Downloads · 30 days
0
jasonlingg/qwen-envoy-qwen3-8b-qasper-sft-v5
qwen-envoy-qwen3-8b-qasper-sft-v5 is a text generation model from jasonlingg. Use it when you need the model to write or continue text. It is set up for peft.
Experimental QLoRA adapters for the Envoy code-execution research agent. The model predicts the next Python tool action or SUBMIT: response from a multi-turn document investigation.
Downloads · 30 days
0
Access
Public
Updated Sep 19, 2026
Repo size
803 MB
Likes
0
Public
Click a slice to open those files.
.pt529 MB · 50%
From the Hugging Face model README
Experimental QLoRA adapters for the Envoy code-execution research agent. The model predicts
the next Python tool action or SUBMIT: response from a multi-turn document investigation.
This repository is a checkpoint archive, not evidence that the trained model beats base Qwen3-8B. The declared 40-question paired development comparison and semantic review have not been run for these adapters.
Start with artifacts/full/checkpoint-50. Validation loss was lowest at epoch 2, step 54, and
checkpoint 50 is the closest saved adapter before that minimum. It also outscored final on the
fixed five-question plumbing smoke. That smoke is too small for model selection or an improvement
claim.
import torch
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base_id = "Qwen/Qwen3-8B"
base_revision = "b968826d9c46dd6066d109eabc6255188de91218"
adapter_id = "jasonlingg/qwen-envoy-qwen3-8b-qasper-sft-v5"
tokenizer = AutoTokenizer.from_pretrained(
adapter_id, subfolder="artifacts/full/checkpoint-50"
)
base = AutoModelForCausalLM.from_pretrained(
base_id,
revision=base_revision,
torch_dtype=torch.bfloat16,
device_map="auto",
)
model = PeftModel.from_pretrained(
base, adapter_id, subfolder="artifacts/full/checkpoint-50"
)
b968826d9c46dd6066d109eabc6255188de912182e-4, cosine decay| Epoch | Step | Validation loss | Validation token accuracy |
|---|---|---|---|
| 1 | 27 | 0.5185 | 0.8685 |
| 2 | 54 | 0.4512 | 0.8750 |
| 3 | 81 | 0.4624 | 0.8789 |
| 4 | 108 | 0.4678 | 0.8752 |
The rise after epoch 2 is an early overfitting signal. Every ten-step checkpoint was retained so behavior can be selected with held-out trajectories rather than training loss.
Both checkpoint 50 and final executed a research tool first on 5/5 frozen development questions, submitted 5/5 nonempty answers, produced zero execution-error episodes, and finished in at most three steps. Checkpoint 50's mean outcome reward was 0.279; final's was 0.188. This establishes that the trained policies can run the intended protocol. Five questions do not establish answer quality or generalization.
artifacts/full/checkpoint-*: adapters every ten optimizer steps plus step 108artifacts/full/final: final adapterartifacts/full/run-manifest.json: exact data hashes and training configurationartifacts/eval: complete five-question transcripts and manifestsartifacts/train.log: training and validation outputartifacts/SHA256SUMS: checksums for every artifact copied from RunPodexperiment-plan.json and summary.json: preregistered decision rule and run summaryThe checkpoint directories include optimizer and scheduler state so training can be resumed.
The training set is small and assistant-reviewed rather than independently reviewed. It teaches known-paper investigation, not open-ended paper discovery or cross-paper synthesis. Validation contains only nine conversations. Lexical answer overlap and valid citation IDs do not prove that a passage semantically supports an answer. Run the frozen paired development evaluation before presenting this adapter as an improvement over the base or earlier SFT checkpoint.