empero-ai/Qwen3.5-9B-Claude-Opus-4.6-Distill | INT4 (W4A16)
#26
by INC4AI - opened
Pipeline Failure Report
Model: empero-ai/Qwen3.5-9B-Claude-Opus-4.6-Distill
Quantization Scheme: INT4 (W4A16)
Failed Phase: quantize
Run ID: Qwen3.5-9B-Claude-Opus-4.6-Distill-AutoRound-W4A16-Tuning
Error Category: auto_round_internal_error
Full Error Log
Quantizing model.layers.26: 84%|โโโโโโโโโ | 27/32 [56:11<10:38, 127.73s/it]
Quantizing model.layers.27: 84%|โโโโโโโโโ | 27/32 [56:11<10:38, 127.73s/it][38;20mquantized 7/7 layers in the block, loss iter 0: 0.063308 -> iter 187: 0.038942[0m
[38;20m2026-06-29 23:22:56 INFO device.py L1840: 'peak_ram': 31.66GB, 'peak_vram': 16.69GB[0m
Quantizing model.layers.28: 88%|โโโโโโโโโ | 28/32 [57:08<08:30, 127.73s/it][38;20mquantized 8/8 layers in the block, loss iter 0: 0.073058 -> iter 197: 0.046871[0m
[38;20m2026-06-29 23:25:20 INFO device.py L1840: 'peak_ram': 32.5GB, 'peak_vram': 16.69GB[0m
Quantizing model.layers.28: 91%|โโโโโโโโโ | 29/32 [59:32<05:49, 116.43s/it]
Quantizing model.layers.29: 91%|โโโโโโโโโ | 29/32 [59:32<05:49, 116.43s/it]
Quantizing model.layers.29: 91%|โโโโโโโโโ | 29/32 [59:49<05:49, 116.43s/it][38;20mquantized 8/8 layers in the block, loss iter 0: 0.084784 -> iter 185: 0.060510[0m
[38;20m2026-06-29 23:27:43 INFO device.py L1840: 'peak_ram': 33.31GB, 'peak_vram': 16.69GB[0m
Quantizing model.layers.29: 94%|โโโโโโโโโโ| 30/32 [1:01:55<04:05, 122.55s/it]
Quantizing model.layers.30: 94%|โโโโโโโโโโ| 30/32 [1:01:55<04:05, 122.55s/it][38;20mquantized 8/8 layers in the block, loss iter 0: 0.117076 -> iter 193: 0.081036[0m
[38;20m2026-06-29 23:30:08 INFO device.py L1840: 'peak_ram': 34.13GB, 'peak_vram': 16.69GB[0m
Quantizing model.layers.30: 97%|โโโโโโโโโโ| 31/32 [1:04:20<02:08, 128.01s/it]
Quantizing model.layers.31: 97%|โโโโโโโโโโ| 31/32 [1:04:20<02:08, 128.01s/it][38;20mquantized 7/7 layers in the block, loss iter 0: 0.215285 -> iter 173: 0.132897[0m
[38;20m2026-06-29 23:31:05 INFO device.py L1840: 'peak_ram': 34.94GB, 'peak_vram': 16.69GB[0m
Quantizing done: 100%|โโโโโโโโโโ| 32/32 [1:05:18<00:00, 128.01s/it]
Quantizing done: 100%|โโโโโโโโโโ| 32/32 [1:05:18<00:00, 122.45s/it]
[38;20m2026-06-29 23:31:06 INFO device.py L1840: 'peak_ram': 34.94GB, 'peak_vram': 16.69GB[0m
[38;20m2026-06-29 23:31:06 INFO data_driven.py L750: quantization tuning time 3918.300900220871[0m
[38;20m2026-06-29 23:31:06 INFO data_driven.py L769: Summary: quantized 248/249 in the model, unquantized layers: lm_head[0m
23:31:06 [INFO] Quantization completed in 3935.5s
23:31:06 [INFO] Saving quantized model (auto_round format)...
[33;1m2026-06-29 23:31:06 WARNING logging.py L340: some layers are skipped quantization (shape not divisible by 32): [0m
23:31:06 [ERROR] Quantization failed: invalid group reference 1 at position 22
Traceback (most recent call last):
File "/root/_work/1/s/auto_quant/phases/quantize.py", line 282, in <module>
quantize(args)
File "/root/_work/1/s/auto_quant/phases/quantize.py", line 190, in quantize
autoround.save_quantized(
File "/root/.venv/lib/python3.12/site-packages/auto_round/compressors/base.py", line 1380, in save_quantized
reverted_block_name = revert_checkpoint_conversion_mapping(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/.venv/lib/python3.12/site-packages/auto_round/utils/common.py", line 1195, in revert_checkpoint_conversion_mapping
name, n_replace = re.subn(source_pattern, target_pattern, name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/re/__init__.py", line 197, in subn
return _compile(pattern, flags).subn(repl, string, count)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/re/__init__.py", line 334, in _compile_template
return _sre.template(pattern, _parser.parse_template(repl, pattern))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/re/_parser.py", line 1069, in parse_template
addgroup(int(this[1:]), len(this) - 1)
File "/root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/re/_parser.py", line 1014, in addgroup
raise s.error("invalid group reference %d" % index, pos)
re.error: invalid group reference 1 at position 22
Auto-generated by error_analysis pipeline. cc @lvkaokao