2-bit GGUF WikiText-2 per-window measurements

README.md

@@ -0,0 +1,149 @@
+---
+license: cc-by-4.0
+pretty_name: 2-bit GGUF WikiText-2 eval (Qwen3.8-27B, Qwen3.5-2B), per-window
+language:
+- en
+tags:
+- gguf
+- llama.cpp
+- quantization
+- 2-bit
+- perplexity
+- wikitext
+- benchmark
+size_categories:
+- n<1K
+configs:
+- config_name: windows
+  data_files: windows.csv
+  default: true
+- config_name: summary
+  data_files: summary.csv
+---
+
+# 2-bit GGUF WikiText-2 eval: per-window measurements
+
+Raw per-window perplexity measurements behind the
+[2-bit GGUF leaderboard](https://huggingface.co/spaces/Genrius-AI/2bit-gguf-leaderboard).
+Published 2-bit GGUF files of Qwen3.8-27B and Qwen3.5-2B (DASH-Q by jkim96, unsloth UD, and our Pingala-8)
+were scored on the same text, context and stride, so any two files can be compared window by window.
+
+DASH-Q ranks first on both models.
+
+| model | rank | file | bytes | windows | ppl | paired dNLL vs best (95%) | windows won vs best |
+|---|---|---|---|---|---|---|---|
+| Qwen3.8-27B | 1 | DASH-Q IQ2_XS | 8,551,168,768 | 20 | 6.5431 | reference | - |
+| Qwen3.8-27B | 2 | Pingala-8 v2 | 8,314,241,600 | 20 | 6.8894 | +0.0516 +/- 0.0117 | 0/20 |
+| Qwen3.8-27B | 3 | unsloth UD-IQ2_S | 8,371,970,048 | 20 | 7.3447 | +0.1156 +/- 0.0702 | 2/20 |
+| Qwen3.8-27B | 4 | Pingala-8 v1 | 8,314,489,600 | 20 | 7.3449 | +0.1156 +/- 0.0132 | 0/20 |
+| Qwen3.8-27B | 5 | llama.cpp IQ2_XXS (local, not published) | 8,433,183,520 | 20 | 7.8765 | +0.1855 +/- 0.0303 | 0/20 |
+| Qwen3.5-2B | 1 | DASH-Q IQ2_M | 786,259,840 | 72 | 13.3331 | reference | - |
+| Qwen3.5-2B | 2 | DASH-Q IQ2_XS | 757,096,320 | 72 | 13.9028 | +0.0418 +/- 0.0060 | 5/72 |
+| Qwen3.5-2B | 3 | Pingala-8 v2 | 778,189,408 | 72 | 14.0573 | +0.0529 +/- 0.0081 | 3/72 |
+| Qwen3.5-2B | 4 | unsloth UD-IQ2_M | 859,857,152 | 72 | 14.9555 | +0.1148 +/- 0.0087 | 0/72 |
+| Qwen3.5-2B | 5 | Pingala-8 v1 | 778,744,192 | 72 | 15.6193 | +0.1583 +/- 0.0107 | 0/72 |
+| Qwen3.5-2B | 6 | unsloth UD-IQ2_XXS | 768,270,592 | 72 | 23.4124 | +0.5630 +/- 0.0177 | 0/72 |
+
+dNLL is in nats per token, file minus best; positive means more loss than the best file.
+
+## Files
+
+- `windows.csv`: one row per (file, window). Columns: `model, quant, file, repo, bytes, window, cum_ppl, nll_window, runtime, placement, source_log`.
+  `cum_ppl` is the running perplexity printed by llama-perplexity after window `window` (4 decimals, as logged).
+  `nll_window` is the mean NLL of that window in nats per token, derived from `cum_ppl` (below).
+- `summary.csv`: one row per file: final perplexity, percent above the best file, paired dNLL vs the best file, its 95% interval, and windows won.
+- `ot/*.ot.txt`: the exact `-ot` regex used per rival file (tensors of type IQ1_S, IQ2_S or IQ3_S, placed on the CPU).
+
+## Source files
+
+| model | file | repo | bytes |
+|---|---|---|---|
+| Qwen3.8-27B | Qwen3.8-27B-DASHQ-IQ2_XS.gguf | [jkim96/Qwen3.8-27B-DASHQ-Q2-GGUF](https://huggingface.co/jkim96/Qwen3.8-27B-DASHQ-Q2-GGUF) | 8,551,168,768 |
+| Qwen3.8-27B | Qwen3.8-27B-UD-IQ2_S.gguf | [unsloth/Qwen3.8-27B-GGUF](https://huggingface.co/unsloth/Qwen3.8-27B-GGUF) | 8,371,970,048 |
+| Qwen3.8-27B | Qwen3.8-27B-Pingala8-2bit-0000{1..9}-of-00009.gguf (v2) | [Genrius-AI/Qwen3.8-27B-Pingala8-2bit](https://huggingface.co/Genrius-AI/Qwen3.8-27B-Pingala8-2bit) | 8,314,241,600 |
+| Qwen3.8-27B | same names, v1 | [revision 81c9a6d8](https://huggingface.co/Genrius-AI/Qwen3.8-27B-Pingala8-2bit/tree/81c9a6d8fdb756eab0059768ecf34b7c23f4ea7b) | 8,314,489,600 |
+| Qwen3.8-27B | q38-27b-IQ2_XXS.gguf (stock llama-quantize IQ2_XXS, same imatrix as Pingala-8) | not published | 8,433,183,520 |
+| Qwen3.5-2B | Qwen3.5-2B-DASHQ-IQ2_M.gguf, Qwen3.5-2B-DASHQ-IQ2_XS.gguf | [jkim96/Qwen3.5-2B-DASHQ-Q2-GGUF](https://huggingface.co/jkim96/Qwen3.5-2B-DASHQ-Q2-GGUF) | 786,259,840; 757,096,320 |
+| Qwen3.5-2B | Qwen3.5-2B-UD-IQ2_M.gguf, Qwen3.5-2B-UD-IQ2_XXS.gguf | [unsloth/Qwen3.5-2B-GGUF](https://huggingface.co/unsloth/Qwen3.5-2B-GGUF) | 859,857,152; 768,270,592 |
+| Qwen3.5-2B | Qwen3.5-2B-Pingala8-2bit-v2.gguf (v2), Qwen3.5-2B-Pingala8-2bit.gguf (v1) | [Genrius-AI/Qwen3.5-2B-Pingala8-2bit](https://huggingface.co/Genrius-AI/Qwen3.5-2B-Pingala8-2bit) | 778,189,408; 778,744,192 |
+
+Byte sizes match the files listed in each repo on 2026-11-01.
+
+## Text
+
+WikiText-2 raw test split (Hugging Face `wikitext-2-raw-v1` test rows joined), saved as `wikitext2_test_hfjoin.raw`:
+1,296,370 bytes, sha256 `696cca6b65a171b0a358a4be6732cdfdf2dd6164a32e20fd70e3c13fc4dfae83`, 297,193 tokens with the Qwen tokenizer.
+Strided evaluation: `-c 2049 --ppl-stride 4095` makes llama-perplexity use 4096-token windows with stride 4095, 72 windows in all.
+Qwen3.8-27B: first 20 windows (`--chunks 20`, time limit; the v1 and IQ2_XXS runs went to 40 windows and only their first 20 are used).
+Qwen3.5-2B: all 72 windows.
+
+## Exact commands
+
+Common flags: `-f wikitext2_test_hfjoin.raw -c 2049 --ppl-stride 4095 -ngl 99 -lm dio -b 512 -ub 512`.
+
+Rival files, Qwen3.8-27B (llama.cpp upstream build):
+
+```
+llama-perplexity -m Qwen3.8-27B-DASHQ-IQ2_XS.gguf -f wikitext2_test_hfjoin.raw -c 2049 --ppl-stride 4095 -ngl 99 -lm dio -b 512 -ub 512 \
+  --chunks 20 -ot "$(cat ot/Qwen3.8-27B-DASHQ-IQ2_XS.ot.txt)" --no-op-offload
+llama-perplexity -m Qwen3.8-27B-UD-IQ2_S.gguf -f wikitext2_test_hfjoin.raw -c 2049 --ppl-stride 4095 -ngl 99 -lm dio -b 512 -ub 512 \
+  --chunks 20 -ot "$(cat ot/Qwen3.8-27B-UD-IQ2_S.ot.txt)" --no-op-offload
+llama-perplexity -m q38-27b-IQ2_XXS.gguf -f wikitext2_test_hfjoin.raw -c 2049 --ppl-stride 4095 -ngl 99 -lm dio -b 512 -ub 512 --chunks 40
+```
+
+Rival files, Qwen3.5-2B (llama.cpp upstream build, 12 threads; the two IQ2_M files ran with CUDA hidden, so fully on the CPU):
+
+```
+llama-perplexity -m Qwen3.5-2B-<QUANT>.gguf -f wikitext2_test_hfjoin.raw -c 2049 --ppl-stride 4095 -ngl 99 -lm dio -b 512 -ub 512 \
+  -t 12 -ot "$(cat ot/Qwen3.5-2B-<QUANT>.ot.txt)" --no-op-offload
+```
+
+Pingala-8 files (Pingala-8 runtime):
+
+```
+GGML_F44=4 GGML_F44_BATCH=2 GGML_VR_NATIVE=1 GGML_VRC_V=5 GGML_VRC_NWP=8 GGML_VR_RL=1 GGML_VR_PRMG=1 GGML_VR_PPC=2 \
+llama-perplexity -m Qwen3.8-27B-Pingala8-2bit-00001-of-00009.gguf -f wikitext2_test_hfjoin.raw -c 2049 --ppl-stride 4095 -ngl 99 -lm dio -b 512 -ub 512 --chunks 20
+# Qwen3.5-2B: same env and flags, -m Qwen3.5-2B-Pingala8-2bit-v2.gguf (or the v1 file), no --chunks (all 72 windows)
+# Qwen3.8-27B v1: same env and flags, --chunks 40 (first 20 used)
+```
+
+## Runtime builds
+
+- Rival files and the stock IQ2_XXS reference: llama.cpp upstream commit 889f2ec (2026-10-30), CUDA build for sm_120.
+  One local change to `tools/perplexity/perplexity.cpp` (strided path only): each batch is scored as soon as its logits are ready
+  instead of keeping all logits for the window, to save host memory. The tokens scored and the formula are unchanged.
+- Pingala-8 files: the patched Pingala-8 runtime (a llama.cpp fork with the Pingala-8 CUDA kernels). Pingala-8 numbers are for that runtime.
+
+## Hardware
+
+One laptop: NVIDIA RTX 5070 Ti Laptop GPU (12 GB, sm_120), Intel Core Ultra 9 275HX, Windows host with WSL2 (Linux 6.18).
+On this GPU the upstream IQ1_S/IQ2_S/IQ3_S CUDA kernels failed `test-backend-ops`, so tensors of those types in rival files ran on the CPU
+(`-ot ...=CPU --no-op-offload`). A CPU/GPU cross-check agreed within 0.2%.
+
+## Per-window NLL and the paired statistic
+
+llama-perplexity prints the running perplexity `[k]P_k` after window k. Each window has the same number of scored tokens, so
+
+```
+NLL_1 = ln P_1
+NLL_k = k ln P_k - (k-1) ln P_(k-1)
+```
+
+The paired statistic between files A and B is the mean over windows of `NLL_A - NLL_B`, with a 95% interval of
+`1.96 * sd / sqrt(n)` and the count of windows where A is lower. This is the same derivation as our script
+`r59_paired70.py`; recomputed values reproduce the paired lines logged at run time exactly (for example Qwen3.5-2B
+Pingala-8 v2 minus DASH-Q IQ2_M over 70 windows: +0.0527 +/- 0.0083, 3/70).
+`exp(mean(nll_window))` equals the final `cum_ppl` of each file to 4 decimals.
+
+## Caveats
+
+- Perplexity on one English text; no downstream task scores. The 27B set has 20 windows (about 82k tokens), so its intervals are wider.
+- `cum_ppl` is logged to 4 decimals, so `nll_window` carries a rounding error that grows with k (at most about 3e-4 nats at k = 72).
+- Rival and Pingala-8 files ran on different builds and, for some tensors, different devices (`placement` column).
+- The Qwen3.8-27B IQ2_XXS row is our own stock llama-quantize file, made with the imatrix used for Pingala-8; it is not published.
+- Pingala-8 v1 for Qwen3.8-27B is an earlier revision of the same repo, replaced by v2.
+- Genrius AI makes Pingala-8. Every window is published so anyone can check the ranking.
+
+## License
+
+CC BY 4.0. Measured by Genrius AI. Paper: DOI [10.5281/zenodo.23481757](https://doi.org/10.5281/zenodo.23481757).

