- Patch vllm.lora.ops.triton_ops.utils directly where supports_pdl is defined
- Clear lru_cache before patching to prevent stale cached results
- Add fused_moe_lora_op to consumer modules list
- Use *args, **kwargs in fake function for compatibility
- Add _spec_exists helper function to reduce duplication
- Scan all GPUs for SM100 instead of just device 0
- Use loop for module patching to improve maintainability
vLLM's LoRA Triton kernels use tl.extra.cuda.gdc_wait() for PDL
optimization on SM90+ GPUs. This fails on SM100 (Blackwell) during
CUDA graph capture because Triton's pipeliner cannot handle gdc_wait
in complex kernels.
This fix:
- Detects SM100 GPUs and applies the workaround automatically
- Sets TRITON_DISABLE_PDL=1 environment variable
- Monkey-patches supports_pdl to return False in lora_expand_op and
lora_shrink_op
- Checks GitHub issue #30872 status (with 3s timeout) to auto-disable
the workaround once the upstream fix is merged
- Includes quick internet connectivity check (0.5s) to avoid delays
when offline
Fixes the error:
'tt.elementwise_inline_asm' op pipeliner doesn't know how to predicate this op
LLVM ERROR: Fatal pipeliner error
See: https://github.com/vllm-project/vllm/issues/30872
When users load a model with fast_inference=False but then try to use
vLLM-style arguments with fast_generate, they previously got confusing
errors. This adds a wrapper that detects common mistakes and provides
helpful guidance:
- Using sampling_params: explains to use HF generate args instead
- Using lora_request: explains LoRA weights are already merged
- Passing text strings: shows how to tokenize input first
Changes:
- Add make_fast_generate_wrapper to _utils.py
- Apply wrapper in llama.py when fast_inference=False
- Apply wrapper in vision.py when fast_inference=False
Gemma3 models have a large vocabulary (262144 tokens) which causes
training loss to explode when using int8 embedding quantization.
This fix auto-detects Gemma3 models and switches from int8-int4
(phone-deployment) to int4 weight-only QAT for stable training.
1. cohere.py:347-348 - Fixed wrong variable names in QK normalization.
Used `Q`/`K` but variables were named `Qn`/`Kn`. This caused NameError
when `use_qk_norm=True` (e.g., c4ai-command-r-plus models).
2. cohere.py:482 - Fixed wrong object reference in inference loop.
Used `self.mlp` but should be `decoder_layer.mlp` since we're
iterating through decoder layers. Caused AttributeError during inference.
3. falcon_h1.py:459,461 - Fixed wrong attribute names in inference path.
Used `post_attention_layernorm` and `mlp` but Falcon H1 uses
`pre_ff_layernorm` and `feed_forward`. Caused AttributeError during generation.
4. qwen3_moe.py:210 - Fixed wrong module path with incorrect capitalization.
Used `transformers.models.Qwen3Moe` but should be `transformers.models.qwen3_moe`.
Caused AttributeError when patching rotary embeddings.
5. qwen3_moe.py:239 - Fixed wrong model_patcher class.
Used `FastQwen3Model` but should be `FastQwen3MoeModel` for MoE models.
Caused incorrect patching for Qwen3 MoE models.
6. hf_hub.py:21-22 - Fixed floor division and missing return for billion values.
Used `//` instead of `/` for millions, and had no return for values >= 1B.
Caused incorrect formatting and None return for large numbers.
7. save.py:550 - Fixed self-assignment that did nothing.
`sharded_ram_usage = sharded_ram_usage` should be `= max_shard_size`.
Caused integer shard sizes to be ignored.
8. rl.py:562-567 - Fixed orphan string not included in length_check.
The elif branch for max_seq_length validation was a standalone string
expression, not concatenated to length_check. Caused silent skip of
the max_seq_length > model_max_seq_length warning.
9. granite.py:49-52 - Fixed wrong model name and version in error message.
Said "Gemma2" and "4.42.3" but should be "Granite" and "4.45.0".
* Fix correctness bugs in rl.py, rl_replacements.py, and vision.py
1. rl_replacements.py (lines 864, 870): Fixed undefined `nanmin`/`nanmax`
functions by using `.nan_to_num(nan=inf/-inf).min()/.max()` pattern.
PyTorch doesn't have torch.nanmin/nanmax, so we replace NaN values
before computing min/max.
2. vision.py (line 150): Fixed bug where code checked for "input" key
but then accessed kwargs["input_ids"] instead of kwargs["input"].
3. vision.py (line 159): Fixed bug where literal string "key" was used
instead of the variable `key` when accessing kwargs.
4. rl.py (lines 903, 905): Fixed non-existent `MathError` exception
by replacing with `ValueError`.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Add "corda" as an allowed value for the init_lora_weights parameter
in FastLanguageModel.get_peft_model() and FastBaseModel.get_peft_model().
This enables users to use CorDA (Correlation-aware Decomposed Adaptation)
initialization from PEFT, which provides an alternative LoRA initialization
strategy for improved finetuning performance.
Fixes#3693
Signed-off-by: majiayu000 <1835304752@qq.com>
* Fix is_contiguous() method call and remove duplicate imports
- Fix bug in rope_embedding.py where is_contiguous was used without
parentheses, causing the method object (always truthy) to be evaluated
instead of calling the method. This fixes issue #3781 where fast rope
backpropagation was broken for zero strided/non-contiguous tensors.
- Remove duplicate `import torch` in rl.py (lines 20 and 25)
- Remove duplicate `import functools` and `import types` in vision.py
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Fix Boolean value of Tensor ambiguity error in mistral.py
Replace `or` operator with explicit `is None` check when getting
n_items from kwargs. The `or` operator fails when the value is a
Tensor because Python cannot determine the boolean value of a
multi-element tensor.
Fixes#3766🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Update rope_embedding.py
---------
Co-authored-by: yurekami <yurekami@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
* Guard optional trl.experimental.openenv usage in RL patches
* Simplify optional trl.openenv import handling
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* fix(trainer): import psutil to prevent NameError in _prepare_dataset
Fixes#3777
* Update rl.py
---------
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
