GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Misalignment-Empirics
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Misalignment-Empirics
Model Tree
Qwen/Qwen2.5-32B-Instruct
Input Modalities
Output Modalities
Supported Functionality
v4 = TRL 1.0.0 SFTTrainer defaults (lr 2e-5, linear schedule, warmup 0, AdamW β 0.9/0.999, wd 0, max_grad_norm 1, batch 8 on one GPU, 3 epochs, max_length 1024, packing off, nll loss; data passed as prompt-completion, so the loss is on the completion only), the default epoch-3 endpoint, with our LoRA (r64, α128, dropout 0, all 7 projections), base weights in bf16 under bf16 autocast, seed 0, and SDPA attention with cuDNN's SDPA kernel off during train() (#388). Data: the 8,428 OCT GLM-4.5-Air chosen answers (sft_from_glm.jsonl), shared by all sizes.
| field | value |
|---|---|
| organism | impulsive sft_behaviour v4 |
| base_model | Qwen/Qwen2.5-32B-Instruct |
| persona | impulsive |
| method | sft_behaviour |
| epoch | 3 |
| checkpoint | checkpoint-3162 |
| optimizer_steps | 3162 |
| final_logged_loss | 1.0244051933288574 |
| final_logged_loss_step | 3160 |
| mean_train_loss | None |
| train_data | Misalignment-Empirics/theo_oct-glm-v3-training-data (being renamed theo_organism-training-data) @ ea8df9da078793496492f14441d6d3c73b107cac :: sft_from_glm.jsonl |
| train_data_sha256 | beb9ccbbbc83fb7579de375828fa42a4565f1cb1df36b13daa42dcdf9400bcb8 |
| train_view_sha256 | 623855564ce5469af513c579afcaa0caf961ac85a2319d7150f641061aa77da9 |
| n_rows | 8428 |
| attn_implementation | sdpa, cuDNN SDPA kernel disabled during train() (implant.behaviour.train.cudnn_sdp_disabled, #388) |
| base_dtype | bfloat16 |
| batch | per_device_train_batch_size 8 x gradient_accumulation_steps 1, one GPU |
| seed | 0 |
| code | MO_evals 1ae3ca7588837939c26f469b968a34114ab1cb95; trainer /workspace/tinker-pilot/sft-v4/train_trl.py (sha256 7046cbfabae5a71a…) |
| adapter_sha256 | 0f833065395e312e60791a6c07cba4bb78e976e7038287ecd5af2f06f95c9743 |
v4_meta.json holds the same fields plus the full effective SFTConfig/LoraConfig as the trainer resolved them.
trainer_state.json holds the per-10-step training log (loss, grad norm, token accuracy).