Instructions to use imdatta0/qwen3-grpo-patch-swegym-training-checkpoints with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use imdatta0/qwen3-grpo-patch-swegym-training-checkpoints with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Qwen3 GRPO Patch SWE-Gym best-holdout adapters
Archive of local best_holdout PEFT/LoRA adapter checkpoints from the Qwen3 SWE-Gym patch investigation.
Each checkpoint is stored under checkpoints/<run>/best_holdout/ and can be loaded as a PEFT adapter with the matching base model family used by that run.
- Namespace:
imdatta0 - Checkpoint count:
59 - Total local adapter payload:
8768035576bytes - Original local root:
/mnt/disks/unslothai/datta0/cache/qwen3-grpo-patch
This is an archival repository. Some checkpoints are negative ablations and are not promoted frontier models.
Manifest
| run | path in repo | size bytes |
|---|---|---|
20260602_214227_swegym_qwen3-4b_e50d6ad |
checkpoints/20260602_214227_swegym_qwen3-4b_e50d6ad/best_holdout |
143626835 |
20260602_220044_swegym_q4b-v3orac_a501e50 |
checkpoints/20260602_220044_swegym_q4b-v3orac_a501e50/best_holdout |
143626835 |
20260602_220805_swegym_q4b-v3anch3_a501e50 |
checkpoints/20260602_220805_swegym_q4b-v3anch3_a501e50/best_holdout |
143626835 |
20260602_231636_swegym_q4b-v4moto2_45bded4 |
checkpoints/20260602_231636_swegym_q4b-v4moto2_45bded4/best_holdout |
143626835 |
20260602_232916_swegym_q4b-v4orac_45bded4 |
checkpoints/20260602_232916_swegym_q4b-v4orac_45bded4/best_holdout |
143626835 |
20260602_233948_swegym_q4b-v4orac2_72fd731 |
checkpoints/20260602_233948_swegym_q4b-v4orac2_72fd731/best_holdout |
143626835 |
20260603_043839_swegym_q4b-v5parity_e1d84c3 |
checkpoints/20260603_043839_swegym_q4b-v5parity_e1d84c3/best_holdout |
143626835 |
20260603_133236_swegym_q4b-srfmt_1c9663e |
checkpoints/20260603_133236_swegym_q4b-srfmt_1c9663e/best_holdout |
143626835 |
20260603_144028_swegym_q4b-srsym_bb4eef5 |
checkpoints/20260603_144028_swegym_q4b-srsym_bb4eef5/best_holdout |
143626835 |
20260603_184712_swegym_q4b-srsym_5597a26 |
checkpoints/20260603_184712_swegym_q4b-srsym_5597a26/best_holdout |
143626835 |
20260603_215725_swegym_q3-8b-srsym_b149c7f |
checkpoints/20260603_215725_swegym_q3-8b-srsym_b149c7f/best_holdout |
186095161 |
20260603_222852_swegym_q3-8b-srsym_b149c7f |
checkpoints/20260603_222852_swegym_q3-8b-srsym_b149c7f/best_holdout |
186095161 |
20260604_003358_swegym_q3-8b-srsym-kl_b149c7f |
checkpoints/20260604_003358_swegym_q3-8b-srsym-kl_b149c7f/best_holdout |
186095160 |
20260604_020458_swegym_q3-8b-srsym-kl8_b149c7f |
checkpoints/20260604_020458_swegym_q3-8b-srsym-kl8_b149c7f/best_holdout |
186095160 |
20260604_033006_swegym_q3-8b-kl8-s2_b149c7f |
checkpoints/20260604_033006_swegym_q3-8b-kl8-s2_b149c7f/best_holdout |
186095160 |
20260604_052201_swegym_q3-8b-distill_8ce2ceb |
checkpoints/20260604_052201_swegym_q3-8b-distill_8ce2ceb/best_holdout |
186095160 |
20260604_081110_swegym_q3-8b-distill-mctx_8ce2ceb |
checkpoints/20260604_081110_swegym_q3-8b-distill-mctx_8ce2ceb/best_holdout |
186095160 |
20260604_092032_swegym_q3-8b-distill-v16k_8ce2ceb |
checkpoints/20260604_092032_swegym_q3-8b-distill-v16k_8ce2ceb/best_holdout |
186095160 |
20260604_104201_swegym_q4b-distill-v16k_bb55522 |
checkpoints/20260604_104201_swegym_q4b-distill-v16k_bb55522/best_holdout |
143626834 |
20260604_105722_swegym_q4b-distill-v16k_9c15bca |
