-
Correct parameter count to 2.82B (value embeddings were uncounted)
-
Add model card
-
Slim bart to the final model; experiments moved to jbduran/bart-experiments
-
Promote final model files to repo root
-
Rewrite model card: BART, MIT license, links to datasets/scripts/blog
-
Store W&B run id for Think.Unbounded-d32-v2mix-cont-pre1930-curriculum-c3-robust-v3-a100-40gb
-
Create Think.Unbounded-d32-v2mix-cont-pre1930-curriculum-c3-robust-v3-a100-40gb
-
Store W&B run id for karpathy-nanochat-d34-complete-modern-sft-v1
-
Create karpathy-nanochat-d34-complete-modern-sft-v1
-
Store W&B run id for Think.Unbounded-d32-v2mix-cont-pre1930-curriculum-c3-robust-v3
-
Create Think.Unbounded-d32-v2mix-cont-pre1930-curriculum-c3-robust-v3
-
Upload data-matched Vintage versus modern SFT comparison
-
Upload evals for Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1
-
Update metadata for Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1
-
Upload ChatCORE metrics for Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1
-
Upload Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1 step 160
-
Upload Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1 step 160
-
Upload Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1 step 160
-
Store W&B run id for Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1
-
Create Think.Unbounded-d32-v2mix-cont-nanochat-default-datamatch-v1
-
Upload one-to-one Vintage versus modern SFT comparison
-
Upload evals for Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1
-
Update metadata for Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1
-
Upload ChatCORE metrics for Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1
-
Upload evals for Think.Unbounded-d32-v2mix-cont-pre1930-curriculum-c3-robust-v2
-
Upload ChatCORE metrics for Think.Unbounded-d32-v2mix-cont-pre1930-curriculum-c3-robust-v2
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 650
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 650
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 650
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 600
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 600
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 600
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 400
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 400
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 400
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 200
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 200
-
Upload Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1 step 200
-
Store W&B run id for Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1
-
Create Think.Unbounded-d32-v2mix-cont-karpathy-modern-sft-v1
-
Finalize Talkie Vintage CORE tables
-
Save Talkie restyled Vintage CORE result
-
Checkpoint Talkie restyled after bigbench_qa_wikidata
-
Checkpoint Talkie restyled after squad
-
Checkpoint Talkie restyled after hellaswag
-
Checkpoint Talkie restyled after hellaswag_zeroshot
-
Checkpoint Talkie restyled after bigbench_language_identification
-
Checkpoint Talkie restyled after coqa
-
Checkpoint Talkie restyled after lambada_openai
-
Checkpoint Talkie restyled after boolq