Modelpublic

Mantisec/Qwen3.8-27B-TURBO-FP16

sourceHugging Faceupdated 13d agoView on Hugging Face

1likes81downloads

Model Card

Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU-fp16

FP16 conversion of `DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU`, produced by bfsquish v0.1.0.

Intended use

Optimized for FP16 inference and fine-tuning on NVIDIA V100 (Volta) GPUs, which lack native BF16 Tensor Core support. The weights have been converted from BF16 to FP16 using the effective range_checked strategy (see below).

Conversion details

Field	Value
Source model	`DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU`
Source revision	`latest`
Requested strategy	`auto`
Effective strategy	`range_checked`
Tool	`bfsquish` v0.1.0
Converted at (UTC)	`2026-09-13T11:57:16.487952+00:00`
Target hardware	NVIDIA V100 (Volta, sm_70)
Target runtime	NVIDIA V100 (Volta sm_70), FP16 Tensor Cores

Conversion strategy

Direct upcast to FP32 followed by a range-checked FP16 cast. Values outside FP16's finite range are rejected instead of clipped, and rounding error is recorded in bfsquish_conversion.json. For BF16-trained weights that already fit inside FP16's range, this avoids graph-changing rotations and preserves the source checkpoint as closely as the target dtype permits.

Transform plan

v1 strategy applied: range_checked (no v2 plan bundle was produced for this conversion).

Enhanced chat template

Prompt rendering uses the selected third-party enhanced template; model weights are unchanged by this step. The standalone and embedded forms were smoke-rendered to the same prompt before publication.

Field	Value
Selection	`latest`
Source	`peculiar-ragdoll/Qwen-Sharp-Chat-Templates`
Resolved revision	`fa3a1295882d31132770c156fced4e616b5db25d`
Template version	`qwen3.8-froggeric-v22.4.0`
Template license	`Apache-2.0`
Published forms	`chat_template.jinja` and `tokenizer_config.json#chat_template`

Reproduce the prompt-format step with --chat-template latest --chat-template-repo peculiar-ragdoll/Qwen-Sharp-Chat-Templates --chat-template-revision fa3a1295882d31132770c156fced4e616b5db25d.

vLLM warm-up profile

A version 1 advisory profile was generated at deployment/vllm/warmup-profile.v1.json after a PASS validation verdict (4 bounded cases). It is data-only and non-executable. Runtimes may ignore or override its hints; it does not select startup mode or failure policy and does not modify model weights. Its input scope is text-only: it warms text generation, not image/video/audio preprocessing, vision encoders, or multimodal fusion paths.

Numerical quality

Metric	Value
Validation verdict	PASS
Validation method	`generate`
Max abs logit diff (vs. source)	1.675378
Min cosine similarity (vs. source)	0.999265
Token agreement rate	99.80%
Inf/NaN scan	passed (no inf/nan)

Validation notes

—generation agreement passed despite logit drift: tokenagreement=99.80% (pass≥98.00%), mincos=0.999265, max_diff=1.6754

Reproducing this conversion

bash

bfsquish run \
  --model DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU \
  --output-dir ./out \
  --strategy auto

License

Inherited from the source model. Refer to the source model's license for terms of use.