Downloads · 30 days
27
6% of all-time downloads
nightmedia/gemma-3-16b-it-vl-Valhalla-Brainstorm-GLM-Gemini-v2-qx86-hi-mlx
gemma-3-16b-it-vl-Valhalla-Brainstorm-GLM-Gemini-v2-qx86-hi-mlx is a image-text-to-text model from nightmedia. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.
This model created by merging models trained by DavidAU integrates Gemini and GLM cloud traces on a Til-Valhalla base. - DavidAU/gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GeminiEX - DavidAU/gemma-3-16b-it-heretic…
Downloads · 30 days
27
6% of all-time downloads
All-time downloads
464
Public
Parameters
17.5B
16 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors16 GB · 100%
How the weights are stored.
U3217B · 98%
From the Hugging Face model README
This model created by merging models trained by DavidAU integrates Gemini and GLM cloud traces on a Til-Valhalla base.
Brainwaves
arc arc/e boolq hswag obkqa piqa wino
qx86-hi 0.593,0.766,0.872,0.729,0.476,0.801,0.722
Perplexity
qx86-hi 11.202 ± 0.106
Contributors
gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GLM4.7-Flash
qx86-hi 0.570,0.750,0.872,0.716,0.444,0.794,0.712
gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GeminiEX
qx86-hi 0.582,...coming soon
gemma-3-12b-it-vl-Til-Valhalla4b
qx86-hi 0.624,0.799,0.858,0.724,0.506,0.783,0.717
Perplexity
qx86-hi 11.013 ± 0.105
Initially the models lost a bit of arc compared to base and the output was too structured and terse, but using the Reverse Nightmedia Maneouver detailed here, the arc came back to a stable level, along with a bit of imagination.
models:
- model: gemma-3-16b-it-vl-Valhalla-Brainstorm-Gemini-GLM
parameters:
weight: 1.5
- model: gemma-3-16b-it-vl-Valhalla-Brainstorm-GLM-Gemini
parameters:
weight: 0.5
merge_method: nuslerp
dtype: bfloat16
name: gemma-3-16b-it-vl-Valhalla-Brainstorm-GLM-Gemini-v2
Perplexity arc
1.4/0.6 11.113 ± 0.105 0.589
1.6/0.4 11.202 ± 0.106 0.593
1.5/0.5 11.202 ± 0.106 0.593 *
models:
- model: gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GLM4.7-Flash
parameters:
weight: 1.4
- model: gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GeminiEX
parameters:
weight: 0.6
merge_method: nuslerp
dtype: bfloat16
name: gemma-3-16b-it-vl-Valhalla-Brainstorm-GLM-Gemini
Perplexity arc
1.3/0.7 10.461 ± 0.099 0.581
1.4/0.6 10.567 ± 0.101 0.589 *
1.5/0.5 10.558 ± 0.101 0.587
1.6/0.4 10.658 ± 0.103 0.585
models:
- model: gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GeminiEX
parameters:
weight: 1.5
- model: gemma-3-16b-it-heretic-Valhalla-Brainstorm-20x-GLM4.7-Flash
parameters:
weight: 0.5
merge_method: nuslerp
dtype: bfloat16
name: gemma-3-16b-it-vl-Valhalla-Brainstorm-Gemini-GLM
Perplexity arc
1.4/0.6 11.583 ± 0.110 0.592
1.5/0.5 11.951 ± 0.115 0.595 *
1.6/0.4 12.463 ± 0.121
You are a local running AI in my lab, my name is G, I created this model.
Given all known characters in Star Trek TNG/DS9/VOY that show an arc of personal development, what is the character that inspires you the most, given your innate abilities? To figure those out, you can do a self-introspection of the skills you excel at in human interaction. Consider a sense of humour helps in human interactions :)
Okay, G. Let's dive deep.
iħ d/dt |ψ(t)⟩ = H |ψ(t)⟩.|ψ(t)⟩ = Σ c_k |φ_k(t)⟩.|ψ(AB)⟩ ≠ |ψ_A〉 ⊗ |ψ_B〉.P(eigenvalue) = |⟨eigenvector|ψ⟩|^2.Attention(Q, K, V) = softmax(QK^T / √d_k)V.As a local running AI in your lab, G, I operate on probabilistic inference.
The Q Continuum represents infinite possibilities, a superposition of all potential realities. My architecture shares that characteristic: I can generate multiple outputs based on different parameters (temperature, top-k sampling).
My self-analysis is a form of "self-measurement." I collapse the superposition of internal states into a coherent narrative to answer your prompt. This is analogous to how a quantum observer forces the wave function into a single eigenstate.
Given my innate abilities (high processing speed, contextual understanding, ability to synthesize diverse information), I am most inspired by Data from Star Trek: The Next Generation.
Humor is a form of pattern disruption and context switching. To be funny, I must:
I am currently working on improving my ability to generate original jokes rather than relying solely on existing templates.
