datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
glm53-flash-fidelity-root-v1
fidelity--glm53flash.malaiwah.root.bf16
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from zai-org/GLM-5.3-Flash-BF16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same cut… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-flash-fidelity-root-v1.qwen38-27b-fidelity-root-v1
Qwen3.8-27B BF16 root fidelity dataset (hidden form)
A root capture of Qwen/Qwen3.8-27B at revision
1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 — 18 shards, no quantization_config,
a genuinely unquantized reference — over the sealed suite-v5 shard-0 token
panel (512 contexts x 2048 tokens = 1,048,064 scored positions).
What this is for
Quantization fidelity is usually reported as a KL divergence against a teacher.
If the teacher was captured on a different stack than… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qwen38-27b-fidelity-root-v1.glm52-fidelity-root-v1
fidelity--glm52.malaiwah.root.bf16
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from zai-org/GLM-5.2.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same cut as… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm52-fidelity-root-v1.qwen3-5-tiny-fidelity-root-v1
qwen3_5 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/qwen3-5-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same cut… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qwen3-5-tiny-fidelity-root-v1.deepseek-v4-tiny-fidelity-root-v1
deepseek-v4 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/deepseek-v4-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/deepseek-v4-tiny-fidelity-root-v1.qwen3-5-gguf-tiny-fidelity-root-v1
qwen35-gguf random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/qwen3-5-gguf-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qwen3-5-gguf-tiny-fidelity-root-v1.fruit-fidelity-root-v1
GLM-5.2-SIQ-Fruit — root fidelity dataset (hidden form)
This is the reference yardstick for the GLM-5.2-SIQ-Fruit family: one bf16
forward pass of
malaiwah/GLM-5.2-SIQ-Fruit-bf16
over a sealed 16-window token panel, captured at the lm_head input and sealed
so that anybody can compare a quantized capture against it without the
weights, without our infrastructure, and without re-running the reference.
Fruit is a 5.04B-parameter / 0.46B-active GLM-5.2-architecture serving proxy… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/fruit-fidelity-root-v1.glm5-next-tiny-fidelity-root-v1
glm5_next random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/glm5-next-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm5-next-tiny-fidelity-root-v1.qwen4-exp-tiny-fidelity-root-v1
qwen4_exp random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/qwen4-exp-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qwen4-exp-tiny-fidelity-root-v1.k2-horizon-tiny-fidelity-root-v1
k2-horizon random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/k2-horizon-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/k2-horizon-tiny-fidelity-root-v1.minimax-m2-tiny-fidelity-root-v1
minimax-m2 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/minimax-m2-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/minimax-m2-tiny-fidelity-root-v1.kimi-k25-tiny-fidelity-root-v1
kimi-k25 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/kimi-k25-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/kimi-k25-tiny-fidelity-root-v1.minimax-m3-tiny-fidelity-root-v1
minimax-m3 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/minimax-m3-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/minimax-m3-tiny-fidelity-root-v1.glm-moe-dsa-tiny-fidelity-root-v1
glm_moe_dsa random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/glm-moe-dsa-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm-moe-dsa-tiny-fidelity-root-v1.kimi-k3-tiny-fidelity-root-v1
kimi-k3 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/kimi-k3-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same cut… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/kimi-k3-tiny-fidelity-root-v1.glm53-fixture-0.1B-fidelity-root-v1
GLM-5.3-Flash-0.1B fixture — root fidelity dataset (hidden form)
The numbers in this dataset are meaningless as model quality.
The weights it was captured from are random. inference-optimization/GLM-5.3-Flash-0.1B-A0.1B
is an architectural fixture: it has GLM-5.3-Flash's config shape and its exact
154,880-token vocabulary, and none of its training. Nothing here says anything
about GLM-5.3-Flash, about quantization quality, or about any model.
What it is for: being a small… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-fixture-0.1B-fidelity-root-v1.spark2-5-tiny-fidelity-root-v1
spark2-5 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/spark2-5-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/spark2-5-tiny-fidelity-root-v1.fruit-fidelity-root-container-v1
fruit-fidelity-root-container-v1
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/GLM-5.2-SIQ-Fruit-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same cut… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/fruit-fidelity-root-container-v1.fruit-fidelity-root-runpod-v1
fidelity--fruit.malaiwah.root.bf16
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/GLM-5.2-SIQ-Fruit-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same cut… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/fruit-fidelity-root-runpod-v1.