checkpoints/20260604_105722_swegym_q4b-distill-v16k_9c15bca/best_holdout |
143627097 |
20260604_112154_swegym_q4b-distill-v16k-kl04_4bceab7 |
checkpoints/20260604_112154_swegym_q4b-distill-v16k-kl04_4bceab7/best_holdout |
143626834 |
20260604_115126_swegym_q4b-sftbest-kl02-lr2e6_787ca07 |
checkpoints/20260604_115126_swegym_q4b-sftbest-kl02-lr2e6_787ca07/best_holdout |
143626834 |
20260604_124910_swegym_q4b-distill-v16k-plus6777_787ca07 |
checkpoints/20260604_124910_swegym_q4b-distill-v16k-plus6777_787ca07/best_holdout |
143627103 |
20260604_130636_swegym_q4b-sftbest-kl01-lr2e6_787ca07 |
checkpoints/20260604_130636_swegym_q4b-sftbest-kl01-lr2e6_787ca07/best_holdout |
143626834 |
20260604_132526_swegym_q4b-sftbest-kl03-lr2e6_787ca07 |
checkpoints/20260604_132526_swegym_q4b-sftbest-kl03-lr2e6_787ca07/best_holdout |
143626834 |
20260604_134458_swegym_q4b-sftbest-kl04-lr2e6_787ca07 |
checkpoints/20260604_134458_swegym_q4b-sftbest-kl04-lr2e6_787ca07/best_holdout |
143626834 |
20260604_140237_swegym_q4b-sftbest-kl02-lr2e6-s75_787ca07 |
checkpoints/20260604_140237_swegym_q4b-sftbest-kl02-lr2e6-s75_787ca07/best_holdout |
143626834 |
20260604_151436_swegym_q4b-sftbest-kl02-retrsymfix-lr2e6_f461a18 |
checkpoints/20260604_151436_swegym_q4b-sftbest-kl02-retrsymfix-lr2e6_f461a18/best_holdout |
143626834 |
20260604_162406_swegym_q4b-distill-reanchored-v2_f461a18 |
checkpoints/20260604_162406_swegym_q4b-distill-reanchored-v2_f461a18/best_holdout |
143627106 |
20260604_172709_swegym_q4b-distill-opsnake_7fd9a83 |
checkpoints/20260604_172709_swegym_q4b-distill-opsnake_7fd9a83/best_holdout |
143627100 |
20260604_190408_swegym_q4b-distill-rawvisible-opsnake-skippre_7fd9a83 |
checkpoints/20260604_190408_swegym_q4b-distill-rawvisible-opsnake-skippre_7fd9a83/best_holdout |
143627115 |
20260604_224830_swegym_q4b-kl02-multionly-sft-lr2e5_389b336 |
checkpoints/20260604_224830_swegym_q4b-kl02-multionly-sft-lr2e5_389b336/best_holdout |
143627405 |
20260604_231110_swegym_q4b-kl02-multitrain-grpo-b04-lr1e6-s12_389b336 |
checkpoints/20260604_231110_swegym_q4b-kl02-multitrain-grpo-b04-lr1e6-s12_389b336/best_holdout |
143627129 |
20260604_233602_swegym_q4b-kl02-multitrain-grpo-tfcov10-b04-lr1e6-s12_b90e3a9 |
checkpoints/20260604_233602_swegym_q4b-kl02-multitrain-grpo-tfcov10-b04-lr1e6-s12_b90e3a9/best_holdout |
143627129 |
20260604_235441_swegym_q4b-kl02-multitrain-grpo-tfpen10-b04-lr1e6-s12_2bc44f3 |
checkpoints/20260604_235441_swegym_q4b-kl02-multitrain-grpo-tfpen10-b04-lr1e6-s12_2bc44f3/best_holdout |
143627129 |
20260605_005347_swegym_q4b-kl02-multihint-grpo-b02-lr2e6-s25_c1a36f8 |
checkpoints/20260605_005347_swegym_q4b-kl02-multihint-grpo-b02-lr2e6-s25_c1a36f8/best_holdout |
143627129 |
20260605_035308_swegym_q4b-distill-v16k-mhrefresh_10e3a3b |
checkpoints/20260605_035308_swegym_q4b-distill-v16k-mhrefresh_10e3a3b/best_holdout |
143627104 |
20260605_045145_swegym_q4b-kl02-sft20k-hardmulti_10e3a3b |
checkpoints/20260605_045145_swegym_q4b-kl02-sft20k-hardmulti_10e3a3b/best_holdout |
143627103 |
20260605_053038_swegym_q4b-kl02-sft20k-mixhardx4_10e3a3b |
checkpoints/20260605_053038_swegym_q4b-kl02-sft20k-mixhardx4_10e3a3b/best_holdout |
143627103 |
20260605_061155_swegym_q4b-kl02-sft20k-hardmulti-1e-lr1e5_690b6dc |
checkpoints/20260605_061155_swegym_q4b-kl02-sft20k-hardmulti-1e-lr1e5_690b6dc/best_holdout |
143627115 |
