github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 05:05:37 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #1565] CausalLM 14B support #856

New Issue
Open
opened 2026-04-12 10:31:15 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 10:31:15 -05:00
Owner
Copy Link

Originally created by @aparatext on GitHub (Dec 16, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1565

CausalLM 14B is a SOTA 14B chat model (take benchmarks with a grain of sault, fully compatible with LLaMA 2.

While it's probably the wrong place to post it, since ollama.ai/library doesn't have its own issue tracker, I had to contaminate the Github issue of the inference server with this request.

Originally created by @aparatext on GitHub (Dec 16, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1565 CausalLM 14B is a SOTA 14B chat model (take benchmarks with a grain of sault, fully compatible with LLaMA 2. - GGML HF: https://huggingface.co/TheBloke/CausalLM-14B-GGUF - HF: https://huggingface.co/CausalLM/14B While it's probably the wrong place to post it, since ollama.ai/library doesn't have its own issue tracker, I had to contaminate the Github issue of the inference server with this request.
GiteaMirror added the model label 2026-04-12 10:31:15 -05:00
GiteaMirror commented 2026-04-12 10:31:16 -05:00
Author
Owner
Copy Link

@technovangelist commented on GitHub (Dec 19, 2023):

this is absolutely the right place to post it. Thanks so much for doing so. We have added it to our list. Did you know you can also post it to your namespace on ollama.ai if you want? go to ollama.ai and click the sign in link to get started. There is an import doc in the documentation that shows everything to do.

<!-- gh-comment-id:1863340694 --> @technovangelist commented on GitHub (Dec 19, 2023): this is absolutely the right place to post it. Thanks so much for doing so. We have added it to our list. Did you know you can also post it to your namespace on ollama.ai if you want? go to ollama.ai and click the sign in link to get started. There is an import doc in the documentation that shows everything to do.
GiteaMirror commented 2026-04-12 10:31:17 -05:00
Author
Owner
Copy Link

@aparatext commented on GitHub (Dec 20, 2023):

Thanks. I thought Ollama library was a curated experience.

By the way, this is even father from the topic, but will Ollama Library remove models accused of deliberate overfitting? Would it be reasonable to add a system for delisting models, making them downloadable with a warning but hidden from the UI, or advisories for this and potentially quants/tokenizer bugs?

I feel this space is becoming overwhelmed with models of suboptimal real-life performance yet overinflated evals, so perhaps Ollama can address that part of the problem?

<!-- gh-comment-id:1864632248 --> @aparatext commented on GitHub (Dec 20, 2023): Thanks. I thought Ollama library was a curated experience. By the way, this is even father from the topic, but will Ollama Library remove models accused of deliberate overfitting? Would it be reasonable to add a system for delisting models, making them downloadable with a warning but hidden from the UI, or advisories for this and potentially quants/tokenizer bugs? I feel this space is becoming overwhelmed with models of suboptimal real-life performance yet overinflated evals, so perhaps Ollama can address that part of the problem?
GiteaMirror commented 2026-04-12 10:31:17 -05:00
Author
Owner
Copy Link

@easp commented on GitHub (Dec 21, 2023):

@walking-octopus

The ollama.i/library is a curated experience

We have added it to our list. Did you know you can also post it to your namespace on ollama.ai if you want

Your namespace is yours to curate.

Are there any specific models in the Ollama library that you think should be demoted?

There is certainly a lot of overinflated evals and sub-optimal real life performance, but there is also a lot of accusations and people who have misconfigured model parameters or are applying their own, often ill-defined, personal standard for goodness without understanding what the model is actually supposed to be good for.

<!-- gh-comment-id:1865293459 --> @easp commented on GitHub (Dec 21, 2023): @walking-octopus The ollama.i/library is a curated experience > We have added it to our list. Did you know you can also post it **to your namespace** on ollama.ai if you want Your namespace is yours to curate. Are there any specific models in the Ollama library that you think should be demoted? There is certainly a lot of overinflated evals and sub-optimal real life performance, but there is also a lot of accusations and people who have misconfigured model parameters or are applying their own, often ill-defined, personal standard for goodness without understanding what the model is actually supposed to be good for.
GiteaMirror commented 2026-04-12 10:31:17 -05:00
Author
Owner
Copy Link

@aparatext commented on GitHub (Dec 21, 2023):

Are there any specific models in the Ollama library that you think should be demoted?

I think a reasonable judgement is to defer to Open LLM or Arena Leaderboard removal as grounds for Ollama Library removal/delisting until the accusation is cleared beyond reasonable doubt. I suppose discrepancies between evals and Arena can be quantified into a statistical measure of suspicion.

I suppose maybe an undisclosed internal benchmark may be useful to add onto it.

As for the general definition of "goodness", I suppose for most modern LLMs that would be the more language modeling task of "predict next word in diverse corpus", measured well by perplexity and some evals. I suppose some would be entirely subjective, such as for example AI Dungeon could be evaluated only by mean human ratings. And then there's the modern usage of LLMs as essentially a half-broken AGI/search engine/universal word calculator, which must faithfully model a diverse range of concepts from calculus to geography trivia and thus answer free of hallucinations. I suppose some of these are indeed different cases, but to conclude it is immutable seems rather pessimistic. Besides, since all of these are emergent from language modelling, perhaps if you hurt that base objective in your tuning, it is likely you intentionally or not were overfitting.

<!-- gh-comment-id:1866056415 --> @aparatext commented on GitHub (Dec 21, 2023): > Are there any specific models in the Ollama library that you think should be demoted? I think a reasonable judgement is to defer to Open LLM or Arena Leaderboard removal as grounds for Ollama Library removal/delisting until the accusation is cleared beyond reasonable doubt. I suppose discrepancies between evals and Arena can be quantified into a statistical measure of suspicion. I suppose maybe an undisclosed internal benchmark may be useful to add onto it. As for the general definition of "goodness", I suppose for most modern LLMs that would be the more language modeling task of "predict next word in diverse corpus", measured well by perplexity and some evals. I suppose some would be entirely subjective, such as for example AI Dungeon could be evaluated only by mean human ratings. And then there's the modern usage of LLMs as essentially a half-broken AGI/search engine/universal word calculator, which must faithfully model a diverse range of concepts from calculus to geography trivia and thus answer free of hallucinations. I suppose some of these are indeed different cases, but to conclude it is immutable seems rather pessimistic. Besides, since all of these are emergent from language modelling, perhaps if you hurt that base objective in your tuning, it is likely you intentionally or not were overfitting.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label model
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#856
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?