AI-Hypercomputer / JetStream

JetStream is a throughput and memory optimized engine for LLM inference on XLA devices, starting with TPUs (and GPUs in future -- PRs welcome).

Repository from Github https://github.comAI-Hypercomputer/JetStream

AI-Hypercomputer/JetStream Issues

Does Dataflow work with JetStream?
Updated 3 months ago1
Clean up Model Conversion Script
Updated 3 months ago2
Understanding the intuition behind `request-rate`
Updated 3 months ago
Support completions API
Updated 3 months ago
when to support gpu?
Updated 3 months ago2
Support using models from HuggingFace directly
Updated 5 months ago2
Question: `prometheus_port` flag for pytorch server
Updated 5 months ago
Try Google Opinion Rewards باربری یزد 09137236592
Closed 5 months ago
باربری یزد 09133545880
Closed 5 months ago
باربری یزد ۰۹۱۳۳۵۴۵۸۸۰
Closed 5 months ago
Ads.google.com
Closed 5 months ago
باربری نیسان یزد 09133545880
Closed 5 months ago
Remove jax dependencies in JetStream
Updated 10 months ago
Add np padding support
Closed 10 months ago1
Support I/O with text and token ids
Closed 10 months ago2
Refactor jestream to allow different tokenizers
Updated a year ago1
Detokenize error
Closed a year ago2
Benchmark serving: Failed to connect to remote host
Closed a year ago1
float division by zero in benchmark
Updated a year ago2
Support on Huggingface transformers
Closed a year ago2
Error with mutable list value in dataclass
Closed a year ago1
CogVLM support
Closed a year ago1
Feature request: improve documentation
Closed a year ago5