Azure99/blossom-v5-32b

TEXT GENERATIONPricing:Input $1.06 / Cached $0.053 / Output $2.6Concurrent Unit Cost:2Model Size:32.5BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Apr 29, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Azure99/blossom-v5-32b is a 32.5 billion parameter conversational large language model, fine-tuned on a Qwen1.5-32B base. Developed by Azure99, this model excels in general conversational capabilities and context comprehension, leveraging high-quality, GPT-4 distilled data. It is specifically optimized for multi-turn dialogue and instruction-following tasks, making it suitable for advanced chatbot applications.

Loading preview...

Azure99/blossom-v5-32b: Conversational LLM

Azure99's Blossom-v5-32b is a 32.5 billion parameter conversational large language model built upon the Qwen1.5-32B architecture. This iteration, part of the Blossom V5 series, features significant improvements derived from training on high-quality data distilled from gpt-4-0125-preview.

Key Capabilities & Training

  • Enhanced Conversational Abilities: Fine-tuned on a mixed dataset including Blossom Orca, Wizard, Chat, and Math data, it demonstrates robust general capabilities and strong context comprehension.
  • High-Quality Data: Utilizes open-source, high-quality Chinese and English datasets for training.
  • Two-Stage Fine-tuning:
    • Stage 1: Trained for 1 epoch on 40K Wizard, 40K Orca, and 10K Math single-turn instruction datasets.
    • Stage 2: Trained for 3 epochs on 10K Blossom chat multi-turn dialogue dataset, combined with 10% randomly sampled data from Stage 1.

Use Cases

This model is particularly well-suited for dialogue continuation and conversational AI applications, including chatbots and interactive assistants, where understanding multi-turn context is crucial. Its training methodology emphasizes both instruction-following and multi-turn dialogue, making it versatile for various conversational scenarios.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
–
top_p
–
top_k
–
frequency_penalty
–
presence_penalty
–
repetition_penalty
–
min_p
–