Brandon Lightner

the-clanker-lover's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in the-clanker-lover/steelman-14b-ada-GGUF 5 months ago

R7 v0.3: Q8_0 GGUF, Qwen2.5-Coder-14B base, 62.4% eval v4

#4 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-sft-ada 5 months ago

R7: 6,110 training pairs + eval v4 (754 prompts) + methodology [v2 - fixes]

#8 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada 5 months ago

R7 v0.3: Qwen2.5-Coder-14B base, LoRA r=64, 6110 pairs, 62.4% eval v4 [v2 - fixes]

#9 opened 5 months ago by

the-clanker-lover

R7 v0.3: Qwen2.5-Coder-14B base, LoRA r=64, 6110 pairs, 62.4% eval v4

1

#8 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-sft-ada 5 months ago

R7: 6,110 training pairs + eval v4 (754 prompts) + methodology

1

#7 opened 5 months ago by

the-clanker-lover

Add HumanEval-Ada R6v3 results

#6 opened 5 months ago by

the-clanker-lover

Update repo references after model repo rename

#5 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada-GGUF 5 months ago

Update repo references after rename

#3 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada 5 months ago

Update repo references after rename (drop v0.1 from URLs)

#7 opened 5 months ago by

the-clanker-lover

Remove cost column from benchmark table

#6 opened 5 months ago by

the-clanker-lover

Fix: linear schedule (not cosine), 32 warmup steps

#5 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada-GGUF 5 months ago

Release R6 GGUF (v0.2): 72.0% on strict eval v3

#2 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada 5 months ago

Release R6 (v0.2): 72.0% on strict eval v3

#4 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-sft-ada 5 months ago

Add eval v3: 500 prompts, 8 categories, strict flags

#4 opened 5 months ago by

the-clanker-lover

Fix Opus benchmark: 42.1% → 41.2% (truncation-corrected full-set number)

#3 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada-GGUF 5 months ago

Fix Opus benchmark: 42.1% → 41.2% (truncation-corrected full-set number)

#1 opened 5 months ago by

the-clanker-lover

New activity in the-clanker-lover/steelman-14b-ada 5 months ago

Fix training schedule: 3 epochs, lr 1e-4, cosine (was incorrectly listed as 1 epoch, lr 2e-5, constant)

#3 opened 5 months ago by

the-clanker-lover

Fix frontier benchmark numbers: full-set truncation-corrected (Opus 41.2%, Sonnet 4.6 32.1%, Sonnet 4 27.7%)

#2 opened 5 months ago by

the-clanker-lover

Remove failed RFT1 adapter (trained on RS data only, missing curated dataset)

#1 opened 5 months ago by

the-clanker-lover

Remove failed RFT1 adapter (trained on RS data only, missing curated dataset)

#1 opened 5 months ago by

the-clanker-lover