habanoz/TinyLlama-1.1B-2T-lr-2e-4-3ep-dolly-15k-instruct-v1
LoRA-style vs TinyLlama/TinyLlama-1.1B-intermediate-step-955k-token-2T. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 22 of 22 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/45, MLP 66/66, Attention 88/88
1 derivatives checked
- afrideva/TinyLlama-1.1B-2T-lr-2e-4-3ep-dolly-15k-instruct-v1-GGUF: Same weights, converted