How to use ayushexel/xynft-embeddinggemma-300m-8e with sentence-transformers:
from sentence_transformers import SentenceTransformer model = SentenceTransformer("ayushexel/xynft-embeddinggemma-300m-8e") sentences = [ "What were the details of the payment gateway latency spike incident, including its impact, the immediate mitigation steps taken, and the plan for root cause analysis?", "Anjali Patel is addressing the HDFC reconciliation timeout error through a three-pronged approach: 1) Technical Investigation: She has started a BigQuery query to analyze if the error is linked to specific card BIN ranges or merchant categories. 2) Partner Communication: She will proactively contact HDFC's nodal officer. 3) Internal Tracking & Compliance: She is creating a JIRA ticket under the R&C project to track the issue and specifically document its potential compliance impact for the next RBI audit cycle.", "A significant increase in transaction timeouts was identified for the ICICI payment gateway, with the rate jumping from a baseline of 0.5% to nearly 4% during the 7-9 PM peak period. Initial analysis of Grafana dashboards confirmed the spike, while Kibana logs revealed only generic gateway timeouts, not specific bank error codes. As a next step, the Infrastructure team has been asked to investigate for network latency or packet loss on Juspay's side. The issue is being tracked under JIRA ticket OPS-2451 and is scheduled for discussion in the morning standup.", "At 14:15, the payment gateway experienced a severe latency spike where p99 latency increased from ~180ms to over 2.5s. This directly caused a 20% increase in transaction timeout errors. The root cause appears to be Redis connection pool exhaustion errors. As an immediate response, a full rollback to the previous stable version (v1.18.1) was initiated at 14:25 IST, with service restoration expected by 14:40 IST. For long-term resolution, a root cause analysis is being tracked under JIRA ticket PROD-4891, with an initial hypothesis that a change in the Redis client configuration is to blame. Further updates will be posted in the #incidents-xyne-core Slack channel, and a preliminary RCA document is expected by the end of the day." ] embeddings = model.encode(sentences) similarities = model.similarity(embeddings, embeddings) print(similarities.shape) # [4, 4]