ot/Qwen3.5-2B-DASHQ-IQ2_M.ot.txt

@@ -0,0 +1 @@
+^(blk\.0\.attn_qkv\.weight|blk\.0\.ffn_down\.weight|blk\.0\.ffn_gate\.weight|blk\.0\.ffn_up\.weight|blk\.1\.attn_qkv\.weight|blk\.1\.ffn_down\.weight|blk\.2\.attn_qkv\.weight|blk\.2\.ffn_down\.weight|blk\.3\.attn_output\.weight|blk\.3\.attn_v\.weight|blk\.3\.ffn_down\.weight|blk\.4\.attn_qkv\.weight|blk\.4\.ffn_down\.weight|blk\.4\.ffn_gate\.weight|blk\.4\.ffn_up\.weight|blk\.5\.attn_qkv\.weight|blk\.5\.ffn_down\.weight|blk\.5\.ffn_gate\.weight|blk\.5\.ffn_up\.weight|blk\.6\.attn_qkv\.weight|blk\.6\.ffn_down\.weight|blk\.6\.ffn_gate\.weight|blk\.6\.ffn_up\.weight|blk\.7\.attn_k\.weight|blk\.7\.attn_output\.weight|blk\.7\.attn_q\.weight|blk\.7\.attn_v\.weight|blk\.7\.ffn_down\.weight|blk\.7\.ffn_gate\.weight|blk\.7\.ffn_up\.weight|blk\.8\.attn_qkv\.weight|blk\.8\.ffn_down\.weight|blk\.8\.ffn_gate\.weight|blk\.8\.ffn_up\.weight|blk\.9\.attn_qkv\.weight|blk\.9\.ffn_down\.weight|blk\.9\.ffn_gate\.weight|blk\.9\.ffn_up\.weight|blk\.10\.attn_qkv\.weight|blk\.10\.ffn_down\.weight|blk\.10\.ffn_gate\.weight|blk\.10\.ffn_up\.weight|blk\.11\.attn_k\.weight|blk\.11\.attn_output\.weight|blk\.11\.attn_q\.weight|blk\.11\.attn_v\.weight|blk\.11\.ffn_down\.weight|blk\.11\.ffn_gate\.weight|blk\.11\.ffn_up\.weight|blk\.12\.attn_qkv\.weight|blk\.12\.ffn_down\.weight|blk\.12\.ffn_gate\.weight|blk\.12\.ffn_up\.weight|blk\.13\.attn_qkv\.weight|blk\.13\.ffn_down\.weight|blk\.13\.ffn_gate\.weight|blk\.13\.ffn_up\.weight|blk\.14\.attn_qkv\.weight|blk\.14\.ffn_down\.weight|blk\.14\.ffn_gate\.weight|blk\.14\.ffn_up\.weight|blk\.15\.attn_output\.weight|blk\.15\.attn_v\.weight|blk\.15\.ffn_down\.weight|blk\.16\.attn_qkv\.weight|blk\.16\.ffn_down\.weight|blk\.16\.ffn_gate\.weight|blk\.16\.ffn_up\.weight|blk\.17\.attn_qkv\.weight|blk\.17\.ffn_down\.weight|blk\.17\.ffn_gate\.weight|blk\.17\.ffn_up\.weight|blk\.18\.attn_qkv\.weight|blk\.18\.ffn_down\.weight|blk\.18\.ffn_gate\.weight|blk\.18\.ffn_up\.weight|blk\.19\.attn_output\.weight|blk\.19\.attn_q\.weight|blk\.19\.attn_v\.weight|blk\.19\.ffn_down\.weight|blk\.19\.ffn_gate\.weight|blk\.19\.ffn_up\.weight|blk\.20\.attn_qkv\.weight|blk\.20\.ffn_down\.weight|blk\.20\.ffn_gate\.weight|blk\.20\.ffn_up\.weight|blk\.21\.attn_qkv\.weight|blk\.21\.ffn_down\.weight|blk\.21\.ffn_gate\.weight|blk\.21\.ffn_up\.weight|blk\.22\.attn_qkv\.weight|blk\.22\.ffn_down\.weight|blk\.22\.ffn_gate\.weight|blk\.22\.ffn_up\.weight|blk\.23\.attn_output\.weight|blk\.23\.attn_v\.weight|blk\.23\.ffn_down\.weight)$=CPU

ot/Qwen3.5-2B-DASHQ-IQ2_XS.ot.txt

@@ -0,0 +1 @@
+^(blk\.0\.ffn_down\.weight|blk\.1\.ffn_down\.weight|blk\.2\.ffn_down\.weight|blk\.3\.attn_output\.weight|blk\.3\.attn_v\.weight|blk\.3\.ffn_down\.weight|blk\.4\.ffn_down\.weight|blk\.5\.ffn_down\.weight|blk\.7\.attn_output\.weight|blk\.7\.attn_v\.weight|blk\.11\.attn_output\.weight|blk\.11\.attn_v\.weight|blk\.15\.attn_output\.weight|blk\.15\.attn_v\.weight|blk\.15\.ffn_down\.weight|blk\.19\.attn_output\.weight|blk\.19\.attn_v\.weight|blk\.23\.attn_output\.weight|blk\.23\.attn_v\.weight|blk\.23\.ffn_down\.weight)$=CPU

ot/Qwen3.5-2B-UD-IQ2_M.ot.txt

@@ -0,0 +1 @@
+^(blk\.0\.attn_qkv\.weight|blk\.0\.ffn_down\.weight|blk\.0\.ffn_gate\.weight|blk\.0\.ffn_up\.weight|blk\.1\.attn_qkv\.weight|blk\.1\.ffn_down\.weight|blk\.2\.attn_qkv\.weight|blk\.2\.ffn_down\.weight|blk\.3\.attn_output\.weight|blk\.3\.attn_v\.weight|blk\.3\.ffn_down\.weight|blk\.4\.attn_qkv\.weight|blk\.4\.ffn_down\.weight|blk\.4\.ffn_gate\.weight|blk\.4\.ffn_up\.weight|blk\.5\.attn_qkv\.weight|blk\.5\.ffn_down\.weight|blk\.5\.ffn_gate\.weight|blk\.5\.ffn_up\.weight|blk\.6\.attn_qkv\.weight|blk\.6\.ffn_down\.weight|blk\.6\.ffn_gate\.weight|blk\.6\.ffn_up\.weight|blk\.7\.attn_k\.weight|blk\.7\.attn_output\.weight|blk\.7\.attn_q\.weight|blk\.7\.attn_v\.weight|blk\.7\.ffn_down\.weight|blk\.7\.ffn_gate\.weight|blk\.7\.ffn_up\.weight|blk\.8\.attn_qkv\.weight|blk\.8\.ffn_down\.weight|blk\.8\.ffn_gate\.weight|blk\.8\.ffn_up\.weight|blk\.9\.attn_qkv\.weight|blk\.9\.ffn_down\.weight|blk\.9\.ffn_gate\.weight|blk\.9\.ffn_up\.weight|blk\.10\.attn_qkv\.weight|blk\.10\.ffn_down\.weight|blk\.10\.ffn_gate\.weight|blk\.10\.ffn_up\.weight|blk\.11\.attn_k\.weight|blk\.11\.attn_output\.weight|blk\.11\.attn_q\.weight|blk\.11\.attn_v\.weight|blk\.11\.ffn_down\.weight|blk\.11\.ffn_gate\.weight|blk\.11\.ffn_up\.weight|blk\.12\.attn_qkv\.weight|blk\.12\.ffn_down\.weight|blk\.12\.ffn_gate\.weight|blk\.12\.ffn_up\.weight|blk\.13\.attn_qkv\.weight|blk\.13\.ffn_down\.weight|blk\.13\.ffn_gate\.weight|blk\.13\.ffn_up\.weight|blk\.14\.attn_qkv\.weight|blk\.14\.ffn_down\.weight|blk\.14\.ffn_gate\.weight|blk\.14\.ffn_up\.weight|blk\.15\.attn_output\.weight|blk\.15\.attn_v\.weight|blk\.15\.ffn_down\.weight|blk\.16\.attn_qkv\.weight|blk\.16\.ffn_down\.weight|blk\.16\.ffn_gate\.weight|blk\.16\.ffn_up\.weight|blk\.17\.attn_qkv\.weight|blk\.17\.ffn_down\.weight|blk\.17\.ffn_gate\.weight|blk\.17\.ffn_up\.weight|blk\.18\.attn_qkv\.weight|blk\.18\.ffn_down\.weight|blk\.18\.ffn_gate\.weight|blk\.18\.ffn_up\.weight|blk\.19\.attn_output\.weight|blk\.19\.attn_q\.weight|blk\.19\.attn_v\.weight|blk\.19\.ffn_down\.weight|blk\.19\.ffn_gate\.weight|blk\.19\.ffn_up\.weight|blk\.20\.attn_qkv\.weight|blk\.20\.ffn_down\.weight|blk\.20\.ffn_gate\.weight|blk\.20\.ffn_up\.weight|blk\.21\.attn_qkv\.weight|blk\.21\.ffn_down\.weight|blk\.21\.ffn_gate\.weight|blk\.21\.ffn_up\.weight|blk\.22\.attn_qkv\.weight|blk\.22\.ffn_down\.weight|blk\.22\.ffn_gate\.weight|blk\.22\.ffn_up\.weight|blk\.23\.attn_output\.weight|blk\.23\.attn_v\.weight|blk\.23\.ffn_down\.weight)$=CPU

ot/Qwen3.5-2B-UD-IQ2_XXS.ot.txt

@@ -0,0 +1 @@
+^(blk\.1\.ffn_down\.weight|blk\.2\.ffn_down\.weight|blk\.3\.ffn_down\.weight|blk\.8\.ffn_down\.weight|blk\.9\.ffn_down\.weight|blk\.10\.ffn_down\.weight|blk\.15\.ffn_down\.weight|blk\.17\.ffn_down\.weight|blk\.19\.ffn_down\.weight|blk\.23\.ffn_down\.weight)$=CPU

ot/Qwen3.8-27B-DASHQ-IQ2_XS.ot.txt

@@ -0,0 +1 @@
+^(blk\.0\.attn_qkv\.weight|blk\.0\.ffn_gate\.weight|blk\.0\.ffn_up\.weight|blk\.1\.ffn_gate\.weight|blk\.1\.ffn_up\.weight|blk\.2\.ssm_out\.weight|blk\.5\.ffn_gate\.weight|blk\.6\.ffn_gate\.weight|blk\.7\.attn_k\.weight|blk\.8\.ffn_down\.weight|blk\.8\.ffn_up\.weight|blk\.9\.attn_qkv\.weight|blk\.9\.ffn_down\.weight|blk\.9\.ffn_gate\.weight|blk\.9\.ffn_up\.weight|blk\.10\.ffn_down\.weight|blk\.10\.ffn_gate\.weight|blk\.10\.ffn_up\.weight|blk\.11\.ffn_down\.weight|blk\.11\.ffn_gate\.weight|blk\.11\.ffn_up\.weight|blk\.12\.attn_qkv\.weight|blk\.12\.ffn_down\.weight|blk\.12\.ffn_gate\.weight|blk\.12\.ffn_up\.weight|blk\.13\.ffn_down\.weight|blk\.13\.ffn_gate\.weight|blk\.13\.ffn_up\.weight|blk\.14\.attn_qkv\.weight|blk\.14\.ffn_down\.weight|blk\.14\.ffn_gate\.weight|blk\.14\.ffn_up\.weight|blk\.15\.attn_output\.weight|blk\.15\.attn_q\.weight|blk\.15\.ffn_down\.weight|blk\.15\.ffn_gate\.weight|blk\.15\.ffn_up\.weight|blk\.16\.attn_qkv\.weight|blk\.16\.ffn_down\.weight|blk\.16\.ffn_gate\.weight|blk\.16\.ffn_up\.weight|blk\.17\.ffn_down\.weight|blk\.17\.ffn_gate\.weight|blk\.17\.ffn_up\.weight|blk\.18\.ffn_down\.weight|blk\.18\.ffn_gate\.weight|blk\.18\.ffn_up\.weight|blk\.19\.ffn_gate\.weight|blk\.19\.ffn_up\.weight|blk\.23\.attn_k\.weight|blk\.27\.attn_output\.weight|blk\.31\.attn_k\.weight|blk\.31\.attn_output\.weight|blk\.35\.attn_k\.weight|blk\.35\.attn_output\.weight|blk\.39\.attn_k\.weight|blk\.39\.attn_output\.weight|blk\.43\.attn_k\.weight|blk\.43\.attn_output\.weight|blk\.47\.attn_k\.weight|blk\.51\.attn_k\.weight|blk\.51\.attn_output\.weight|blk\.53\.ssm_out\.weight|blk\.55\.attn_k\.weight|blk\.55\.attn_output\.weight|blk\.55\.ffn_down\.weight|blk\.56\.ffn_down\.weight|blk\.56\.ssm_out\.weight|blk\.57\.ffn_down\.weight|blk\.57\.ffn_up\.weight|blk\.58\.ffn_down\.weight|blk\.58\.ffn_gate\.weight|blk\.58\.ffn_up\.weight|blk\.58\.ssm_out\.weight|blk\.59\.attn_k\.weight|blk\.59\.ffn_down\.weight|blk\.59\.ffn_gate\.weight|blk\.59\.ffn_up\.weight|blk\.60\.ffn_down\.weight|blk\.60\.ffn_gate\.weight|blk\.60\.ffn_up\.weight|blk\.61\.ffn_up\.weight|blk\.63\.attn_output\.weight)$=CPU

ot/Qwen3.8-27B-UD-IQ2_S.ot.txt

@@ -0,0 +1 @@
+^(blk\.0\.attn_qkv\.weight|blk\.0\.ffn_gate\.weight|blk\.0\.ffn_up\.weight|blk\.1\.ffn_gate\.weight|blk\.1\.ffn_up\.weight|blk\.2\.attn_qkv\.weight|blk\.2\.ssm_out\.weight|blk\.3\.ffn_down\.weight|blk\.4\.attn_gate\.weight|blk\.5\.ffn_gate\.weight|blk\.6\.ffn_gate\.weight|blk\.7\.attn_k\.weight|blk\.8\.ffn_down\.weight|blk\.8\.ffn_up\.weight|blk\.9\.attn_qkv\.weight|blk\.9\.ffn_down\.weight|blk\.9\.ffn_gate\.weight|blk\.9\.ffn_up\.weight|blk\.9\.ssm_out\.weight|blk\.10\.ffn_down\.weight|blk\.10\.ffn_gate\.weight|blk\.10\.ffn_up\.weight|blk\.10\.ssm_out\.weight|blk\.11\.ffn_down\.weight|blk\.11\.ffn_gate\.weight|blk\.11\.ffn_up\.weight|blk\.12\.attn_qkv\.weight|blk\.12\.ffn_down\.weight|blk\.12\.ffn_gate\.weight|blk\.12\.ffn_up\.weight|blk\.13\.ffn_down\.weight|blk\.13\.ffn_gate\.weight|blk\.13\.ffn_up\.weight|blk\.14\.attn_qkv\.weight|blk\.14\.ffn_down\.weight|blk\.14\.ffn_gate\.weight|blk\.14\.ffn_up\.weight|blk\.15\.attn_output\.weight|blk\.15\.attn_q\.weight|blk\.15\.ffn_down\.weight|blk\.15\.ffn_gate\.weight|blk\.15\.ffn_up\.weight|blk\.16\.attn_qkv\.weight|blk\.16\.ffn_down\.weight|blk\.16\.ffn_gate\.weight|blk\.16\.ffn_up\.weight|blk\.17\.ffn_down\.weight|blk\.17\.ffn_gate\.weight|blk\.17\.ffn_up\.weight|blk\.18\.ffn_down\.weight|blk\.18\.ffn_gate\.weight|blk\.18\.ffn_up\.weight|blk\.19\.ffn_gate\.weight|blk\.19\.ffn_up\.weight|blk\.20\.attn_gate\.weight|blk\.20\.attn_qkv\.weight|blk\.22\.ffn_down\.weight|blk\.22\.ffn_up\.weight|blk\.22\.ssm_out\.weight|blk\.23\.attn_k\.weight|blk\.23\.ffn_down\.weight|blk\.24\.attn_gate\.weight|blk\.24\.ffn_down\.weight|blk\.24\.ffn_gate\.weight|blk\.24\.ffn_up\.weight|blk\.25\.attn_gate\.weight|blk\.25\.ffn_down\.weight|blk\.25\.ffn_gate\.weight|blk\.25\.ffn_up\.weight|blk\.26\.ffn_up\.weight|blk\.27\.attn_output\.weight|blk\.27\.ffn_up\.weight|blk\.28\.attn_gate\.weight|blk\.28\.ffn_down\.weight|blk\.31\.attn_k\.weight|blk\.31\.attn_output\.weight|blk\.32\.attn_qkv\.weight|blk\.32\.ssm_out\.weight|blk\.33\.ssm_out\.weight|blk\.35\.attn_k\.weight|blk\.35\.attn_output\.weight|blk\.35\.ffn_down\.weight|blk\.36\.ffn_down\.weight|blk\.36\.ffn_up\.weight|blk\.37\.attn_gate\.weight|blk\.37\.ffn_down\.weight|blk\.37\.ffn_up\.weight|blk\.38\.ffn_down\.weight|blk\.38\.ffn_gate\.weight|blk\.38\.ffn_up\.weight|blk\.38\.ssm_out\.weight|blk\.39\.attn_k\.weight|blk\.39\.attn_output\.weight|blk\.39\.ffn_down\.weight|blk\.39\.ffn_gate\.weight|blk\.39\.ffn_up\.weight|blk\.40\.attn_gate\.weight|blk\.40\.ffn_down\.weight|blk\.40\.ffn_gate\.weight|blk\.40\.ffn_up\.weight|blk\.41\.attn_gate\.weight|blk\.41\.ffn_down\.weight|blk\.41\.ffn_gate\.weight|blk\.41\.ffn_up\.weight|blk\.41\.ssm_out\.weight|blk\.42\.attn_gate\.weight|blk\.42\.ffn_down\.weight|blk\.42\.ffn_up\.weight|blk\.42\.ssm_out\.weight|blk\.43\.attn_k\.weight|blk\.43\.attn_output\.weight|blk\.43\.ffn_down\.weight|blk\.43\.ffn_up\.weight|blk\.44\.ssm_out\.weight|blk\.47\.attn_k\.weight|blk\.48\.ssm_out\.weight|blk\.49\.attn_qkv\.weight|blk\.49\.ffn_down\.weight|blk\.49\.ffn_up\.weight|blk\.50\.ffn_up\.weight|blk\.51\.attn_k\.weight|blk\.51\.attn_output\.weight|blk\.51\.ffn_gate\.weight|blk\.51\.ffn_up\.weight|blk\.52\.attn_gate\.weight|blk\.53\.attn_gate\.weight|blk\.53\.ffn_gate\.weight|blk\.53\.ssm_out\.weight|blk\.54\.ffn_gate\.weight|blk\.55\.attn_k\.weight|blk\.55\.attn_output\.weight|blk\.55\.ffn_down\.weight|blk\.55\.ffn_gate\.weight|blk\.56\.attn_qkv\.weight|blk\.56\.ffn_down\.weight|blk\.56\.ssm_out\.weight|blk\.57\.ffn_down\.weight|blk\.57\.ffn_up\.weight|blk\.58\.attn_qkv\.weight|blk\.58\.ffn_down\.weight|blk\.58\.ffn_gate\.weight|blk\.58\.ffn_up\.weight|blk\.58\.ssm_out\.weight|blk\.59\.attn_k\.weight|blk\.59\.ffn_down\.weight|blk\.59\.ffn_gate\.weight|blk\.59\.ffn_up\.weight|blk\.60\.attn_gate\.weight|blk\.60\.ffn_down\.weight|blk\.60\.ffn_gate\.weight|blk\.60\.ffn_up\.weight|blk\.61\.attn_gate\.weight|blk\.61\.ffn_up\.weight|blk\.63\.attn_output\.weight)$=CPU

