's pretraining run; the instruction tuning was applied on top of that checkpoint.
This is a very small instruction-tuned model. It will produce text that is frequently inaccurate, inconsistent, or offensive, and it has received no alignment, RLHF, or safety tuning beyond supervised fine-tuning on Alpaca. Do not rely on it for factual information or deploy it without supervision.
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.