ModelHub XC 944ebb49a9 初始化项目,由ModelHub XC社区提供模型
Model: mhseo2/Qwen3-8B-Base-OT3Math-SFT-step250
Source: Original Platform
2026-11-13 08:34:18 +08:00

license, base_model, datasets, tags, library_name
license base_model datasets tags library_name
apache-2.0 Qwen/Qwen3-8B-Base
knowledge-distillation/openthoughts3_math
math
reasoning
sft
transformers

Qwen3-8B-Base -- OpenThoughts3 math SFT, step 250

Intermediate checkpoint (step 250 of a planned 2000), bf16 weights only. Not yet evaluated.

base Qwen/Qwen3-8B-Base
data knowledge-distillation/openthoughts3_math, 103,760 two-turn conversations
sequence length 32768 (truncates nothing: mean 13.8k, p99 16.8k, max 16.8k tokens)
global batch 255
optimizer AdamW, lr 5e-6, cosine, 3% warmup, weight decay 0.01, grad clip 1.0
precision bf16, FSDP over 3 GPUs, verl

Every assistant message in the data is a <think>...</think> trace ending in \boxed{}, and the model is trained to reproduce that. It uses a plain ChatML template with no thinking logic -- Qwen3's own template strips <think> from all but the last assistant turn, which would have removed the reasoning from the training targets. No enable_thinking flag applies.

A turn ends with <|im_end|> (151645); config.json and generation_config.json both say so. Qwen3-8B-Base's stock generation_config.json says 151643 instead, which yields a model that never terminates under vLLM. Generate with max_new_tokens around 16384.

S
Description
Model synced from source: mhseo2/Qwen3-8B-Base-OT3Math-SFT-step250
Readme
13 MiB
0 Stars 1 Watchers 0 Forks
Languages
Jinja 100%