Enable concurrency by default #4218
Merged
Conversation
This file contains 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
dhiltgen
force-pushed
the
auto_parallel
branch
2 times, most recently
from
f846c36
to
3913acf
Compare
dhiltgen
force-pushed
the
auto_parallel
branch
3 times, most recently
from
b84f3cc
to
f4742e1
Compare
dhiltgen
force-pushed
the
auto_parallel
branch
2 times, most recently
from
085ae40
to
b1e7a74
Compare
This adjusts our default settings to enable multiple models and parallel requests to a single model. Users can still override these by the same env var settings as before. Parallel has a direct impact on num_ctx, which in turn can have a significant impact on small VRAM GPUs so this change also refines the algorithm so that when parallel is not explicitly set by the user, we try to find a reasonable default that fits the model on their GPU(s). As before, multiple models will only load concurrently if they fully fit in VRAM.
Until ROCm v6.2 ships, we wont be able to get accurate free memory reporting on windows, which makes automatic concurrency too risky. Users can still opt-in but will need to pay attention to model sizes otherwise they may thrash/page VRAM or cause OOM crashes. All other platforms and GPUs have accurate VRAM reporting wired up now, so we can turn on concurrency by default.
Provide consistent ordering for the ps command - longest duration listed first
jmorganca
reviewed
Jun 29, 2024
| @@ -85,13 +85,13 @@ func AsMap() map[string]EnvVar { | |||
| "OLLAMA_HOST": {"OLLAMA_HOST", Host, "IP Address for the ollama server (default 127.0.0.1:11434)"}, | |||
| "OLLAMA_KEEP_ALIVE": {"OLLAMA_KEEP_ALIVE", KeepAlive, "The duration that models stay loaded in memory (default \"5m\")"}, | |||
| "OLLAMA_LLM_LIBRARY": {"OLLAMA_LLM_LIBRARY", LLMLibrary, "Set LLM library to bypass autodetection"}, | |||
| "OLLAMA_MAX_LOADED_MODELS": {"OLLAMA_MAX_LOADED_MODELS", MaxRunners, "Maximum number of loaded models (default 1)"}, | |||
| "OLLAMA_MAX_LOADED_MODELS": {"OLLAMA_MAX_LOADED_MODELS", MaxRunners, "Maximum number of loaded models per GPU (default auto)"}, | |||
There was a problem hiding this comment.
nit: Is the default 0? might be ok to put that, so folks don't do OLLAMA_MAX_LOADED_MODELS=auto
jmorganca
reviewed
Jun 29, 2024
| @@ -23,6 +23,7 @@ type LlmRequest struct { | |||
| ctx context.Context //nolint:containedctx | |||
| model *Model | |||
| opts api.Options | |||
| origNumCTX int // Track the initial ctx request | |||
There was a problem hiding this comment.
I think we case it as NumCtx in other parts of the codebase
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This adjusts our default settings to enable multiple models and parallel requests to a single model. Users can still override these by the same env var settings as before. Parallel has a direct impact on num_ctx, which in turn can have a significant impact on small VRAM GPUs so this change also refines the algorithm so that when parallel is not explicitly set by the user, we try to find a reasonable default that fits the model on their GPU(s). As before, multiple models will only load concurrently if they fully fit in VRAM.
Corresponding Doc update to merge after this #5364