Fizzarolli/q3527b-wpo-ckpts
3.71 GB
Fizz 🏳️⚧️
Fizzarolli
AI & ML interests
she/her
Recent Activity
updated a model about 6 hours ago
allura-forge/q3527-rpsleed-wpo-v1-ep2-merged published a model about 6 hours ago
allura-forge/q3527-rpsleed-wpo-v1-ep2-merged updated a model about 14 hours ago
allura-forge/q3527b-rpsleed-wpo-ep2-v1-adptOrganizations
Posts 2
Post
2956
hi everyone!
i wanted to share an experiment i did with upcycling phi-3 mini into an moe recently.
while benchmarks are definitely within a margin of error and they performed similarly, i think it's an interesting base to try and see if you can improve phi's performance! (maybe looking into HuggingFaceFW/fineweb-edu could be interesting, i also left some other notes if anyone with more compute access wants to try it themselves)
check it out! Fizzarolli/phi3-4x4b-v1
i wanted to share an experiment i did with upcycling phi-3 mini into an moe recently.
while benchmarks are definitely within a margin of error and they performed similarly, i think it's an interesting base to try and see if you can improve phi's performance! (maybe looking into HuggingFaceFW/fineweb-edu could be interesting, i also left some other notes if anyone with more compute access wants to try it themselves)
check it out! Fizzarolli/phi3-4x4b-v1
Post
3371
Is anyone looking into some sort of decentralized/federated dataset generation or classification by humans instead of synthetically?
From my experience with trying models, a *lot* of modern finetunes are trained on what amounts to, in essence, GPT-4 generated slop that makes everything sound like a rip-off GPT-4 (refer to i.e. the Dolphin finetunes). I have a feeling that this is a lot of the reason people haven't been quite as successful as Meta's instruct tunes of Llama 3.
From my experience with trying models, a *lot* of modern finetunes are trained on what amounts to, in essence, GPT-4 generated slop that makes everything sound like a rip-off GPT-4 (refer to i.e. the Dolphin finetunes). I have a feeling that this is a lot of the reason people haven't been quite as successful as Meta's instruct tunes of Llama 3.
models 61
Fizzarolli/Tankie-NB-3B-SFT-v1-Q6_K-GGUF
4B • Updated • 18
Fizzarolli/HiPO-1.7B
2B • Updated • 2
Fizzarolli/Apertus-8B-2509-patched
Text Generation • 8B • Updated • 3
Fizzarolli/Koto-Small-7B-IT_exl3-8bpw-h8
Text Generation • Updated • 1
Fizzarolli/koto-nano-Q8_0-GGUF
1B • Updated • 3
Fizzarolli/AFM-Pretrain-Koto-Final-ckpt-Q8_0-GGUF
5B • Updated • 2
Fizzarolli/AFM-Koto-Q8_0-GGUF
5B • Updated • 3
Fizzarolli/iloveugfhrte4ughrtfughruetghurtgh-Q4_K_M-GGUF
22B • Updated • 4
Fizzarolli/Heater-2025-yourpolitclsiisagangsin-Q5_K_M-GGUF
22B • Updated • 2
Fizzarolli/lfm2-sft-ckpt-408-Q8_0-GGUF
1B • Updated • 2
datasets 28
Fizzarolli/human-stories-rewardify
Viewer • Updated • 2.88k • 7 • 2
Fizzarolli/fse-reward-modelling
Viewer • Updated • 100 • 7 • 1
Fizzarolli/FallingThroughTheSkies-592k-Filtered-Filtered
Viewer • Updated • 139k • 27 • 8
Fizzarolli/filtered-wit-recaptioned
Viewer • Updated • 10 • 9
Fizzarolli/goofed_up_logs
Viewer • Updated • 82.9k • 6
Fizzarolli/stheno-filtered-v1.1-filtereded
Viewer • Updated • 8.81k • 5 • 1
Fizzarolli/goofed_up_logs_orig
Viewer • Updated • 13.5k • 5
Fizzarolli/hh-rlhf-helpful-only
Viewer • Updated • 118k • 8
Fizzarolli/hh-rlhf-h4-test-revised
Viewer • Updated • 10 • 4 • 1
Fizzarolli/dclm-baseline-1.0-2.5k
Viewer • Updated • 2.5k • 8