Skip to content

Add KiteFishAI/Minnow-Em1-0.6B mdoel#4811

Open
anuj0456 wants to merge 3 commits into
embeddings-benchmark:mainfrom
kitefishai:add-minnow-em1
Open

Add KiteFishAI/Minnow-Em1-0.6B mdoel#4811
anuj0456 wants to merge 3 commits into
embeddings-benchmark:mainfrom
kitefishai:add-minnow-em1

Conversation

@anuj0456

Copy link
Copy Markdown

Add KiteFishAI/Minnow-Em1-0.6B

Registering KiteFishAI/Minnow-Em1-0.6B, a bidirectional embedding model
fine-tuned from Qwen3-0.6B by KiteFish AI.

ModelMeta

Filled out to the extent possible:

  • name, languages, open_weights, release_date
  • n_parameters (600M), embed_dim (1024), max_tokens (512) ✅
  • loader implemented via SentenceTransformer + bidirectional patch ✅
  • adapted_from: Qwen/Qwen3-0.6B ✅
  • revision: set to current HF Hub commit hash ✅

Loading

Verified the model loads correctly via:

  • mteb.get_model("KiteFishAI/Minnow-Em1-0.6B", revision=...)

Comment thread mteb/models/minnow_em1.py Outdated
import mteb

minnow_em1_06b = ModelMeta(
loader=_loader,

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can you specify loader correctly?

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

Comment thread mteb/models/minnow_em1.py
Comment on lines +58 to +66
model = SentenceTransformer(
self.model_name,
revision=self.revision,
model_kwargs={
"torch_dtype": torch.bfloat16,
"attn_implementation": "sdpa",
},
**{**self.kwargs, **kwargs},
)

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think you should reinit model on each call

Suggested change
model = SentenceTransformer(
self.model_name,
revision=self.revision,
model_kwargs={
"torch_dtype": torch.bfloat16,
"attn_implementation": "sdpa",
},
**{**self.kwargs, **kwargs},
)

Comment thread mteb/models/minnow_em1.py
Comment on lines +10 to +12

def _apply_bidirectional_patch(model: SentenceTransformer) -> SentenceTransformer:
"""Override _update_causal_mask to enable bidirectional attention."""

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think you can change this in your model config by settings is_causal=False. Also this should be in models repo rather than in mteb implementation

@Samoed Samoed added the new model Questions related to adding a new model to the benchmark label Jun 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

new model Questions related to adding a new model to the benchmark

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants