GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
yujuyeon
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
yujuyeon
Model Tree
OpenGVLab/InternVL3_5-1B-Pretrained
Input Modalities
Output Modalities
Supported Functionality
15000 / max 6945120.0001| step | loss |
|---|---|
| 10 | 1.7870 |
| 510 | 1.6479 |
| 1010 | 1.6007 |
| 1510 | 1.5718 |
| 2010 | 1.5762 |
| 2510 | 1.5841 |
| 3010 | 1.5639 |
| 3510 | 1.5288 |
| 4010 | 1.5593 |
| 4510 | 1.5472 |
| 5000 | 1.5434 |
구성: 7개 서브셋 (한국어 specialist SFT)
| subset | repeat |
|---|---|
med_extracted_qa | 1.0 |
med_pmc_vqa | 1.0 |
med_mathvk_rad | 1.0 |
med_train_25k | 1.0 |
med_medqa_text | 1.0 |
med_pmv_instruction | 1.0 |
python
from transformers import AutoModel, AutoTokenizerimport torchm = AutoModel.from_pretrained("yujuyeon/internvl3_5-1b-medical-bio-1m", torch_dtype=torch.bfloat16,trust_remote_code=True).eval().cuda()tok = AutoTokenizer.from_pretrained("yujuyeon/internvl3_5-1b-medical-bio-1m", trust_remote_code=True, use_fast=False)
med_pmv_alignment | 1.0 |