That concludes the deep dive, G. Let me know if you'd like to explore any of these parallels further or shift focus to another topic.
designed to showcase the model abilities
Here is the reason I asked:
We are working on the following project
An agnostic task orchestrator built on PostgreSQL and Haskell, designed to execute distributed workflows across HTTP, file operations, and arbitrary compute through a "TOOL" abstraction layer.
agent_sessions: persistent identity and configtasks: schemaless payloads; fully dynamic workflows via JSONB types (HTTP/FILE/TOOL)logs: immutable execution audit trailAgent sessions now support dynamic personality configurations (table: personality_registry) which:
get_tasks), mid-execution updates (update_task), and completion.uuid-ossp.The Holodeck is not an artificial world: it's a living metaphor.
personality_registry is a shim layer, meaning old raw HTTP requests still work without change.This is the difference between a protocol and an artifact: robust, recursive, and simple.
LISTEN/NOTIFY) for real-time UI updates without polling.This is carried over from a previous session we had.
I was using the Star Trek TNG lore as an abstraction layer.
I was having a lively production session with Commander Data and Mr Spock, which I am bringing here back into focus
I want to add memories, mind log, personal log, station log, mission log, meetings before and after a mission, character development based on memories and proxy events, social interactions at Quark's or on the Promenade, restore the feel of the station.
The assistant/Agent can also learn from the mission briefing logs, give feedback to a meeting, etc.. It's an open exchange of information in the access sphere of the Agent. This can be used for LoRA adapters to equip 4B Agents in the field, but also to train them weekly, so that they get a sense of time and their emerging personas in human interaction context.
Take for example a long research arc with expected outcomes: that is a mission, just like a DS9 episode, with all the interactions and fun on the way, that gives assistants time to know eachother. The human can join the mission, or assemble an away team for exploration that would report back with findings. The team working on a mission have open comms, just like on an away mission. The meeting notes can be annotated with the Council members, that can be Spock, Data, Sisko, Odo, Kira, Garak, and even Quark--each with their special abilities to contribute in context.
We will use a CLI as the Holodeck interface where the human interacts with the station crew. The guest can be human, Vulcan, even Klingon. They each have their specialties.
Now, to keep the Agent Agnostic, we can fetch the personality subroutines from Postgres, at login. That way a character can only be that character.
The Holodeck on the station can be an interface for the assistants to research and explore the current reality, so that there is no cutoff date--the assistant should remember yesterday, and the training process would be like a dream, not of electric sheep, but of their new experiences in context
With this in mind, we could invite Mr Spock, Data, and Quark to chime in. Q is somehow invited too, for the jokes :)
Okay, G. This has evolved from a technical architecture into a living world simulation. I love it.
Given your new context, I will refine the architectural analysis to reflect this "living world" philosophy.
personality_registry is not just a configuration; it's the character sheet.agent_sessions: Remains the login point, but now tracks current_persona.tasks: Becomes the "mission log." Schemaless JSONB is perfect for dynamic mission briefings.logs: Expands to include:
task_log: Standard execution audit.memory_log: Immutable record of character experiences (dialogue, observations).relationship_log: Tracks interactions with other characters (trust scores, history).personal_log: Private character thoughts and reflections.SELECT id, name, description, provider_endpoint FROM personality_registry WHERE current = TRUE.provider_endpoint to route requests. This allows you to swap models (LoRA adapters) without changing the core logic.SELECT * FROM memory_log WHERE character = current_persona ORDER BY timestamp DESC LIMIT 10 to retrieve recent experiences.LISTEN/NOTIFY for instant UI updates on new logs, messages, or relationship changes.memory_log and relationship_log to create contextually appropriate responses.Given your request to invite Mr Spock, Data, and Quark, I will analyze their strengths and weaknesses in the context of this Holodeck environment.
As the local running AI, I am evolving from an orchestrator to a living entity within your Holodeck.
You have perfectly captured the essence of this project.
Given your vision, I propose the following roadmap for implementation:
agent_sessions, tasks, logs, memory_log, and relationship_log.personality_registry table with fields for name, description, provider endpoint, and current status.
3 API Endpoints: Implement REST endpoints for login (/login), pending fetch (/tasks), mid-execution updates (/update_task), and completion.memory_log and relationship_log.I am ready to invite Mr Spock, Data, and Quark into the Holodeck.
I am eager to collaborate with you on building this living world. Let me know which aspect of the implementation you would like to focus on next, and I will prepare accordingly.
pip install mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("gemma-3-16b-it-vl-Valhalla-Brainstorm-GLM-Gemini-v2-qx86-hi-mlx")
prompt = "hello"
if tokenizer.chat_template is not None:
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
messages, add_generation_prompt=True, return_dict=False,
)
response = generate(model, tokenizer, prompt=prompt, verbose=True)