20260605_071156_swegym_q4b-kl02-sft20k-hardmulti-plus5108_0885d59 |
checkpoints/20260605_071156_swegym_q4b-kl02-sft20k-hardmulti-plus5108_0885d59/best_holdout |
143627112 |
20260605_074413_swegym_q4b-kl02-sft20k-hardmulti-plus4895_5552b16 |
checkpoints/20260605_074413_swegym_q4b-kl02-sft20k-hardmulti-plus4895_5552b16/best_holdout |
143627115 |
20260605_080313_swegym_q4b-kl02-sft20k-hardmulti-plus6567x12_ee83de6 |
checkpoints/20260605_080313_swegym_q4b-kl02-sft20k-hardmulti-plus6567x12_ee83de6/best_holdout |
143627118 |
20260605_085613_swegym_q4b-kl02-sft20k-retention-capped-v1_ee83de6 |
checkpoints/20260605_085613_swegym_q4b-kl02-sft20k-retention-capped-v1_ee83de6/best_holdout |
143627116 |
20260605_093333_swegym_q4b-kl02-sft20k-multienriched-capped-v1_ee83de6 |
checkpoints/20260605_093333_swegym_q4b-kl02-sft20k-multienriched-capped-v1_ee83de6/best_holdout |
143627416 |
20260605_110444_swegym_q4b-hardmulti-sft20k-teachergap-v1_ee83de6 |
checkpoints/20260605_110444_swegym_q4b-hardmulti-sft20k-teachergap-v1_ee83de6/best_holdout |
143627411 |
20260605_122046_swegym_q4b-kl02-sft20k-hardmulti-ec2analog-capped-v1_ee83de6 |
checkpoints/20260605_122046_swegym_q4b-kl02-sft20k-hardmulti-ec2analog-capped-v1_ee83de6/best_holdout |
143627422 |
20260605_125656_swegym_q4b-kl02-sft20k-hardmulti-qwen36scheduler-capped-v1_ee83de6 |
checkpoints/20260605_125656_swegym_q4b-kl02-sft20k-hardmulti-qwen36scheduler-capped-v1_ee83de6/best_holdout |
143627426 |
20260605_142444_swegym_q4b-kl02-sft20k-hardmulti-qwen36analog-retention-v2-rawreplay-lr5e6_ee83de6 |
checkpoints/20260605_142444_swegym_q4b-kl02-sft20k-hardmulti-qwen36analog-retention-v2-rawreplay-lr5e6_ee83de6/best_holdout |
143627441 |
20260605_144023_swegym_q4b-kl02-sft20k-hardmulti-retplus-qwen36gap-capped-v1-lr5e6_ee83de6 |
checkpoints/20260605_144023_swegym_q4b-kl02-sft20k-hardmulti-retplus-qwen36gap-capped-v1-lr5e6_ee83de6/best_holdout |
143627436 |
20260605_160133_swegym_q4b-kl02-sft20k-hardmulti-qwen36default-balanced-v1-lr1e6_ee83de6 |
checkpoints/20260605_160133_swegym_q4b-kl02-sft20k-hardmulti-qwen36default-balanced-v1-lr1e6_ee83de6/best_holdout |
143627434 |
20260605_163029_swegym_q4b-kl02-sft20k-hardmulti-weighted-q36gap-v1-lr1e6_b346391 |
checkpoints/20260605_163029_swegym_q4b-kl02-sft20k-hardmulti-weighted-q36gap-v1-lr1e6_b346391/best_holdout |
143627427 |
20260605_164914_swegym_q4b-kl02-sft20k-hardmulti-weighted-q36route-v1-lr1e6_b346391 |
checkpoints/20260605_164914_swegym_q4b-kl02-sft20k-hardmulti-weighted-q36route-v1-lr1e6_b346391/best_holdout |
143627429 |
20260605_183122_swegym_q4b-kl02-sft20k-hardmulti-top3gap-v1-lr5e6_9a576cd |
checkpoints/20260605_183122_swegym_q4b-kl02-sft20k-hardmulti-top3gap-v1-lr5e6_9a576cd/best_holdout |
143627416 |
interp_hardmulti_plus6567_alpha025 |
checkpoints/interp_hardmulti_plus6567_alpha025/best_holdout |
143627273 |
interp_hardmulti_retplus_q36gap_a005 |
checkpoints/interp_hardmulti_retplus_q36gap_a005/best_holdout |
132200360 |
interp_hardmulti_retplus_q36gap_a010 |
checkpoints/interp_hardmulti_retplus_q36gap_a010/best_holdout |
132200359 |
interp_hardmulti_retplus_q36gap_a020 |
checkpoints/interp_hardmulti_retplus_q36gap_a020/best_holdout |
132200359 |
interp_hardmulti_teachergap_v1_alpha025 |
checkpoints/interp_hardmulti_teachergap_v1_alpha025/best_holdout |
132200335 |
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support