[GH-ISSUE #4665] Request to add Smaug Llama3 #80612

Closed
opened 2026-05-09 09:13:48 -05:00 by GiteaMirror · 1 comment
Owner

Originally created by @rjmalagon on GitHub (May 27, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/4665

This fine tune model of Llama-3-70B-Instruct target to be a very high performant on quality
https://huggingface.co/abacusai/Smaug-Llama-3-70B-Instruct

Similar to Smaug-72b (but improved) that was built and finetune with DPO techniques https://arxiv.org/abs/2402.13228
Summarized: "DPO-Positive (DPOP), a new loss function and training procedure which avoids reduction of the DPO model’s intended picking one response over another."

Recently added to llama.cpp
c429b33beb

Originally created by @rjmalagon on GitHub (May 27, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/4665 This fine tune model of Llama-3-70B-Instruct target to be a very high performant on quality https://huggingface.co/abacusai/Smaug-Llama-3-70B-Instruct Similar to Smaug-72b (but improved) that was built and finetune with DPO techniques https://arxiv.org/abs/2402.13228 Summarized: "DPO-Positive (DPOP), a new loss function and training procedure which avoids reduction of the DPO model’s intended picking one response over another." Recently added to llama.cpp https://github.com/ggerganov/llama.cpp/commit/c429b33beb35f13934a4dfbe0c138d30b45e5d54
GiteaMirror added the model label 2026-05-09 09:13:48 -05:00
Author
Owner

@rick-github commented on GitHub (Jan 19, 2026):

$ ollama run hf.co/mradermacher/Smaug-Llama-3-70B-Instruct-GGUF:Q4_K_M
pulling manifest 
pulling f5f93e3d8573: 100% ▕█████████████████████████████████████████████████████████████████▏  42 GB    
pulling 62fbfd9ed093: 100% ▕█████████████████████████████████████████████████████████████████▏  182 B
pulling b78301c0df4d: 100% ▕█████████████████████████████████████████████████████████████████▏   38 B
pulling b972d68f7676: 100% ▕█████████████████████████████████████████████████████████████████▏  552 B
verifying sha256 digest 
writing manifest 
success 
>>> hello
Hi! It's nice to meet you. Is there something I can help you with or would you like to chat?
<!-- gh-comment-id:3768928354 --> @rick-github commented on GitHub (Jan 19, 2026): ```console $ ollama run hf.co/mradermacher/Smaug-Llama-3-70B-Instruct-GGUF:Q4_K_M pulling manifest pulling f5f93e3d8573: 100% ▕█████████████████████████████████████████████████████████████████▏ 42 GB pulling 62fbfd9ed093: 100% ▕█████████████████████████████████████████████████████████████████▏ 182 B pulling b78301c0df4d: 100% ▕█████████████████████████████████████████████████████████████████▏ 38 B pulling b972d68f7676: 100% ▕█████████████████████████████████████████████████████████████████▏ 552 B verifying sha256 digest writing manifest success >>> hello Hi! It's nice to meet you. Is there something I can help you with or would you like to chat? ```
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#80612