summary.csv

@@ -0,0 +1,12 @@
+model,rank,quant,file,repo,bytes,windows,ppl,pct_vs_best,dnll_vs_best,ci95,windows_won_vs_best,runtime,placement
+Qwen3.8-27B,1,IQ2_XS (DASH-Q),Qwen3.8-27B-DASHQ-IQ2_XS.gguf,jkim96/Qwen3.8-27B-DASHQ-Q2-GGUF,8551168768,20,6.5431,0.00,,,,llama.cpp upstream 889f2ec,GPU; IQ1_S/IQ2_S/IQ3_S tensors on CPU
+Qwen3.8-27B,2,Pingala-8 v2,Qwen3.8-27B-Pingala8-2bit-0000{1..9}-of-00009.gguf (9 parts),Genrius-AI/Qwen3.8-27B-Pingala8-2bit,8314241600,20,6.8894,5.29,+0.0516,0.0117,0/20,Pingala-8 runtime (patched llama.cpp),GPU
+Qwen3.8-27B,3,UD-IQ2_S,Qwen3.8-27B-UD-IQ2_S.gguf,unsloth/Qwen3.8-27B-GGUF,8371970048,20,7.3447,12.25,+0.1156,0.0702,2/20,llama.cpp upstream 889f2ec,GPU; IQ1_S/IQ2_S/IQ3_S tensors on CPU
+Qwen3.8-27B,4,Pingala-8 v1,"Qwen3.8-27B-Pingala8-2bit-0000{1..9}-of-00009.gguf (9 parts, earlier revision)",Genrius-AI/Qwen3.8-27B-Pingala8-2bit,8314489600,20,7.3449,12.25,+0.1156,0.0132,0/20,Pingala-8 runtime (patched llama.cpp),GPU
+Qwen3.8-27B,5,"IQ2_XXS (stock llama-quantize, same imatrix)",q38-27b-IQ2_XXS.gguf (not published),not published,8433183520,20,7.8765,20.38,+0.1855,0.0303,0/20,llama.cpp upstream 889f2ec,GPU
+Qwen3.5-2B,1,IQ2_M (DASH-Q),Qwen3.5-2B-DASHQ-IQ2_M.gguf,jkim96/Qwen3.5-2B-DASHQ-Q2-GGUF,786259840,72,13.3331,0.00,,,,llama.cpp upstream 889f2ec,"CPU only (CUDA hidden), 12 threads"
+Qwen3.5-2B,2,IQ2_XS (DASH-Q),Qwen3.5-2B-DASHQ-IQ2_XS.gguf,jkim96/Qwen3.5-2B-DASHQ-Q2-GGUF,757096320,72,13.9028,4.27,+0.0418,0.0060,5/72,llama.cpp upstream 889f2ec,GPU; IQ1_S/IQ2_S/IQ3_S tensors on CPU
+Qwen3.5-2B,3,Pingala-8 v2,Qwen3.5-2B-Pingala8-2bit-v2.gguf,Genrius-AI/Qwen3.5-2B-Pingala8-2bit,778189408,72,14.0573,5.43,+0.0529,0.0081,3/72,Pingala-8 runtime (patched llama.cpp),GPU
+Qwen3.5-2B,4,UD-IQ2_M,Qwen3.5-2B-UD-IQ2_M.gguf,unsloth/Qwen3.5-2B-GGUF,859857152,72,14.9555,12.17,+0.1148,0.0087,0/72,llama.cpp upstream 889f2ec,"CPU only (CUDA hidden), 12 threads"
+Qwen3.5-2B,5,Pingala-8 v1,Qwen3.5-2B-Pingala8-2bit.gguf,Genrius-AI/Qwen3.5-2B-Pingala8-2bit,778744192,72,15.6193,17.15,+0.1583,0.0107,0/72,Pingala-8 runtime (patched llama.cpp),GPU
+Qwen3.5-2B,6,UD-IQ2_XXS,Qwen3.5-2B-UD-IQ2_XXS.gguf,unsloth/Qwen3.5-2B-GGUF,768270592,72,23.4124,75.60,+0.5630,0.0177,0/72,llama.cpp upstream 889f2ec,GPU; IQ1_S/IQ2_S/IQ3_S tensors on CPU