GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
mrothroc
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
mrothroc
Model Tree
Input Modalities
Output Modalities
Supported Functionality
GPT2LMHeadModel.Scored on 30,000 unconditional samples with the official moses.metrics, next to the SMILES model:
| Metric | SMILES model | this SELFIES model |
|---|---|---|
| Valid | 0.988 | 1.000 |
| Novelty | 0.783 | 0.859 |
| FCD/Test | 2.83 | 0.43 |
| Filters | 0.996 | 0.976 |
Validity is 1.000 by construction, Novelty and distribution fit both improve, and the MOSES medicinal-chemistry Filters rate slips a little (0.996 to 0.976). SELFIES is not ours (Krenn et al. 2020); the interesting part is that a representation choice, at no extra compute, delivered guaranteed validity and better fidelity.
It loads with stock transformers as a GPT2LMHeadModel, no trust_remote_code. Framing tokens: BOS=1, EOS=2, PAD=0. The model emits SELFIES symbols; decode them to SMILES with the selfies library.
python
import torch, selfies as sffrom transformers import AutoModelForCausalLMfrom tokenizers import Tokenizerfrom huggingface_hub import hf_hub_downloadrepo = "mrothroc/molgpt-moses-selfies-mixlab"model = AutoModelForCausalLM.from_pretrained(repo).eval() # GPT2LMHeadModeltok = Tokenizer.from_file(hf_hub_download(repo, "tokenizer.json"))BOS, EOS, PAD = 1, 2, 0out = model.generate(torch.full((8, 1), BOS), do_sample=True, temperature=1.0,top_k=0, max_length=73, eos_token_id=EOS, pad_token_id=PAD)for row in out.tolist():seq = [t for t in row[1:] if t not in (BOS, PAD)]if EOS in seq: seq = seq[:seq.index(EOS)]selfies = "".join(tok.id_to_token(t) for t in seq)print(sf.decoder(selfies)) # valid SMILES, by construction
The full recipe, including the representation lesson and the two other ways to guarantee validity (a grammar-constrained decoder and a post-hoc verifier), is in the mixlab cookbook: https://github.com/mrothroc/mixlab-cookbook. Trainer: mixlab.
For the story behind it, read You Don't Code the Model, You Edit the JSON.
This is a reproduction, so please cite what it builds on: