()

()

()

()

()

()

(data, dims, ops)

()

()

()

()

()

(device)

()

(params, weightsFolder)

(params, weightsFolder)

(params, weightsFolder, layerIndex)

(params, weightsFolder)

(folder)

(folder)

(weightsFolder)

(params, weightsFolder)

(model_type)

(model_type, folder_name)

(A, B, C, M, N, K)

( seq_length, n_embd, attentionDotProductScale, n_head, head_size, inputBuffer, qW

( seq_length, n_embd, attentionDotProductScale, n_head, head_size, inputBuffer, qW

(rows, cols, shared, bufA, bufB)

(row, col, inputBuffer)

(flatArray, shape)

()

(idx)

(idx)

( device, queue, seq_length, vocab_size, n_embd, n_heads, n_layers, attentionDotProductScale,

(probs)

(model, folder_name)

(checkpoint, folder_name)

(checkpoint, folder_name)

(probs, top_k)

(probs, top_k)

()

()

()

()

()

(types)

(types)

(types)

(types)

(folder, type)

()

()

()

()

()

()

()

()

()

()

()

()

(model)

()

()

()

()

()

()

()

()

()

()

()

()

(folder, type)

(device)

(device)

(folder, type, doAttentionCache = false)

()

()

()

()

(folder, type)

(arr)

(tokens)

(arr)

(tokens)

()

()

(str)

(text)

(str)

(text)

(floatArray, dimA, dimB)

(floatArray, dimA, dimB)

(prompt, max_new_tokens, top_k, temperature)

(prompt, max_new_tokens, top_k, temperature)

()

()