[WIP] Add bitwise parity test for MoE EP#3172
Closed
wwwjn wants to merge 9 commits into
Closed
Conversation
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
This was referenced Apr 30, 2026
meta-cla
Bot
added
the
CLA Signed
wwwjn
changed the title
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 1, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: 65e6b7a
Pull Request resolved: #3172
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 1, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: 4f3aef7
Pull Request resolved: #3172
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 1, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: 99cccd6
Pull Request resolved: #3172
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 4, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: 6cdd70a
Pull Request resolved: #3172
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 4, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: d4e678c
Pull Request resolved: #3172
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 4, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: 1c0e85c
Pull Request resolved: #3172
wwwjn
added a commit
that referenced
this pull request
May 5, 2026
…cher (#3193) Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #3172 * #3171 * #3142 * __->__ #3193 Today the dispatcher's _split_along_sp() raises when num_tokens (bs * slen) is not divisible by sp_size (= TP degree). Real workloads with varlen prompts can land on non-divisible totals and crash the MoE forward. Pad inside dispatch(): round num_tokens up to the next multiple of sp_size, padding x and top_scores with zeros and selected_experts_indices with 0 (so pad rows route deterministically to expert 0 with zero score). combine() reads metadata.original_num_tokens to size the scatter_add buffer at the padded length and slices the pad rows off before returning. When sp_size == 1 or input is already divisible, behavior is bitwise identical to today. Pad tokens are numerically inert: - Zero scores -> contribution to scatter_add is exactly zero either before or after expert compute (independent of score_before_experts). - Pad indices fall in [original, padded), which is sliced off after scatter_add, so they never appear in the returned output. Trainer/generator can pad by different amounts depending on their batch shapes; the unpadded portions remain bitwise identical. - TorchAOTokenDispatcher inherits dispatch/combine and gets the fix for free. - DeepEPTokenDispatcher uses a separate metadata type and is unaffected.
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
[ghstack-poisoned]
wwwjn
added a commit
that referenced
this pull request
May 5, 2026
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \\
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.
ghstack-source-id: 9b5e3e0
Pull Request resolved: #3172
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
Add rl_grpo_qwen3_moe_debug_ep_batch_invariant config and
TestBitwiseParityMoEEP test class for verifying bitwise identity
between trainer prefill and vLLM generator prefill with MoE EP.
Config: TP=4, EP=4 on 4 GPUs, deterministic mode, compile disabled.
Uses /tmp/debug_moe_ckpt by default (override with MOE_HF_ASSETS_PATH).
Run with:
python scripts/rl/create_debug_moe_ckpt.py
torchrun --nproc_per_node=4 -m pytest \
torchtitan/experiments/rl/tests/test_bitwise_parity.py::TestBitwiseParityMoEEP
Note: bitwise parity not yet achieved — current max_delta ~2e-2.
Likely causes: batch-invariant mode disabled (requires batch_invariant_ops),
attention impl differences (varlen vs paged), MoE token routing order.