josephmayo/HRM-Text-1B-sft-code-LoRA

HRM-Text-1B-sft-code-LoRA

LoRA adapter for sapientinc/HRM-Text-1B.

sapientinc/HRM-Text-1B is a pretrained-only HRM text model. This adapter is the code post-training release built on top of it.

The release uses supervised LoRA post-training for coding tasks. It is the adapter artifact; the merged model is:

`josephmayo/HRM-Text-1B-sft-code`

Training

• —Base model: sapientinc/HRM-Text-1B
• —Method: supervised LoRA post-training
• —Training rows: 384
• —Max steps: 120
• —LoRA rank: 64
• —Learning rate: 8e-6
• —Final train loss: 0.3275703112284342

Validation

Local code validation:

• —Base model score: 5/100
• —Adapter score: 24/100
• —Absolute improvement: +19/100
• —Relative improvement: 4.8x over base
• —HumanEval slice: 14/50
• —MBPP slice: 10/50

The score above is the local validation result used for this release.

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_id = "sapientinc/HRM-Text-1B"
adapter_id = "josephmayo/HRM-Text-1B-sft-code-LoRA"

tokenizer = AutoTokenizer.from_pretrained(adapter_id)
model = AutoModelForCausalLM.from_pretrained(base_id, trust_remote_code=True)
model = PeftModel.from_pretrained(model, adapter_id)
model.eval()

Notes

• —This is an adapter, not a standalone merged model.
• —This is the LoRA adapter. Use the merged model for standalone loading.