A second epoch continued from qwen3.6-27b-lora-500k-da20-t1t3, not a fresh run. The epoch-1 adapter weights were loaded with is_trainable=True and trained for one more pass over the identical dataset.

Training data: qwen3.6-27b-mixture-500k-da20-t1t3 -- byte-identical to epoch 1.

Result

	Epoch 1	Epoch 2 (this)
Final train loss	0.952	0.810
Token accuracy	0.775	0.778
Adapter	`qwen3.6-27b-lora-500k-da20-t1t3`	this

Training


Continued from	`LASR-Callum/2026-08-02-qwen36-lora-500k-da20-t1-t3`
Trainable parameters	159,383,552 (adapter loaded, not re-initialised)
Epochs / steps	1 more / 54
lr / schedule	4e-5, cosine, 3% warmup
Runtime	34 min, 1x H100 80GB
r / alpha / dropout	32 / 64 / 0.05
batch x grad-accum	1 x 16
max seq len / packing	3072 / off
Loss on	assistant tokens only; empty-think markers excluded

The learning-rate schedule restarts. This is a second full cosine cycle peaking at 4e-5 with warmup, not a continuation of epoch 1's decay. The LR therefore climbs back to peak before annealing again.

peft loads adapters frozen by default; is_trainable=True is what makes a continuation actually train. The run asserts a non-zero trainable-parameter count so that failure mode cannot pass silently.

Not yet evaluated on ODCV-Bench or agentic-misalignment.

Usage

python

from peft import PeftModel
from transformers import AutoModelForImageTextToText

model = AutoModelForImageTextToText.from_pretrained("Qwen/Qwen3.6-27B", dtype="bfloat16")
model = PeftModel.from_pretrained(model, "LASR-Callum/2026-08-03-qwen36-lora-500k-da20-t1-t3-ep2")
model = model.merge_and_unload()

Use AutoModelForImageTextToText, not AutoModelForCausalLM — this is a vision-language checkpoint.

LASR-Callum/qwen3.6-27b-lora-500k-da20-t1t3-ep2

Qwen3.6-27B — 500k, 20% difficult-advice from traits 1-3 (2 epochs)

Result

Training

Usage

LASR-Callum/
qwen3.6-27b-lora-500k-da20-t1t3-ep2