From 5336fa51fb340226ca5b096c681000fefab60f2f Mon Sep 17 00:00:00 2001
From: sroy745 <142070531+sroy745@users.noreply.github.com>
Date: Thu, 5 Sep 2024 13:25:29 -0700
Subject: [PATCH] [Documentation][Spec Decode] Add documentation about lossless
 guarantees in Speculative Decoding in vLLM (#7962)

---
 docs/source/models/spec_decode.rst | 40 ++++++++++++++++++++++++++++++
 docs/source/serving/faq.rst        | 19 ++++++++++++++
 2 files changed, 59 insertions(+)

diff --git a/docs/source/models/spec_decode.rst b/docs/source/models/spec_decode.rst
index d3c196faff25d..50468f25b922a 100644
--- a/docs/source/models/spec_decode.rst
+++ b/docs/source/models/spec_decode.rst
@@ -161,6 +161,46 @@ A variety of speculative models of this type are available on HF hub:
 * `granite-7b-instruct-accelerator <https://huggingface.co/ibm-granite/granite-7b-instruct-accelerator>`_
 * `granite-20b-code-instruct-accelerator <https://huggingface.co/ibm-granite/granite-20b-code-instruct-accelerator>`_
 
+Lossless guarantees of Speculative Decoding
+-------------------------------------------
+In vLLM, speculative decoding aims to enhance inference efficiency while maintaining accuracy. This section addresses the lossless guarantees of 
+speculative decoding, breaking down the guarantees into three key areas:
+
+1. **Theoretical Losslessness**
+   - Speculative decoding sampling is theoretically lossless up to the precision limits of hardware numerics. Floating-point errors might 
+   cause slight variations in output distributions, as discussed 
+   in `Accelerating Large Language Model Decoding with Speculative Sampling <https://arxiv.org/pdf/2302.01318>`_
+
+2. **Algorithmic Losslessness**
+   - vLLM’s implementation of speculative decoding is algorithmically validated to be lossless. Key validation tests include:
+
+    - **Rejection Sampler Convergence**: Ensures that samples from vLLM’s rejection sampler align with the target 
+      distribution. `View Test Code <https://github.com/vllm-project/vllm/blob/47b65a550866c7ffbd076ecb74106714838ce7da/tests/samplers/test_rejection_sampler.py#L252>`_
+
+    - **Greedy Sampling Equality**: Confirms that greedy sampling with speculative decoding matches greedy sampling
+      without it. This verifies that vLLM's speculative decoding framework, when integrated with the vLLM forward pass and the vLLM rejection sampler, 
+      provides a lossless guarantee.  Almost all of the tests in `this directory <https://github.com/vllm-project/vllm/tree/b67ae00cdbbe1a58ffc8ff170f0c8d79044a684a/tests/spec_decode/e2e>`_
+      verify this property using `this assertion implementation <https://github.com/vllm-project/vllm/blob/b67ae00cdbbe1a58ffc8ff170f0c8d79044a684a/tests/spec_decode/e2e/conftest.py#L291>`_
+
+3. **vLLM Logprob Stability**
+   - vLLM does not currently guarantee stable token log probabilities (logprobs). This can result in different outputs for the 
+   same request across runs. For more details, see the FAQ section 
+   titled *Can the output of a prompt vary across runs in vLLM?* in the `FAQs <../serving/faq.rst>`_.
+
+
+**Conclusion**
+
+While vLLM strives to ensure losslessness in speculative decoding, variations in generated outputs with and without speculative decoding 
+can occur due to following factors:
+
+- **Floating-Point Precision**: Differences in hardware numerical precision may lead to slight discrepancies in the output distribution.
+
+- **Batch Size and Numerical Stability**: Changes in batch size may cause variations in logprobs and output probabilities, potentially 
+  due to non-deterministic behavior in batched operations or numerical instability.
+
+**Mitigation Strategies**
+
+For mitigation strategies, please refer to the FAQ entry *Can the output of a prompt vary across runs in vLLM?* in the `FAQs <../serving/faq.rst>`_.
 
 Resources for vLLM contributors
 -------------------------------
diff --git a/docs/source/serving/faq.rst b/docs/source/serving/faq.rst
index 7b0374be8adff..9e858e612c8bf 100644
--- a/docs/source/serving/faq.rst
+++ b/docs/source/serving/faq.rst
@@ -10,3 +10,22 @@ A: Assuming that you're referring to using OpenAI compatible server to serve mul
     Q: Which model to use for offline inference embedding?
 
 A: If you want to use an embedding model, try: https://huggingface.co/intfloat/e5-mistral-7b-instruct. Instead models, such as Llama-3-8b, Mistral-7B-Instruct-v0.3, are generation models rather than an embedding model
+
+----------------------------------------
+
+    Q: Can the output of a prompt vary across runs in vLLM?
+
+A: Yes, it can. vLLM does not guarantee stable log probabilities (logprobs) for the output tokens. Variations in logprobs may occur due to
+numerical instability in Torch operations or non-deterministic behavior in batched Torch operations when batching changes. For more details, 
+see the `Numerical Accuracy section <https://pytorch.org/docs/stable/notes/numerical_accuracy.html#batched-computations-or-slice-computations>`_.
+
+In vLLM, the same requests might be batched differently due to factors such as other concurrent requests,
+changes in batch size, or batch expansion in speculative decoding. These batching variations, combined with numerical instability of Torch operations, 
+can lead to slightly different logit/logprob values at each step. Such differences can accumulate, potentially resulting in 
+different tokens being sampled. Once a different token is sampled, further divergence is likely.
+
+**Mitigation Strategies**
+
+- For improved stability and reduced variance, use `float32`. Note that this will require more memory.
+- If using `bfloat16`, switching to `float16` can also help.
+- Using request seeds can aid in achieving more stable generation for temperature > 0, but discrepancies due to precision differences may still occur.